Human Capital Development

The Importance of Causal Inference – Medical Scheme Racial Bias Allegations

The struggle against racism ought to be data-driven, at least

South Africa’s apartheid system was or is the epitome of racism. It was a system that stripped the black person of their dignity. The 1994 declaration of freedom (not independence) was supposed to usher in a new dawn, a new dispensation that guaranteed true freedom for every inhabitant of the country.

Many non-white people especially the black community feel that not enough has been done to rid the country of the remnants of this evil system. As suggested by the Julius Malema led Economic Freedom Fighters (EFF), the 1994 transaction was cosmetic and did not empower the black person to thrive and be the master of their own destiny. Many in the African National Congress (ANC) and the Democratic Alliance (DA) will dispute the EFF claims and point to the emergence of the “black diamonds”, a black middle class that did not previously exist.

This article focuses on the racial bias struggles of black professionals in the work place as typified by the claims by black medical providers. Many “black diamonds” in corporate and entrepreneurial settings report that they have been prejudiced by the unchanged discriminatory system. They claim that 1994 was a conduit to a more subtle form of apartheid that was pivoted on further economic control by the white minorities. Some disgruntled black professionals have even gone to the extent of accusing Nelson Mandela of selling out at the negotiating table. Did black South Africans get a bad deal?

It is my belief that these accusations should not be ignored. Has corporate South Africa truly transformed? Is racism still as bad as it was pre-1994? Has the economic system changed? What needs to be done in order to address this issue?

Recently, I received the Section 59 Investigation report summary on racial bias in fraud, waste, abuse (FWA) identification and outcomes in the medical insurance sector. This report is as a result of the submissions of racial bias of medical schemes against Black and Indian medical practitioners. The investigation was established in terms of Section 7(a)(b)(c)(d), 8(a) and (k) and 9(2) of the Medical Schemes Act, 131 of 1998.

I was excited as I read the report and this led me to search the length and breadth of the internet to find the arguments made by both the medical schemes and black medical practitioners. It was with great relief when I realised that the Adv Tembeka Ngcukaitobi-led commission had done a sterling job of documenting all the submissions for public consumption on

The high-level investigation outcomes were as follows:

  1. Some of the current procedures followed by medical schemes to enforce their rights in terms of section 59 of the Act are unfair. The commission also found that Black providers are unfairly discriminated against on the grounds of race.
  2. The commission stressed that they had not found evidence of deliberate unfair treatment – the evidence shows the unfair discrimination is in the outcomes. The Constitution regards the form of unfairness that they found to exist as constituting unfair racial discrimination.
  3. The commission did not have power to find anyone guilty. Nor were they appointed to investigate the veracity of each individual claim of unfair treatment and unfair discrimination. But they believed that they would be failing in their duty if they ignored degrading, humiliating and distressing impact of racism against the individuals who testified before them. A part of their function was to provide a platform for the expression of individual experiences of racial discrimination and other forms of unfair treatment.
  4. The commission does not believe that they have covered each and every possible complaint of medical providers against schemes. Their mandate was not exhaustive thus they cannot claim to have explored all possible manifestations of racial discrimination and unfair procedures. But they received sufficient data and information to make informed and reliable conclusions of the patterns of conduct by the schemes.

Given the submissions and the investigation report, it dawned on me that, in as much as the verdict that there was racial discrimination in the outcomes of FWA processes, the submissions by the medical practitioners and their representatives were not good enough.

It is my view that the commission had to stretch itself as much as possible to try and help the medical practitioners make their point of racial discrimination. Given the submissions, I was pleasantly surprised that the commission actually concluded that they had found racial bias in the medical scheme processes. The black man in me was very happy with the outcome but I quickly thought about how we can implement measures to avoid such weak submissions in the future.

The submissions by black medical practitioners through forums like Solutionist Thinkers, Independent Community Pharmacy Association among others were not data-driven. Their submissions lacked the rigour required to revolutionise the industry. I believe that the testimonies of the medical practitioners appealed to the humanity of the commissioners thus their conclusion.

I expected better quantitative and qualitative research to prove beyond reasonable doubt that there is racial discrimination in the FWA process. The burden of proof was with the medical practitioners thus it was a no-brainer that they should have gone for the jugular with their presentations. No stone should have been left unturned.

Being black, I am not oblivious to the existence of racism in its overt and covert nature. I am a victim of discrimination in its many ugly forms.

