From one event to a whole career of events
Every rung you have climbed so far has rested, quietly, on a single yes-or-no event: the insured life dies, once, and the policy pays. The whole machinery of the [[force-of-mortality|force of mortality]], the life table, and net premiums is built to price that one irreversible event. Welcome to morbidity, where that comfortable assumption falls apart. Here the event of interest is not death but *illness, injury, and disability* — and unlike death, illness can strike more than once, vary enormously in how much it costs, and, crucially, it can end. People recover. A morbidity model has to track not just whether you fall ill, but how deep you fall and whether you climb back out.
That single difference — recovery — reshapes everything. In life insurance the survival curve only ever falls, because no one un-dies. In health, a person can move from *healthy* to *sick* and back to *healthy* repeatedly across a single year, each round generating doctor visits, prescriptions, maybe a hospital stay. So morbidity is naturally a model of *transitions between states*, not a one-way slide toward an absorbing end. You will meet that machinery formally in continuance tables later in this rung; for now, just hold the picture of a life that bounces between health and illness, with the insurer on the hook for the costly side of every bounce.
Cracking the claim cost into frequency × severity
How much does it cost to insure one person's health for a year? Staring at the raw number — say, $4,200 of claims — tells you almost nothing about *why*. So health actuaries almost never quote a single figure; they crack the [[health-claim-cost|health claim cost]] open into two independent questions through the [[frequency-severity-decomposition|frequency–severity decomposition]]. First, *how often* does this person claim in a year — the claim frequency? Second, *when they do claim, how big is it* — the claim severity? Multiply the average of each together and you have the expected annual cost. It is the single most important sentence in health pricing: cost equals frequency times severity.
One member, one year of medical cover:
expected frequency = 6.0 claims per year (doctor visits, scripts, a procedure)
average severity = $700 per claim (mean cost when a claim occurs)
expected claim cost = frequency * severity
= 6.0 * $700
= $4,200 per member per year
Same $4,200 can come from very different worlds:
12 claims * $350 = $4,200 (many small claims)
1 claim * $4,200 = $4,200 (one rare, huge claim)
--> identical average, wildly different RISK.Splitting the cost this way is not just tidy bookkeeping — it lets the two halves be modelled with the right tools, which you have already met. Frequency is a *counting* problem, so it is typically handled with a claim frequency distribution like the Poisson, which answers "how many events in a year?" Severity is an *amount* problem, modelled with a positive, often right-skewed distribution where a few catastrophic claims sit far out in the tail. Treating them separately also lets you reason about policy design cleanly: a deductible mostly knocks out the *frequency* of small claims, while a high-cost specialty drug attacks the *severity* side. One lever per question.
Why health risk runs hotter than life risk
Step back and compare the two trades, because their differences explain why [[health-vs-life-actuarial-work|health and life actuarial work]] feel like different professions. Life risk is *long-horizon*: a whole-life policy can run sixty years, so the actuary's deepest worry is the slow drift of mortality and the force of interest discounting decades of future cash flows. Health risk is *short-horizon*: most medical and disability cover is priced and repriced annually. That sounds easier — and in the discounting sense it is, since a single year barely needs present-value gymnastics — but the short horizon hides a sharper danger.
Health risk is also *more frequent* and *more volatile*. A 40-year-old's annual probability of dying might be well under one percent — most years, a block of life policies simply pays nothing. But that same person will visit a doctor several times a year, fill prescriptions, perhaps land in hospital. Claims arrive constantly, in a wide range of sizes, and they swing hard year to year with flu seasons, a new blockbuster drug, an economic shock that pushes people to defer or rush care. Where a life actuary watches a quiet, slow-moving curve, a health actuary watches a noisy signal that can lurch double digits in a single year. More events does *not* mean more predictable on a percentage basis — it means a faster, jumpier world to forecast.
Disability cover sharpens the contrast even further. A [[disability-income-insurance|disability income]] policy does not pay a lump sum on a single event; it pays a monthly cheque for as long as the claimant stays disabled — which might be three months or thirty years. So its cost depends not only on *how often* people become disabled (an incidence rate, the morbidity cousin of a death rate) but on *how long they stay there before recovering or dying* (a continuance, or termination, rate). Two morbidity dials at once — getting sick, and getting better — and the second one can dwarf the first.
Medical trend: the thing that keeps everyone up at night
Here is the danger the short horizon was hiding. Because health cover is repriced every year, the actuary's forecast leans almost entirely on one fragile number: [[medical-trend|medical trend]] — the annual rate at which the *cost per member* grows. Trend is itself a frequency × severity story compounded over time: prices for the same services rise, *and* people use more services, *and* the mix shifts toward newer, pricier treatments. Stack those together and a health plan's cost can climb 6, 8, even 10 percent a year while general inflation sits far lower. Get next year's trend assumption wrong by even two points, and a plan priced to break even quietly turns into a loss.
Put numbers on it to feel the sting. Suppose this year's cost is $4,200 per member, and you price next year assuming 6% trend, so you collect for $4,452. If trend actually lands at 8%, real cost is $4,536 — a shortfall of $84 per member. That looks trivial, until you multiply it across 50,000 members: $84 × 50,000 ≈ $4.2 million of unfunded claims, born entirely from being two percentage points light on a single forecast. Trend compounds and then multiplies across the whole membership, which is why a small miss is a multi-million-dollar event, not a rounding error.
Staying honest about what morbidity models can and cannot do
It is tempting to treat morbidity tables like the crisp life tables of the survival rung, but stay honest about the difference. "Sick" and "recovered" are not as clean as "alive" and "dead." Where death is unambiguous and reported almost perfectly, a morbidity claim depends on whether the person *chose* to seek care, whether a doctor *coded* it a certain way, and where exactly the policy *draws the line* on disability. The same underlying health can produce very different claim counts under two different benefit designs. So a morbidity rate is always entangled with behaviour and definitions in a way a mortality rate is not — which is why raw claim data must be cleaned and adjusted before it can be trusted.
Two more cautions worth carrying forward. First, frequency × severity is an *average*, and an average is silent about the tail: a block can sit right on its expected claim cost yet be one premature-baby NICU stay or one transplant away from a brutal year. The decomposition prices the centre of the distribution; capital and reinsurance, which you will meet later, exist to cover the far edge it ignores. Second, last year's morbidity is a *starting point*, never a destiny — and because health is repriced annually, the actuary gets the gift and the curse of frequent feedback: errors surface fast, but so does the temptation to chase noise as if it were signal.
Pull it together and the shape of this rung is clear. Health actuarial work is mortality's restless cousin: faster, noisier, repriced every year, and driven by morbidity — the rates at which people fall ill, run up cost, and recover. The single tool you carry into every following guide is the decomposition: cost is frequency times severity, dragged forward each year by medical trend. Hold that, stay humble about the tail and about behaviour, and the rest of the rung — continuance tables, disability, long-term care, health reserves — becomes a series of variations on this one honest theme.