Mediclinic Southern Africa (Pty) Ltd and Another v Competition Commission (172/CAC/Feb19) [2020] ZACAC 3; [2020] 1 CPLR 66 (CAC) (6 February 2020)

THE COMPETITION APPEAL COURT OF SOUTH AFRICA

JUDGMENT

Case No: 172/CAC/Feb19 In the matter between MEDICLINIC SOUTHERN AFRICA (PTY) LTD FIRST APPELLANT MATLOSANA MEDICAL HEALTH SERVICES (PTY) LTD SECOND APPELLANT And THE COMPETITION COMMISSION RESPONDENT

Coram: Rogers, Victor and Vally JJA

Heard: 14 & 15 October 2019

Delivered: 6 February 2020

[Non-confidential version]

Rogers JA (Victor JA concurring)

Introduction

[1] The appellants appeal against the Tribunal’s prohibition of their large merger. The first appellant is Mediclinic Southern Africa (Pty) Ltd (‘Mediclinic’), the second appellant Matlosana Medical Health Services (Pty) Ltd (‘Matlosana’). In terms of the merger transaction Mediclinic will gain control of Matlosana. The respondent, the Competition Commission (‘Commission’),

recommended that the merger be prohibited and now resists the appeal.

[2] Mediclinic owns a multidisciplinary hospital in Potchefstroom (‘MC Potch’). MC Potch is one of 50 hospitals which Mediclinic owns in South Africa. Matlosana owns two multidisciplinary hospitals in Klerksdorp called Wilmed and Sunningdale and a psychiatric hospital called Parkmed.

[3] Potchefstroom and Klerksdorp, both of which are in the North West Province (‘NWP’), are just under 50 km apart (the

travelling time is 41 minutes), Potchefstroom lying to the east of Klerksdorp.

[4] It was and is common cause that Parkmed’s services are not in the same product market as those provided by the three multidisciplinary hospitals mentioned above, and that the acquisition by Mediclinic of control over Parkmed does not raise any competition

or public interest concerns. The contentious issues concern Mediclinic’s acquisition of control over Wilmed and Sunningdale, to which I shall refer collectively as the targets. All references hereafter to hospitals are to multidisciplinary hospitals unless otherwise stated.

Background

[5] There are three large corporate hospital groups in South Africa: Netcare, Life Healthcare (‘Life’) and Mediclinic. Many independent hospitals are affiliated to the National Health Network (‘NHN’), a non-profit company. Historically NHN has been permitted, by way of an exemption granted in terms of s 10 of the Competition Act 89 of 1998, to negotiates tariffs and other benefits with medical schemes on behalf of its affiliated hospitals. In November 2018 this exemption was expanded to include procurement on behalf of affiliated hospitals. (I shall refer to this part of the exemption as the procurement exemption.) Matlosana’s Klerksdorp hospitals form part of NHN. Nationally, the numbers of hospitals and beds operated by these four groups, and by unaffiliated independents, are as follows (the national market share, by number of beds, is given in brackets):

· Netcare: 54 hospitals/10 004 beds (24,9%);

· NHN: 62 hospitals/6611 beds (24,7%);

· Life: 57 hospitals/7987 beds (21,3%);

· Mediclinic: 50 hospitals/7164 beds (20,3%).

· Unaffiliated: 3065 beds (8,8%).

· Total beds: 34 831.

[6] The targets have 247 beds. If they are acquired by Mediclinic, the latter’s national market share by beds will increase by about 0,7%.

[7] Apart from MC Potch, there is one other multidisciplinary hospital in Potchefstroom, MooiMed, which forms part of NHN. The drive-time

between MC Potch and MooiMed is five minutes. Apart from the targets, there is one other multidisciplinary hospital in Klerksdorp, Life Anncron (‘Anncron’), which belongs to Life. The drive-time between Wilmed and Anncron is five minutes.

[8] A hospital bill (‘cost per event” – CPE) comprises three components:

(a) Fees for ward and theatre time according to the hospital’s tariff. Hospitals’ tariffs differ. Although one hospital may have a higher tariff than a competitor, the tariff component of the former’s bill might in comparable cases be lower than the latter’s because of more efficient time management (ie less time spent in theatres and wards).

(b) Ethicals (drugs) supplied. Ethicals are by legislation subject to a single exit pricing (‘SEP’) regime, the effect of which is that the same drug supplied by two competing hospitals will appear in their respective bills at an identical unit cost. In comparable cases one hospital’s bill might nevertheless have a lower ethicals component (and thus be more efficient) than another’s because

the same drugs are used in lower but adequate quantities or because satisfactory cheaper substitutes (generic as against patented

versions) are used.

(c) Surgicals consumed or supplied. Surgicals (which includes prostheses) are not subject to the SEP regime. One hospital may be able to bill surgicals at a lower cost than a competitor because it has procured the same surgicals at lower prices (superior bargaining with suppliers), or because the same surgicals are used in lower quantities, or because satisfactory cheaper alternatives are used.

[9] In the case of the targets, their CPE on an average weighted basis comprises a tariff component of 68%, an ethicals component of 11% and a surgicals component of 21%. CPE at other hospitals is broadly similar but there would naturally be differences.

[10] Although traditionally a hospital bill is made up of numerous line items for each service and item supplied (the fee-for-service (‘FFS’) model), medical schemes and hospitals sometimes negotiate alternative reimbursement models (‘ARMs’), a typical example of which would be a fixed fee for a certain type of procedure. An ARM allows the scheme to share some of the cost risk with the hospital, since if in a particular case the procedure turns out – perhaps because of complications – to be unusually expensive, the hospital still only gets the ARM fee. From the hospital group’s perspective, it will want to be reasonably confident that, averaged out over all the procedures covered by the ARM, the fees payable in terms thereof will be profitable, even if some of the procedures turn out to cost more than the ARM fee.

[11] More than 95% of patients who receive services from private hospitals are insured by medical schemes. In respect of this group, price negotiation occurs not between hospital and patient but between hospital and scheme. The hospital (or hospital group) typically

negotiates its tariffs annually with each scheme. Where a scheme option permits a patient a choice between two or more hospitals,

quality of care and patient experience may play a role in the patient’s choice.

[12] For the fewer than 5% of patients who are uninsured, each hospital has its own tariff, with hospital managers having more or less discretion to grant discounts. In relation to these patients, therefore, hospitals compete not only on quality of care and patient experience but on price.

[13] Most schemes are national, ie have members throughout South Africa. Each scheme typically offers a variety of benefit options. For richer options, members are not restricted in their choice of hospital. So in an area served by two or more hospitals, the member can choose any one of them without financial prejudice.

[14] For low-cost options, however, a scheme constructs a network of designated or preferred service providers (although there is a distinction, for present purposes I can refer to them collectively as DSPs). In order to be covered or avoid co-payment, the member must go to a DSP. Since usually only one of the several hospitals serving the area where the patient resides is a DSP, the patient has no choice.

[15] Typically a DSP network comprises one or two anchor groups out of the four large groups (Netcare, Life, Mediclinic, NHN) with filler hospitals, drawn from the other groups and from unaffiliated hospitals, to provide coverage in areas not served by the anchor(s).

Hospital groups compete with each other to be included as anchors. The attraction to a hospital group of inclusion is the increased

volumes arising from the fact that scheme members must use that group’s hospitals in order to enjoy cover. The trade-off is that the scheme will expect the group to offer a tariff discount. It is this discount which enables the scheme to offer a low-cost option to its members. Network negotiations, like tariff negotiations, occur annually. Even with discounted tariffs, however, low-cost options are generally not independently sustainable; there is some element of subsidisation from richer schemes.

[16] Since schemes compete with each other for members, a scheme would usually want to offer its low-cost members reasonably accessible (ie not too distant) DSP hospitals. Accessibility from the scheme’s perspective is also important for those services falling within the legislatively defined range of ‘prescribed minimum benefits’ (PMBs), because where a member needs treatment constituting a PMB, the scheme must cover the cost in full unless the member fails to use a reasonably accessible DSP. For this purpose, the Council for Medical Schemes (‘CMS’) regards 50 km as the outer limit of reasonable accessibility. This means that if the scheme does not have a DSP within 50 km of the member’s residence, the scheme will need to cover the full cost of a member’s use of a non-DSP hospital within the 50 km radius, even though the scheme has no negotiated discounted tariff with the non-DSP hospital.

[17] Low-cost options are a small part of medical scheme business. In the case of Bonitas, one of the large national schemes which received some attention in the Tribunal’s hearing, 66 000 (9,4%) of its 700 000 members are on its three low-cost options. This is typical of other large schemes.

[18] Because of the features summarised above, it has been recognised in previous cases concerning hospital mergers that the geographic market for hospital services has a national and a local dimension:

(a) In the provision of services to insured patients, there is a national market in which hospitals compete with each other in their tariff and network negotiations with schemes. For patients on low-cost options, there is typically no local competition for patients since they are confined to a single DSP hospital. For richer options, where members have a choice of hospitals, local competition is limited to quality of care and patient experience. There is no local pricing competition.

(b) In the provision of services to uninsured patients, the market is wholly local, proximate hospitals competing with each other for patients on price, quality of care and patient experience.

(See Phodiclinics (Pty) Ltd & others and Protector Group Medical Services (Pty) Ltd & others (122/LM/Dec05) [2007] ZACT 17 paras 27-29 and 117; Life Healthcare Group (Pty) Ltd v Amabubesi Hospitals (Pty) Ltd & another (11/LM/Mar10) [2010] ZACT 40 para 5.)

[19] It does not necessarily follow, from the preceding paragraph, that, in the case of the provision of services to insured patients, a change in the local landscape may not have national pricing effects. One of the issues in the present case is whether the increased

presence which Mediclinic would enjoy in Potchefstroom and Klerksdorp as a result of the merger would make it a practical necessity for schemes to include Mediclinic as the anchor, or one of the anchors, in their DSP networks in order to provide members with a reasonably accessible hospital in that area. Regional dominance giving rise to this practical necessity would eliminate or reduce Mediclinic’s incentive to offer discounts as a quid pro quo for its inclusion in networks. This would result in higher tariffs for low-discount options and higher premiums for low-cost members.

Tribunal’s decision

[20] The Tribunal, to whose careful and comprehensive reasons I pay tribute, made the following key findings.

[21] First, that the product market is the market for private multidisciplinary acute inpatient hospital services. This definition excluded single-discipline hospitals, day hospitals and day-case services provided by multidisciplinary hospitals. (The appellants dispute the exclusion of day cases.)

[22] Second, that the geographic market comprised the combined municipal areas of Matlosana and JB Marks,[1] an area which the parties and the Tribunal called MaJB. The largest towns in the these two municipalities are Klerksdorp and Potchefstroom respectively. (The appellants dispute that Klerksdorp and Potchefstroom are in the same geographic market. In what follows, when I refer to Klerksdorp I mean the greater Klerksdorp area, ie the City of Matlosana, which encompasses Klerksdorp, Stilfontein, Orkney

and Hartbeesfontein.

[23] Third, that the hospitals providing the defined services in MaJB, and their market shares, are MC Potch (31%), MooiMed (13%), Wilmed and Sunningdale (32% combined) and Anncron (24%). The merger will result in Mediclinic’s market share in MaJB rising from

31% to 63%. (If the Tribunal’s definition of the services and geographic markets are right, these percentages are not in issue.)

[24] Fourth, that the tariff component of the target hospitals’ bills for medical schemes (ie for insured patients) will, upon implementation of the merger, immediately increase by […]%. This is because Mediclinic will immediately implement its own scheme tariffs at the targets, and its scheme tariffs are on average […]% higher than the targets’ scheme tariffs. Assuming no change in the targets’ current efficiencies in ward and theatre times and in their use of ethicals and surgicals, this will cause the targets’ CPE to rise by […]%, given that the tariff component makes up […]% of the targets’ CPE ( […]). (If the assumption of unchanged efficiency is sound, which the appellants dispute, this calculation of the increase in the targets’ CPE is common cause.)

[25] Fifth, that the tariff component of the target hospitals’ bills for uninsured patients will rise substantially, because (a) the

targets’ tariffs for uninsured patients are […]%-[…]% lower than Mediclinic’s, and (b) the targets’

hospital managers currently have the power to grant greater discretionary discounts than Mediclinic hospital managers – up

to […]% as against […]%. Although the arithmetic was not done by the Tribunal, this would mean – again assuming no change in the targets’ current efficiencies – that CPE for uninsured patients would rise by […]% – […]%. before taking into account discretionary discounts which might amplify the increase. (The appellants again dispute the assumption of unchanged efficiency.)

[26] Sixth, that reliance could not be placed on actuarial evidence which purported to show, by way of comparison between selected Mediclinic hospitals and the target hospitals, that under Mediclinic’s control the targets were likely to achieve efficiencies in ward and theatre times, in the use of ethicals, and in the procurement and use of surgicals. Both sides presented actuarial evidence. The appellants’ evidence was (a) that the effect of its higher tariffs on the targets’ CPE would be significantly counteracted by more efficient ward and theatre times, reducing the effect of tariff increases on the targets’ CPE from […]% to […]%; (b) that, through Mediclinic’s more efficient management of ethicals use, there would be a […]%

decrease in the ethicals component of the targets’ CPE, which would reduce the targets’ CPE by […]% ([…]); and (c) that, through Mediclinic’s more efficient procurement and use of surgicals, there would be a […]% decrease in the surgicals component of the targets’ CPE, which would reduce the targets’ overall CPE by […]% ([…]). Taking all these efficiencies into account, the targets’ CPE for schemes would, under Mediclinic’s control, be […]% lower than it currently is. Even if one disregarded all the efficiencies in surgicals, the efficiencies in ethicals would be sufficient to neutralise the efficiency-adjusted effect of the tariff increase. (The Tribunal’s rejection of the actuarial evidence, and of the factual evidence which underpinned it, is hotly contested.)

[27] Seventh, that even if were accepted that Mediclinic hospitals are currently more efficient than the targets in their use of ethicals, a merger is not necessary to enable these efficiencies to be achieved; and that, in particular, the targets could improve their use of cheaper generics without having to be under Mediclinic’s control. (The appellants dispute this conclusion.)

[28] Eighth, that although Mediclinic’s size has hitherto enabled it to procure surgicals at lower prices than Matlosana, the procurement exemption will enable NHN-affiliated hospitals to achieve the same procurement efficiencies, given that NHN has a larger national market share. In this regard, the procurement exemption only applies to NHN-affiliated hospitals which are ‘small businesses’ or ‘firms owned or controlled by historically disadvantaged persons’, but there is a two-year grace period during which non-compliant hospitals may benefit from the procurement exemption. The targets are not (or were not, when the Tribunal made its decision) compliant. The Tribunal felt unable to predict whether Matlosana would become compliant, and thus confined its finding of improved procurement efficiencies for the targets to the two-year grace period (1 November 2018-31 October 2020). (The appellants

dispute that NHN will be able, within the two-year period, to achieve equivalent procurement efficiencies. They say the targets are unlikely to achieve more than half of Mediclinic’s current procurement efficiency.)

[29] Ninth, that a merger implies that the merged firms’ pricing decisions will be coordinated to maximise profit, that the present merger will result in a highly concentrated market in MaJB, that barriers to entry are high, and that the transaction will substantially prevent or lessen competition in the relevant market. Although the Tribunal did not expressly state that the likely price increases at the targets would be a reflection of a substantial lessening of competition (‘SLC’), this seems to have been its view. (Apart from disputing that prices will increase rather than decrease, the appellants challenge the view that any price increases

which eventuate will be the consequence of an SLC.)

[30] Tenth, that although the evidence was limited, the targets seemed to be performing better than MC Potch, and that the merger was likely to lead to a deterioration in non-price competition as reflected in patient experience. (The appellants dispute this.)

[31] Eleventh, that, in addition to resultant adverse pricing and quality effects at the target hospitals, the merger would give Mediclinic regional dominance in MaJB which would make it difficult for schemes to exclude Mediclinic’s hospitals in Potchefstroom and Klerksdorp in their DSP networks; that Mediclinic could use this regional dominance to minimise discounts and/or to compel inclusion of its hospitals in areas where it faces more competition, to the prejudice of competing hospitals in those areas; and that this, too,

represented harm flowing from an SLC. (The appellants dispute this analysis.)

[32] Twelfth, that the remedies tendered by the appellants in respect of the insured market were inadequate:

(a) The Tribunal mentioned two remedies, the so-called MMHS-minus remedy and the Mediclinic-plus remedy. Since the merging parties themselves did not support the MMHS-minus remedy, the Tribunal did not give it detailed consideration.

(b) The Mediclinic-plus remedy was that post-merger Mediclinic would, for five years, apply a specified discount at the targets against the Mediclinic tariffs, the discount set at a figure to ensure that the […]% increase due to the tariff increase was wholly neutralised after taking into account any net balance of efficiencies likely to be achieved by Mediclinic.

(c) The merging parties proposed, conservatively against themselves (in their view), a discount of three percentage points, which (i) disregarded the alleged efficiencies in regard to theatre and ward time and ethicals and (ii) assumed that but for the merger the targets could during the grace period achieve about half of the surgicals procurement efficiencies which Mediclinic could achieve. (The arithmetic proceeded thus: […]% [tariff increase in CPE, disregarding time efficiency] – […]% [Mediclinic’s surgicals procurement efficiency] + […]% [assumed surgicals procurement efficiency achieved by targets during grace period] = […]%).

(d) The Tribunal rejected the merging parties’ contention that the targets could only achieve half the procurement efficiencies. The Tribunal considered, further, that a remedy limited as to period was inappropriate but that a remedy in perpetuity would be impractical. The remedy in any event in any event did not address the adverse consequences flowing from regional dominance. (The appellants maintain that, to the extent that it is found on appeal that there is likely to be an increase in CPE at the targets, the remedy remains appropriate. The Tribunal could have lengthened the period of the remedy and increased the extent of the discount, depending on its findings.)

[33] Thirteenth (and finally), that the remedy proposed by the merging parties in regard to uninsured patients was likewise inappropriate. In that regard the merging parties proposed that Mediclinic be required to maintain the target hospitals’ tariffs and discount policies for a period of five years, subject only to annual escalation by not more than the Consumer Price Index (‘CPI’). This remedy the Tribunal held to be inappropriate because of its limited duration and because of the difficulty in determining what discounts the targets would have granted had they not been taken over by Mediclinic.

[34] The Commission’s counsel supported the Tribunal’s reasoning in all material respects. It is not unlikely that their

able representation of the Commission was of much assistance to the Tribunal and found expression in the Tribunal’s decision. I thus mean no disrespect to their submissions if, in what follows, I focus mainly on the Tribunal’s reasoning and the appellants’

criticism of it.

The product market

[35] In regard to the Tribunal’s first and second key findings, the appellants counsel’s principal argument was that even if those findings were correct, the appeal should succeed. The appellants nevertheless addressed written and oral submissions as to why these findings were wrong. In my view the market definition questions must be addressed if the appeal is to be determined on a principled basis. Furthermore, the arguments about market definition, particularly the extent of the local geographic market, are closely allied to predictions about SLC and harm to consumer welfare.

[36] The economists agreed that outpatient services provided by the hospitals (ie cases not involving admission to a ward and theatre)

should not be included in the product market. They differed, however, as to whether day cases should be included in the product market. Although there is no uniform definition of a day case, it refers in a general sense to a hospital admission where the patient is in hospital for less than a day and (perhaps) does not need an overnight bed. The sorts of procedures performed as day cases could often be done at a day clinic rather than at a multidisciplinary hospital.

[37] The Tribunal excluded day cases from the product market for the reason that although the multidisciplinary hospitals could readily reallocate resources from multi-day cases to single-day cases, day clinics could not, without substantial capital investment and regulatory approvals, reallocate their resources from single-day cases to multi-day cases. In my opinion the Tribunal’s

reasoning was correct. A similar view has been taken by the United Kingdom’s Competition and Market Authorities (see Ashford St Peter’s NHS Foundation Trust/Royal Surrey County NHS Foundation Trust UK CMA Final report 16 September 2015 (‘Ashford’) paras 5.18-5.22).

[38] Indeed, and despite the appellants’ submissions, the merging parties’ economist, Prof Nicola Theron, agreed with the Commission’s economists (the lead being Prof Liberty Mncube) that day clinics cannot compete with the bulk of services provided by multidisciplinary hospitals and should not be included in the product market. She noted, however, that the day clinics nevertheless pose a competitive restraint on a subset of services offered by the multidisciplinary hospitals. Although Prof Theron contended that day-case services, when offered by multidisciplinary hospitals, should be included in the product market, I agree with Prof Mncube’s view that it would be illogical then to exclude day clinics, which compete with the multidisciplinary hospitals for day-case patients.

[39] In the light of my conclusion on the geographic market, it is unnecessary to determine precisely how the day-case exclusion should be defined. The evidence indicated that some of the day cases performed at multidisciplinary hospitals could not readily be performed at day clinics.

