The drug that is not the drug

Start with what actually happens, because the framing of this problem as a criminal matter is the reason hospitals are so bad at it.

A patient is started on an antihypertensive. At review the pressure has not moved. The dose is increased. At the next review it still has not moved, so a second agent is added, and somewhere in the notes appears a sentence about adherence, because the most common reason a medicine does not work is that it was not taken. Eventually the patient is labelled resistant, investigated for secondary causes, and managed as a more complex case than they are.

Every decision in that sequence is correct. Every note is defensible. And if the tablets contained no active ingredient, not one word of it would have been different, because there is no step in normal clinical reasoning at which somebody asks whether the drug was real. The possibility is not in the differential.

So the failure is silent by construction. It does not produce a complaint, an incident form or a coroner’s question. It produces slightly worse outcomes, spread thinly across a lot of patients, attributed to the disease and occasionally to the patient. This site has a name for that shape already. It is the third kind of cost: not measured at all, so never weighed against anything.

Which has a direct consequence for the strategy. Because the problem cannot be detected reliably at the bedside, almost all of the work has to happen before the product reaches a patient. The bedside is where you find out you were wrong, and by then the number of people it happened to is a question of throughput rather than of vigilance.

Put throughput into that and the scale of it becomes clear immediately.

The middle figure is the one that ends the argument for relying on clinical vigilance. A lot is consumed in under a fortnight at an ordinary dispensing rate, so by the time a pattern of non-response is visible at all, the physical evidence has left the building and the patients who received it have gone home.

The third figure is the automation case stated in the only terms that matter. With lot level custody you can name every patient who received the implicated batch in the time it takes to make coffee. Without it, your recall is a press release addressed to three and a half thousand people, almost none of whom were affected, and you will never know which ones were.

None of these numbers say anything about how often this happens. They say what it costs when it does, and they are arithmetic on a stated dispensing rate rather than a claim about frequency.

"Fake drugs" covers four distinct failures, and the reason to separate them is that a control which works on one may do nothing whatever about another.

Two of them are worth dwelling on. Degraded product is the category that hospitals create for themselves: a genuine, correctly manufactured medicine held above its storage temperature for weeks is, at the bedside, indistinguishable from a falsified one. In a market with an unreliable grid and an ambient temperature well above the storage specification, this is not an edge case. It means the power design and the pharmacy are the same subject, which is not how either of them is usually staffed.

Diverted product is the other one, and it belongs with the bad actors rather than with procurement. Genuine medicine, genuine specification, broken provenance: stolen or rerouted stock, often originally donated or bought with public money, re-entering the market at a price that looks like a bargain. The tell is the price. An unexplained bargain in a market this tight is information, and the instinct to treat it as a procurement win is exactly the instinct being exploited.

There is direct evidence from this market, and it is more useful than the national prevalence arguments because it tests the thing a buyer can actually act on.

Researchers bought over twelve hundred medicines from retail pharmacies across all six geopolitical zones of Nigeria and tested a subset by high performance liquid chromatography against a ninety to a hundred and ten per cent acceptance range for active ingredient. Price turned out to be the strongest observable predictor of quality among everything they looked at, which is worth pausing on: not the look of the pharmacy, not the packaging. The price.

In the cheapest band, more than half the samples failed. Among cheap analgesics it was nearly four in five. So procurement on lowest unit cost, which is how most procurement is judged, is not thrift. It is a coin toss conducted on behalf of somebody who is ill and does not know it is happening.

And then the part that gets left out of the summary. In the dearest band, a third still failed. Paying more shifts the odds substantially and it does not make the problem go away, which means a purchasing policy built on buying the expensive one is not a quality system, it is a slightly better guess. Everything else on this page exists because of that second number.

It is worth understanding the mechanism rather than treating this as a story about criminals, because the mechanism is what tells you which interventions can possibly work.

A buyer standing at a counter cannot tell a good tablet from a bad one. Neither can a hospital storekeeper. The quality is real, it matters enormously, and it is invisible at the moment of purchase. In a market where quality cannot be verified by the buyer, a seller who invests in quality cannot charge for it, because the buyer has no way of telling that seller apart from one who simply says the same thing more cheaply. So the premium is unavailable, the honest supplier’s margin disappears, and the market drifts toward whatever is cheapest to produce.