I know that if I try to buy a house in some leafy suburb of Cape Town or Johannesburg, I am most likely to be charged more than white buyers. I know that my white colleagues who do the same job as I do or are even subordinates are likely to get paid more than I do. The problem I have with these lived experiences of mine is that, there is no data to use in addressing this. It is true but it is tough to argue for in a court of law.

The black medical doctors suffered from the same deficiency in their submissions. They feel they have been racially profiled but do not have sufficient data, both quantitative and qualitative to bring about the necessary change needed in the sector.

On the other hand, the medical aid schemes had troves of data and hired experts to craft science-backed submissions which can stand stress tests in a court of law. I must say the submissions by the medical schemes focused too much on the quantitative and data analytics but were void of the qualitative and humane approach.

My argument in this article is: Black people have to do more in collecting, analysing and presenting racial discrimination data.

As black people, we have to build robust knowledge bases to complement our current efforts in fighting racism. We have to be evidence-driven in our approach to fighting for justice.

I know our ancestors relied on oral tradition to transmit information but it is our duty to supplement that with properly researched documents that are accessible to all. I believe the lack of records contributed a lot to our colonisation.

There are too many unanswered questions that I believe the medical practitioners could have used in their arguments. In my mission to help, I have listed a few:

  • Given that the PCNS database does not use demographic variables like race, how did the medical practitioners qualify/quantify the statement that black providers were being targeted? (At least the commission through Dr Zaid Kimmie constructed a racial classifier using surnames)
  • In their submissions medical practitioners did not prove the correlation or causal link between race and FWA outcomes.
  • The medical practitioners did exhibit an understanding of the structure of medical aids. They referred to medical aids as “white-owned”, which is contrary to the fact that medical aids are non-profit organisations that are voluntary associations owned by the members. They should have been armed with this information and used it to forward compelling arguments, instead they gave arsenal to the medical schemes to accuse them of ignorance.
  • If Discovery is a monopoly, have the medical professionals written evidence-driven research papers and also submitted their concerns to the relevant authorities, in this case, the Competition Commission for adjudication? From the publicly available documents, there is no evidence in the submissions. (I stand to be corrected).
  • The accusation that, “If one investigates, then the others follow”, can easily be proven a case by case time series analysis. Did the practitioners provide compelling evidence for this hypothesis?
  • The medical practitioners could have done a better job at supporting their claim in the submission that, “The system disadvantages ethical practices”. Until supported by evidence, it is a spurious allegation.
  • Last time I checked, South Africa is a free country and the constitution emphasises on freedom of association, thus I baffled by the statement in the providers’ submission that says, “We are coerced to join medical schemes”.
  • Did the medical practitioners request data on the distributions of fees for certain functions to see if cases of “charging too much” were within 3 standard deviations from the mean, assuming the medical schemes use that dispersion measure to detect outliers
  • Have the medical practitioners asked for inclusion in the drafting of remedial actions when FWA accusation is made? They could propose alternative requirements when it comes to proof. Currently medical schemes require things like clinical notes, files, proof of purchase and others.
  • The laws of this country are clear that medical practitioners cannot be forced to acknowledge debt. This is something that every practitioner should be talking to their lawyers and making sure that every transactions with the medical schemes is above board.

Now that I have made my points about the weaknesses of the medical practitioners’ submissions, I now want to look at the commission’s analysis as presented by Dr. Kimmie. I do not envy him for the task that he had in this commission.

Given the poor submissions by the medical practitioners, he did well. The only technical report that I have is the one he published on the 19th of November 2019 and am not sure if there was a subsequent report that incorporated the feedback from the various medical schemes.

Dr. Kimmie did a good job of taking the very broad brief and narrowed it to two specific questions. The initial brief was:

Assist with the interpretation of the algorithms and data used by the various medical schemes and administrators to identify Fraud, Waste and Abuse (FWA) among medical service providers.

The revised brief was:

  1. Is there an explicit racial bias in the algorithms and methods used to identify FWA?
  2. Are the outcomes of the FWA process racially biased? In particular, were Black providers identified as having committed FWA at a higher than expected rate.

From a methodology perspective, I believe the commission did not have enough data to make informed decisions, I suspect that the practitioner testimonies swayed them to come to the conclusions they did. The problem of racial bias is a wicked one which cannot afford to be oversimplified. I must say, I am not privy to his full analysis and the data, I merely used what is publicly available.