[40] It is, however, convenient to mention, here, a complicating feature of the day-case issue. Different hospitals have different tariffs for what can broadly be described as day cases. The day-case tariff is usually lower than the ordinary tariff. There is, however, no uniformity as to what a ‘day case’ is.

(a) The targets only have a special day-case tariff for ward time. A patient who is admitted and discharged on the same date benefits from the day-case ward tariff. The duration of such an admission could thus be anything short of 24 hours. There is no discounted tariff for theatre time.

(b) Mediclinic has two day-case definitions with related tariffs. In relation to a list of specified procedures, a patient whose stay is less than 23 hours benefits from discounted ward and theatre tariffs. This is so even though the patient is admitted on one date and discharged on the next. Short stays not qualifying for these tariffs benefit from a different reduced ward tariff, the qualification being the same as at the targets – admission and discharge on the same date. For this latter class of short-stay patients, there is, as with the targets, no reduced theatre tariff. I shall refer to these qualifications as the 23-hour rule and the same-date rule.

Although Mediclinic’s tariffs are generally higher than the targets’, patients who qualify under its 23-hour rule will pay lower tariffs than equivalent patients at the targets. I shall return to this when considering pricing effects in the context of public interest as distinct from SLC.

[41] The Tribunal did not address another limitation proposed by the Commission’s economists, namely the exclusion of services in disciplines where there was no overlap between MC Potch and the targets. They listed, as excluded specialities, paediatric surgery,

pulmonology, cardiology, cardiothoracic surgery, gastroenterology, nephrology, rheumatology, vascular surgery, neurosurgery and

oncology.[2] These exclusions are potentially significant when one considers the evidence about the geographic market, namely evidence of the movement of patients and doctors between Klerksdorp and Potchefstroom. To the extent that such movement relates to services which can only be obtained in one town or the other, such evidence does not necessarily indicate that Klerksdorp and Potchefstroom are in the same local geographic market when it comes to services available in both towns.

[42] For example, neither of the Potchefstroom hospitals has an oncology unit whereas Klerksdorp hospitals do. Potchefstroom and Klerksdorp might thus be in the same local geographic market in the provision of oncology services but not in relation to general surgery, which is available at all the multidisciplinary hospitals in the two towns.

[43] However, since neither side invited us to delve into this question, it is preferable to disregard the suggested exclusion. This might be justified on the basis that hospitals in one town could perhaps, without great inconvenience or cost, reallocate resources to introduce a speciality currently available only in the other town. Stated differently, all the hospitals in both towns are at least potential competitors in all specialities (cf Ashford paras 5.23-5.28). As against this, the very introduction in, say Potchefstroom, of a new discipline currently only offered in Klerksdorp might fragment the geographic market, since Klerksdorp might lose all its drawing power in Potchefstroom in that discipline.

The local geographic market

[44] The obtaining of medical services by patients is generally not a discretionary matter. If the required services are not available

nearby, the patient must go further afield. However, and for obvious reasons, patients far prefer a hospital which is reasonably

accessible. While patients in outlying towns which do not have a multidisciplinary hospital may have no choice but to travel 50 km or more to a town having such a hospital, one would not ordinarily expect patients in a town well supplied with hospitals to travel to another town. Quite apart from convenience, admission in another town increases the overall cost, since the patient, and family members wishing to visit the patient, have to fund the cost of travel without assistance from their medical scheme. The Commission’s economists noted that patients want to be hospitalised near their homes and family. It must be borne in mind, in this regard, that the exclusion of day cases from the product market means that what is now under consideration is the likely behaviour of patients who will have to stay in hospital one or more nights.

[45] Evidence was led[3] to show that MC Potch (2015-2017 data) and MooiMed (2010-2014 data) drew only […]% and […]% respectively of their patients from Klerksdorp. They drew […]% and […]% respectively of their patients from Potchefstroom itself. Other significant contributors to MC Potch (all falling outside MaJB) were Fochville (56 km – […]%), Carltonville (51 km – […]%) and Parys (54 km – […]%) to the south and east of Potchefstroom, and Lichtenburg (154 km – […]%) to the north. MooiMed drew […]% of its patients from Parys and […]% from Fochville.

[46] Wilmed and Sunningdale (2015-2017 data) drew […]% and […]% respectively of their patients from Klerksdorp. Only […]% and […]% respectively came from Potchefstroom. In Wilmed’s case, Lichtenburg to the north (105 km) was the source of […]% of admissions while Bothaville to the south (64 km) and Wolmaransstad to the west (86 km) contributed […]% and […]% respectively. In Sunningdale’s case, Lichtenburg ([…]%) and Bothaville ([…]%) contributed more than or the same number of patients as Potchefstroom, Wolmaransstad a shade less.

[47] Anncron in Klerksdorp (2010-2014 data) drew only […]% of its patients from Potchefstroom (this dropped to […]% in 2015). Around […]% came from Klerksdorp. Lichtenburg, Trotsville (Wolmaransstad) and Ottosdal (to the west) and Bothaville each contributed more patients than Potchefstroom.

[48] The relatively high figure of […]% that Wilmed draws from Potchefstroom is likely to be a result of the fact that Potchefstroom

does not have a hospital offering oncology and neurosurgery.

[49] Although the economists on both sides offered detailed information and analysis of patient flows, the Tribunal found the evidence to be of little value, since it was backward-looking and did not tell one how patients in one town would respond to a ‘small but significant and non-transitory increase in price’ (‘SSNIP’) imposed by a hypothetical monopolist of the hospitals in that town. Would a SSNIP by the hypothetical monopolist in Klerksdorp cause Klerksdorp residents to switch in sufficient numbers to Potchefstroom hospitals to make the SSNIP unprofitable for the monopolist? If so, one can conclude that Klerksdorp and Potchefstroom are in the same geographic market.

[50] The Tribunal excluded, from the geographic market, hospitals falling outside the MaJB area. Although outlying towns such as Lichtenburg, Wolmaransstad, Bothaville and Parys contribute a significant portion of the patients admitted to the hospitals in Potchefstroom and Klerksdorp, there was no evidence showing what other hospitals are used by residents of these towns. Hospitals in Virginia, Kroonstad, Vanderbijlpark, Vereeniging, Brits, Rustenburg, Mahikeng, Vryburg and Frankfort are all likely to draw what I can call rural patients from the same outlying towns.

[51] This suggests that there is more than one local geographic market implicated by the merger. There is a broader market in which hospitals in Potchefstroom, Klerksdorp and other largish towns compete with each other for outlying rural patients, and a narrower

market in which hospitals compete for patients in Klerksdorp and Potchefstroom. Although the merging parties’ economist, Prof Theron, may not quite have articulated the existence of two local markets, this notion seems to be behind the following formulation of her position in the economists’ joint minute:[4]

‘Prof Theron considers that there are two levels of the relevant market: national and local. In the local relevant market, the data and relevant evidence indicate that there is very limited competition between the Potchefstroom and Klerksdorp hospitals. In a larger market including both towns, the market has to be expanded to include areas to the west and east of Klerksdorp and Potchefstroom and the hospitals to which patients in these areas travel’.

[52] The extent of the broader local market was not established in evidence and it was not shown that Mediclinic’s acquisition of the targets would affect competition in the broader local market. This was not the case the Commission set out to prove.

[53] The contentious question is the narrower local market, the one concerned with the residents of Klerksdorp and Potchefstroom. Do the hospitals in Potchefstroom compete for patients in Klerksdorp and vice versa? As I have said, the evidence shows that hitherto there has been no significant movement of patients between the two towns except in relation to disciplines not available in one or the other town.

[54] In a submission to the Commission, Discovery said that it did not think that the targets competed with MC Potch. GEMS’ observation,

that in a rural setting a significant number of patients might be willing to travel 50 km or more, may be true of residents of towns which do not have hospitals, but may not be true for the residents of towns like Potchefstroom and Klerksdorp. It is significant that most medical schemes have network hospitals in both towns,[5] suggesting that they are aware that members from one of the towns would find it most unpalatable to be forced to use a hospital in the other town. Anncron, in its written response to the Commission,[6] stated that its catchment area was about 150 km², but it did not name Potchefstroom as being part of the catchment area. The size of the catchment area was attributed by Anncron to the fact there were no private hospitals in the catchment area, a statement which would obviously be incorrect if Potchefstroom were part of it.

[55] Let us consider first the case of the insured patients, making up more than 95% of admissions. In their case, price is a non-issue,

since it is negotiated nationally between the hospitals and schemes, and the merger will not materially affect national market shares. So the SSNIP test has to be reformulated as a lowering of quality equivalent in significance to a SSNIP. Would such a lowering of quality by a hypothetical monopolist in Klerksdorp cause patients to go to Potchefstroom hospitals and vice versa?

[56] Even this reformulation has an air of unreality about it. Day cases, which fall outside the product market, constitute a significant part of a multidisciplinary hospitals’ business – in Mediclinic’s case, around […]% of admissions[7]. It is difficult to envisage a lowering of quality tainting only overnight cases. Our hypothetical monopolist is not a monopoly owner of the day clinics and day hospitals in Klerksdorp. Since the hypothetical monopolist would stand to lose business to his

Klerksdorp day-case competitors if standards dropped, this would serve to maintain hospital standards in general.

[57] Be that as it may, one must suppose a lowering of standards by the hypothetical monopolist equivalent to a SSNIP. Would this cause patients to divert to Potchefstroom in sufficient numbers to make the SSNIP unprofitable? There are two features of the market which militate against an affirmative answer. First, quality of care (or ‘patient experience’, an expression used in the evidence), unlike price, is imprecise and lacking in transparency. Second, one is dealing with services which most individual consumers need only infrequently. The individual consumer has limited experience for making comparisons.

[58] In the case of a commodity which consumers regularly buy, it is reasonable to suppose that they will become aware of a significant

price increase and will look around for a cheaper substitute. In the case of hospital care, it is difficult for a patient to make

comparisons. Assuming the patient finds his or her second experience in the same hospital less satisfactory than the first, how will the patient know that anything better is to be had at another town (assuming he or she ever needs to be admitted again to a hospital)? One can ask another hospital what its prices are; one cannot sensibly ask whether it offers better patient care than another hospital. And if the patient requires to be admitted to hospital on a third occasion, standards may have changed once again at the hospital he or she used on the second occasion.

[59] One must remember that the SSNIP, reformulated as a lowering of quality, does not mean a catastrophic collapse of standards, only a ‘small but significant’ drop in quality (what this amounted to, in the case of degradation of quality, was not the subject of expert evidence). Hospitals and the nursing profession are subject to legislative regulation. It would not be a reasonable application of the test to postulate a deviation from standards such as would constitute actionable negligence or professional misconduct. One must also bear in mind that the degradation of service in our hypothetical enquiry is not a lowering of care by the doctors who treat the patients, only a degradation of service in ancillary hospital care. The most important consideration for patients is the reputation and skill of their doctors. A patient may want nurses who are friendly and food which is palatable, but this is hardly critical for a hospital stay which, for most patients, is unlikely to exceed a week or two.

[60] Many patients will be guided by their doctors in their choice of hospital (cf Netcare Hospital Group (Pty) Ltd and Community Hospital Group (Pty) Ltd (68/LM/Aug06) [2007] ZACT 83 para 31). The evidence in the present case showed that although specialists in the one town sometimes consult at hospitals in the other town, they hardly ever perform procedures there. Time and convenience are important to busy doctors. A small lowering of hospital standards at Klerksdorp is unlikely to cause a Klerksdorp specialist to start referring patients to hospitals in Potchefstroom.

The SSNIP to be postulated is not, I repeat, one of such dimensions that a specialist could not properly continue to perform procedures

at the hypothetical monopolist’s hospitals.

[61] I thus am not surprised to find that a recent working paper by two members of America’s’ Federal Trade Commission begins:[8]

‘As economists have known since Hotelling (1929), demand declines rapidly with distance in retail and health care markets. For example,

Gowrisankaran et al (2015) find that a five minute increase in travel time to a hospital reduces demand between 17 and 41 percent.’

And later:

‘The incentives that health care providers have to improve quality depend upon the degree to which patients are willing to substitute towards higher quality facilities. Because patient distance to facility is typically the most important variable explaining patient choices, re­searchers have typically examined the marginal rate of substitution (MRS) between quality and distance (Tay, 2003; Romley and Goldman, 2011; Chandra et al, 2016; Gaynor et al, 2016). In general, the literature has found that patients are not willing to travel very far to go to a higher quality hospital. For example, Romley and Goldman (2011) find that a baseline pneumonia patient in the LA area would be willing to travel 2.9 miles farther to go from a hospital at the 25th percentile of quality to one at the 75th percentile of quality, and Chandra et al (2016) find that the average heart attack (AMI) patient will travel 1.8 miles for a hospital with a 1 percentage point higher risk-adjusted

survival rate. . .’

[62] As appears from the above working paper, merging parties in hospital transactions often argue for a broader geographic market by pointing to significant patient flows to more distant hospitals. If a regulator accepts a more broadly defined market, the merging parties may be able to contend that there is no significant increase in concentration. Arguments in favour of a broader geographic market, based on patient flow, tend to overlook the ‘silent majority’ of patients who may not be willing to travel further afield in response to a SSNIP.

[63] The ‘silent majority fallacy’ was explained in a scholarly article which the Tribunal cited in discounting patient-flow data,[9] even though the primary concern in the article is that such data tends to overstate rather than understate the extent of geographic

markets. On the tendency of patient-flow analysis to yield overbroad geographic markets, see Federal Trade Commission & another v Penn State Hershey Medical Center & another 838 F.3d 327 (2016) (‘Hershey’) at 340 and FTC & another v Advocate Health Care Network & others (US Court of Appeals, 7th Circuit, No 16-2492, 2016 (‘Advocate Health’) pp 15-18. For an academic discussion of these cases, see Levine and Hasty ‘Analyzing the Geographic Market in Hospital Mergers: Travel patterns take a backseat to payer response’ Competition Policy International,

December 2016.

[64] In Hershey the District Court, based on significant patient flows into the Harrisburg area from outside, accepted the merging parties’ broader geographic market – a market comprising 19 hospitals with drive-times of up to 65 minutes. On appeal the Third Circuit held that this analysis ignored the fact that 91% of patients from inside the Harrisburg area did not travel to hospitals outside the Harrisburg area and had median drive-times of 15 minutes. Similarly, in Advocate Health an appeal against a broader market succeeded before the Seventh Circuit. The correct geographic market was no larger than the so-called North Shore Area (‘NSA’) of Chicago. Part of the evidence in support of this conclusion was that 73% of patients in NSA obtained their hospital services within that area and that 80% of those patients had drive-times of no more than 20 minutes and journeys of less than 15 miles (about 24 km).

[65] If, in the present case, there were evidence of significant patient flows from Klerksdorp to Potchefstroom, the ‘silent majority fallacy’ would caution against assuming that a significant number of Klerksdorp non-travellers would start travelling to Potchefstroom in response to a SSNIP. As a fact, though, there are no significant patient flows from Klerksdorp to Potchefstroom. We do not know precisely what percentage of patients in Klerksdorp area obtain their hospital services outside of Klerksdorp but we know that only a very small number of them (by my calculation, around […]%[10]) use Potchefstroom hospitals. So for the types of admissions that can be handled in Klerksdorp and Potchefstroom hospitals (as distinct from highly complex work which might be performed only in the tertiary hospitals of major metropolitan centres), one can safely say that around […]% of the residents of Klerksdorp use Klerksdorp hospitals, not Potchefstroom hospitals. (Compare this to the 91% of the Harrisburg residents in Hershey who did not travel outside the Harrisburg area and the 73% of NSA residents in Advocate Health who did not travel outside the NSA.)

[66] Ultimately one must take a practical and common-sense view of the matter. It strikes me as quite unrealistic to conclude that a modest decline in the quality of ancillary hospital care at the hospitals of a hypothetical monopolist in Klerksdorp, unaccompanied by any decline in the standard of care provided by the doctors in those hospitals, would cause Klerksdorp residents to seek admission to hospitals in Potchefstroom, at considerable inconvenience and cost to themselves and their treating doctors, or at any rate to do so in sufficient numbers to deter the monopolist.

[67] The premise underlying the Tribunal’s criticism of historic patient-flow data is that the current absence of significant movement between the towns might be accounted for by the fact that healthy competition inhibits any of the hospitals in the two towns from significantly increasing prices or lowering quality. If a hypothetical monopolist in one of the towns were to increase prices or lower quality to an extent meeting the SSNIP test, would patterns change to an extent rendering the SSNIP unprofitable? I have already given reasons why I do not think this at all likely for the insured market.

[68] I wish only to add that the evidence did not show that there was a close equivalence of quality at the five hospitals located in the two towns. The Tribunal said that although the evidence was limited, the target hospitals seemed to be performing better than MC Potch in relation to patient experience, and that the merger was likely to lead to a lowering of quality at the targets. If that finding was justified, one may ask why more Potchefstroom residents do not seek admission to Klerksdorp hospitals. It is true that they currently have the option of MooiMed but that hospital may not have the capacity or range of specialities to absorb significant numbers of dissatisfied MC Potch patients. The more probable answer, in my opinion, is that despite small but not insignificant variations in standards at the hospitals in question, these variations do not cause patients in one town to seek hospital admission in the other town except perhaps in very small numbers.

[69] In the case of the uninsured market, the SSNIP test can be posed as a price increase by the hypothetical monopolist. There was evidence that some uninsured patients obtain quotes from hospitals but no evidence that uninsured patients in the one town seek quotes from hospitals in the other town. Although patients can to an extent compare prices, the quoted prices are rates for theatre and ward time. The patient’s overall bill will be affected by efficiencies in the management of theatre and ward times and in the acquisition and use of surgicals and ethicals. Whether one hospital will in fact turn out cheaper than another just because its quoted tariffs are lower is something patients cannot know in advance (cf Netcare Hospital Group (Pty) Ltd and Community Hospital Group (Pty) Ltd (68/LM/Aug06) [2007] ZACT 83 para 65).

[70] Although uninsured patients may be reasonably price-sensitive, are they so sensitive that a modest price increase of, say, 5% in the hospital bill will cause them to forsake the hospitals in their own town for a hospital in the other town? From a purely financial point of view, the patients would have to weigh, against the increased bill, the cost of travel for themselves and their visitors. Then there are the non-financial considerations – convenience, the time which the patients and their visitors would have to spend on the road and the preferences of their treating doctors.

[71] Although some uninsured patients seek quotes, the evidence did not establish that uninsured patients are, as a class, highly

price-sensitive. The evidence did not show that most or even a significant number of uninsured patients ask for quotes before choosing a hospital. According to the evidence, MooiMed’s tariffs for uninsured patients are significantly lower than MC Potch’s. Of course, Mediclinic maintains that overall its CPE is lower than NHN hospitals on account of superior efficiencies, but this would not be apparent from competing quotations. So if uninsured patients were highly price-sensitive, one would expect to find that MooiMed treats a disproportionately high number of uninsured Potchefstroom patients. If such evidence had been adduced, it might have shed some light on the answer to the hypothetical monopolist test, because it would have shown that, at least within the same town, price matters to a significant extent. Although the data must have existed, there was no evidence of a disproportionate bias of uninsured Potchefstroom patients towards MooiMed.

[72] There was also no evidence that, among the relatively small number of Potchefstroom patients who are admitted to the target hospitals, there is a significant proportion of uninsured patients attracted by the targets’ lower tariffs. I would also have expected information to have been available about one-hospital towns in South Africa which could have served as a proxy for the hypothetical monopolist.

[73] MC Potch’s private tariffs are higher than those of MooiMed and the targets by an amount which exceeds the small price increase postulated by the SSNIP test. If Mediclinic’s higher tariffs have not been shown to have caused a disproportionate flow of uninsured Potchefstroom patients to MooiMed and the targets, it must be doubtful whether a more modest SSNIP would have that effect. I have already identified the factors which would influence uninsured patients in choosing a hospital in their home town despite a modest increase in price.

[74] Mr Hendrik Steenkamp, the target hospitals’ manager, testified that he did not regard the Potchefstroom hospitals as competitors. His view was that Potchefstroom had good specialists, good general practitioners (‘GPs’) and good hospitals, ‘so we do not interfere’. He rather introduces his network of specialists to doctors in the rural areas to the west of Klerksdorp.

[75] Steenkamp also testified that specialists do not commute significantly between hospitals in the two towns. He was not aware of any Klerksdorp specialists who admitted patients to Potchefstroom hospitals, though he knew of certain Klerksdorp specialists – a dermatologist (Dr […]), a plastic surgeon (Dr […]) and an oncologist (Dr […]) – who occasionally consulted in Potchefstroom. He refuted the assertion of Ms Susanna van Reenen (the MooiMed hospital manager who was called by the Commission) that a certain Dr […] of Potchefstroom admitted patients at Wilmed. He explained the isolated and particular circumstances which had caused Dr […] (a urologist) and Dr […] (an orthopaedic surgeon) of Potchefstroom to perform or participate in the performance of one or two procedures in Klerksdorp.

