Topic 26 of 76 · Economic analysis types
Return on Investment (ROI)
ROI is the ratio of net gain to money invested. It is the metric engineering and finance already share — health economics adds the discipline that makes an ROI claim survive scrutiny: declared perspective, comparator, horizon, and benefit categories.
Why it matters
ROI is the lingua franca of budget holders, and public health uses it too: the landmark Masters et al. review found a median ROI of 14.3:1 for public health interventions (every £1 returns ~£14 to the wider economy and health system) — a number widely used to argue prevention spending. But that 14:1 is a societal, long-horizon figure; a hospital CFO's ROI is payer-perspective and 1–3 years. Most ROI fights are actually undeclared-perspective fights.
The math
ROI = (Benefits − Costs) / Costs (often × 100%)
Payback period = Costs / annual net benefit
An ROI claim is under-specified without four declarations:
- Perspective — whose benefits count? (see analysis perspective)
- Comparator — versus what alternative? (see opportunity cost)
- Horizon — over how long, and discounted?
- Benefit class — cash-releasing, capacity, or qualitative? (see cash-releasing vs non-cash-releasing)
Worked example
E-rostering system, cost £500,000 over 3 years.
Cash-releasing: agency shift reduction £450,000
Capacity: ward-manager admin time freed £600,000 (valued, not banked)
Qualitative: staff satisfaction, safety not monetized
Strict financial ROI = (450,000 − 500,000)/500,000 = −10%
Economic ROI = (1,050,000 − 500,000)/500,000 = +110%
Both numbers are true. A vendor quoting "+110% ROI" to a CFO who can only bank £450k will lose trust; presenting both, labeled, wins it. The same split protects an internal champion when finance audits the benefits two years later.
Software engineering connection
Every tooling proposal has an ROI slide; almost none declare the four parameters. The most common failure is category-blending: capacity gains (developer minutes) presented as financial return. Structure AI/platform ROI as the worked example above — cash line, capacity line, qualitative line — and add sensitivity analysis on the soft numbers. For the P&L reality check on AI ROI specifically, see AI return on investment.
Pitfalls
- Perspective laundering: societal benefits over a decade quoted to a budget holder with a 12-month horizon.
- Gross instead of net: "returns £3M" on £2M spend is 50% ROI, not 300%.
- Ratio maximization: tiny denominators produce spectacular ROIs on trivial investments; rank portfolios by NPV or net monetary benefit, use ROI as a screen.
- No benefits audit: forecast ROI without benefits realization tracking is a promise, not a result.
Sources
- Masters R, et al. "Return on investment of public health interventions: a systematic review." J Epidemiol Community Health 2017. https://pmc.ncbi.nlm.nih.gov/articles/PMC5537512/
- HM Treasury Green Book. https://www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-government/the-green-book-2020