Medical Coding: Kenya's Quietest, Most Expensive Problem

Every few weeks, a new headline tells us SHA is hemorrhaging billions to fraud. Then a counter-headline tells us most of those "fraudulent" claims were actually just clerical errors. The Cabinet Secretary blames hospitals. Hospitals blame the algorithm. Patients are left wondering whether their cover will actually pay out when they need it.

Almost nobody is naming the actual mechanism underneath all of this: medical coding. Or rather, the near-total absence of it.

If you work in a hospital, an insurance desk, or a policy office in Kenya right now, this is the newsletter for you.

---

What Medical Coding Actually Is

Medical coding is the practice of translating a clinical encounter a diagnosis, a procedure, a complication into a standardized alphanumeric code that any system, anywhere, can read the same way.

A patient who comes in with a ruptured appendix and gets an appendectomy isn't just "the appendix case" in a well-coded system. They're a specific ICD-10 diagnosis code, paired with a specific procedure code, attached to a specific tariff line. That code means the same thing whether it's read by a claims officer in Nairobi, an auditor in Mombasa, or a researcher compiling national disease burden data five years from now.

Coding is the connective tissue between three things that currently don't talk to each other properly in Kenya's health system: what actually happened to the patient, what the facility is claiming for, and what the regulator or insurer is willing to pay.

Why Now

Kenya didn't suddenly discover it has a coding problem. The problem has existed for decades, quietly, while claims were small, manual, and absorbed by NHIF's slower, more forgiving processes. What changed is SHA.

SHA digitized claims adjudication and ran it at national scale, with automated, often AI-driven flagging of "irregular" claims. That exposed the coding gap overnight. A system that found Health Cabinet Secretary Aden Duale revealing that the government lost at least Sh11 billion through fraudulent and irregular medical claims under SHA is also, by independent analysis, a system where an audit of Sh10 million in claims found an 8.8% rejection rate attributable purely to clerical errors — not fraud. Extrapolated nationally across counties, missions, and private hospitals, that scale of loss from clerical error alone could run into billions of shillings in legitimate revenue that hospitals never recover.

Those two numbers, fraud and clerical error, are being reported as if they're in competition. They're not. They are very often the same underlying failure described by two different people with two different incentives to describe it that way.

This isn't theoretical for the people on the ground either. Teachers and civil servants brought onto SHA cover have publicly described facilities allocating sharply reduced per-visit amounts and rejecting coverage that was previously guaranteed under their old scheme, including for conditions like premature births and chronic illness. The Kenya Medical Association's own leadership has pushed back hard on the blanket fraud narrative, arguing that an automated system that treats a clerical error the same as a criminal act risks destroying the working trust between the state and providers, and is already pushing some private facilities to quietly withdraw from SHA altogether. There's even a court case now testing whether SHA's claims and adjudication functions have proper statutory grounding at all, with a hearing date that's already come and gone this month and more directions still pending.

So "why now" has a precise answer: Kenya built a national, automated, high-stakes claims system on top of a coding foundation that was never built to hold that weight.

It Is Not Just ICD

This is the part that gets flattened in most conversations, including, frankly, in some of KMA's own public messaging. People hear "coding" and think "ICD-10 lookup." That's one layer. There are at least four others sitting underneath it, and Kenya is weak in most of them simultaneously.

Diagnosis coding (ICD-10/11) tells you what was wrong with the patient. This is the layer everyone talks about. Kenya's use of it remains inconsistent and largely confined to mortality and surveillance reporting to the Ministry of Health rather than billing — South Africa is still the only country on the continent with a properly structured national standard governing how ICD-10 should actually be applied in clinical coding.

Procedure coding tells you what was done to the patient — the surgery, the imaging, the therapy. Kenya has no equivalent of CPT or a national procedure classification system. Without it, a claim can say "surgery performed" without specifying which one, at what complexity, with what inputs. That ambiguity is exactly where the C-section example lives: Cabinet Secretary Duale's own audit flagged a private facility reporting that all 500 of its maternity claims were C-section deliveries, against a clinically expected rate closer to 10–15%. That's not an ICD problem. That's a procedure-coding and clinical-governance problem.