[76] Mr Blake van Aswegen, the MC Potch hospital manager, likewise testified that specialists do not commute between the two towns.

Klerksdorp GPs are not part of MC Potch’s doctor referral network. MC Potch’s referral manager does not visit the Klerksdorp doctors; this decision was taken in 2012 because practically it was not considered worthwhile to chase business in Klerksdorp, given that it had three hospitals. Potchefstroom patients are sometimes referred to Klerksdorp specialists for neurosurgery and

oncology, those being disciplines not available in Potchefstroom. The Klerksdorp oncologists (Dr […] and Dr […]) occasionally consult at MC Potch on an outpatient basis (he estimated once every three months), but MC Potch is primarily served by a visiting Vanderbijlpark oncologist. There are occasional ENT and urology referrals to Klerksdorp because there is only one Potchefstroom specialist in each of those disciplines. Because the two Potchefstroom gynaecologists are in high demand and have long waiting lists, gynaecological patients needing or wanting quicker assistance might be referred elsewhere, including Klerksdorp, though some of Van Aswegen’s staff drove as far as Welkom to see a gynaecologist.

[77] For reasons explained earlier, the fact that Potchefstroom patients go to Klerksdorp for specialities not available in Potchefstroom does not show that Potchefstroom and Klerksdorp are generally in the same local market. The same is true where demand for specialities in Potchefstroom exceeds supply. Since medical services are generally not discretionary, patients have to go further afield if they cannot be helped locally. The more distant hospital in such a case is not a competitive constraint on the local hospital which ex hypothesi is operating at full capacity and thus has no need to compete for patients.

[78] Steenkamp and Van Aswegan were cross-examined with reference to corporate documents, and the Tribunal preferred to place weight on these documents than on patient-flow data. In Steenkamp’s case, he was taken to a document prepared for Matlosana by […] in connection with the proposed disposal of its hospitals. Prospective buyers were told that Matlosana’s ‘key markets’

included various towns in addition to Klerksdorp, one of these being Potchefstroom. In a section dealing with competition, Anncron and MC Potch were identified. With reference to MC Potch, the document stated that because Klerksdorp had more specialists than

Potchefstroom, ‘many patients travel to Klerksdorp for medical treatment’.

[79] To some extent, this document contains the puffery one might expect in a sales pitch. More importantly, one cannot cherry-pick. If one is to attach significance to the phrase ‘key markets’, one must then consider all the towns so described, including Lichtenburg, Mahikeng and Bothaville, as part of the ‘catchment area’ though these fall outside the geographic market defined by the Tribunal. Furthermore, it is in fact the truth that in Wilmed’s case it attracts a significant number of patients from Potchefstroom, but this is because it offers specialities not available in Potchefstroom. The Potchefstroom hospitals do not compete for those patients. The statement that ‘many patients’ travel to Klerksdorp for treatment should not be understood as conveying that patients who could obtain the required specialist services in Potchefstroom nevertheless choose to go to Klerksdorp.

[80] There was another Matlosana document, to which the Tribunal did not refer, which is in my opinion a more accurate statement of the position. Matlosana commissioned a report from a consultancy firm, Fernridge, as part of an application for more beds. Fernridge, in its report of April 2015, identified the targets’ catchment area as the area from which the targets drew 85%-90% of their patients. The catchment area excluded Potchefstroom. Although the relatively distant Lichtenburg was included, the report noted that patients from that town went to other locations as well, including Rustenburg. Regarding the exclusion of Potchefstroom, the report stated that Klerksdorp and Potchefstroom were ‘large Central Place towns, each servicing their respective markets with multiple private hospitals’.[11] Although many people commuted between the towns for work purposes, there was no need to do so for shopping or medical services, except for specialised treatment. In connection with the same application, Matlosana prepared a schedule setting out the percentages of patient-support from towns outside the catchment area, Potchefstroom being one such town.[12]

[81] In Van Aswegan’s case, he was taken to a Mediclinic motivation document of February 2015 requesting approval for more beds and theatres at MC Potch. This document does not support the view that Klerksdorp and Potchefstroom are in the same local market. The document understandably identifies MooiMed as MC Potch’s biggest competitor, with the ‘primary catchment area’ being said to include (apart from Potchefstroom) Parys, Fochville, Carltonville, Viljoenskroon, Ventersdorp and Lichtenburg.

Significantly, Klerksdorp is not mentioned as part of MC Potch’s catchment area. What the document says is that because of a shortage of beds at MC Potch, doctors practising there frequently admit their patients to other hospitals in Potchefstroom or refer them to Klerksdorp ‘50 km away’. The referral of patients to Klerksdorp is in the context of capacity constraints in Potchefstroom, and the author’s statement of the distance between the two towns was evidently intended to convey that such referral is inconvenient and undesirable. Van Aswegan testified that the additional beds were likely to be operational by April 2019; thereafter one would expect referrals to Klerksdorp to diminish.

[82] All the hospitals in Potchefstroom and Klerksdorp have in recent times taken steps to increase their capacity. Prof Theron disputed Dr Mncube’s proposition that the hospitals were capacity-constrained in relation to patients from the towns where they were located, and explained her view with reference to bed numbers and occupancy rates.[13] The Tribunal did not make a finding that the hospitals in either of the towns were capacity-constrained.

[83] Van Aswegan was also referred to Mediclinic’s non-binding letter of intent to acquire Matlosana. Mediclinic stated in this letter that Matlosana would be a ‘perfect fit’ to enhance Mediclinic’s business and would ‘complement our existing network of hospitals’, as already demonstrated by the fact that certain doctors who worked in Matlosana hospitals also worked at Mediclinic hospitals. Steenkamp said that he had no involvement in the letter. His evidence, as the person with direct knowledge of MC Potch’s activities, was that some Klerksdorp specialists consulted in Potchefstroom but did not admit or perform procedures there. The statement that the Klerksdorp hospitals would complement Mediclinic’s existing network does not shed light on the geographic market; Mediclinic spoke of its network of hospitals, and said that some doctors who worked at Matlosana hospitals also worked at Mediclinic hospitals (note in each case the plural ‘hospitals’).

[84] Discovery’s view was that, given the distance between Klerksdorp and Potchefstroom, it did not expect that MC Potch competed for patients in Klerksdorp.[14] The Tribunal did not mention Discovery’s opinion, but cited GEMS’ submission to the Commission and the evidence of its chief healthcare officer, Dr V Gqola. It is clear, however, from the statement in the submission that GEMS viewed a distance of nearly 50 km as significant, but thought that in a rural setting it is not ‘untoward’ for patients to travel this distance to access healthcare. Dr Gqola repeated this opinion in her evidence, saying that 50 km was not an unreasonable distance to travel in a rural area, as distinct from an urban area where there were more hospitals: “This is a region [the NWP] with very few hospitals.’[15]

[85] GEMS’ opinion is no doubt true in relation to patients living on farms and in small rural towns. Many of them might have to travel even more than 50 km to reach their closest hospital. This does not tell one very much about the willingness of patients living in Klerksdorp and Potchefstroom to travel between the two towns, bearing in mind the presence in each of those towns of several large hospitals. Dr Gqola testified that only 2,3% of GEMS’ Potchefstroom admissions came from Klerksdorp and only 3,8% of its Klerksdorp admissions came from Potchefstroom. She had examined data relating to doctors who practised at the Klerksdorp hospitals and the Potchefstroom hospitals, ‘and we didn’t see any crosslink in terms of our claims data’.[16] Elsewhere she stated the truism that patients’ choice of hospital tends to be based on their doctors.[17]

[86] As I understand the Commission’s economists’ reports and evidence, they constructed the candidate market in two stages. First, they determined the ‘catchment areas’ of the hospitals in the MaJB area, using the 80% threshold laid down in the NHS guidelines in the United Kingdom. Second, they included any hospital which drew at least 1% of its patients from the catchment areas of MC Potch and the targets, provided such hospital drew patients from the catchment areas of all three subject hospitals.

[87] I have not found it easy to grasp the second stage of the Commission’s analysis. As to the first stage, even in the United Kingdom the 80% catchment-area approach is only a rough starting point. In a small densely populated country with a fairly

even distribution of hospitals, this is likely to result in narrow geographic markets. In Ashford the 80% catchment areas for the three hospitals in question resulted in maximum drive-times for patients of 15 minutes, 20 minutes and 25 minutes respectively (see para 5.37).

[88] This approach, however, does not appear to me to be helpful when determining the geographic market in rural South Africa, where

patients in small country towns may have no choice but to travel long distances to obtain hospital services but where patients in larger rural centres may have several hospitals on their doorstep. The Commission’s economists appear to have started from the premise that the target market should be constructed with reference to the generalised catchment areas of MC Potch and the targets. Furthermore, they did not construct their catchment area for each hospital by plotting the sources of patients in decreasing order of magnitude until they reached 80% and by drawing an irregular isochrone map with reference to these localities (contrast Ashford paras 5.41-5.42 and the Fernridge report previously mentioned). Instead they assumed that the catchment area was distributed in radiating circles around each town, with localities in the expanding radius being included until one reached 80% of the hospital’s

admissions. The result, as Econex correctly observed, was to include localities which made smaller contributions to the hospital while excluding other localities which made larger contributions.

[89] Most importantly, the Commission’s approach ignored the important distinction between the hospital options open to, and thus the travelling behaviour of, rural patients on the one hand and the residents of Klerksdorp and Potchefstroom on the other. On the Commission’s radius approach, Klerksdorp and Potchefstroom were inevitably each found to have a catchment area which included the other, because each town drew patients from further afield than 47 km (the distance between Klerksdorp and Potchefstroom). In relation to Klerksdorp, this wrongly assumed that the options open to, and thus the likely travelling behaviour of, patients in Potchefstroom was the same as the options open to, and thus the likely behaviour of, patients in (say) Fochville.

[90] The Commission’s counsel referred to evidence from the Commission’s economists that a hypothetical medical scheme could not market a health plan to employers in the MaJB area without including a hypothetical monopolist of hospitals located in that area. No doubt the Commission’s answer to that question is correct but the hypothesis shows no more than that the market for providing hospital services to the residents of Klerksdorp and Potchefstroom is not wider than the hospitals in those two towns (and that is the point which the Commission’ economists were making in their report[18]). The hypothesis does not answer the crucial question, namely whether Klerksdorp and Potchefstroom are one or two geographic markets,

since, in the case supposed, the hypothetical monopolist is a monopoly provider in Klerksdorp, in Potchefstroom, and thus in MaJB.

[91] The more relevant question is whether a hypothetical medical scheme could market a health plan to employers in Klerksdorp without

including a hypothetical monopolist of Klerksdorp hospitals. In other words, could such a scheme successfully market its health plan if all it could offer prospective members were hospitals in Potchefstroom? The answer is surely no. Asking the question I have indicated is just another way of saying that one must test market definition by starting with the narrowest candidate market and working one’s way out, not the other way round.

[92] Thus in Hershey and Advocate Health the appeal courts did not ask whether a scheme which wanted to market a health plan in the broader market proposed by the merging parties could exclude a hypothetical monopolist of hospitals in the broader area – the answer was obviously no. The important question was whether a scheme wishing to market a health plan in the narrower area (the Harrisburg area in Hershey, the NSA area in Advocate Health) could exclude a hypothetical monopolist of hospitals in the narrower area. The answer was no. The fact that so few residents within the narrower areas travelled to hospitals outside those areas was one of the considerations in reaching the answer.

[93] The Commission’s counsel submitted that medical schemes would not have opposed the merger if they thought Klerksdorp and Potchefstroom were separate geographic markets. However, the medical schemes which responded to the Commission’s request for comment did not uniformly oppose the merger. Discovery, Medihelp, Selfmed, Hosmed and Bankmed said they had no concerns about the merger. Polmed saw pros and cons.

[94] GEMS’ written submission recorded a concern at the increase in Mediclinic’s national market share, adding that its concern was limited (though not negated) by the fact that Anncron provided strong competition for patients in Klerksdorp. In oral evidence, on the other hand, GEMS’ Dr Gqola stated that an increase of 0,8% in Mediclinic’s national market share would not really change the dynamics of tariff negotiations, and that although GEMS had given factual responses to the Commission’s questions about regional market shares,

‘it doesn’t really have relevance in the GEMS sense, because when we negotiate tariffs we negotiate on a national basis, and in terms of member access, the market shares won’t have an impact’.[19]

[95] Bonitas, Barloworld and the Old Mutual Staff Medical Aid Fund (‘OMSMAS’) are administered by Medscheme, and their

points of concern were expressed in identical language. I shall deal with this later in relation to the question of regional dominance.

[96] A final observation regarding the geographic market is this. There was no evidence to show how many Klerksdorp patients would need to divert to Potchefstroom to render a SSNIP of 5% unprofitable for a hypothetical Klerksdorp monopolist.

[97] I thus consider that the Tribunal erred in holding that the relevant local market included both Klerksdorp and Potchefstroom. The Tribunal should have held that only the Klerksdorp hospitals compete for Klerksdorp patients and only the Potchefstroom hospitals compete for Potchefstroom patients. A hypothetical monopolist in one of those towns could profitably engage in a SSNIP. At any rate, there was not robust evidence to the contrary.

Implications of market definition

[98] Because Potchefstroom and Klerksdorp do not fall in the same local market, the merger will not give rise to an SLC in relation to the local market within the meaning of s 12A(1). There are currently three multidisciplinary hospitals in Klerksdorp, one owned by Life, two owned by Matlosana. If the merger is implemented, the market structure in the Klerksdorp market will be unaffected. The two Matlosana hospitals will simply have a different owner. The new owner will have no greater market power within the Klerksdorp market than Matlosana currently has.

[99] Accordingly, any post-merger price increases at the target hospitals will not be a consequence of an SLC. The same is true of the Tribunal’s prediction that there will be a deterioration of quality at the targets. The enquiry into local price increases and deterioration of quality will thus have to take place in the context of an assessment as to whether the merger, despite not giving rise to an SLC in the local market, is unjustifiable ‘on substantial public interest grounds’ (s 12A(1)(b)), having regard to the factors set out in s 12A(3). I deal with this later in my judgment.

Leveraging regional dominance

[100] The Tribunal found that the merger would give Mediclinic regional dominance in MaJB, which Mediclinic could exploit, at a national level, in its negotiations with schemes. In short, so it was held, schemes would find it difficult to exclude Mediclinic from their DSP networks, given the dominant position it would enjoy in Potchefstroom and Klerksdorp. As I have already said, this theory of harm is applicable only to schemes’ low-cost options, comprising less than 10% of the insured market.

[101] To the extent that the Tribunal’s reasoning depended on its finding that Klerksdorp and Potchefstroom are part of a single geographic market, it fails at the threshold. To the extent that it is a self-standing competition concern, the evidence in support of this theory of harm was not compelling. Almost all schemes, including their low-cost options, are national, even though they may attract more members from some areas than from others. No scheme characterised by a regional focus on MaJB was identified (apart perhaps from AngloGoldAshanti). In the case of national schemes, the NWP as a whole makes up only 3,5% of the insured market. The largest town in the NWP is Rustenburg. Potchefstroom and Klerksdorp collectively are likely to account for less than 2% of the insured market.

[102] For reasons I have explained, schemes want good coverage for members. Where possible, members should have access to a hospital within a 50 km radius. Although Potchefstroom and Klerksdorp are just under 50 km apart, the weight of the evidence was that it would not be regarded as reasonable for a scheme to have a DSP in only, say Klerksdorp, on the basis that its members in Potchefstroom would forfeit coverage or have to make co-payments unless they travelled to Klerksdorp for treatment.[20] I thus accept that a scheme would usually want a DSP in each of Klerksdorp and Potchefstroom.

[103] However, it is not the case that a scheme wanting a network with a DSP in each of Klerksdorp and Potchefstroom will, if the merger is approved, have no choice but to include Mediclinic as an anchor. In Potchefstroom the position post-merger will be unchanged. A scheme could choose MooiMed in preference to MC Potch. In Klerksdorp a scheme could include Anncron in preference to Wilmed or Sunningdale. Anncron is a large hospital, equivalent in size to Wilmed. To put the matter colloquially, the merger will not cause Mediclinic to be the ‘only show in town’.

[104] For a scheme which would in any event have selected Life as its sole anchor, nothing will change, since in Klerksdorp Anncron will be the DSP, while in Potchefstroom the scheme, as is currently the case, will be able to select MooiMed or MC Potch as a filler. For a scheme which would in any event have selected Life and NHN as dual anchors, it will continue to have Anncron as its DSP in Klerksdorp and MooiMed as its DSP in Potchefstroom. For a scheme which would in any event have selected Mediclinic as an anchor, it will now have a Mediclinic DSP in Klerksdorp and in Potchefstroom.

[105] There is no hospital group (in which I include NHN) with a complete national footprint. The merger will affect Mediclinic’s

national footprint only marginally. The fact is that Mediclinic has no representivity in the Eastern Cape and a weak presence in KwaZulu Natal. Netcare has no representivity in Mpumalanga and the Northern Cape, while Life has no facilities in Limpopo or the Northern Cape. In national negotiations, schemes always have to consider a second anchor or filler hospitals in order to secure reasonable access for all their members. There is no group with such a wide and evenly distributed national presence as to permit its appointment as a sole and exclusive DSP. (Mr Buys testified that Mediclinic and Life are often appointed as co-anchors because their facilities complement each other in terms of geographic spread.[21]) It is simply not plausible that, by having two additional hospitals in a single town in a province which makes up a tiny

percentage of the insured market, Mediclinic will become a practically compulsory inclusion in all DSP networks.

[106] The Tribunal discerned, in negotiations which took place between Mediclinic and Bonitas in 2012, an instance of the sort of leverage which regional dominance conferred. The negotiations related to Bonitas’ low-cost option, BonCap. The evidence disclosed the following. Mediclinic’s 52 hospitals in South Africa had formally been part of the BonCap network. In 2006 Bonitas unilaterally excised 41 of the Mediclinic hospitals from the network, retaining only 11 hospitals as fillers in towns where Bonitas did not have a DSP. Since the discounts which hospital groups give for inclusion in a network have, as their quid pro quo, the prospect of additional volumes, this action left Mediclinic disgruntled, since the hospitals retained on the network were in towns in which Mediclinic in any event had the only hospitals. Mediclinic ended up tolerating this state of affairs for six years because relatively few BonCap members used the 11 retained hospitals.

[107] In 2012, however, Bonitas notified Mediclinic that it intended implementing a new network of hospitals in 2013 and invited a revised proposal based on an upfront discount. Mediclinic made the granting of a […]% discount conditional on the reinstatement of its hospitals (not necessarily all of them) in areas where it could expect to gain volumes. As Mediclinic’s witness, Mr Roland Buys, put it, Mediclinic in 2012 was not trying to force its way onto a network as a newcomer, it was ‘actually trying to buy back volume’ of which it had been unilaterally deprived.

[108] In my view, these negotiations are unremarkable, and reflect the stance that all hospital groups, including NHN, take in network

negotiations. Unless a hospital group has the prospect of achieving increased volumes in towns served by two or more hospitals, it has no commercial incentive to grant discounts on the ordinary tariffs it has negotiated with the scheme. The rationale of discounts is subverted where a scheme wants to use a group only in those towns where the group’s hospitals are the sole hospitals, since the group does not need to offer discounts to attract all the patients from that town.

[109] These negotiations do not reflect a leveraging of regional dominance, if by this is meant power conferred by having a majority though not substantially all of the hospital beds serving a particular subset of the population. The ‘dominance’ which Mediclinic had in the ‘regions’ served by the 11 hospitals which Bonitas retained on the BonCap network was dominance in individual towns by virtue of Mediclinic having the only hospital in each of those towns. No broader regional dominance was involved.

[110] Each of the four hospital groups is likely to have some hospitals in one-hospital towns. More importantly, though, for present purposes is the fact that the present merger will not result in Mediclinic owning all the hospitals in either Klerksdorp or Potchefstroom. So neither Klerksdorp nor Potchefstroom will be among the towns where Mediclinic could say to a scheme, ‘If you want my monopoly hospitals in towns X, Y and Z on your network at a discount, you will have to include some of my non-monopoly hospitals where I face competition and where I can expect to gain volume by being appointed the DSP.’ And on the assumption that the 11 hospitals implicated in the BonCap negotiations remain monopoly hospitals, Mediclinic is already able to exercise the sort of leverage which worried the Tribunal. It was not shown that the addition of a twelfth town to the list of monopoly towns would increase its leverage.

