Health Economics Metrics

Topic 21 of 76 · Economic analysis types

Cost-Utility Analysis (CUA)

CUA is cost-effectiveness analysis with a generic, preference-weighted outcome — almost always the QALY (or DALY averted). Because the outcome unit is universal, CUA can compare interventions across completely different diseases.

Why it matters

A national health service must choose between a cancer drug, a mental-health app, and a surgical robot from one budget. Natural units can't compare them; QALYs can. CUA is therefore the reference-case method at NICE and most HTA bodies: its output — cost per QALY, judged against a threshold — is the closest thing health policy has to a universal exchange rate. If you want your software funded instead of something else, CUA is the arena.

The math

ICUR = ΔCost / ΔQALYs      (the ICER with QALYs as the effect unit)

ΔQALYs = Σ (duration_i × utility_i)_new − Σ (duration_i × utility_i)_old

Utilities from validated instruments (EQ-5D); costs and QALYs both discounted at 3.5% (NICE reference case); uncertainty via PSA.

Worked example

A CBT app for moderate anxiety vs waiting list for face-to-face therapy, per patient:

Costs:  app licence + support        £250
        therapy displaced            −£680   (40% of users no longer need it)
        ΔC = 250 − 680 = −£430 (saves money)

QALYs:  6 months at utility 0.76 instead of 0.68 while waiting
        ΔE = 0.5 × (0.76 − 0.68) = +0.04 QALYs

ΔC < 0 and ΔE > 0: the app dominates — better and cheaper, no ratio needed. Had the therapy-displacement assumption been only 10%, ΔC = 250 − 170 = +£80, and ICUR = 80 / 0.04 = £2,000/QALY — still far below £20,000. The case survives even with the key assumption slashed: that is what a robust CUA looks like (and the tornado diagram proves it).

Software engineering connection

CUA's deep idea — one composite, preference-weighted unit to compare unlike things — is the pattern for comparing unlike engineering investments (security vs developer experience vs reliability). The honest options are either a defensible composite unit (rare) or an explicit cost-consequence table (usual). What CUA warns against is the fake composite: a weighted "impact score" whose weights were tuned after the fact to make the preferred option win. Health economics spent decades standardizing utility elicitation precisely so weights precede the comparison.

Pitfalls

  • Utility gains below the instrument's sensitivity (see minimal clinically important difference in patient-reported outcomes) — tiny ΔE times large populations is a classic laundering trick.
  • Missing comparator care displacement — the biggest cost term for digital products is often what they replace.
  • Mapping non-preference scores to utilities with unvalidated crosswalks.

Sources