A fifty-four-year-old woman, whom the practice will call Mrs Eze, presents on a humid Thursday afternoon at a busy general clinic in Enugu with two days of pain between her shoulder blades, a heaviness she describes as sitting on her upper back rather than her chest, and a nausea she has been treating at home with the antacid her sister-in-law swears by. She is a schoolteacher, twenty years in the classroom, not overweight, no cigarettes, and she mentions — because the clinician asks, and because it seems the more urgent complaint — that the pain flares when she turns to reach for something on the passenger seat of the car. The clinician, who has seen forty patients since the morning and has four still waiting outside the door, examines the paraspinal muscles, finds a tender band along the right side of the thoracic spine, and reaches, reasonably and in under six minutes, for the diagnosis that fits ninety-five per cent of the back pain that walks through that door: musculoskeletal strain. She is sent home with a muscle relaxant and paracetamol. She returns thirty-one hours later, by tricycle, in the middle of an inferior myocardial infarction that had been declaring itself, atypically and quietly, from the very first visit.
The diagnosis that killed the first consultation was not a rare one. It was one of the most common presentations in medicine. The problem was never that the cardiac event was hard to recognise. The problem was that, once the tender band along the spine was found, the search was over. The most dangerous diagnosis is not the difficult diagnosis. It is the diagnosis that was never placed on the list at all — the one closed off, early and confidently, by an impression that fit the first few facts and then stopped taking in new ones.
This piece is about how that happens. Not about the rare presentation or the exotic tropical illness that no one could have been expected to know, but about the ordinary, everyday mechanism by which a competent clinician, working in good faith, arrives at a wrong answer and never realises there was a question still open. It is about the architecture of diagnostic error — and, more usefully, about how a medical system can be built to catch the error the individual mind cannot catch on its own.
The fast mind is usually right, which is exactly the problem
The clinician who diagnosed Mrs Eze's back strain was not careless. He was using the same cognitive machinery that made him a good doctor for the several thousand patients before her. The mind of an experienced clinician runs, most of the time, on pattern recognition — the fast, intuitive, almost pre-verbal process that the literature calls System 1. A patient walks in, a handful of features register, and a diagnosis presents itself before the conscious reasoning has even begun. This is not a flaw. It is the thing that makes an experienced doctor faster and, on the aggregate, more accurate than a novice grinding through a checklist. The fast mind is right the overwhelming majority of the time.
The trouble is that the fast mind does not announce, when it is wrong, that it is wrong. It produces the same sensation of certainty for the correct answer and the catastrophic one. And so the failure mode of the experienced clinician is not that he cannot reason carefully — it is that he does not switch into the slow, effortful, deliberate reasoning that the literature calls System 2, because the fast answer already felt complete. There was nothing to alert him. The certainty was the whole problem.
Several well-described biases sit inside that failure, and it is worth naming them plainly, because a clinician who can name them in himself is a clinician who can begin to guard against them.
Anchoring is the tendency to fix on the first piece of information and weight everything afterward against it. Mrs Eze mentioned that the pain flared when she turned in the car seat. That single detail — mechanical, reassuring, easy — became the anchor, and every subsequent finding was interpreted in its light. The nausea was assumed to be the antacid's territory. The interscapular location was folded into the musculoskeletal story. Nothing that arrived after the anchor was strong enough to move it, because the mind was no longer weighing the evidence freshly; it was checking the evidence against a conclusion it had already reached.
Premature closure is the companion failure: the search stops the moment a satisfactory answer appears. Once "back strain" felt sufficient, the differential — the list of what else this could be — was never built. The clinician did not consciously reject a cardiac cause. He never considered one. The list had one item on it, and the item was plausible, and the door closed. Premature closure is, on the diagnostic-error literature, one of the single largest contributors to serious missed diagnoses, precisely because it is invisible from the inside. You cannot feel yourself failing to consider something.
Availability bias is the tendency to reach for the diagnosis that comes most easily to mind — usually the one seen most recently or most often. In a clinic where the last two hundred back-pain patients had muscular strain, the two-hundred-and-first is overwhelmingly likely to be read as the two-hundred-and-first case of muscular strain. The base rate is real and it is not stupid to lean on it. But the base rate is a statement about the population, not about the person in the chair, and the atypical cardiac presentation is exactly the case where the base rate quietly betrays you.