[111] Van Reenen expressed concern that, because schemes supposedly look for hospitals which can provide a ‘one-stop shop’, ie hospitals covering the full range of specialities, schemes are likely – if the merger is approved – to appoint Mediclinic as the DSP in Potchefstroom and Klerksdorp, thus prejudicing MooiMed and Anncron. Life did not, in its response to the Commission’s

queries, express any such concern. Van Reenen conceded that she did not have personal experience of scheme negotiations. Mediclinic’s

witnesses disagreed with the ‘one-stop shop’ thesis. And since Matlosana’s hospitals and MooiMed are currently part of NHN, and would on Van Reenen’s version currently offer a ‘one-stop shop’, one would expect to find that all or almost all low-cost options have NHN as their DSP, yet this is not the case.

[112] It is also necessary to remind oneself that competition regulation does not have as its object to protect individual competitors per se but rather to safeguard the process of competition. It was not shown that either MooiMed or Anncron would cease to be viable hospitals if they were deprived of the relatively small number of insured patients on schemes’ low-cost options.

[113] The Tribunal stated that according to the written submission by Bonitas’ Dr Jenni Noble, Mediclinic wielded negotiating power ‘inter alia through its demographic exclusivity in several areas’ (Dr Noble’s expression was ‘geographic exclusivity’[22]) and that if agreement were not reached Mediclinic would ‘typically threaten to charge members cash upfront at private rates’. The scheme might have to back down in the face of such demands. (As I mentioned earlier, identical language was used in the submissions

made by Barloworld and OMSMAS, which are relatively small schemes. Since Medscheme is the administrator of these three schemes, it is a fair inference that the concerns were formulated by the administrator on behalf of the schemes.) Dr Noble did not testify but this theme was repeated by Bonitas’ chief operating officer, Mr Marion, in his testimony.

[114] Dr Noble’s statements formed part of a longer paragraph in which she said that negotiations with Mediclinic were ‘significantly

more technical’ than with NHN. This was ‘largely because of Mediclinic’s support structures and resources, central

availability of data and ability to create complex reimbursement models’. Mediclinic had more data than Bonitas, ‘including

quality data at hospital level, line-item data underlying ARMs, underlying cost information and on-the-ground operational and management information’: ‘These all add to Mediclinic’s negotiation power.’ Then followed the statement that, ‘in addition’, Mediclinic wielded its negotiation power through geographic exclusivity in several areas. The ‘threats’ which she described were those which flowed from the totality of Mediclinic’s ‘negotiation power’.

[115] By contrast, said Dr Noble, NHN has historically not had similar negotiation power, the reason being that until recently it has ‘not had centralised data or the ability to implement [ARMs]’. Its ability to procure collectively also weakened its position as a group.

[116] The central thrust of Dr Noble’s comparison between Mediclinic and NHN was that Mediclinic is a more efficient and savvy negotiator. One can well understand that Bonitas and its administrator, Medscheme, prefer to negotiate with groups whose technical support and command of data are, in the opinion of Bonitas and its administrator, weaker than their own, but there is nothing objectionable about a hospital group which has good support structures and resources, centrally available data and the ability to

create complex ARMs. Ignorance is not competitively efficient.

[117] As to the additional element of ‘geographic exclusivity in several areas’, Mediclinic is indeed an exclusive hospital

provider in certain towns. So, too, on my understanding of the evidence, is each of the other hospital groups, including NHN. This

is part of the negotiating dynamic between schemes and hospital groups. But for present purposes it is sufficient to observe that the approval of the merger will not result in Mediclinic having geographic exclusivity in Klerksdorp or in MaJB.

[118] I make one final point in relation to the identical concerns expressed by Bonitas, Barloworld and OMSMAS. In Bonitas’ case, the targets were not DSPs on its low-cost options at the time it made its submission.[23] It is thus puzzling that Bonitas should have thought that the merger would have adverse regional effects for it. The targets are among the Klerksdorp DSPs for the two smaller schemes but so too is Anncron. This was also the position for Bonitas by the time its chief operations officer, Mr Kenneth Marion, testified in May 2018.[24]

[119] The Tribunal also mentioned a submission made by FedHealth, subsequent to the conclusion of oral evidence, in which FedHealth said that the remedy at that time proposed by the merging parties did not address the issue of networks.[25] According to FedHealth, Mediclinic’s stance on network discounts had historically been ‘that they will offer minimal

if any network discount for hospitals in areas where they do not stand to gain in volumes’. FedHealth thought that in Klerksdorp

Mediclinic would not stand to gain much in additional volumes, ‘as the only other close competitor in the area is [Anncron] and, to a minimal extent, [MooiMed]’.

[120] What FedHealth seems to have been saying is that the owner of the target hospitals in Klerksdorp could not expect to gain volume from Anncron if it were appointed the DSP in Klerksdorp. This assertion was not canvassed in evidence and I find nothing in the record to suggest that it is correct. Furthermore, if FedHealth’s statement were true, it would apply as much to the current owner of the targets as it would to Mediclinic. If Mediclinic’s negotiating tactics differ from NHN’s, this has nothing to do with the effect of the merger on the state of competition in Klerksdorp.

[121] In the light of Dr Noble’s comparison between the negotiating capacities of Mediclinic and NHN, one must perhaps bear in mind, when assessing the submissions made by schemes to the Commission, the possibility that some of them (or their administrators) might believe that it would be in their best interests for the targets to remain in the hands of a less savvy negotiator. If Mediclinic negotiated in the manner suggested by Dr Noble and Mr Marion, and if this were unusual, one might have expected schemes such as Discovery, Medihelp, Polmed, Selfmed and Hosmed to have expressed similar concerns, but they did not.

[122] The Tribunal relied on the statements I have discussed above (from Ms van Reenen, Dr Noble, Mr Marion and FedHealth) without explaining the mechanism by which the concerns expressed by these parties would come to pass, having regard to the uncontentious facts of the case (eg that schemes and hospital groups negotiate nationally; that no group has a complete national footprint; that in small towns it is not unusual for there to be only one hospital; that Klerksdorp and Potchefstroom collectively are only a tiny fraction of hospital groups’ business; and that if the merger is allowed Mediclinic will not be the only hospital in either Klerksdorp or Potchefstroom). The ipse dixit of a witness such as Ms van Reenen – that it would be difficult, post-merger, for a scheme to exclude the Mediclinic hospitals in Klerksdorp and Potchefstroom – has no weight if the witness does not explain (and have the knowledge and experience to explain) why this should be so.

[123] The Tribunal thus erred in finding that the merger would confer regional dominance on Mediclinic and that this would give rise to an SLC in negotiations between hospital groups and schemes for the construction of DSP networks.

Price effects and the public interest

[124] It is not in dispute that, if the merger is allowed, Mediclinic will (subject to any condition requiring it to do otherwise) immediately implement its tariffs at the target hospitals. In the case of schemes, this means the tariffs Mediclinic has already negotiated with the schemes, including discounted tariffs for low-cost options where Mediclinic is a DSP. In the case of uninsured patients, this means Mediclinic’s private tariffs.

[125] The fact that the tariffs will immediately increase is not a consequence of an enhancement in Mediclinic’s market power.

Mediclinic’s negotiated scheme tariffs and its private tariffs apply uniformly to all its hospitals in South Africa. Mediclinic will have no greater pricing power in Klerksdorp than Matlosana and NHN currently have. The tariffs which Mediclinic will implement are tariffs which have already been negotiated or set for uninsured patients and are thus set at levels uninfluenced by the merger. In post-merger scheme negotiations, Mediclinic’s marginal increase in national market share will not give it greater pricing power.

[126] Similarly, uninsured tariffs are set by Mediclinic nationally. Although the extent of discounts is probably affected by local

competition, my view of the geographic market means that Mediclinic will have no greater pricing power post-merger than Matlosana

currently has. The discounts which Matlosana currently offers, and those which Mediclinic post-merger will offer, would have regard

almost exclusively to Anncron’s competitive position in Klerksdorp.

[127] The Tribunal said that one of the Commission’s theories of harm was that Mediclinic’s increased market power in MaJB would lead to an increase in prices,[26] and the Tribunal discussed price effects, including the actuarial evidence, in the context of this theory of harm,[27] touching only very briefly on the same subject in the context of public interest.[28] If the Tribunal was of the view that price increases would be the result of increased market power or an SLC, it plainly erred. However, the Tribunal’s assessment of the evidence remains relevant to the correct enquiry, namely whether there are price effects justifying prohibition of the merger as a matter of public interest.

[128] The fact that price effects must, in the present case, be assessed in the context of public interest rather than SLC has an important effect on the evidence dealing with the relative efficiency of the targets and Mediclinic. Where it is shown that a merger is likely to substantially prevent or lessen competition in a relevant market, it is for the merging parties to establish that the merger is likely to result in technological, efficiency or other pro-competitive games which will be greater than, and will offset, the effects of any prevention or lessening of competition (s 12A(1)(a)(i)). Because no SLC has been shown in the present case, the merging parties do not attract this onus. In the context of public interest, we are trying to ascertain what prices are likely to prevail at the targets if the merger is allowed, and whether (assuming such prices to be higher than they would otherwise have been) this is a sufficient basis to prohibit the merger on public interest grounds. The efficiencies which were the subject of factual and actuarial evidence are simply part of this predictive exercise.

[129] We were not addressed on questions of onus and sufficiency of proof in relation to the prohibition of a merger on public interest

grounds. It seems to me that in absence of evidence that a particular harm, which is substantial, may eventuate if the merger is approved, the prohibition of the merger cannot be ‘justified’ within the meaning of s 12A(1). I leave open the question whether this requires the likelihood of harm to be established on a balance of probability or whether it suffices that the danger of such harm is reasonably possible.

The insured market

[130] If average CPE at the targets goes up post-merger because of Mediclinic’s higher tariffs, the harmful effects, in the case of the insured market, will take the form of higher claims from scheme members using the targets. Members (the patients) will only be prejudiced if premiums increase because of higher claims. Since the Klerksdorp patients who use the target hospitals are likely to represent a very small percentage of a scheme’s national population, and since scheme premiums are set nationally, it is doubtful whether a modest CPE increase at the targets (say by 5%) will ever translate into increased premiums. It is also improbable that the slight increase in cost for the schemes will have any material effect on their competitiveness.

[131] One must also bear in mind that, in the case of schemes, the cost of claims is affected by the presence and extent of ARMs. Schemes only negotiate ARMs where they think these will be more beneficial to them than FFS claims. Because ARMs transfer a measure of risk to the hospital, a hospital owner with multiple hospitals is more likely to accept an ARM than a firm which has only one or two hospitals, since it has more facilities over which to spread the risk. To a large group, the fact that some of the procedures covered by the ARM turn out to be substantially more expensive than the norm is unlikely to matter if it does a large number of procedures costing less than the norm. For an owner like Matlosana, by contrast, a few costly procedures could materially affect the profitability of the ARM.

[132] Furthermore, the negotiating and establishing of ARMs is not straightforward. In her report, the merging parties’ economist,

Dr Nicola Theron, said:

‘For all ARMs, providers [hospitals] and funders [schemes] require a large amount of detailed, accurate data on utilisation and cost to implement. ARMs are administratively and operationally complex; adequate systems must be in place to achieve the efficiencies offered by such payment arrangements. In addition, for providers to establish ARMs without exposing themselves excessively to payment risk, providers must be able to spread risk across an adequate number of facilities.’

[133] This view was confirmed by Mr Marion, Bonitas’ chief operations officer, who was called as a factual witness by the Commission. He said that Bonitas had no ARMs with the target hospitals although a few were planned for April 2018. It was ‘early days”, he said, to determine whether ARMs could be rolled out more extensively with NHN hospitals in the future. NHN had acknowledged that there were shortcomings in its management of data. He preferred not to comment on how this might pan out in the future. In regard to Mediclinic, Bonitas had extensive ARMs prior to 2015 but reduced these because it found that many of them were proving more expensive than FFS claims.

[134] If the merger is approved, the targets will be absorbed into the ARMs which Mediclinic has negotiated with various schemes, including Discovery. Mediclinic’s superior capacity to offer ARMs could well offset any modest national increase in FFS claims brought about by the implementation of its higher tariffs.

[135] Discovery reported that slightly more than […]% of the total revenue it paid to Mediclinic was by way of ARMs whereas the figure for NHN hospitals was only […]%. According to Mr Buys, […]% of Mediclinic’s admissions took place under ARMs, and this accounted for […]% of total hospital revenue. Discovery, South Africa’s largest medical scheme, told the Commission that ARMs are one of the key mechanisms for managing utilisation of hospital services. A key driver of health care inflation has been increased utilisation. Given that Mediclinic was ‘far more collaborative’ than NHN in ARMs and volume-based discounting, Discovery expected the merger to be to the scheme’s advantage.[29] Discovery later said that it was ‘less worried’ about NHN hospitals joining Mediclinic ‘since members will end up paying less. . . Hospitals will be more efficient and can get better services.’[30]

Uninsured patients

[136] The position is different for uninsured patients. Although their number in Klerksdorp is likely to be small, perhaps only […] admissions per year, they can be viewed as a vulnerable class. Medical care is not a discretionary item. Health care is a fundamental right guaranteed by s 27(1) of the Constitution. Among the Competition Act’s purposes, as listed in s 2, are to provide consumers with competitive prices and product choices, and to advance the social welfare of South Africans. In the preamble one reads that the Act was passed inter alia to ‘provide for markets in which consumers have access to, and can freely select, the quality and variety of goods and services they desire’.

[137] Although private patients will be able, in most specialities, to turn to Anncron if they find Mediclinic’s prices at the targets unacceptable, their choice might effectively be inhibited by price increases at the targets. Furthermore, if Mediclinic’s prices for uninsured patients rise, Anncron will have less incentive to grant discounts. Overall, prices for uninsured patients in Klerksdorp might thus go up.

[138] Section 12A(1) permits the Tribunal to prohibit a merger, even if it will not cause an SLC, where the merger ‘cannot be justified on substantial public interest grounds’. In determining whether a merger can or cannot be justified on substantial public interest grounds, the Tribunal must assess the factors listed in s 12A(3). For present purposes, only s 12(3)(a) is germane, namely the effect the merger will have on ‘a particular industrial sector or region’.

[139] The appellants’ counsel argued that s 12(3)(a) required the Tribunal ‘to consider the effect of the merger upon a sector or region as self-standing phenomena, rather than the effect upon competitors or consumers in a particular sector or region’ (emphasis in the heads of argument). I reject that submission. The public interest is concerned

with people, not abstractions. Klerksdorp is a region. If prices for uninsured patients increase at two or perhaps all three of the town’s multidisciplinary hospitals because of the merger, that is an effect which the merger will have on Klerksdorp.

[140] If, in the case of insured patients, the merger eventually gave rise to an increase in premiums, this too would be an effect of the merger upon a region. The appellants’ counsel submitted that a premium increase would be national, not local, and that this would thus not be an adverse effect of the merger on a ‘region’. Again I find this argument too narrow. First, it is not necessary that the harm engaging the public interest be purely regional. It suffices that the harm flows from an effect which the merger will have on a region. Second, and even if the first point were incorrect, South Africa as a whole qualifies as a region. It would be odd indeed if the Tribunal had the power to prohibit a merger causing harm in a part of South Africa but lacked such a power where the merger will cause harm in the whole of South Africa.

[141] It is thus necessary to analyse whether the Tribunal was right to find that the merger will cause hospital prices at the target

hospitals to be higher under Mediclinic control then they would be under their current control. It is necessary to bear in mind, in this regard, that the Tribunal only made this finding in respect of the two-year grace period from 1 November 2018 to 31 October 2020. Once the likely price effects of the merger have been ascertained, it will be necessary to consider whether any likely price increase, enduring for a period of two years, is sufficient to constitute a ‘substantial’ public interest grounds justifying

prohibition of the merger.

The non-actuarial evidence

[142] The appellants case in the Tribunal was that although Mediclinic’s tariffs were higher than the targets’, Mediclinic would be more efficient than Matlosana in the management of ward and theatre time, in the procurement and use of surgicals, and in the use of ethicals. The net effect of the superior efficiencies, so the appellants contended, was that, despite the tariff increases, the targets’ post-merger CPE would be less, not more, than it currently is. The evidence in support of this case comprised testimony from factual witnesses about Mediclinic’s countrywide systems for achieving efficiencies, and actuarial evidence which sought to quantify the likely effects of the tariff increase and efficiencies.

[143] Mediclinic has developed sophisticated tools for collecting and analysing hospital data and for reporting results to its hospital

managers. The data is highly granular, and enables a hospital manager to see each item of cost associated with the performance of any given procedure by any given specialist and to compare this with the costs associated with the performance of the same procedure by other specialists in the same discipline at Mediclinic hospitals around the country. Insofar as ethicals and surgicals are concerned, Mediclinic carefully ranks them so as to promote the use of equally efficacious cheaper alternatives.

[144] The culture within the Mediclinic group is that hospital managers constantly engage with specialists in order to influence their choices in favour of efficacious cheaper alternatives. Where the hospital costs associated with the particular specialist are materially out of kilter, the detailed benchmarking data enables the hospital manager to intervene. Mediclinic has found that in general specialists do not like to be more expensive than their peers and will modify their conduct if shown hard data to this effect.

[145] In regard to procurement (applicable to surgicals only, given that SEP applies to ethicals), Mediclinic’s scale allows it to achieve better prices from suppliers than smaller hospital owners. Until the procurement exemption came into force, NHN hospitals could not procure collectively and thus could not achieve this procurement efficiency. Even with the coming into effect of the exemption, one would not expect NHN hospitals to achieve the same procurement benefits. Although NHN will be entitled to bargain centrally in the procurement of surgicals, the choice of surgicals remains that of the individual hospitals. Because Mediclinic’s 50 hospitals are under single control, and because Mediclinic’s systems are designed to identify the most efficient surgical products (ie balancing efficacy with cost), it is likely to purchase a smaller range of items in larger volumes. NHN hospitals, being individually owned, are likely to be fragmented in their choice of products, so that one will have a larger range of items in smaller volumes.

[146] Mediclinic’s procurement advantage is not confined to its size. It also benefits from an international presence, with interests in operations in Switzerland and Dubai. Buys testified that Mediclinic’s insights into international pricing has significantly enhanced its negotiating power.

[147] The Tribunal found that NHN supplied data to individual hospitals which, with sufficient diligence and effort by hospital managers, could be used to achieve similar results to that which Mediclinic claimed. The evidence as a whole leaves me with quite the opposite impression. The NHN data is simply not sufficiently granular to be deployed as Mediclinic does. Steenkamp said that he once tried to encourage a physician at one of his hospitals to use more generics but the specialist ‘wiped the floor with me by saying that he knows what’s best for the patient’. Steenkamp did not have detailed data to persuade the physician otherwise.

[148] Although Steenkamp’s general philosophy has been to refrain from interfering in specialists’ protocols, in one case he was forced to do so when […] implemented a global fee for anthroplasty (hip and knee replacements). Each hospital had to determine how to apportion the globular fee between itself and the specialists involved. This required Steenkamp to obtain the costs his hospital was incurring for such operations. He collected and analysed six months’ data and discovered that one of the four surgeons was generating hospital costs considerably higher than the other three and that this related to one specific item the surgeon was using. This exercise, which involved three managers, took several weeks. This was the only occasion on which he was able to engage with a specialist about inefficiency.

[149] Although the anthroplasty exercise did not engage the managers full time, it is clear that the exercise was laborious because granular data was not readily available. It would be a hugely time-consuming business if a similar exercise were to be done on an ongoing basis for each of the hundreds of procedures which a multidisciplinary hospital regularly performs.

[150] Mediclinic has been able to develop its tools in this field because it is cost-effective for a large group to do so. The efficiency

gains, when multiplied across its 50 facilities, repay the expense involved in developing and maintaining the data systems. Steenkamp

testified that it was not commercially viable for Matlosana, with its two multidisciplinary hospitals, to develop and maintain comparable systems.

[151] Particularly when considering the public interest, the relevant question is not whether theoretically the targets could drill down to obtain and analyse data which in Mediclinic’s case is available at the press of a button. The question is whether practically it is at all likely to happen. The answer is that Matlosana does not regard it as commercially feasible to expend resources in doing so. This assessment was not shown to be wrong or unreasonable and is in any event a commercial judgement to be made by the owner of the business. There is no reason to believe that, absent the merger, utilisation efficiencies of the kind which Mediclinic’s systems encourage will be achieved at the targets.

The Tribunal’s approach

[152] Because the Tribunal rejected all the actuarial evidence, and because the Tribunal was unpersuaded that Mediclinic would in fact achieve efficiencies in the management of theatre and ward time and in the use of ethicals and surgicals, the only efficiency which it granted Mediclinic was in the procurement of surgicals. Because the Tribunal assumed no change in the choice of surgicals, it had regard solely to the lower prices which Mediclinic could, by virtue of its scale, achieve for the same surgicals as the targets historically purchased.

