Fee-for-service payment
A Socratic walk-through of fee-for-service payment — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can paying a hospital for every treatment it delivers leave its patients no healthier?
Fee-for-service sounds like the fairest arrangement anyone could design. You pay for what you get. A hospital that does more work is paid more; one that does nothing is paid nothing. No one is rewarded for idleness, and nobody has to trust a lump sum to be spent well.
And yet health systems built on it spend enormous sums without their populations coming out measurably healthier for the extra. So the puzzle is not that anyone is dishonest. It is that a payment rule which seems to reward effort exactly can produce a great deal of activity and very little health. Where does the gap open?
Reasoning it through
REASONING #Start with a small, uncomfortable question: when a payment rule is written, what exactly is being bought?
Under fee-for-service the answer is a countable event. A consultation. A scan. A stent placed, a night in a bed, a test resulted. Each has a code and a price, and revenue is the sum of the codes. Now ask the question that any incentive analysis has to ask: what can the person inside the system do to make that number larger? The honest answer is that they can produce more countable events. Nothing in the rule asks whether the patient ended up better.
Notice this is not the same as saying the events are worthless. Most of them help. The trouble is subtler. Consider two clinicians facing the same borderline case — a scan that would probably be normal, a procedure with modest expected benefit. Both are acting in good faith; medicine is full of genuine uncertainty and there is usually an argument for doing the thing. The payment rule does not create the ambiguity. It just leans on it, every time, in one direction, for years.
Now follow the sharper edge of it. What happens to a hospital that succeeds at prevention — that keeps its diabetic patients out of the ward, that avoids the amputation? Under fee-for-service, its revenue falls. The work that produced the health is either unpaid or paid least, and the failure it averted was the profitable part. That is the crux: the rule pays for the repair and not for the absence of the damage.
There is a second asymmetry underneath it. The person who chooses the treatment is largely the person who is paid for it, and the patient cannot easily judge whether it was needed. Economists call the resulting slack supplier-induced demand, and the empirical fingerprint is well documented: how much care a patient receives varies enormously between regions with similar populations, and the high-spending regions do not show correspondingly better outcomes.
So what would you do about it? The obvious move is to invert the rule — pay a fixed amount per person per year and let the provider keep what it does not spend. Capitation. But run the same analysis on it and notice the incentive turns cleanly over: now the cheapest patient is the untreated one, and the exposure is underprovision rather than excess. Pay per admission instead, as diagnosis-related groups do, and the pressure moves again — shorter stays, more admissions, and a quiet incentive to record each case as the sicker version of itself.
That is the pattern worth extracting. Every payment rule pays for something measurable, and whatever it counts is what will be maximised. Health itself is unusually hard to count: it arrives late, it is noisy, it depends on the patient's behaviour and circumstances as much as on the clinic, and it cannot be cleanly attributed to any one actor. Which is why value-based schemes, honestly, have delivered mixed and mostly modest results so far — the measurement problem they must solve is real and unsolved, not an oversight.
The analogy
THE ANALOGY #Paying a hospital fee-for-service is like paying a mechanic by the part fitted. He is not a crook, and most of the parts genuinely needed fitting — but on every borderline judgement his interest and yours point opposite ways, and the day he tunes your engine so well that it needs nothing next year is the day he earns nothing from you.
a car has one owner who can seek a second opinion and eventually notice a pattern of unnecessary work, whereas a patient usually cannot judge the necessity of care at all, is often not the one paying, and rarely sees the counterfactual version of themselves who was left alone.
Clarifying the model
THE MODEL #The tempting misreading is that fee-for-service fails because of greed, and that the fix is a more ethical profession. But hold the mechanism steady: the rule produces its effect through thousands of defensible decisions at the margin of genuine uncertainty, made by people who would each pass an ethics review. That is precisely what makes it powerful and hard to see.
The second misreading is that some better rule exists which has no failure mode. It does not. Each rule is a choice about which error to make: fee-for-service buys access and volume at the risk of doing too much; capitation buys restraint at the risk of doing too little; case rates buy efficiency per episode at the risk of more episodes and inflated coding. This is why real systems blend them, and why the interesting question is never "which rule is right" but "which error can this population least afford".
And one honest caveat about the evidence. Regional variation in spending is suggestive, not conclusive — some of it reflects genuinely sicker or poorer populations rather than induced demand. The claim that survives that objection is the narrower one: the payment rule reliably tilts the borderline decisions, and no one has yet written a rule that counts health directly.
A picture of it
THE PICTURE #How to readBegin at the input at the top, follow the two labelled branches out of the diamond, and notice they are not treated alike. The solid loop from revenue back to the decision is the whole mechanism — volume funds capacity, and capacity presses back on the next borderline judgement. The branch marked as a hazard is the one that helps patients and earns nothing. The dotted arrows are the only ones pointing at health, and neither of them carries money.
What became clearer
WHAT CLEARED #A payment rule is not a way of financing care; it is a way of specifying what will be maximised. Fee-for-service specifies countable services, so services are what a system reliably produces — including at the margin where the benefit is thin, and excluding the prevention that would have made them unnecessary. The failure needs no bad actor, only a long run of defensible decisions all leaning the same way. And since every alternative rule counts something else that is also not health, the design question is which direction of error a system can best tolerate.
Where to go next
ONWARD #- Why upcoding follows any rule that pays by category rather than by unit.
- Whether bundled payment for a whole episode of care escapes the trap, or only moves the boundary the gaming happens at.
- How salaried systems substitute waiting lists for volume as their characteristic failure.
Key terms
TERMS #| Term | What it means |
|---|---|
| Fee-for-service | payment for each discrete billable service, so revenue rises with volume. |
| Capitation | a fixed payment per enrolled person per period, regardless of services used. |
| Diagnosis-related group | a fixed payment per admission set by the case's diagnostic category rather than by services rendered. |
| Supplier-induced demand | care generated by the provider's own recommendation where the patient cannot judge necessity. |
| Upcoding | recording a case in a higher-paying category than the care actually delivered warrants. |
Every term the collection defines is gathered in the glossary.