Confirmation bias completes the set. Once the impression forms, the clinician unconsciously seeks the findings that support it and discounts the ones that do not. The tender paraspinal band confirmed the story and was noted. The diaphoresis that Mrs Eze did not think to mention, and was not asked about, was never entered into evidence at all. The mind, having decided, goes looking for agreement.
None of these is a mark of a bad doctor. Every one of them is a feature of a normal, well-trained, competent human mind operating under time pressure. That is the uncomfortable centre of the whole subject: diagnostic error is not, in the main, a problem of ignorance or negligence. It is a problem of ordinary cognition working exactly as it is built to work, in the small percentage of cases where working as built produces the wrong answer.
What Nigerian medicine adds to the ordinary risk
If the biases above are universal — and they are, documented in teaching hospitals from Boston to Birmingham — then Nigerian clinical conditions do not create the risk so much as amplify it. Four structural features of how medicine is commonly practised here take the ordinary vulnerability of the human mind and remove most of the safeguards that might otherwise have caught it.
The first is the absence of continuity. In the fragmented, visit-by-visit model that most Nigerian patients experience, no single clinician holds the patient's baseline. Mrs Eze had never seen the clinician who examined her, and would likely never see him again. There was no one in the room who knew that this was a woman who did not complain, who had never before presented with pain of any kind, whose arrival at a clinic was itself a signal that something was wrong. A doctor who knows the baseline notices when the story does not fit the person. A doctor meeting the patient for the first time has only the story, and the story, on that first afternoon, was a plausible lie.
The second is the fragmentation of the record. Each doctor, in the common case, starts blind. The previous consultations, the old ECG, the note that a cousin had died suddenly at fifty of something cardiac — all of it lives in the patient's memory, or in a paper card in a drawer at a different facility, or nowhere at all. The clinician is asked to reason from a slice of information and cannot see the pattern that the full record would have made obvious. Fragmented records do not merely inconvenience; they structurally guarantee that each new clinician re-enters the same fog the last one was in, and re-runs the same risk of anchoring on whatever the patient happens to say first.
The third is volume and pressure. A clinician seeing forty or sixty patients in a session is not, whatever his intentions, going to switch every one of them from fast intuition into slow deliberate reasoning. The system rewards the quick label. It has to; the queue outside the door is real, and the patients in it are also sick, and the time spent second-guessing the plausible diagnosis is time taken from someone else who needs it. Under that load, System 2 becomes a luxury the clinician cannot spend freely, and the fast mind — usually right, occasionally catastrophic — does the work almost unassisted.
The fourth is the prestige gradient. In a hierarchical clinical culture, the junior doctor, the nurse, or the pharmacist who quietly suspects the senior's diagnosis is wrong faces a real cost in raising it. The consultant has spoken. To question the consultant is to risk looking foolish, or insolent, or both, in front of colleagues one has to work with for years. And so the observation that might have reopened the case — the nurse who noticed the patient was grey and sweating in the waiting area, the junior who thought the pain sounded cardiac — dies unspoken. The steepest hierarchies are the ones in which the fewest challenges to a wrong first impression ever get voiced. The prestige that is meant to signal competence becomes, in this specific way, a mechanism for suppressing the correction that competence would have welcomed.
The defences are structural, not heroic
The instinct, faced with all of this, is to demand that clinicians simply be more careful — to think harder, to always consider the cardiac cause, to never anchor. That instinct is understandable and almost entirely useless. You cannot solve a problem of ordinary cognition by exhorting people to have extraordinary cognition. The clinician who missed Mrs Eze's infarction was already trying his best; "try harder" was not available to him as an intervention. The defences that actually work are not properties of the heroic individual. They are properties of the system the individual works inside.
The first defence is the second opinion, made routine rather than exceptional. A second opinion is not an accusation that the first doctor was incompetent. It is the structural cure for premature closure — a fresh mind that has not yet anchored, building the differential from the facts rather than checking the facts against a conclusion. The reason second opinions catch errors is precisely that the second clinician does not carry the first one's anchor. In a culture where asking for one is treated as an insult to the first doctor, the single most effective safeguard against missed diagnosis is socially expensive to deploy, and so it is deployed too rarely and too late. A well-designed practice makes the second opinion cheap, normal, and frictionless — something the patient is offered, not something the patient has to fight for.