[153] On this basis, Mediclinic’s acquisition costs for the same surgicals resulted in a saving of […]% on the surgicals

component of the targets’ CPE. Because surgicals constituted […]% of the targets’ CPE, the overall effect of this saving on CPE was […]. Conversely, the tariff component of the targets’ CPE would rise by […]%. Since the tariff component constituted 68,2% of the targets’ CPE, the overall effect of this increase on CPE was […].

[154] This exercise, which was referred to in argument as a pure price comparison or a price-only difference, indicated that Mediclinic’s

superior procurement efficiency would cause a decrease in the targets’ CPE of […]%. This, however, did not take into

account the procurement efficiencies which the targets could expect to achieve as from November 2018 through the NHN procurement

exemption. The Tribunal found that NHN, being roughly equal in size to Mediclinic, would likely achieve the same procurement efficiencies

as Mediclinic, so that at least for the two-year grace period the targets, like Mediclinic, would be able to reduce overall CPE by […]%. Overall, therefore, at least during the grace period the targets’ CPE would be […]% higher under Mediclinic’s

ownership than under Matlosana’s.

[155] Even disregarding utilisation efficiencies, this exercise in my view takes an unrealistic view of the procurement efficiencies which the targets are likely to achieve as members of NHN. First, there is the fact that central procurement by NHN will still be procurement of a fragmented basket of products associated with the 62 individually-owned hospitals making up the NHN group. Second, the Tribunal’s assumption leaves out of account the benefits associated with Mediclinic’s international presence and knowledge. And third, it does not seem plausible that NHN could achieve maximum scale efficiencies without a lead time of some

months. NHN has not hitherto done procurement. There was no evidence as to how it would go about it. NHN would presumably need to hire a procurement team which would then have to get to grips with the range of products to be ordered, start establishing relationships with suppliers and gain experience.

[156] The more likely scenario, therefore, is that NHN-affiliated hospitals would take some time, perhaps up to a year, to reach achieve the full benefits of the procurement exemption and that once attained these would be more modest than Mediclinic’s. Since the Tribunal felt unable to predict whether Matlosana would become compliant within the two-year grace period, one may legitimately ask whether a net price increase at the targets of, say, 3% for one year only, would be a sufficient basis to prohibit the merger on public interest grounds.

[157] Furthermore, an assessment of the public interest should take into account what may happen after the two-year grace period expires. As matters currently stand, Matlosana will cease to benefit from the procurement exemption in November 2020. Thereafter, and in the long term, pricing will (ignoring efficiencies) be marginally more expensive if the merger were prohibited (by around […]%). Only if Matlosana becomes compliant will there be an ongoing beneficial effect from the procurement exemption, though such benefit would then be reasonably substantial (3% or more).

[158] Since Matlosana does not qualify as a ‘small business’ as defined in the Competition Act read with the National Small Business Act 102 of 1996, it could only become compliant if control were sold to historically disadvantaged persons (‘HDPs’). Of the parties which submitted non-binding indicative offers for Matlosana’s shares, the vendors pursued discussions with three, one of which,

[…], qualified as HDP-owned. Its indicative offer was the lowest of the three. One cannot discount the possibility that, if the present merger were prohibited, control of Matlosana would pass into HDP ownership, but one would expect the vendors to sell to the highest bidder or to retain their shares if they were dissatisfied with the highest price. The latter is a distinct possibility, given that Mediclinic’s offer was substantially higher than the offers of the other two bidders who were invited to submit letters of intent.

Quantification of efficiencies

[159] All of this disregards the efficiencies which Mediclinic alleges it could achieve. I have already explained why, contrary to the Tribunal’s assessment, one would expect Mediclinic’s superior tools for collating, analysing and deploying data to achieve utilisation efficiencies. Even if the targets coincidently happened to be as efficient as a reasonable selection of Mediclinic comparator hospitals, Mediclinic’s superior data systems could be expected to further enhance the targets’ efficiency.

[160] In my opinion the actuarial evidence pointed firmly to a conclusion that the targets are not as efficient as comparable Mediclinic

hospitals. The Tribunal erred in discounting the actuarial evidence. While it is plainly impossible to predict with precision the effect which Mediclinic’s efficiency initiatives will have on the targets, the actuarial evidence provided rough guidance as to the likely parameters of such effect.

[161] There is a point that needs to be emphasised at the outset of the discussion about the actuarial evidence. The inclusion or exclusion of day cases in the definition of the product market is potentially relevant to the question whether the merger would likely give rise to an SLC. Based on my view of the geographic market, I have concluded that there will be no likely SLC. What I am addressing now is whether the merger will give rise to price increases at the targets and whether this is a substantial public interest ground for prohibiting the merger. In this exercise, there is no reason to exclude day cases (howsoever defined) from the analysis; public interest is as concerned with the costs borne by day-case patients as by multi-day patients.

[162] Actuarial evidence was presented by Mr Barry Childs of Insight Actuaries & Consultants (‘IAC’) and by Mr Zaid Saeed of Alexander Forbes (‘AF’). While both witnesses were duly qualified to provide expert assistance to the Tribunal, it is not unfair to observe that Childs’ experience in general, and in the healthcare sector in particular, was significantly more extensive than Saeed’s. He obtained his honours degree in 1998 and became a Fellow of the Institute of Actuaries in 2004. He has a post-graduate diploma in Health Economics. His work experience over the period 1998 to August 2014, when he joined IAC as joint CEO, included employment with Medical Rescue International, NBC Employee Benefits, Liberty Healthcare, Discovery Health and Lighthouse Actuarial Consulting. He chairs the healthcare committee of the Actuarial Society and is a member of its NHI subcommittee. IAC has among its current and recent clients many medical schemes, administrators and other entities involved in health care, including

Discovery Health, GEMS, Selfmed, Medshield, Medscheme, Sanlam Healthcare Management and the CMS. IAC consults to all four major

hospital groups, including NHN. Saeed by contrast obtained his degree in 2012 and is currently a student member of the Actuarial Society. His employment at AF Health began in August 2013.

[163] Quite what effect the implementation of Mediclinic’s systems will have on the target hospitals’ efficiencies cannot be predicted with certainty. The best method of quantifying the differences was thought to be a comparison between existing Mediclinic hospitals and the targets, but there was a dispute as to what Mediclinic hospitals should be included in the comparison.

[164] Childs selected a group of Mediclinic hospitals which he regarded as providing a dataset that was both representative and statistically reliable. A single-hospital comparison would not be statistically reliable because of idiosyncrasies. His criteria for selection were that the Mediclinic comparators should be (a) inland hospitals (inland regions differ from coastal regions in their disease profile and co-morbidities); (b) located in a radius of 35 km – 250 km from Johannesburg and Pretoria (large urban centres are characterised by a higher density of specialists performing more complicated cases and trying experimental techniques). Seven Mediclinic hospitals met these criteria, one of which (naturally) was MC Potch. Childs then examined whether these seven hospitals were comparable to the targets in terms of the proportions of their admissions in the main specialities and found that they were.

[165] In a note of 25 September 2018[31] Childs mentioned a further, though fortuitous, advantage of his seven-hospital methodology. It turns out that MC Potch did not have billings during the relevant period for some of the procedures performed at the target hospitals. A comparison with MC Potch alone thus required that these procedure codes at the targets to be excluded from the analysis. With Childs’ methodology, there was a comparison which embraced all the targets’ procedure codes.

[166] Childs testified that Mediclinic played no part in the selection of the hospitals. It was not put to Childs that he chose selection criteria with a view to confining his basket to Mediclinic’s more efficient hospitals. It was also not put to him that his selection criteria were inappropriate.

[167] Saeed said that he was ‘comfortable with the wider hospital selection from a statistical credibility point of view’, but observed that two of the seven Mediclinic hospitals were smaller, and that if one excluded those two hospitals from the Mediclinic data, and the Sunningdale hospital (which is also small) from the targets’ data, the results shifted ‘quite noticeably’. Childs’ response was that although IAC had not set out to select big or small hospitals, he had been quite pleased to find that his objective criteria resulted in the inclusion of two smaller Mediclinic hospitals, since one of the two target hospitals was also small. The actuaries calculated an average CPE for the two target hospitals collectively. On this basis, it does not strike me as unacceptable that Childs’ basket of seven hospitals should have included two smaller ones. One is concerned with the overall impact of the merger on hospital prices in Klerksdorp.

[168] In consultation with the Commission, AF in its second report focused on a comparison between the targets and MC Potch. Saeed testified that, in the context of merger assessment, the comparator hospitals should be close competitors of the targets. The Commission in defining the geographic market had identified MC Potch as the only Mediclinic hospital that was a close competitor of the targets. MC Potch was also likely to represent a similar patient demographic.

[169] Quite apart from the lack of statistical reliability in a one-hospital comparison, there are two reasons why Saeed’s approach is not compelling. First, I have found that MC Potch is not in the same geographic market as the targets, which undermines Saeed’s

primary basis for focusing on that hospital. Second, I do not understand why the likelihood of Mediclinic’s initiatives resulting in efficiency improvements at the targets should be thought to depend solely on the success Mediclinic has had with those

initiatives in its Potchefstroom hospital. If initiatives yield better results at some hospitals than at others, this is more likely

to be a result of the management of the individual hospitals and the willingness of particular specialists to modify their treatment

patterns. I think Prof Theron was right when she said that the dispute between the actuaries about the comparator hospitals did not ‘speak to the competition economics of the market’, and was a separate exercise of selecting comparable hospitals.[32]

[170] I thus consider that the single-hospital comparison is unhelpful and that the seven-hospital comparison can provide useful guidance on likely pricing effects. In my opinion, Childs’ results should have been taken into account, bearing in mind that any attempt at precision would have been spurious.

[171] Childs’ initial analysis was done for the 2015 calendar year, later extended to include 2014 and 2016. Childs explained that, in order to compare like with like, one had to make case-mix adjustments to datasets of the target hospitals and the seven Mediclinic hospitals. IAC’s large quantity of data enabled it to generate case-mix indices (‘CMIs’) for the three components of CPE – tariff, surgicals and ethicals. Every admission at the targets and the seven Mediclinic hospitals was allocated to a so-called DRG category, ie categories determined by IAC’s Diagnosis-Related Grouper (‘DRG’). DRG allocation is based on clinical codes, age and gender. In the present case, around 1450 DRG categories were involved. The CMI ranks the relative cost of each DRG category in relation to a value of one. These cost relationships were calculated from a broader set of data than the nine hospitals involved in the Mediclinic/Matlosana CPE comparison in order to ensure stability.

[172] It was also necessary to exclude outliers from the datasets which might distort the results, an adjustment referred to as ‘trimming’. Childs did so using the interquartile method, which removed about 5% of admissions by volume. This trimming method was applied separately to each of the DRG categories. Apart from trimming, Childs excluded neonatal, transplant and critical-care (eg long-term ventilation) admissions due to their cost volatility and low volumes. Childs excluded outpatient cases from his data, but included all genuine inpatient admissions, whether or not they were classified by Mediclinic as day cases.

[173] This, broadly speaking, was the exercise Childs did in his first report.[33] The conclusion was that Mediclinic CPE was […]% higher in regard to tariff items, […]% lower in respect of surgicals and […]% lower in respect of ethicals. Bearing in mind the relative weighting of these three components of CPE ([…]), overall Mediclinic CPE was […]% lower than the targets.

[174] Although not explicitly stated in the first IAC report, my understanding from subsequent reports is that cases charged in terms of ARMs were excluded from the analysis. This was certainly AF’s approach from the beginning, and the final sets of results from both actuaries were explicitly presented on this basis. Since the actuarial analyses were based on comparisons of cost per line items, and since ARMs are characterised by an absence of detailed line items, they could not have featured in the exercise.

[175] In a second report[34] Childs analysed detailed line-item data for those surgicals and ethicals, the use of which was common to the targets on the one hand and the Mediclinic hospitals on the other, to arrive at pure comparisons and case-mix adjusted comparisons. The results were broadly in line with, and thus validated, the results for ethicals and surgicals generated by the methodology used in the first report. The second report reflected that, for surgicals, Mediclinic was […]% cheaper than the targets, of which […]% was attributable to lower prices per unit and […]% to lower volumes used. For ethicals, Mediclinic was […]% cheaper, of which […]% was attributable to lower prices per unit (given SEP, this would mean cheaper generics as against more expensive patent drugs) and […]% to lower volumes used.

[176] The next report was AF’s first report[35]. This report reviewed the seven-hospital methodology. AF attempted to replicate IAC’s results, and also performed an independent set of calculations, using a different set of CMIs, trimming methods and case selection. AF agreed that case mix and trimming adjustments were appropriate, but there were differences as to precisely how this should be done.

[177] While AF agreed with the need to include Mediclinic day cases which were genuine inpatient events, they excluded those which (so they believed) involved no ward or theatre time. This largely neutralised the superior efficiencies in ward time reflected in IAC’s results, although AF confirmed that Mediclinic remained more efficient in theatre time.

[178] AF was able to replicate IAC’s trimming method and was satisfied that IAC’s trimming results were reasonable. AF explained an alternative trimming method, adopting AF’s benchmarks for low-cost (R1000) and high-cost (R100 000)

admissions. Some of the high-cost cases which IAC excluded from the outset (eg transplants) were, in AF’s methodology, excluded as part of the trimming adjustment.

[179] Replicating IAC’s methodology as best it could, AF calculated that the seven Mediclinic hospitals’ average CPE was […]% lower than the targets. With AF’s alternative trimming methodology, the difference reduced to […]%. With AF’s trimming methodology and its alternative CMIs, the difference dropped to […]%. With a further adjustment for differences in cases included and excluded (particularly the exclusion of the cases supposedly involving no theatre and ward time), the Mediclinic hospitals’ CPE emerged as […]% higher. In other words, only if all of AF’s alternative methods were accepted was the Mediclinic CPE higher than the targets’ CPE.

[180] IAC’s third report[36] contained minor corrections for errors brought to light by AF’s first report. This did not change IAC’s actual results and conclusions. Childs later explained in oral evidence that IAC had incorrectly populated a field which had resulted in certain day cases supposedly having involved no time in the ward or theatre. This was not in fact the case, and once the field was correctly populated the overall results remained the same. I understand Saeed to have acknowledged this in his oral evidence, but from his perspective it became irrelevant in view of the approach taken by AF in its second report, which was to exclude all day cases.

[181] Before IAC delivered its substantive response to AF’s first report, AF delivered a second report[37] in which it extended its analysis to cover the period 2013-2016 and applied its own methodology rather than peer-reviewing the IAC methodology. Although AF’s methodology was in the main unchanged, its second report purported to exclude day cases on the basis that the Commission’s economists contended that day cases did not form part of the product market. In the event, and because only those cases coded by Mediclinic under the 23-hour rule were clearly identifiable as day cases, only they were excluded. Cases billed by Mediclinic and the targets in terms of their respective same-date tariffs were not excluded.[38] The 23-hour rule cases were excluded from the Mediclinic data although they were genuine hospital admissions. And although AF still

showed results for IAC’s basket of seven Mediclinic hospitals, the emphasis shifted to a comparison between the targets and MC Potch – a change explained by AF as having occurred ‘in consultation with’ the Commission.

[182] IAC’s substantive response to AF’s first and second reports was contained in its fourth report.[39] In regard to AF’s first report, IAC pointed out that AF had incorrectly removed the admissions which were the subject of the minor corrections made in IAC’s third report. Those admission should not, said IAC, have been removed, since the initial mistake had been a labelling error rather than an inclusion error. The reversal of this incorrect exclusion by AF negated AF’s final scenario (scenario six), under which Mediclinic became […]% more expensive than the targets rather than between […]% and […]% cheaper. (As Childs later said in oral evidence, ‘scenario six shouldn’t exist’.)

[183] IAC criticised the one-hospital comparison because small datasets are accompanied by high volatility. This weakness in a one-hospital

comparison was acknowledged by Saeed in oral evidence.

[184] In regard to AF’s second report, IAC questioned the removal of day cases, and pointed out the significant impact it had on the results. In oral evidence, Childs explained that AF had removed all cases which Mediclinic’s data labelled as day cases. This was in his view incorrect, because day-patients (as distinct from outpatients) are simply a subset of inpatients. He testified that IAC routinely did this kind of analysis for medical schemes and had never been asked to separate out day cases:[40]

‘When medical schemes look at this kind of data, they look at acute hospitals overall in their experience... [T]hey don’t see them as separate. [W[hen they look at acute hospitals, they see day patients and overnight patients the same. We’ve never been asked to do an analysis that wholly or partially carves out day cases from a cost per admission adjusted efficiency analysis.’

[185] Childs also pointed out in his oral evidence that AF had removed day cases in an inconsistent fashion, since a significant

number of admissions were excluded from Mediclinic’s data in circumstances where similar admissions at the targets were not

excluded. If day cases were to be removed, this had to be done on a consistent basis for both datasets.

[186] The removal of Mediclinic’s deeply discounted 23-hour-rule cases biased the results against Mediclinic. Childs regarded AF’ exclusion of Mediclinic’s day cases as quantitatively the most significant difference between IAC’s and AF’s analyses. Unsurprisingly, Saeed acknowledged that it would be a point of concern if the differing coding practices of Mediclinic and the targets resulted in inconsistent exclusions and inclusions.

[187] As I explained earlier, for purposes of the public interest analysis there is no reason to remove day cases (ie cases where patients are admitted to hospital and incur ward and theatre time, but whose stay lasts less than a set period, whether it be a same-date rule or a rule set with reference to 24 hours, 23 hours or 12 hours). (Indeed, outpatient prices might also have been relevant to the public interest analysis, but no evidence in that respect was presented by either side.) In oral evidence, Saeed seemingly shared Childs’ view as to how medical schemes see matters and explicitly explained the exclusion of day cases with reference to a competitive assessment:[41]

‘I think I concur with [Childs] on that point, in that over the course of a standard actuarial analysis for, let’s say one of our medical schemes, we would group day cases with hospital costs, because that’s where those admissions are serviced. But I think it’s important just to consider the context in which the analysis is being performed now, ... which is considering the competitive impact of the services that are included in the analysis.’

[188] In AF’s first report they had not disagreed with IAC’s inclusion of day cases constituting genuine inpatient events; their only disagreement had been in respect of those cases which seemingly involved no theatre or ward time. From the limited cross-examination

allowed, it appears that AF’s change of stance in their second report was solely on account of the Commission’s view regarding the definition of the product market for purposes of the competitive assessment.[42] Dr Mncube confirmed that this was the Commission’s economists’ instruction to AF.[43]

[189] During the ‘hot tub’ evidence of the actuaries, in which the economists were granted an opportunity to ask questions, the Commission’s lead economist asked Childs whether he would accept that the inclusion or exclusion of day cases in a competition

setting was an argument for the economists rather than the actuaries. Childs acknowledged that actuaries are not experts in the intricacies of competition economics:[44]

‘However, what I would like to bring to bear into the discussion is extensive consulting to the purchasers of these services at acute hospitals and what I can tell you [is] that in those cases – and presumably these purchasers of these kinds of services consider . . . the competitive nature of the services that they are buying –, they don’t separate out day cases from acute hospitals. . . From a competition point of view, if that is the basis for the argument, then I would defer to the economists. On determining whether or not the hospitals are comparatively efficient or not I would express my view that they should be included.’

In the context of pricing effects as a public policy consideration, Childs’ concluding sentence is plainly right.

[190] IAC squarely took issue with AF’s alternative trimming methodology. If AF’s static-trim method were used, the floor and ceiling values should at least be adjusted for inflation. There was, however, a more important objection of principle. A static-trim method is typically used when one wants to remove distortions in hospital data unadjusted for case-mix. Since DRGs and CMIs already quantify differences between typically expensive and typically inexpensive admissions, a more robust trimming method was possible: ‘By determining trim points for each DRG, anomalous admissions in each category are removed, rather than removing large claims en masse.’ In short, too much information was unnecessarily discarded in AF’s trimming methodology.

[191] In the limited time allowed to counsel to lead and cross-examine the actuaries, the merging parties’ counsel invited Childs briefly to explain his preference for IAC’s trimming method, which he did. The Commission’s counsel did not devote any cross-examination to this issue, and Saeed was not led on it. In my opinion, IAC’s reports on this aspect, amplified by the oral evidence, make out a cogent case for using the more nuanced IAC trimming method than the blunter AF method.

[192] IAC also questioned AF’s alternative CMIs. AF derived its weighting ratios from the data of the nine hospitals directly

involved in the comparisons. As IAC explained in its first report, IAC had access to a larger universe of data to establish these

weightings. (In oral evidence, Childs referred to IAC’s data as covering about 4,5 million medical scheme admissions.) IAC recognised that AF might be hampered by not having access to such data, but was concerned that the limited data used by AF was too small to derive stable case weights.

[193] The comparative merits of the CMIs were not debated at any length in the oral evidence. In principle, it seems to me that the cost relationships between various procedures are likely to be more reliable when derived from a larger data universe than a smaller one. For purposes of calculating a CMI, there is no merit in focusing only on the nine hospitals whose costs were under consideration in the present case. The relative complexities of medical procedures, as reflected in relative cost, is likely to be a national phenomenon.