If I was in Dr. Kimmie’s shoes, on top of what I have achieved,  I would be asking the following questions:

  • Why did I classify juristic persons as non-Black as my preferred classification method? A provider could have practitioners of different races, a white practitioner could have a practice called Phakama Pharmacy – is that a black pharmacy? (I would have removed the contentious ones). I know you did an alternative test with the strict racial classification as well and obtained a risk ratio of 1.38, so that’s good and shows that the result is not due to classification error.
  • If I could use the geocoding and surnames to estimate race and ethnicity, am I convinced that the profit-seeking medical schemes would not have done that already? Don’t medical scheme analysts/investigators have access to multiple systems that they can use to easily triangulate practice coordinates and use the surname for classification? Did I use Bayesian Improved First Name Surname Geocoding (BIFSG)?
  • How did I adjust for non-white people with “white” surnames and vice-versa?
  • Why did I classify missing values as non-Black? What is the effect of my decision?
  • The evidence from chi-square test of association is composed of two types of associations (causal and non-causal) thus it is merely a correlation of sorts. Why didn’t I pursue establishing causal effects whilst taking confounders into consideration?
  • Did I remember to map out the full FWA workflow from the trigger (tip or system alert) until decision? Where in the workflow are medical scheme likely to establish the identity of the provider?
  • Given that tip-offs constitute on average are about 53% of FWA triggers, should I have focused on tip-off dynamics to make sure that medical scheme do not escape with statements like, “this is out of our control”?
  • Why didn’t I dwell more on the automated fraud detection system to see the parameters used to trigger investigations? Do medical schemes force practitioners to be in a certain pre-defined fees range? If so, is that legal?
  • Given that majority of FWA triggers are tip-offs, are clients of black practitioners more likely to snitch?
  • Is variable X driving tip-offs and the outcome? Isn’t that selection bias? Was there any adjustment for some or all the confounders?
  • What drives tip-offs?
  • Do tip-offs mention practitioner names, location and race?
  • What is the composition of the investigating teams at medical schemes?
  • Should investigators at medical schemes undergo an implicit racial bias test?
  • What is the Average Treatment Effect among black practitioners (ATT) subjected to the FWA investigations?
  • What is the ATE?
  •  What was the unit of analysis? Practitioner or Practitioner-Medical Scheme interactions?
  • How did I account for the difference in black and no-black numbers in the population for contextual purposes?
  • Did I expect black and non-black practitioners to have the same risk ratio?
  • Is an 87% race classification accuracy rate (as reported by GEMS) good enough for the purposes?

Now that I have looked at Dr. Kimmie’s analysis, for completeness, I would like to browse through the submissions by the various medical schemes in defense of accusations of FWA racial bias towards black practitioners. Their submissions are in response to Dr. Kimmie’s report of November 2019.

Let me start with GEMS as submitted by Insight Actuaries and Consultants.  Of all the medical schemes reports, this one left me feeling racially abused. It had no tact and exhibited elements of ignorance of what was presented by in the report.

Their illustrative example to highlight the importance of exposure is under the belt. It is always important to be ethical when coming up with hypothetical situations. Their example spatially isolated black practitioners reminiscent of the apartheid era. In getting their point across it was crucial that they use examples that don’t remind us of the ugly past and make a mockery of the racial bias issue at hand. Exposure is an important point but they should have illustrated in a racially-sensitive manner.

GEMS is spot on when it comes to the assertion that, if 40% of medical practitioners are black and they have 60% of the interactions then the probability to perpetrate FWA is higher by default. It would be interesting to hear how Dr. Kimmie adjusted for these weights.

GEMS should not have lingered too long on the point of Corporatized and State Practices as this was addressed in November report whereby Kimmie used the strict racial classification and his risk ratio was even lower than the 1.47 touted by GEMS.

GEMS did a good job with the desktop audit which resulted in the establishment of an 87% racial classification accuracy rate. This can be used as a baseline for future studies around this topic.

The Bonitas one was the least informative document. The authors focused on their corporate structure. At best it looked like an advertisement for their services. No meaningful technical feedback was provided by the medical scheme.

Discovery estimates about R1.7bn is lost annually on fraudulent claims.

