The number that quietly eats a health plan
You already met the short-horizon danger in passing: because [[health-insurance-and-medical-expense|medical cover]] is repriced every single year, next year's premium leans almost entirely on a forecast of how fast cost will grow. That growth rate has a name — [[medical-trend|medical trend]] — and this guide is about why it, more than morbidity, more than any single large claim, is the thing that dominates health pricing. Set the right premium for the risk you have *today* and you can still go broke, simply because the same risk costs more *tomorrow*. Trend is not noise around the forecast; trend *is* the forecast.
Actuaries quote trend on a per-member basis so it cleans out changes in how many people you cover. The workhorse unit is PMPM — cost *per member per month*. Take a block running at 400 PMPM today; if trend is 7%, you project next year at 400 × 1.07 = 428 PMPM, or 5,136 per member per year. The arithmetic is trivial. The whole craft lies in defending that one multiplier, because, as you saw with the [[frequency-severity-decomposition|frequency–severity decomposition]], the same headline cost can be reached a dozen different ways — and so can the same headline trend.
Three engines: price, utilisation, and new technology
Trend is not one thing rising; it is the *compound* of separate forces, and a good actuary refuses to forecast it as a single blob. The first engine is unit price: the same MRI scan, the same hospital day, the same molecule of a drug simply costs more this year than last — driven by wages, by hospital and provider contracts renegotiated upward, and by drug-list prices. The second engine is utilisation: holding prices fixed, people consume *more* care — more visits, more imaging, more procedures per illness — partly as populations age and chronic conditions accumulate, partly as the practice of medicine grows more intensive.
The third engine is the sly one: new technology and mix shift. Medicine does not just get pricier — it changes *what* it offers. A new gene therapy, a new class of obesity or cancer drug, a new robotic surgery arrives and is genuinely better, so it gets used; but it sits at the expensive end, and as care *mixes* toward it the average cost climbs even if no single price moved. This is why medical trend behaves unlike trend in almost any other product: a car this year is much the same car as last year, but a course of treatment this year may be a thing that did not exist before — and could not be priced before.
- Unit price up 3.0% — the same scan, hospital day, or drug molecule simply costs more (a factor of 1.030).
- Utilisation up 2.5% — at those same prices, people use more visits, imaging, and procedures (a factor of 1.025).
- Technology and mix shift up 1.5% — care tilts toward newer, pricier treatments even if no single price moved (a factor of 1.015).
- Compound them, do not add them: 1.030 × 1.025 × 1.015 − 1 = 7.16% total trend, a touch above the naive 7.0% sum — so 400 PMPM today projects to about 428.6 PMPM next year.
Leveraging: how a fixed deductible makes trend worse
Here is a subtlety that catches beginners every year, and it follows straight from policy design. The member usually pays a fixed dollar deductible before the insurer pays anything. That dollar figure does not rise with trend — but the bills do. So as total cost grows, the deductible covers a *smaller and smaller fraction* of the bill, and the insurer's share grows *faster* than the total. This is called leveraging, and it means the insurer's PMPM trend is higher than the underlying medical trend even though nothing about the disease changed.
Underlying medical trend = 8%, member deductible fixed at 500: This year total bill = 5,000 member pays 500 insurer pays 4,500 Next year total bill = 5,400 member pays 500 insurer pays 4,900 total cost trend = 5,400 / 5,000 - 1 = 8.0% INSURER cost trend = 4,900 / 4,500 - 1 = 8.9% --> the insurer's share trends FASTER than the bill (leveraging)
Leveraging is why a health actuary cannot simply borrow a published medical-trend figure and bolt it onto the premium. The same economy-wide trend produces a *different* insurer trend for a plan with a 500 deductible than for one with a 5,000 deductible — the higher the fixed member contribution, the harder the leverage bites. It is also why managed-care tools matter: a [[managed-care-and-provider-networks|provider network]] that negotiates unit prices down, or steers utilisation toward cheaper settings, is attacking two of the three engines directly. Trend is not weather you merely observe; some of it is yours to bend.