Clinical documentation is the narrative layer that justifies the codes. Coding is only as good as the chart behind it. If the clinician's note doesn't support the complexity being billed, no coding standard in the world will save the claim on appeal — and right now, undertrained health records staff are often coding off incomplete charts, which is precisely how legitimate care turns into a "rejected claim."

Case-mix and tariff logic is what actually determines payment. Codes feed into a payment model — fee-for-service, DRG-style bundled payment, capitation. Kenya's tariff structure under SHA is still maturing, which means even a perfectly coded claim can be paid inconsistently because the rules connecting code to cash aren't yet stable or transparent.

Audit and governance is the feedback loop that should catch errors and fraud separately, with separate consequences. Right now Kenya is running that loop through a single automated gate that outputs one verdict — "rejected" — without reliably distinguishing why. That's the single biggest design flaw exposed by everything happening with SHA this year.

Treat coding as "just ICD" and you fix one-fifth of the problem while leaving the other four-fifths to keep generating headlines.

The Implications

For hospitals and clinicians, this is existential, not academic. Every undercoded or ambiguously documented claim is revenue the facility cannot recover, and every claim that gets auto-flagged as "fraud" rather than "error" is a reputational and potentially regulatory risk for the clinician who signed it. The facilities best positioned to survive the SHA transition are not necessarily the ones providing the best care; they're the ones with the most disciplined coding and documentation discipline. That is an uncomfortable but accurate statement about where competitive advantage now sits in Kenyan healthcare delivery.

For insurers and SHA itself, an adjudication system that cannot reliably separate fraud from error is not actually solving the fraud problem. It's just redistributing financial risk onto providers in a way that erodes the provider network SHA needs to function. An automated fraud-detection system is only as trustworthy as its false-positive rate, and right now that rate appears to be high enough to be generating its own public legitimacy crisis, on top of the legal challenge already testing SHA's claims authority in court.

For regulators and professional bodies — KMPDC, the Clinical Officers Council, KMA — this is a jurisdiction-defining moment. Whoever sets the coding and documentation standard effectively sets the terms of how clinical work gets valued and paid for. KMA's public clash with the Ministry over the fraud narrative this year was, underneath the politics, a fight over exactly this: who gets to define what counts as an error versus a crime.

For policymakers, the implication is that UHC's financial sustainability depends on infrastructure most conversations treat as a back-office detail. You cannot build accurate national disease burden data, reliable case-mix-based financing, or a credible fraud framework without a functioning coding system underneath all three. Kenya is currently trying to run a national insurance scheme on a foundation that was designed for paper-based mortality statistics.

For Kenya's specialist and clinical officer workforce, coding competence is quietly becoming a career asset rather than a clerical afterthought. Health informatics, claims adjudication, and clinical coding are emerging as legitimate horizontal career tracks — particularly relevant for clinical officers navigating career progression within a system that badly needs people who understand both the clinical and the administrative side of a claim.

Where This Goes Next

None of this gets fixed by a single circular from the Ministry. It needs a national coding standard with real teeth (not just an ICD-10 mandate, but procedure coding and documentation standards too), an audit function that is structurally separate from the payment-denial function, and a professional body KMA, KMPDC, or a joint structure willing to own coding competence as a core part of clinical training rather than treating it as something health records officers handle in the back room.

The countries that got this right — South Africa most notably on the continent didn't get there by mandating a code list. They got there by building the documentation, training, and audit ecosystem around the code list. Kenya has the code list. It does not yet have the ecosystem.

That gap is where the next two years of SHA headlines are going to keep coming from, unless something structural changes underneath them.

---

What's your experience been as a clinician, an administrator, or a patient with how claims get coded and adjudicated under SHA? I'd like to hear it, particularly if you've been on the receiving end of a rejection you believed was wrong.

LGM Health Consulting works with health facilities, insurers, and policymakers navigating exactly this terrain from coding readiness audits to claims governance design. If this is a live problem for your organization, let's talk.