The second defence is team medicine — the plain fact that a differential is safer when more than one mind builds it. When the case is discussed rather than decided, when the nurse who saw the patient sweating can say so without cost, when the junior's dissenting thought is invited rather than tolerated, the list of considered diagnoses grows longer and the odds of the fatal omission fall. This requires deliberately flattening the prestige gradient inside the room. The consultant who says, at the start, "tell me what I might be missing" and means it, converts the hierarchy from a suppression mechanism into a safety mechanism. The diagnosis that one confident mind closed too early is exactly the kind of thing a second, less-invested mind is positioned to reopen.
The third defence is continuity — a named clinician who knows the patient's baseline and therefore notices when the story does not fit. Continuity is not a comfort feature. It is a diagnostic instrument. The doctor who has seen a patient across years carries, in memory, the reference range for that specific human being — knows that this is a woman who minimises, that this level of complaint from her means something it would not mean from a habitual complainer, that a symptom out of character is itself a finding. Continuity is the structural counterweight to the fragmented record: it puts back into the room the longitudinal knowledge that visit-by-visit medicine strips out, and it gives the system the one thing that most reliably catches the wrong frame — a person who can say, "this is not like her."
The fourth defence is AI-assisted diagnosis, used correctly, which is the part most easily misunderstood. The value of a well-built clinical AI is not that it is a smarter doctor. It is that it is a tireless generator of "have you considered—". Where the human mind, having anchored, stops widening the differential, a decision-support tool fed the same presentation will surface the cardiac cause alongside the musculoskeletal one, not because it is wiser but because it does not experience the sensation of certainty that shuts the human search down. The machine does not anchor. It has no ego invested in the first impression, no forty patients waiting, no consultant to defer to. Used as a challenge to the confident mind rather than a replacement for it — a second reader that widens the list rather than narrowing it — AI becomes a structural check against premature closure of exactly the kind Mrs Eze's case needed. The critical condition, and it is not negotiable, is that a named human clinician remains accountable for the decision. The tool's job is to reopen the question; the clinician's job is to answer it. An AI that is trusted to close the case is simply a faster way to make the same error. An AI that is used to force the case back open, with a human deciding, is the closest thing medicine has to a reliable defence against its own confident blind spots.
Safety is a property of the system, not the brilliance of the individual
The lesson of the missed infarction is not that the clinic needed a better doctor. It needed a better system — one in which the plausible-but-wrong first impression would have been challenged before the patient walked back out into the afternoon. Every one of the defences above is a way of building, into the structure of care, a mechanism for reopening the question that a single confident mind closed too early. The second opinion reopens it with a fresh clinician. Team medicine reopens it with a fresh voice. Continuity reopens it with a person who knows the patient well enough to sense the wrong note. AI reopens it with a tireless list of alternatives the human stopped generating. None of these depends on any individual being more brilliant than the ordinary excellent clinician the system already has. They depend on the system being designed to distrust its own first impressions.
This is the uncomfortable, freeing truth at the centre of patient safety. The most dangerous diagnosis is the one never considered, and no individual clinician, however good, can be relied upon to always consider everything, because the human mind is not built to feel the weight of what it has failed to imagine. The realistic response is not to demand impossible vigilance from individuals. It is to build systems that assume the individual will occasionally close the search too early, and that are engineered, structurally, to open it back up.
What Kinedic is building
The Kinedic model, operating from Mabushi in Abuja with clinical anchoring at Brookfield Clinics some six hundred metres away, is designed around exactly this problem. Continuity is the foundation: a named physician who holds the patient's baseline over time, so that a symptom out of character registers as a finding rather than a footnote. A low threshold for a second opinion is built into the process rather than treated as a rebuke, so that the cure for premature closure is available before it is needed rather than after harm has been done. AI decision-support is deployed as a deliberate challenge to the first impression — a widener of the differential, with a named human clinician accountable for every decision — rather than as an oracle to be trusted. And continuity of the record means that a wrong initial frame can be revisited, because the information required to revisit it has not been scattered across four facilities and a drawer.
None of this is about being cleverer than the missed diagnosis. It is about building a practice in which the question the confident mind closed too early gets reopened as a matter of routine.
If you are weighing the model — for yourself, a parent, or your organisation — start a conversation with us. The first conversation is private and costs nothing.