Why trend is so brutally hard to predict
If trend were a steady 7% you could just bake it in and sleep. The agony is that it *moves*, and it moves for reasons you cannot see in your own data until they have already happened. A blockbuster drug clears approval and floods in. A pandemic first *suppresses* utilisation as people avoid care, then unleashes a rebound of deferred surgeries the year after. A recession pushes some members to skip care and others to rush it before they lose coverage. Each of these can swing a year's trend by points — and the actuary must commit to a number *before* the year begins, then live with it. This is the distinctive ache of [[health-vs-life-actuarial-work|health versus life work]]: a mortality table drifts gently over decades, but trend can lurch in twelve months.
Worse, the feedback you get is delayed and dirty. Claims take months to be submitted and paid, so when you finally observe a year's experience it is already history, and part of it is still missing — the [[loss-and-premium-trend|claims you must trend and develop to ultimate]] before they even mean anything. So the health actuary is forever steering by a foggy rear-view mirror: estimating a trend that will not be confirmed until long after the rates it sets are locked. A model can smooth the past beautifully and still be blindsided by next year, because medical trend is genuinely about events that have not occurred yet.
And the stakes on that one number are asymmetric and large. Take a block at 400 PMPM, priced for next year at 7% trend (428 PMPM) but coming in at 9% (436 PMPM): a 2-point miss is 8 PMPM, which on 60,000 members is 8 × 12 × 60,000 ≈ 5.8 million of unpriced cost in a single year. Under-forecast trend and the plan bleeds; over-forecast it and you price yourself out of the market and lose the members instead. There is no safe side to err on — only the discipline of splitting trend into its engines, watching each one, and pricing with honest humility about the part you cannot yet see.
The loss ratio: the thermometer regulators read
Once a year has run, how do you tell whether your trend bet paid off? The single most-watched gauge is the [[loss-ratio-and-medical-loss-ratio|loss ratio]]: incurred claims divided by earned premium. Collect 100 of premium, pay 82 in claims, and the loss ratio is 82% — at a glance, the share of every premium dollar that flowed back out as care. If trend ran ahead of what you priced, claims swell and the loss ratio climbs toward and past 100%, the unmistakable fingerprint of a trend miss. It is a thermometer: it tells you the patient is feverish, not yet *why*.
In several markets the loss ratio is not merely watched but *regulated*, as the medical loss ratio (MLR) — the share of premium spent on medical claims plus quality-improvement activities, which law may require to clear a minimum floor (commonly 80% for individuals and small groups, 85% for large groups). The intent is consumer protection: it caps how much of every premium dollar an insurer may keep for administration and profit. The teeth are a rebate — if an insurer's MLR falls *below* the floor, it must refund the shortfall to policyholders. An insurer earning 10 million in premium with 7.6 million of claims has a 76% MLR; against an 80% floor it is 4 points short and must rebate roughly 400,000 back to members.
Pulling the rung together
Step back and the picture locks in. The [[health-claim-cost|health claim cost]] you decomposed earlier into frequency × severity does not sit still — every year it is dragged forward by medical trend, which is itself the compound of three engines: unit price, utilisation, and new technology shifting the mix. Leveraging amplifies it on the insurer's slice, the short annual horizon means you must guess it before the year starts, and its delayed, dirty feedback makes it the single hardest and highest-stakes number a health actuary touches. Then the loss ratio, and its regulated cousin the MLR, close the loop — telling you after the fact whether the bet held, and quietly aiming the whole industry's incentives back at the engines of trend.
Carry one honest reflex out of this guide: when a health number surprises you — a premium jump, a rate filing, a plan that lost money — never stop at "costs went up." Ask which engine moved. Was it price, was it utilisation, was it a new technology shifting the mix, or was it leverage on the insurer's share? That diagnostic habit, paired with the humility that no model fully sees next year's medical world, is exactly what makes a health actuary worth their salt.