[194] The upshot was that IAC, in its fourth report, saw no reason to depart from the conclusions expressed in its earlier reports.

[195] I have touched on relevant aspects of the actuaries’ oral evidence. They were subsequently asked to submit supplementary

reports to deal with various day-case scenarios and to set out the results for the calendar years 2014, 2015 and 2016. This resulted in AF’s third report and IAC’s fifth and sixth reports. In their final report, IAC conveniently set out the results presented by IAC and AF on the various scenarios, using their respective methodologies.

[196] Childs and Saeed’s calculations yielded the following results. Although calculations for other day-case variants were done, I only reproduce those relating to day cases defined as admissions under 24 hours. A minus percentage in the following table indicates that the Mediclinic hospital(s) are cheaper:

[All figures have been omitted from the table because of confidentiality claims]]

IAC % difference AF % difference Comparison 2014 2015 2016 Avg 2014 2015 2016 Avg All day cases included MC7 ⁓ Targets MC7 ⁓ Wilmed MC7 ⁓ Sunningdale MC Potch ⁓ Targets MC Potch ⁓ Wilmed Day cases (≤ 24 hrs) excluded MC7 ⁓ Targets MC7 ⁓ Wilmed MC7 ⁓ Sunningdale MC Potch ⁓ Targets MC Potch v Wilmed

[197] In the seven-hospital comparison, IAC and AF both found that Mediclinic was cheaper than the targets collectively in each of the preceding three years, whether day cases were included or excluded – by averages ranging from […]% to […]%. Based on the exclusion of day cases, defined as admissions for 24 hours or less, the average difference is […]% (AF) or […]% (IAC).

[198] There were three other scenarios for the day-case exclusion: admissions under 12 hours; admissions under 23 hours; and the

exclusion of admissions based on each hospital’s definition of day cases. All these scenarios showed that the seven Mediclinic

hospitals were cheaper than the targets, the most modest difference being based on the exclusion of admissions under 23 hours, where the average difference was […]% (AF) or […]% (IAC).

[199] The calculations reflect that the seven Mediclinic hospitals were on average significantly cheaper than Wilmed but more expensive than Sunningdale. Since larger hospitals tend to have higher CPEs than smaller ones, this is unsurprising. The two smaller hospitals in IAC’s basket of seven were probably as or more efficient than Sunningdale. Discovery’s CPE analysis[45] was that Sunningdale was less efficient than Mediclinic hospitals of comparable size while Wilmed’s efficiency was more or less the same as Mediclinic hospitals of comparable size (Discovery did not undertake the further refining criteria used by IAC). The comparison between the seven Mediclinic hospitals and the targets collectively is the most important one.

[200] I have, for the sake of completeness, included the comparisons between MC Potch and the targets collectively and between MC Potch and Wilmed (the latter two hospitals being of roughly equivalent size). With the inclusion of day cases, Wilmed is marginally more efficient than MC Potch. With the exclusion of day cases (defined as 24 hours or less), MC Potch is significantly more efficient, the difference being […]% (AF) or […]% (IAC).

[201] I have explained my reasons for preferring IAC’s methodology to AF’s and why day cases should be included. I am surprised that, when day cases under 24 hours are excluded, the percentage by which Mediclinic is more efficient goes up rather than down, since I would have expected Mediclinic’s discounted tariff under the 23-hour rule to have resulted in Mediclinic being at its most efficient when all admissions, including this class of day case, were counted. Subsequent to the hearing of the appeal, the appellants’ counsel were asked to deal with this apparent anomaly in a written note. Since the Commission was not amenable to the appellants’ counsel including any material in the written note which could not be derived from the record, counsel’s somewhat cryptic note has not clarified the matter.

[202] Be that as it may, based on the seven-hospital comparator, the inclusion of all day cases and IAC’s methodology, one may reasonably expect that the net effect of the merger and the implementation of Mediclinic’s efficiency initiatives will be that, despite the implementation of Mediclinic’s higher tariffs, CPE at the targets will fall by about […]%.

Conclusions

[203] Since the factual evidence about Mediclinic’s efficiency initiatives was compelling, there is no reason to be sceptical about the figures reflected in the above table. Of course, these figures do not take account of the beneficial effect for the targets of the procurement exemption. On the most meaningful of the actuarial comparisons, IAC calculated Mediclinic to be cheaper by […]%.

The Tribunal found Mediclinic to be more efficient than the targets in procurement by an amount which would reduce the targets’ CPE by […]%. On IAC’s primary scenario (including all day cases), CPE at the targets under Mediclinic control would thus only rise by […]% if one assumes that without the merger the targets could achieve procurement efficiencies for two years of […]%. (On the day-case exclusion, which was the approach taken by the Tribunal and the Commission, the targets would be cheaper under Mediclinic control even if one assumes that the targets could achieve procurement efficiencies of […]%.)

[204] On AF’s calculations of the primary scenario (including all day cases), the beneficial effect for the targets of the procurement efficiency would have to exceed […]% before one could find that the merger will increase the targets’ CPE during the grace period. It is doubtful that the procurement would yield an efficiency of […]%. If it did, it would only be for a portion of the grace period. So on AF’s calculations there is a possibility, though not a very large one, that for part of the grace period the targets’ CPE under Mediclinic control will be higher than under Matlosana’s control by an amount not exceeding […]%.

[205] I do not lose sight of the fact that, just as it might take some time for NHN to attain its maximum procurement benefit from the procurement exemption, so not all of Mediclinic’s efficiencies could be achieved immediately. The greater part of its efficiencies in relation to surgicals is a procurement efficiency rather than a utilisation efficiency. The procurement efficiency will be achieved immediately because Mediclinic’s procurement systems are mature and in place. There will be a lag in achieving utilisation efficiencies on tariff items, surgicals and ethicals, since this requires the collation and analysis of data and engagement with specialists. Dr Smuts said that it took about three months for any initiative to yield positive results. It could be significantly

longer.

[206] The beneficial effects for schemes of ARMs are not accounted for in the above quantification. Since schemes have negotiated ARMs more extensively with Mediclinic than with NHN, the post-merger implementation of these ARMs at the targets could be expected to have some further beneficial, though unquantified, effect on the cost of scheme claims

[207] In my opinion, the Tribunal erred in finding that there were substantial public interest grounds for prohibiting the merger on the strength of price effects.

Public interest and quality of care

[208] As with price effects, the Tribunal discussed the possibility of a post-merger deterioration of quality at the targets as something

which would supposedly be brought about by a decrease in competition.[46] The evidence the Tribunal discussed, however, did not suggest that the supposed deterioration in quality would be the result of an SLC. Rather, the debate was whether, as matters currently stood, the targets or Mediclinic was doing better in the sphere of quality care, the assumption being that if Mediclinic was doing worse than the targets, this poorer quality of care would automatically (like the tariffs) be imposed on the targets. Once again, this is not a matter of possible harm flowing from an SLC. Such relevance as it has arises in the public interest assessment.

[209] As the Tribunal observed, there was limited evidence on the differences between the quality of service at MC Potch and the

targets. All the same, the Tribunal thought that the targets were performing better than Mediclinic.

[210] The evidence was indeed meagre. There is currently no agreed way of measuring quality of care or patient experience. Furthermore, the Tribunal’s focus on MC Potch was misconceived. At a national level, Mediclinic has ‘comprehensive and globally benchmarked systems’.[47] It is these systems that will be implemented at the targets. If it be so that management at MC Potch has fallen below the mark, that tells one nothing about what will happen at the targets, since MC Potch management will have nothing to do with the management of the targets. Indeed, no change of personnel is envisaged, and Mr Steenkamp will remain as the hospital manager. There is thus no reason to think that standards will decline. If they are already good, Mediclinic’s sophisticated systems can only help them to get better.

[211] In the context of the hypothetical monopolist test, I observed that it was unrealistic to suppose a decline of standards affecting only overnight care, something which had to be postulated in view of the exclusion of day cases from the product market. Quality of care, I would expect, typifies a hospital as a whole. The Tribunal did not find that there would be any lessening of competition in relation to day cases. The risk of losing patients to Anncron and day hospitals would, I think, be sufficient to prevent an otherwise ‘dominant’ Klerksdorp hospital owner from allowing standards to drop at the target hospitals.

[212] At any rate, the evidence fell far short of showing a material decline of standards as a reasonable possibility. The prohibition of the merger in the public interest was not justified on this ground.

Conditions

[213] Since the Tribunal’s factual findings on SLC and public interest cannot be sustained, we are at large to consider whether

approval of the merger should be conditional or unconditional. Because I find no SLC, the question is whether the possibility of a slightly increased CPE at the targets during the grace period calls for a condition.

[214] In respect of insured patients, the appellants proposed the so-called Mediclinic-minus remedy – a 3% discount against Mediclinic’s scheme tariffs, with a five-year duration. In argument the appellants’ counsel said that the merging parties had not been, and were not now, wedded to the discount percentage or the duration proposed – a different percentage and/or period could be inserted in accordance with the Tribunal’s (and now this court’s) findings.

[215] Since the tariff component is about […]% of CPE, a 3% discount off Mediclinic’s tariffs would reduce overall CPE by […]%. Based on AF’s calculations of the seven-hospital comparison, Mediclinic will achieve efficiencies at the targets of around […]%. If the procurement exemption were to result in the targets achieving procurement efficiencies of […]%, they could notionally be […]% cheaper if the merger were prohibited. The suggested discount of 3% will largely neutralise this possibility. Since I think it unlikely that the targets will in fact achieve procurement efficiencies of […]%, and since I consider IAC’s calculation of the likely efficiency gains to be more cogent than AF’s, the condition is likely to result in CPE at the targets being lower than it would be without it.

[216] As to the duration of the condition, the grace period expires at the end of October 2020. The appellants proposed the condition at a time when the grace period had just started to run. If we impose a longer duration than the remaining extent of the grace period, we would be making an allowance – admittedly of limited duration – for the possibility that the targets might, but for the merger, have become HDP-compliant. Since I do not think that this is very likely, and since they are unlikely to achieve procurement efficiencies of the same magnitude as Mediclinic, a five-year duration is likely to ensure that, for slightly more than four years following the expiry of the grace period, the targets will be materially cheaper as a result of the merger than if it had been prohibited, and that once the condition lapses the targets will still be cheaper though not to the same extent.

[217] The remedy which the appellants proposed for uninsured patients was that, for five years from the implementation of the merger,

Mediclinic would continue to apply the target hospitals’ base tariff for uninsured patients, escalating annually by no more than CPI, and would continue for the same period to apply the target hospitals’ discount policy. The proposed condition requires the target hospitals to furnish the base tariff and discount policy within five days of approval of the merger.

[218] For reasons which will be apparent, this condition is, on the most plausible scenarios, likely to result in the target hospitals being materially cheaper for five years following the implementation of the merger than they would have been had the merger been prohibited. At any rate, they could not be worse off.

Order

[219] The following order is made:

(a) The Tribunal’s decision of 30 January 2019, prohibiting the merger between the appellants, and the certificate of prohibition issued pursuant thereto, are set aside.

(b) The merger is approved subject to the conditions contained in annexure ‘X’ to this judgment, save that

(i) the ‘Approval Date’ for purposes of annexure ‘X’ shall be the date of this judgment, ie 6 February 2020;

(ii) the initiatives contemplated in clause 3.3.1 of annexure ‘X’ shall include, but not necessarily be limited to, those specified in annexure ‘Y’ hereto.

(c) The respondent shall pay the appellants’ costs of appeal, including the costs attendant on the employment of two counsel.

Vally JA (dissenting)

[220] I have had the privilege of reading the judgment of my brother Rogers JA, which my sister Victor JA concurs with. It is with regret that I record my disagreement with its conclusions and with the order proposed. I would dismiss the appeal with costs. My approach and my reasoning are elucidated here.

[221] The first appellant, Mediclinic (Mediclinic) and the second appellant, Matlosana Medical Health Services (MMHS), are in the

business of providing private medical care in South Africa. Mediclinic, however, is a significant player in that area of business, whereas MMHS is not. Mediclinic owns and manages 48 multi-disciplinary private hospitals around the country, whereas MMHS owns and manages only two such hospitals: Wilmed Park Private Hospital (Wilmed) and Sunningdale Hospital (Sunningdale). Both Wilmed and Sunningdale are located in Klerksdorp. Some 50 kms away lies Potchefstroom. There Mediclinic owns and operates one hospital:

Mediclinic Potchefstroom (MC Potch). The travelling time between Potchefstroom and Klerksdorp is approximately 40 minutes. The two appellants (the merging parties) seek to merge their respective businesses. The practical effect of merger would be that Wilmed and Sunningdale (the target hospitals) would become part of Mediclinic and therefore sister hospitals to MC Potch. Mediclinic would

effectively own and manage all three hospitals.

[222] In terms of the Competition Act 89 of 1998 (the Act) this merger would constitute ‘a large merger’ which required the merging parties to notify the respondent, the Competition Commission (the Commission) of their intention to do so. The Commission decided not to support the proposed merger. Relying, amongst others, on the provisions of s 12A of the Act[48], it submitted to the Competition Tribunal (the Tribunal) that the proposed merger be prohibited. After a lengthy hearing the Tribunal,

on 29 January 2019, granted an order prohibiting the proposed merger. It handed down its reasons for the order on 22 March 2019. Its conclusion reads:

‘… we conclude that the proposed transaction is likely to substantially prevent or lessen competition in the relevant market. Since no appropriate remedies were tendered that would effectively address the competition concerns, we prohibit the proposed transaction.’[49]

[223] It is this conclusion that is under attack in this appeal.

[224] Given its sheer size, Mediclinic is capable of negotiating substantial discounts from all its suppliers, and theoretically is able to negotiate (or offer) tariffs and other benefits to medical schemes and to uninsured patients that MMHS cannot on its own match. This advantage, again theoretically, would automatically accrue to MC Potch and would redound to the disadvantage of Wilmed and Sunningdale (assuming for the moment that they are in direct competition with each other). However, over the years both Wilmed and Sunningdale have managed to hold their own in competing with MC Potch. The main reason for MMHS being able to match Mediclinic lies in the fact that it is a member of a non-profit company, the National Health Network (NHN). The NHN is a conglomeration of separate, disparate, independent private hospitals (unlike Mediclinic which is a conglomerate in its own right) which negotiates tariffs and benefits with medical schemes on behalf of all its members. While the matter was before the Tribunal, the NHN had managed to secure a conditional exemption to procure goods on behalf of all its members. This effectively means that the members of NHN would in the near future be acting collectively (through the NHN) to procure goods, such as surgicals and ethicals, from their respective suppliers. The purpose of securing the exemption is to acquire the benefit of size in the market place that the individual members of the NHN lack. Prior to the exemption each member of the NHN such as MMHS was required to procure these goods on its own. As mentioned above, given its puny size in comparison to Mediclinic, MMHS would not be able to match the discounts Mediclinic would have secured. By the time the exemption was secured the actuaries employed by Mediclinic, as well as the Commission, had already completed their analyses of the respective efficiencies of MC Potch and Wilmed and Sunningdale. They had, understandably, not taken note of the new situation.

[225] At the inception of the Tribunal hearing the merging parties proposed a remedy to deal with the objections of the Commission to the merger. This remedy was withdrawn after some witnesses highlighted certain deficiencies therein. After the lengthy process of receiving evidence was complete, and on the eve of the day for which argument was set, the merging parties proposed a new set of remedies. This was done without any prior notice to the Commission or the Tribunal. It necessitated a postponement of the hearing.

[226] The Tribunal, rightly in my view, expressed its strong disapproval of the manner in which the merging parties conducted themselves. The merging parties provided an inadequate explanation as to why they only brought the proposal at such a late stage in the hearing,

especially after the witnesses had already testified. The consequence was that the proposal could not be put to the various witnesses for their comment. The Tribunal believed, again correctly, that this made it very difficult for it to assess the utility or value of the proposal. Nevertheless, the merging parties were allowed to table a final proposal just before the hearing concluded. They tabled two alternative sets of proposals. The Tribunal refers to them as the ‘Mediclinic minus remedy’ and the ‘MMHS plus remedy’.

[227] To assess the potential impact of the proposed merger it is of course necessary to scrutinise the relevant product and geographic markets within which the merging parties operate.

[228] The parties had agreed that the product market was the provision of services by private, multi-disciplinary, acute, inpatient hospital services. Outpatient or day-care services were not regarded as part of the relevant market. There was a controversy about this exclusion. The merging parties wanted the services to be included. However, the evidence presented by each party’s economist was harmonious on the issue – that it should be excluded. I do not believe that much should be made of the controversy. The Tribunal explored the evidence, engaged with the submissions of the merging parties and correctly concluded that on the evidence the exclusion of the day-care services from the product market was appropriate. There was no misdirection on its part. Moreover, the Tribunal pointed out that even if the day-cases were included its conclusion regarding the effect of the proposed transaction would not change. Here too, I can find no misdirection on its part.

The geographic market

[229] We know that Wilmed and Sunningdale are located in Klerksdorp which is 50kms away from MC Potch. There are two more multi-disciplinary

private hospitals in the vicinity: Mooimed in Potchefstroom and Life Anncron in Klerksdorp. To assist the Tribunal to decide what the relevant geographic market is the Commission offered a view that focussed on municipal demarcations. In this regard it invited the Tribunal to hold that the said market consists of three local municipalities, namely the city of Matlosana and JB Marks local

municipalities (conveniently referred to as the MaJB area). The said area covers both Potchefstroom and Klerksdorp. The choice of the MaJB area was motivated by three factors: (i) it should cover only multi-disciplinary private hospitals; (ii) it should cover all such hospitals that enjoy at least one percent of the total number of patients in the area where the three hospitals – Wilmed, Sunningdale and MC Potch – operate; and (iii) it should cover the area where any other multi-disciplinary private hospital operates which also competes with all three of these hospitals.

[230] The merging parties on the other hand offered a view that at one level is very broad and at another very narrow. The broad view is that hospitals compete at a national level when it comes to determining the tariffs that should be charged, especially to medical schemes, and therefore the geographic market is national. While at the narrow level their perspective was that Potchefstroom and Klerksdorp were separate geographic markets as the patients they seek to attract tend not to (for convenience reasons) travel outside of their respective localities. Since Potchefstroom and Klerksdorp each constitute a separate locality the hospitals located in one do not compete with hospitals located in the other. The non-price competition factors are, therefore, localised. They added that as medical schemes are required to provide their members with reasonable access to a Designated Service Provider (DSP) these schemes are less likely to require their members to travel from Potchefstroom to Klerksdorp (and vice versa) to obtain their required services from a DSP. Hence, Potchefstroom and Klerksdorp constituted separate geographic markets. They nevertheless conceded that medical schemes regarded a reasonable distance of travel to a DSP as being 50km, which is the distance between Potchefstroom and Klerksdorp. They contended further that if the Tribunal was minded to adopt the perspective of the Commission and extend the area (to the MaJB area) then it would be more appropriate to widen the area to include localities east of Klerksdorp and west of Potchefstroom as patients are drawn from these areas. The MaJB area includes only urban localities. The expanded area would include rural localities.

[231] The Tribunal took special note of certain documentary evidence received from the merging parties, namely their strategic documents. It held that these were the best guide in establishing what the relevant geographic market was. The Tribunal thoughtfully scrutinised these documents. The one document prepared revealed that MC Potch had understood and regarded the hospitals in Klerksdorp as its

competitors. Another document specifically identified Wilmed and Sunningdale as competitors to MC Potch. Hence, the merging parties’

own documents contradicted their proposal that the geographic market be viewed narrowly by treating Potchefstroom and Klerksdorp as separate and distinct markets. Why they would say one thing to each other internally and another to the Tribunal was never explained.

There was though an averment by one of the merging parties’ witnesses to the effect that MC Potch management did not perceive

Wilmed and Sunningdale as its competitors because of their geographical distance. The averment was essentially discredited during cross-examination. In the face of all this evidence the Tribunal took the view that the most reliable source of what the merging parties regarded as the relevant geographic market was their own strategic documents ‘since they were prepared based on the commercial realities at the time and not for purposes of the merger proceedings’.[50]

[232] It is not uncommon for a competition regulatory body to determine the issue of the appropriate geographic market for merger cases by having regard to the SSNIP (small but significant and non-transitory increase in price) test. The test focusses on a hypothetical

monopolist that is able to increase its price without undermining or threatening its profits. In the present case it is aimed at

establishing the distance customers (patients in this case) are willing to travel to off-set the increase in price or deterioration

in non-price factors such as drop in quality of service. When questioned by a member of the Tribunal as to whether it would be

reasonable for patients to travel from Klerksdorp to Potchefstroom (and vice versa) to access a hospital there should the hospital in Klerksdorp (or in Potchefstroom) raise its prices, one of the merging parties’ witnesses conceded that it was reasonable for the patient to bear the inconvenience of the travel in order to overcome the burden of the increased price.