Discovery’s response to the FWA racial bias claims was meaty. The company took the reader on a journey, from defining FWAs to their methodologies. They defined technical jargon that allowed the reader to follow without too much friction. They outlined their FWA identification process clearly and their statistical analysis methodologies for coming up with a risk score was well articulated.

A question that should have been asked by Dr. Kimmie and the commission is whether the tip-off incentives offered to whistle-blowers by Discovery has statistically significant effect on the rate and veracity of the tip-offs, in the light that this is the most popular trigger for FWAs. Who is likely to be enticed by this offer such that they will exploit it?

It will be interesting to know what kind information the whistle-blowers share with the medical scheme and whether this is full of proxies for gender and race.

As I stated before, the medical practitioners did a bad job in stating their case as they did not have data to back themselves up. Discovery did not waste this opportunity to trivialize their case in light of the poor evidence presented. Discovery wrote:

DH conducts over 4,000 investigations per year Data provided to the panel to substantiate allegations of racial profiling are based on unrepresentative and biased samples

  • HealthMan data: based on 320 investigations over approximately 8 years vs >11,000 for DH
  • SAPPF data: based on 22 investigations from 2013 to 2019
  • SASOP data: based on 120 cases. The period of investigation was unclear. – Between 2012 and 2019, DH investigated 297 Psychiatrist practices of which 60% were valid.
  • SA Optometric Association data: based on 8 cases from 2019 – From January to June 2019 DH conducted 155 investigations into Optometrist practices of which 66% were valid.
  • Dental Professions Association data: based on 66 investigations in a retrospective survey that they performed on members (the period was not provided in the study) – In 2018 DH conducted 180 investigations into Dental practices of which 73% were valid.

Given Discovery’s 35 000+ medical provider database, these samples are easy to trivialize. I believe these institutions that are fighting on behalf of the black medical practitioners should improve in this regard.

As highlighted in my questions to Dr. Kimmie above, the “Coloured” population is almost impossible to detect and inter-marriages present a real challenge. I concur with Discovery that the analysis can only be at an aggregated level and cannot be used at the granular statistics levels.

Although Discovery’s feedback was comprehensive, they should not hide behind, “most of our FWAs are triggered by tip-offs, and we don’t know what is happening there”. It is imperative that they establish the key drivers and other dynamics around tip-offs. This will lead to massive gains in the overall fight against FWA.

Medscheme also responded to the report on racial profiling analysis by Dr. Kimmie. Medscheme like all other medical schemes agrees with the findings for question one. There was no evidence of explicit racial profiling in the design or implementation of systems used to identify potential FWAs.

From the data provided by Medscheme it can be seen that 81.71% of whistleblowers are black. Is this a case of there are more black people in the country or black people have a higher propensity towards whistle-blowing?

They show that 72.38% of providers with whistle-blower complaints are black. Do black providers generally serve black patients? What could this be attributed to?

Medscheme has the highest risk ratio (above 3) for both compulsory FWA investigation cases and whistleblower/industry referrals. The statistical racial bias is proportionately higher in the compulsory FWA cases over which Medscheme claims to have no explicit or implicit influence (tip-offs).

In conclusion, I have taken a very long journey and looked at all the stakeholders and their respective submissions.

The medical schemes are adamant that the methodology used by Dr. Kimmie is inadequate as it fails to establish a causal link between race and FWA outcomes. The responses from medical schemes were based on statistical analysis and solely relied on data and did not do enough to acknowledge the context that we find ourselves in. A more humane approach is needed from corporate South Africa in dealing with the cries of the oppressed who are in a system they don’t feel part of.

To the organizations representing black medical practitioners, a lot still needs to be done on the evidence gathering, analysis and presentation front. Their submissions were horrible. For the fight against racism to flourish we have to work hard and not be lazy. Its important to leave no stone unturned in the pursuit of justice. Technology now allows us to be data-informed in our decision making so let us exploit this transition.

As the fight continues, remember, the burden of proof lies with the accuser not the medical schemes.

*Disclaimer: I was not paid to write this. I did not do this under duress.

The struggle against racism ought to be data-driven, at least South Africa’s apartheid system was or is the epitome of racism. It was a system that stripped the black person of their dignity. The 1994 declaration of freedom (not independence) was supposed to usher in a new dawn, a new…
Relevance - 100%
Accuracy - 100%
Clarity - 100%


User Rating: 5 ( 1 votes)

Leave a Reply

No apps configured. Please contact your administrator.