That is not a moral failure of a country. It is what happens to any market in which the buyer cannot verify the good, and it is the same structure as the page before this one. A family cannot verify a hospital, so it buys a country instead. A pharmacist cannot verify a tablet, so they buy a price instead. Both are the same disease presenting in two organs, and both have the same cure, which is making quality verifiable by somebody with a reason to check.

The useful consequence is that an operator does not have to wait for the market to improve. Verification is something a single buyer can install on their own side of the transaction, unilaterally, and the moment they do, price stops being the only signal available to them.

Seven layers. Read the last column of each, because a strategy that cannot fill that column in has not been examined, it has been purchased.

The layers are ordered deliberately. The first two decide most of the outcome and cost the least to implement, because they are policy rather than equipment. The middle three are the ones with budget lines attached, which is why they are the ones that get presented as the strategy. The last two are detection rather than prevention, and they only exist if the record system and the cabinets were bought as one thing.

The second layer is the one that will actually be tested. Refusing the spot market is easy to agree to in a meeting and hard at two in the morning when a patient needs something that is not on the shelf, and every sourcing policy in this market is eventually offered that exact trade. Which is why it cannot stand on its own.

A single accredited source with a firm rule against buying anywhere else is the strongest control on this page and the most fragile, and the failure mode is not corruption. It is a shortage.

When the shelf is empty and the patient is in the building, one of three things happens. Somebody buys from a distributor who is not on the list, which is the moment the entire chain of custody argument evaporates. Or the family is sent out to a pharmacy on the road to buy it themselves, which is common, which is worse, and which the hospital does not record as a procurement decision at all even though that is exactly what it is. Or the patient goes without.

So the sourcing policy is only as strong as the stock policy that protects it, and this is the point at which just in time stops being a general efficiency argument and becomes a safety one. Safety stock on the restricted lines is not working capital tied up in a store room. It is the thing that prevents a three in the morning purchase from an unverified source, and it should be costed and defended in exactly those terms.

Which gives the fourth instance of the sentence this study keeps arriving at. Beds, people, fuel, and now medicines. The buffer looks like waste, is not waste, and survives the cost round only where somebody wrote down what it was for.

It also produces one rule that is not negotiable and costs nothing to state. No inpatient, and no family member of an inpatient, is ever sent outside the building to buy a medicine. If it is not on the shelf that is the hospital’s failure to solve, not the family’s errand to run, and the moment it becomes their errand the hospital has outsourced its supply chain to a road.

Every layer above leaks, so the question that finishes the strategy is what happens to the product that got through all of them. Today, in almost every hospital anywhere, the answer is nothing. It reaches patients, it does not work, and the record of it not working is written in a hundred separate clinical notes that are never compared.

The mechanism that changes that is simple to state and takes a deliberate decision at build time to make possible. Join the prescribing record to the batch record, then watch for clusters. If patients on a particular lot of a particular product are failing to respond at a rate the disease does not explain, that is a supply question before it is a clinical one, and somebody has to own asking it.

This is only possible if the record system and the dispensing cabinets were specified as one system. Two systems that each hold half of it will produce a project to join them that runs for three years and is descoped in year two, which is the ordinary fate of this exact integration. It is cheap at design stage and expensive at every point afterwards, which puts it with everything else on this site that has to be decided before the concrete.

It should be said plainly that this is a weak detector on its own. It is slow, it needs volume, and it is measured in exposed patients rather than in days. It is the last layer precisely because everything it catches is something six better controls already missed. But it is the only one pointed at the failure that actually reaches the bedside, and at present the data it needs is sitting in two systems that do not speak.

Four weaknesses, at the same size as the strategy.

The last one is the one to sit with. Nobody knows how big this problem is in Nigeria, and that is not a gap in this page, it is a gap in the world. Published prevalence runs from around five per cent to twenty five per cent, the regulator’s own working figure sits in the middle of that, the last formal national survey is well over a decade old, and at least one government agency has put the figure far higher than any of them. Those numbers cannot all be right and there is no way from the outside to tell which is.

Which turns the usual conclusion upside down. A hospital that tests what arrives is not only protecting its patients. After two years it is holding the only prevalence data that is actually about its own supply chain, measured rather than estimated, and it knows something about this market that the regulator does not. On a site that keeps insisting the instrument is the product, that is not a side effect. It is arguably the main one.