[233] The Tribunal, however, did not leave the matter there. It looked at the evidence presented to it by some of the medical schemes about the geography and demography of Potchefstroom and Klerksdorp. This evidence demonstrated that it was reasonable to expect patients to travel a distance of 50 kms in the event of a SSNIP – in other words should a hypothetical monopolist in Potchefstroom increase its price or should it allow the quality of its service to deteriorate, the patient is likely to travel to Klerksdorp for the service in order to mitigate the effect of a price increase or deterioration in service.

[234] On the basis of the results of the SSNIP test, the contents of the strategic documents of the merging parties and the evidence of some of the medical schemes, the Tribunal came to the conclusion that the relevant geographic market was the MaJB area.

[235] I simply cannot see where it went wrong in this regard.

Tariffs charged by Mediclinic and MMHS

[236] The next issue considered by the Tribunal was the impact of the proposed transaction on the tariffs charged to the insured and uninsured patients. Insured patients pay whatever tariffs their respective medical schemes have secured through negotiation with MMHS and Mediclinic. The tariffs in themselves are not reflective of the true cost of the services provided by the three hospitals. The best measure of true cost is referred to as a cost per event (CPE). It consists of the cost of theatre time, accommodation, ethical and surgical consumables. This will be dealt with later.

[237] It was not disputed that Mediclinic’s tariffs applicable to medical schemes were higher than those of the NHN (which is applied by Wilmed and Sunningdale) and in some cases Mediclinic’s tariffs were significantly higher. However, taking note of the proportion of tariffs on overall costs, it was found that the overall charge by Mediclinic is approximately […]% higher than NHN. As for the uninsured patients, MMHS grants larger discounts to these patients than Mediclinic. The merging parties’ internal documents confirmed this. In fact, MMHS’s tariffs for uninsured patients are […]% - […]% lower than that of Mediclinic. The merging parties proposed that it be made a condition of the approval that the MMHS tariffs be retained at Wilmed and Sunningdale for a period of 5 years post the merger. The Tribunal was not persuaded that holding on to the prices of MMHS for 5 years would be sufficient to mitigate the very real increase in tariffs that would eventuate once the merger is approved. It came to the conclusion that the proposed merger would result in a price increase for insured (especially those on low cost suites) as well as uninsured patients, but that it would have a particularly weighty adverse effect on the uninsured patients. Of importance for purposes of this appeal though is that the evidence supporting this conclusion was indisputable.

[238] The Tribunal is invested with inquisitorial powers when examining the potential effects of a merger. It exercised these powers to great effect in this matter. It ordered the Commission to undertake a market investigation on the behavioural remedies proposed by the merging parties to establish the concerns and views of customers. The Commission was only able to assess the views of some of the medical schemes. It was not able to assist with regard to views of uninsured patients. This is understandable. The task of establishing the views of uninsured patients would be a near impossible (if not altogether impossible) one. Uninsured patients are disparate and not informed enough to hold a view on the issue. They do not have anyone speaking on their behalf. They do not know the intricate details of the private medical care market. Medical schemes on the other hand are the exact opposite. They are well-informed, have extensive experience in the field of private medical care and speak for the large numbers of patients who are fortunate enough to access private medical care as they are the members of these schemes. The medical schemes that were willing to assist the Commission in this regard were Bonitas, Barloworld Medical Scheme, Old Mutual Staff Medical Aid Fund (Old Mutual), Fedhealth, AngloGold Ashanti Health (Pty) Ltd (AngloGold), Government Employees Medical Scheme (GEMS), Bankmed, Discovery, The South African Police Medical Aid (Polmed), Hosmed Medical Scheme (Hosmed), Selfmed Medical Scheme (Selfmed) and Medihelp. The evidence

received from them was not controversial. It revealed the following:

a. Bonitas raised concerns about the impending tariff increase and the growth of concentration in the hospital sector.

b. Barloworld reiterated what was stated by Bonitas but was a bit more explicit about its fear that Mediclinic would in time abuse its increased market power. It articulated its concern in the following terms:

‘In the event that a negotiation agreement [with regard to future tariffs] is not reached, Mediclinic will typically threaten to charge members upfront at private rates. In an effort to minimise any access or financial impact on its members Barloworld may have to back down to Mediclinic [sic] demands in these circumstances.’[51]

Barloworld was also more explicit in expressing its strong reservations about the highly concentrated nature of the private hospital market in South Africa and the obvious impact of increasing the concentration the proposed merger would have on this already unsatisfactory situation. It was not only concerned about the increase in the bargaining position of Mediclinic but was equally anxious about Mediclinic imposing its tariffs on all future patients of Wilmed and Sunningdale.

c. Old Mutual made the same point as Bonitas and Barloworld.

d. AngloGold is particularly important in the scheme of things. It owns and operates mines in the Klerksdorp and Potchefstroom areas. It employs more than 10 000 employees (the majority of whom are mineworkers) many of whom belong to its medical scheme. The scheme is open to category 4 to 8 employees. It revealed that Mediclinic was unwilling to give the same discount that MMHS grants to it. Mediclinic’s tariff is […]% higher than the discounted tariff it had secured with NHN. Hence, should it lose the benefits of the tariffs that are applicable at Wilmed and Sunningdale the cost of medical care for its employees will increase with a concomitant detrimental effect on their healthcare. It suggested that should the merger be approved it be on the basis that the tariffs set at Wilmed and Sunningdale continue permanently with the necessary annual adjustments for inflation.

e. GEMS, too, indicated that it was concerned about the higher tariffs that would result should the merger be approved. It was equally concerned about the reduction of competition in the Klerksdorp and Potchefstroom areas. On the issue of the remedies proposed by the merging parties it stated that:

‘There is no clear remedy to the reduction in competition in the Klerksdorp and Potchefstroom region. Nor is there a clear remedy to the increase in Mediclinic market power.’[52]

The point being that the remedies proposed will not address what would be a fundamental long-term problem, which has associated problems of price increase and quality decrease in the long term.

f. Discovery, the biggest private medical scheme in the country, was somewhat ambiguous in its response. In 2016 its Principal Officer indicated that it had no concerns about this specific proposed merger, but warned about the creeping mergers in the hospital sector generally. In 2018 on the other hand its new Principal Officer indicated that it was concerned about the adverse effects especially on the price of medical care, which it said was not adequately addressed by the proposed remedies of the merging parties.

g. Polmed made its submissions through a teleconference. Its response was ambiguous as well as strange. It expressed a concern about the high level of concentration in the private hospital market while at the same time said that the proposed merger would increase competition.

h. Hosmed, Medihelp and Selfmed stated that they had no concerns about the proposed merger. Hosmed recorded that the impact of the merger on its members would be small since only a small number of its members utilise the hospitals in the areas - MC Potch, Wilmed and Sunningdale.

[239] This evidence reveals that the major medical schemes and the ones most affected by the proposed merger were anxious about its detrimental effect on the cost of healthcare for their members. The smaller ones, whose members were not as significantly affected,

demonstrated a lack of interest in the proposed merger.

[240] Noting the evidence received from these schemes the Tribunal moved on to focus on whether the proposed merger would, as alleged by some of the medical schemes, result in an increase in concentration in the relevant market. It is common ground that MC Potch held […]% of the geographic market (MaJB area) while Wilmed and Sunningdale collectively held […]% of the same market. Combined they would hold […]% of the market. This, the Tribunal held, would result in ‘significantly [increasing] concentration in the relevant marker and leads to a highly concentrated relevant market.’[53]

[241] The concern for the Tribunal, borne out by the evidence before it, was the consequences that such a large concentration of market power in the hands of Mediclinic would have for the users of private health care services in the MaJB area. The prospect of this increase was a source of anxiety for the medical schemes that had a significant number of members in the area. This was based on their experience in negotiations with Mediclinic on rates and tariffs.

a. Bonitas was explicit in expressing its anxiety, which it said was borne out by its experience in negotiations with Mediclinic. It said that it was a practice of Mediclinic to take full advantage of its market power in a region to extract concessions from it, or to refuse to give it discounts in regions where its market power was not dominant. Mediclinic would demand that Bonitas increase the patient load at other Mediclinic facilities, failing which it would not offer any discounts in the area where it commands significant market power. Thus, assuming Bonitas was able to secure better rates at hospitals in other areas, it would be forced to encourage its members to utilise Mediclinic’s services in these areas, failing which Mediclinic would not offer any discounts to its members in the MaJB area. It would also threaten to demand cash upfront from the patients in the MaJB area if its demands were not met. And this, if implemented, would mean that Bonitas’s members would be severely prejudiced as Mediclinic would, assuming the merger was approved, command such extensive market power in the MaJB area. The prejudice would be real, especially since most of Bonitas’s members would not have the funds to pay Mediclinic in cash and then seek reimbursement from Bonitas. Bearing in mind that it is the healthcare of the member that is the focus here, the extent of the prejudice could be devastating. In short, Bonitas’s concern was that it would be on the receiving end of a hard bargain driven by Mediclinic in future negotiations because of its very strong position in the MaJB area. The consequence is that its members would effectively have to bear the costs of the increased price that, in its view, would in all likelihood eventuate.

b. Fedhealth echoed the sentiment:

‘The increased level of concentration, with lessening of competition will strengthen Mediclinic’s negotiation power’[54]

And:

‘… Mediclinic’s stance on network discounts has historically been that they will offer minimal if any network discount for hospitals in areas where they do not stand to gain in volumes. It is therefore anticipated that this merger will result in Mediclinic offering poor network discounts, but Fedhealth would be obliged to include these hospitals on their networks for member access, which can impact on member contributions.’[55]

[242] Further the economist called by the Commission bore testimony to this[56] and one of the merging parties’ witnesses conceded it.[57]

[243] In essence, the evidence unquestionably revealed that the increased market power of Mediclinic in the MaJB area is likely to have a detrimental effect on (i) the choices available to patients in the MaJB area, and (ii) on the discounts offered to medical schemes in areas where Mediclinic’s market power is not so substantial. The Tribunal recognised this by concluding that the proposed merger would on the one hand restrict choice in the MaJB area and, on the other hand, it ‘may potentially also have adverse effects on consumers outside of the [MaJB] market’[58] In my judgment there is no quarrel with these conclusions.

[244] The substantial growth in the market power of Mediclinic was also, understandably, a source of anxiety for one of the competitor

hospitals in the Potchefstroom area, namely Mooimed Hospital (Mooimed). Mooimed indicated that should the merger be approved it

(Mooimed) would find it difficult to retain its designated service provider (DSP) status with the medical schemes:

‘As a result of the proposed merger it is highly unlikely that any of the independent hospitals in the area would be considered for DSP and PSP [preferred service provider] arrangements in future. Currently some independent hospitals with NHN tariffs have been allocated DSP contracts with many low-cost options of medical schemes and in the event that the proposed merger takes place these hospitals may lose their DSP or PSP status. The DSPs or PSPs are likely to be awarded to Mediclinic and the patients would go to these facilities at an increased cost relative to a similar arrangement with an NHN hospital.’[59]

[245] Mooimed went further and pointed out that the increased market share that Mediclinic would secure post the merger would result in it (Mooimed) being unable to attract the specialists required for it to continue operating. It showed that Mediclinic used its present market power to subdue specialists into operating from its premises and on its terms (fundamental to which is that they must fill its beds) and to prevent them from operating in a competitor hospital, or from directing some of their patients to a competitor hospital such as Mooimed:

‘[…]’ [60]

[246] Mooimed indicated that should the merger be approved, Mooimed would in all probability close down.[61] This would be the result of the increased market power of Mediclinic as Mooimed would not be able to attract the specialists necessary to sustain the hospital. Patients would be forced to follow the specialists who in turn would ensure that the business was directed

to Mediclinic. This evidence was not discredited. It showed that apart from eliminating the competition that Mediclinic currently

faces from Wilmed and Sunningdale, the proposed merger could further eliminate the competition it faces from Mooimed. This consequence

would be particularly deleterious as we are dealing with healthcare here.

The CPE

[247] Having established that the tariffs of Mediclinic are higher (substantially, in some cases) than those of MMHS, the Tribunal

proceeded to examine what the parties maintained was a more accurate measure of health care costs, the CPE. The CPE involves a reasonably simple calculation. However comparisons between CPEs at or between various hospitals are notoriously difficult. They can be markedly different between hospitals belonging to the same group. The factors that affect the CPE are: doctor behaviour, cost of surgicals, cost of ethicals, hospital management and even in some cases patient responses to different therapies. No single factor can explain the divergences. However, it is common ground that Mediclinic enjoys a lower cost of surgicals and ethicals than Wilmed and Sunningdale in its CPE. The price of ethicals are strictly regulated. Each ethical is subject to a Single Exit Price (SEP), ie a supplier selling to two different purchasers has to sell it to both at the same SEP. No price discrimination (whether by way of cash discounts or any other form of discounts) of any sort is tolerated. The only way to reduce the cost of ethicals used in any medical procedure (‘event’ as captured in the CPE) is to substitute patented ethicals with generic ones, where generics are available. Applicable legislation

encourages but does not compel generic substitution. The Tribunal noted that the cost efficiencies that may exist by the decision of Mediclinic to use generics was helpful in pointing out that there was room for cost-cutting measures to be introduced at Wilmed and Sunningdale, but it was not a factor that was merger-specific and therefore was of neutral value.

[248] The cost of surgicals according to the undisputed evidence was […]% cheaper at Mediclinic than it was at MMHS. Should the merger be approved and assuming that Mediclinic transfers these lower costs to the ultimate bill of a patient, it would translate into a cost saving of […]% of the overall costs of healthcare at MMHS as surgicals contribute […]% of the said overall costs. The merging parties placed heavy emphasis on this potential saving, and in particular strenuously contended that while Mediclinic’s

tariffs were higher, this saving in surgical costs would off-set the higher tariff. The conclusion they invited the Tribunal to draw from this is that the merger would ultimately benefit the general public.

[249] The lower price of surgicals at Mediclinic is a result of it utilising its significantly larger buying power than these independent hospitals to its advantage. MMHS, we will recall, relies on NHN to negotiate its tariffs with medical schemes. However, NHN was not able to negotiate with suppliers of surgicals for the independent hospitals that belong to it. Thus these hospitals suffered the disadvantage of size vis-a-vis large hospital groups such as Mediclinic. More recently, and after the potential effect of these lower surgical costs was factored into the analysis of the actuary employed by the merging parties, as noted above NHN was able to secure from the Commission an exemption from the prohibition, thus allowing it to now negotiate prices of surgicals on behalf of all its members. MMHS will no longer be negotiating with these suppliers on its own and will defer the task to NHN, who acting on behalf of all its members is likely to achieve a reduction in prices of surgicals. This was referred to as a relevant counterfactual by some witnesses and was treated as such by the Tribunal.

[250] The merging parties claimed that the exemption would not produce a reduction in prices of surgicals paid by MMHS as the exemption

contained caveats, such as it applying to small businesses only and to businesses that are ‘controlled or owned by historically disadvantaged’ persons. Neither Wilmed nor Sunningdale meet these criteria. The Tribunal rejected both contentions on the grounds that neither of these criteria are precisely set out in the exemption note nor are they immutable. More important for the Tribunal was the fact that Mediclinic accepted that centralised procurement of surgicals has reaped it significant benefits, which demonstrated that the same benefits could be conferred on MMHS once NHN takes advantage of the exemption it secured from the Commission. Whether approached from the perspective of inferential logic (albeit dealing with a prospective future event) or through the lens of a counterfactual assumption this conclusion in my judgment is both coherent and realistic. I see no error there.

[251] Essentially, the Tribunal understood the total evidence on this aspect as revealing that in the very near future the advantage

enjoyed by the larger hospital groups in securing lower prices for surgicals would soon end. Applying this understanding the Tribunal

came to the conclusion that the lower price of surgicals at Mediclinic would not endure for long and so its intended benefit for

Wilmed and Sunningdale post-merger cannot be accepted as a given.

[252] In sum, the Tribunal came to the conclusion that the merging parties overstated the post-merger potential for cost savings in ethicals and surgicals for Wilmed and Sunningdale. It therefore did not accept that the CPEs of these two hospitals would likely decrease because of the merger. But it did not leave the issue there. It considered the actuarial evidence placed before it by the parties.

The evidence of the actuaries on the CPEs

[253] Turning its attention to the evidence of the actuaries employed by both sides regarding their respective calculations of the divergent CPEs at Wilmed and Sunningdale on the one hand, and some of the Mediclinic hospitals on the other hand, the Tribunal critically examined the different methodologies and comparisons made by these actuaries. Noting the importance of differences in sizes of hospitals in the private healthcare sector, and differences in approaches of doctors at each hospital, the Tribunal was alert to the fact that the conclusions drawn by the actuaries were sensitive to the choice of hospitals used as comparators.

[254] The actuaries appointed by the two sides were not able reach an understanding of which particulars hospitals’ CPEs should

constitute a best fit as comparators. The actuary appointed by Mediclinic chose the CPEs of seven Mediclinic hospitals as ideal

comparators to the CPEs of Wilmed and Sunningdale. The CPE of MC Potch was one of them. The actuary appointed by the Commission chose to compare the CPE of MC Potch to that of Wilmed and Sunningdale. The two actuaries also disagreed on how to factor the day-cases into their respective analyses. Another factor that had to be included in the analyses was the impact of collective procurement of surgicals in the future by MMHS. Neither of the actuaries factored this into their analyses.

[255] Since both parties placed considerable emphasis on the question of the CPEs of Wilmed and Sunningdale vis-a-vis the CPEs achieved by some Mediclinic hospitals, the Tribunal examined the issue very carefully. It took particular note of what medical schemes’ views of CPEs were and what they regarded as appropriate comparator hospitals when drawing conclusions on the performance of a particular hospital. In this regard it sought to establish from the medical schemes how they perceived the CPEs of Wilmed and Sunningdale vis-a-vis the CPE of MC Potch. The evidence it received was that three of the medical schemes, Discovery, Bonitas and GEMS understood the CPEs of Wilmed and Sunningdale as being better than that of MC Potch, while Medihelp and Polmed took the opposite view.

[256] At the hearing a substantial amount of intellectual energy was consumed by parties criticising each other’s actuaries’ methodologies and conclusions. The Tribunal scrutinised these criticisms carefully, took particular note of the concessions made by the actuary appointed by Mediclinic and came to the conclusion that his analyses contained an inherent ‘flaw’ in that it was based on a comparator that was not appropriate. The flaw ruined the utility of his analyses altogether.[62] Similarly, with the actuary appointed by the Commission the Tribunal found that his selection of comparator hospitals was not appropriate.[63]

[257] Taking note of the views of the medical schemes, the robust disagreement between the respective actuaries as to the appropriate comparators, the virulent criticisms mounted by each side of the other’s actuaries’ methodologies and conclusions, and the failure of the actuaries to take note of the conditional exemption that NHN had secured to procure surgicals collectively,[64] the Tribunal came to what clearly is a very sensible conclusion, that no weight should be attached to the actuarial calculations of either side.[65] Bearing in mind what is required of an independent expert testifying in a quest to assist the Tribunal in its determination of the issues before it[66], I believe the Tribunal was correct to find that the evidence of the two actuaries was of no value.

The non-price effect of the proposed merger

[258] Factual evidence was presented by both parties concerning the non-price effect of the proposed merger. The dispute between the parties was on whether the proposed merger would result in a deterioration of factors such as clinical quality or patient experience. The evidence was extremely limited in scope and at times based on subjective perceptions. However, it was common cause that Wilmed in particular has succeeded in earning a reputation for providing a quality of care that is superior to that of MC Potch. On the basis of this evidence the Tribunal concluded that - despite the evidence being very limited in scope as well as hinging substantially on subjective perceptions - both Wilmed and Sunningdale provided better quality of care and achieved greater patient satisfaction than MC Potch. From this conclusion (and assuming all things remain equal[67]) the Tribunal extrapolated that ‘from a non-price competition perspective, the proposed transaction will likely lead to a deterioration in patient experience at [Wilmed and Sunningdale] if the merger is implemented.’ [68]

[259] There can be little doubt that the Tribunal’s conclusion is an inference drawn from the very limited evidence at its disposal. I am not convinced that it is a correct conclusion given the meagre and insubstantial evidence that was placed before it. There was however evidence, (again in the form of an inference drawn from the fact that the proposed merger would doubtlessly increase the market power of Mediclinic), that apart from the possibility of increased prices for patients, there would be a concomitant decrease in the incentive to improve patient experience or even the quality of the healthcare once Mediclinic secures dominance. This is so especially since the patient experience and quality of care it provides has been found not to match that of Wilmed. It is not an illogical inference but, in my view, not much weight should be attached to this.

Barriers to entry

[260] It was generally accepted by both the merging parties and the Commission that the barriers to entry in the private multi-disciplinary hospital market is very high. Not only is it extremely onerous and time consuming to secure a licence to operate a private hospital but it is also very difficult for an existing private medical hospital to get a licence to increase the number of beds it is allowed to hold. It is notorious that the process can take many years and even a decade before the licence is secured.[69]

[261] In these circumstances it is highly unlikely that any independent, private multi-disciplinary hospital would be entering the MaJB market in the near future and replace Wilmed or Sunningdale should the merger be approved. The Tribunal was particularly sensitive to this fact.

The proposed remedies

[262] While not admitting that the proposed merger would substantially reduce the competition for private healthcare services in the MaJB area, or even result in an increase in tariffs for insured and uninsured patients, in September 2018 the merging parties proposed a remedy which was aimed at mitigating any of the adverse effects of a possible lessening of competition.

[263] The proposed remedy reads:

‘After the implementation of the merger and for a period of three years, Mediclinic shall ensure that the base tariff which it applies in respect of services at the target hospitals for each Medical Scheme which reimburses Mediclinic on a fee for services basis, shall be the base tariff which it applies in respect of those services at all other Mediclinic hospitals for that Medical Scheme, discounted by […]%.[70].

[264] The Tribunal ordered the Commission to seek out the views of medical schemes as to the viability of the proposed remedy. Nine medical schemes responded, with seven indicating that they had significant difficulty with the proposal. Most of them pointed out the inherent dangers of increased market power that arose from ‘creeping mergers’ in the healthcare industry. This refers to where a series of takeovers have taken place which individually raise no anticompetitive concerns but when taken collectively have shown to have significant anti-competitive effects. An individual merger may not substantially raise the market power of the merged entity but over time the merged entity can acquire a very significant increase in its market power by methodically taking over one entity at a time. It is a conveyor belt moving towards greater market power and market domination. It essentially involves a gradual accumulation of market power and has been successfully utilised by the private hospitals in the country since the late 1990’s. After receiving the responses from the medical schemes the proposal was dispensed with. It was replaced by the merging parties on 7 January 2019 with two new possible remedies. These have been characterised by the Tribunal as (i) the MMHS plus tariff remedy, and (ii) Mediclinic minus tariff remedy.

[265] The proposed MMHS plus remedy reads:

‘Following the Implementation Date, and for the remainder of that calendar year, Mediclinic shall ensure that the tariff which it applies in respect of services at the Target Hospitals for each Medical Scheme (or particular option, as the case may be) that reimburses Mediclinic on a fee for service basis, shall not exceed by more than 3% the tariff which at that stage applies to those services at the Target Hospitals in respect of that Medical Scheme (or option, as the case may be) in terms of NHN 57/58 Tariff Schedule.’[71]

[266] The remedy depends upon Mediclinic having access to NHN confidential tariff files which the NHN was not prepared and cannot be forced to release to Mediclinic. The merging parties had no answer to this. The Tribunal found the remedy to be unviable. It was neither practical nor enforceable. On this basis it rejected it. I, too, would come to the same conclusion.

[267] In the alternative, the Mediclinic minus tariff remedy was proposed. The relevant portion reads:

‘3.1 Insured Patients at [Wilmed and Sunningdale]

3.1.1 Mediclinic shall ensure that the tariff which it applies in respect of services at the Target Hospitals for each Medical Scheme (or particular option, as the case may be) that reimburses Mediclinic on a fee for service basis, shall be the tariff which it applies in respect of those services at all other Mediclinic hospital for that Medical Scheme (or option, as the case may be) in terms of the Mediclinic’s 57/58 Tariff Schedule, discounted by 3%

…

3.1.6 [The 3% discount referred to in 3.1.1 above] shall be applicable from the Implementation Date and for a period of (5) five years subject to paragraph 3.1.7 below

3.1.7 At any time during the (5) five year period indicated in paragraph 3.1.6 above, the application of the Conditions is this paragraph 3.1 to any Medical Scheme (or option) shall be suspended, varied or terminated by agreement between Mediclinic and the Medical Scheme concerned.

3.2. Uninsured patients of [Wilmed and Sunningdale]

3.2.1 Within five days after the Approval Date [Wilmed and Sunningdale] shall furnish Mediclinic with the base tariff and discount policy which are currently applied in respect of uninsured patients at [Wilmed and Sunningdale]

3.2.2 Upon the Implementation Date and for a period of 5 (five) full years thereafter, Mediclinic shall ensure that in respect of uninsured patients at Wilmed and Sunningdale:

3.2.2.1 The base tariff which it applies shall be the base tariff which is currently applied in respect of uninsured patients at [Wilmed and Sunningdale], escalated at the commencement of each calendar year by no more than CPI [consumer price index]; and

3.2.2.2 Discounts on the base tariffs referred to in paragraph 3.2.2.1 above shall be offered in accordance with the discount policy which is currently applied in respect of uninsured patients at [Wilmed and Sunningdale]’[72]

[268] In addition, Mediclinic will honour all alternative reimbursement mechanism/model (ARM) and Designated Service Provider (DSP) contracts that the Wilmed and Sunningdale have with the medical schemes.

[269] The substance of the remedy for insured patients is that Mediclinic would for a period of five years give a 3% discount on its tariffs to any medical scheme that reimburses it for providing services at Wilmed or Sunningdale to that medical scheme’s members. The medical scheme and Mediclinic are free to negotiate an alternative system of reimbursement within the five year period. The reason the discount is fixed at 3% is because Mediclinic believes that on the analysis of its actuary the procurement savings it has achieved over Wilmed and Sunningdale is […]%, but given that NHN has secured an exemption it is likely that this procurement saving would translate to only a 3% advantage for Mediclinic. On this understanding it offered a 3% discount to the medical schemes for a period of five years. The Tribunal rejected the assumption that the exemption would only reduce the advantage of Mediclinic’s procurement costs by 3%. On the available evidence it came to the conclusion that a more realistic outcome would be that the advantage would disappear altogether. [73] Hence, the Tribunal found a discount of 3% (‘size of discount’) to be wholly inadequate.

[270] The Tribunal had a more fundamental problem with the proposed remedy, which is that if the merger were to be approved the market would fundamentally change. The change would be long-term if not permanent. In the words of the Tribunal:

‘… the proposed remedy is not only inappropriate in terms of the size of the discount off [sic] the tariff, it is also flawed in principle because it does not address the source of the competitive harm. It does not take the likely post merger change in bargaining dynamics as a result of the proposed transaction into account and does not address the issue of post merger regional dominance in the relevant market. Since the proposed behavioural remedy fails to address the source of the competitive harm resulting from the proposed transaction, at a principle or absolute level, even without considering the further elements, we find that the proposed remedy is not appropriate.’[74]

[271] The Tribunal found two further problems with this proposed remedy, namely (i) the duration of five years was wholly inadequate and, (ii) the policing of the remedy was impractical.

[272] On the first issue, it took note of the views of medical schemes, some of whom (Discovery, Bankmed and Fedhealth) submitted that the discount should not be time-restricted at all. Others (GEMS, Bonitas and Momentum Health) submitted that the period of five years is acceptable while one (Polmed) requested that it be for a period of seven years. The Tribunal came to the conclusion that given the extremely high barriers to entry, no period (five or seven years) would suffice, especially since the product in question is the healthcare of the populace.

[273] There can be no doubt that the merger would fundamentally alter the private healthcare landscape in the MaJB area for a considerable

length of time, if not permanently. All the adverse effects of this increased market power of Mediclinic (referred to above) would then have to be borne by the populace that rely on the healthcare services in that area.

[274] The proposed remedy would at best, assuming the merging parties are correct that its procurement advantage (which according to it would only be 3%) would still prevail despite the exemption secured by NHN, result in medical scheme members retaining the tariffs of Wilmed and Sunningdale for a period five years post the merger. Thereafter they would be at the mercy of Mediclinic.

[275] On the second issue, the Commission submitted that effective monitoring of such an order would involve the employment of independent auditors and actuarial experts and as a result would be impractical. The merging parties were not able to gainsay this.

[276] The substance of the proposed remedy for uninsured patients of Wilmed and Sunningdale is that they would be charged the same tariff that these two hospitals charge them now, but this rate would only prevail for a period of five years. Moreover, it would increase annually at a rate no higher than the CPI. They would also receive any discounts that Wilmed and Sunningdale offer but, again, only for a period of five years.

[277] The same problems identified with the aspect of the proposed remedy applicable to insured patients apply in this case. Hence, it, too, was found to be inappropriate and, in my view, rightly so.

Constitutional importance of healthcare and the Public Interest

[278] Finally, of fundamental importance is the nature of the service that forms the subject-matter of this case: healthcare. Every individual needs healthcare: it is basic. It is a protected right in terms of the Constitution of the Republic of South Africa, Act 108 of 1996 (the Constitution). Section 27 of the Constitution provides that ‘(e)veryone has a right to have access to health care services, including reproductive health care.’ The Tribunal was acutely aware of this and incorporated the constitutional protection of access to healthcare services into its consideration. It did so as part of its focus on the public interest.[75]

[279] The evidence, in my view, demonstrates that the proposed merger would undermine rather than advance the constitutional right of the populace in the MaJB area to healthcare. This is because the proposed merger would make access to healthcare in that area more rather than less onerous. It would therefore not be in the public interest to approve the proposed merger.

Conclusion

[280] In the light of s 12A of the Act two questions were posed in this matter: on a conspectus of all the evidence is there a likelihood of a substantial lessening of competition should the proposed merger be approved? If not, is there a public interest consideration that militates against approving the merger? Both questions were squarely addressed by the Tribunal in the carefully considered reasons it has provided.

[281] The Tribunal came to the conclusion that the proposed merger would result in a substantial reduction in competition in the provision of healthcare services in the MaJB area, which (i) in all likelihood would cause serious and possibly irreversible harm to patients in that area, (ii) could harm patients in other areas where Mediclinic’s market power was not substantial. There is coherence and consistency in the logic of the Tribunal. But, and more importantly, its conclusions are ensconced in the constellation of the evidence. In my judgment the conclusions reached are correct. I, therefore, find no reason to disturb its order.

[282] On the analysis above I would dismiss the appeal with costs of two counsel.

APPEARANCES

For Appellant J Butler SC and M Norton SC Instructed by Cliffe Dekker Hofmeyr Inc 11 Buitengracht Street Cape Town For Respondent N H Maenetje SC (with him Y Ntloko) Instructed by Competition Commission DTI Campus, Block C, Mulayo Building 77 Meintjies Street Sunnyside, Pretoria

[1] The Tribunal recorded (para 111 of its reasons) that the Commission contended that the relevant geographic market was no broader than the ‘MaJB’ area, 'consisting of Ditsobotla, City of Matlosana and JB Marks local municipalities' (my underlining). Although this is indeed what appears in para 24 of the

Commission's economists' first report at 9/869 (where they summarise their conclusions), it is clear from the detailed discussion in that report of the geographic market that the economists identified the geographic market as no broader than the City of Matlosana and JB Marks local municipalities, (see paras 69 and 70 at 9/894-5), which explains the acronym they chose. These two local municipalities form part of the Dr Kenneth Kaunda District Municipality. Although the economists considered the possibility of including the third local municipality forming part of the Dr Kenneth Kaunda District Municipality, viz the Maquassi Hills Local Municipality, they did not consider including the Ditsobotla Local Municipality (Lichtenburg) in the candidate market – Distobotla forms part of the Ngaka Modiri Molema District Municipality. In the slides which Dr Mncube presented as part of his oral evidence, the MaJB area was again defined as comprising the City of Matlosana and JB Marks local municipalities (see at 26/2658 and 26/2663-4). In argument, the appellants' counsel took it for granted that this was the geographic market defined by the Commission. I thus take the reference to Ditsobotla to be an oversight.

[2] Paras 40-41 at 9/783.

[3] The data comes from Tables 2 and 3 of Econex’s report at 12/1175 and 1177.

[4] Para 2.3 at 12/1247.

[5] Econex report para 89 at 12/1187; Theron 40/4104-5; Steenkamp 36/3751.

[6] At 4/380.

[7] Econex report para 34 at 12/1167-8.

[8] Raval and Rosenbaum Why is Distance Important for Hospital Choice? Separating Home Bias from Transport Costs (Working Paper, June 2018) (http://www.devesh-raval.com/distance.pdf).

[9] Elzinga and Swisher ‘Limits of the Elzinga-Hogarty Test in Hospital Mergers: The Evanston Case’ (2011) Vol 18 No 1 International Journal of the Economics of Business 133-146.

[10] Econex’s tables 2 and 3 at 12/1175 and 1177 contain admission numbers. Table 2 lists the Klerksdorp patient admissions (I include Orkney) over a three-year period (2015-2017) at MC Potch ([…]), Sunningdale ([…]) and Wilmed ([…]). In this segment of Klerksdorp patients, only […]% travel to Potchefstroom. This data does not take into account the Klerksdorp patient admissions at Anncron and MooiMed. The data in that regard in table 3 is for a five-year period (2010-2014). This shows […] Klerksdorp admissions at Anncron and […] Klerksdorp admissions at MooiMed. On the assumption that the travel patterns would not have been markedly different in 2015-2017, one could take three-fifths of the Klerksdorp admissions in table 3 (ie […] for Anncron, […] for MooiMed) in order to add them to the Klerksdorp admissions in table 2. This results in total Klerksdorp patient admissions over a three-year period of […], of which only […] ([…]%) were admitted to Potchefstroom hospitals.

[11] 14/1489.

[12] 15/1503.

[13] 40/4109-4110; 41/4167-4169.

[14] 6/545.

[15] 32/3361.

[16] 32/3351-2.

[17] 33/3372.

[18] 9/899.

[19] 32/3360.

[20] Steenkamp 36/3751-2; Van Reenen 28/2921-3.

[21] 34/3547-8.

[22] Para 2.10 at 6/600.

[23] Para 2.8 at 6/599.

[24] 31/3222-3.

[25] 23/2404.

[26] Para 151 at 44/4499.

[27] Paras 161-298 at 44/4502-4535.

[28] Paras 455-458 at 44/4574-5.

[29] 6/536.

[30] 6/542-543.

[31] 12/1282.

[32] 40/4131

[33] I treat IAC’s report dated 21 November 2016 as its first report, which is how IAC and AF referred to it. IAC had prepared an earlier but substantially similar report dated 31 August 2016.

[34] Dated 27 April 2017.

[35] Dated 8 June 2017.

[36] Dated 3 October 2017.

[37] Dated 4 April 2018.

[38] See IAC’s presentation at 26/2728.

[39] Dated 25 April 2018.

[40] Transcript 38/3958, 39/3968

[41] Transcript 39/3970

[42] Transcript 39/4027-4031.

[43] 41/4206.

[44] Transcript 39/4020-4021.

[45] At 6/535-536.

[46] Paras 299-312 at 44/4535-8: The Tribunal headed this part of its decision, ‘Non-price competition’.

[47] This is how the Tribunal in para 300 (44/4535) described Mediclinic’s submission. The Tribunal did not express disagreement with this as a general description.

[48] The section reads: ‘12A. Consideration of mergers (1) Whenever required to consider a merger, the Competition Commission or Competition Tribunal must initially determine whether or not the merger is likely to substantially prevent or lessen competition, by assessing the factors set out in subsection (2), and— (a) if it appears that the merger is likely to substantially prevent or lessen competition, then determine— (i) whether or not the merger is likely to result in any technological, efficiency or other pro-competitive gain which will be greater than, and offset, the effects of any prevention or lessening of competition, that may result or is likely to result from the merger, and would not likely be obtained if the merger is prevented; and (ii) whether the merger can or cannot be justified on substantial public interest grounds by assessing the factors set out in subsection (3); or (b) otherwise, determine whether the merger can or cannot be justified on substantial public interest grounds by assessing the factors set out in subsection (3). (2) When determining whether or not a merger is likely to substantially prevent or lessen competition, the Competition Commission or Competition Tribunal must assess the strength of competition in the relevant market, and the probability that the firms in the market after the merger will behave competitively or co-operatively, taking into account any factor that is relevant to competition in that market, including— (a) the actual and potential level of import competition in the market; (b) the ease of entry into the market, including tariff and regulatory barriers; (c) the level and trends of concentration, and history of collusion, in the market; (d) the degree of countervailing power in the market; (e) the dynamic characteristics of the market, including growth, innovation, and product differentiation; (f) the nature and extent of vertical integration in the market; (g) whether the business or part of the business of a party to the merger or proposed merger has failed or is likely to fail; and (h) whether the merger will result in the removal of an effective competitor. (3) When determining whether a merger can or cannot be justified on public interest grounds, the Competition Commission or the

Competition Tribunal must consider the effect that the merger will have on— (a) a particular industrial sector or region; (b) employment; (c) the ability of small businesses, or firms controlled or owned by historically disadvantaged persons, to become competitive; and (d) the ability of national industries to compete in international markets.’

[48] The section reads:

‘12A. Consideration of mergers

(1) Whenever required to consider a merger, the Competition Commission or Competition Tribunal must initially determine whether or not the merger is likely to substantially prevent or lessen competition, by assessing the factors set out in subsection (2), and—

(a) if it appears that the merger is likely to substantially prevent or lessen competition, then determine—

(i) whether or not the merger is likely to result in any technological, efficiency or other pro-competitive gain which will be greater than, and offset, the effects of any prevention or lessening of competition, that may result or is likely to result from the merger, and would not likely be obtained if the merger is prevented; and

(ii) whether the merger can or cannot be justified on substantial public interest grounds by assessing the factors set out in subsection (3); or

(b) otherwise, determine whether the merger can or cannot be justified on substantial public interest grounds by assessing the factors set out in subsection (3).

(2) When determining whether or not a merger is likely to substantially prevent or lessen competition, the Competition Commission or Competition Tribunal must assess the strength of competition in the relevant market, and the probability that the firms in the market after the merger will behave competitively or co-operatively, taking into account any factor that is relevant to competition in that market, including—

(a) the actual and potential level of import competition in the market;

(b) the ease of entry into the market, including tariff and regulatory barriers;

(c) the level and trends of concentration, and history of collusion, in the market;

(d) the degree of countervailing power in the market;

(e) the dynamic characteristics of the market, including growth, innovation, and product differentiation;

(f) the nature and extent of vertical integration in the market;

(g) whether the business or part of the business of a party to the merger or proposed merger has failed or is likely to fail; and

(h) whether the merger will result in the removal of an effective competitor.

(3) When determining whether a merger can or cannot be justified on public interest grounds, the Competition Commission or the

Competition Tribunal must consider the effect that the merger will have on—

(a) a particular industrial sector or region;

(b) employment;

(c) the ability of small businesses, or firms controlled or owned by historically disadvantaged persons, to become competitive; and

(d) the ability of national industries to compete in international markets.’

[49] The Tribunal’s Reasons at [460].

[50] Tribunal’s Reasons, at [137].

[51] Record, pp 612-613.

[52] Record, at 577.

[53] Tribunal’s Reasons, at [208].

[54] Fedhealth’s response to the Commission on the proposed conditions, Record, at p 2403.

[55] Id at p 2404.

[56] Viva voce evidence of Dr Liberty Mncube (Dr Mncube), Record, at p 4099.

[57] Viva voce evidence of Mr Roland Theodore Buys (Mr Buys) Record, at pp 3521 – 3522.

[58] Tribunal’s Reasons, at [342].

[59] Witness statement of Ms Sussana Catarina van Reenen (Ms van Reenen), Record, at 705.

[60] Witness statement of Ms Sussana Catarina van Reenen, Record, at 709. See also her viva voce evidence, Record, at pp 3341 – 3342.

[61] Viva voce evidence of Ms van Reenen, Record, at pp 3343. The Act, per s 12A(2)(h), commands that this fact be drawn into the mainstream of the analysis.

[62] Tribunal’s Reasons, at [271], [275], [278], [279] and [280].

[63] Tribunal’s Reasons at [294].

[64] The Tribunal referred to this factor as ‘the relevant counterfactual’. According to it the actuaries had to factor it into the analysis as the exemption would in all probability (according to the common cause factual evidence received by the Tribunal) result in MMHS matching the efficiencies enjoyed by Mediclinic because of the advantage it has by virtue of its size vis-à-vis MMHS. See Tribunal’s Reasons at [267].

[65] Tribunal’s Reasons at [294].

[66] See my judgment in Twine v Naidoo [2018] 1 All SA 297 (GJ) at [18].

[67] ‘ceteris paribus’ in the words of economists.

[68] Tribunal’s Reasons at [312].

[69] See viva voce evidence of Ms van Reenen, Record, at 3339ff.

[70] Record, at p 2328.

[71] Record, at 2341.

[72] Record, at p 2335.

[73] Tribunal’s Reasons, at [407].

[74] Tribunal’s Reasons, at [408].

[75] Tribunal’s Reasons, at [441].