Every figure on this page comes from a synthetic cohort of 1,000 simulated participants. These are not findings from a clinical trial or an observational study. They do not establish the effectiveness, cost-effectiveness, affordability, or equity impact of any real intervention.
Question
Healthcare systems have to decide whether an intervention that appears to improve outcomes is also a reasonable use of limited resources. A result that looks favourable for one person can still be difficult to afford across a large eligible population. This project turns that problem into a reproducible workflow for a hypothetical digital mental health intervention.
Design
The cohort is stored as data/synthetic_cohort.csv. It is not a hospital, a country, a service, or a trial. Two illustrative access groups, Higher_access and Lower_access, support descriptive summaries. They are not validated measures of inequality, deprivation, or digital exclusion.
The workflow has four parts that answer different questions. Cost-effectiveness compares incremental cost with approximate 12-month QALYs. Budget impact explores eligible population, uptake, cost per participant, implementation cost, and assumed offsets. The equity component describes group differences without claiming that access caused them. Deterministic sensitivity analysis varies intervention cost and incremental QALYs.
Results
The incremental improvement rate from the current implementation is 0.054. Its meaning depends on the definition used in the code. Values are rounded for presentation.
Net monetary benefit is λΔQ − ΔC. The threshold of €30,000 per QALY is a modelling assumption in this portfolio. It is not a recommendation for a health system and not an official decision threshold. A positive incremental NMB is conditional on those inputs.
Budget impact
The conceptual model is B = N × U × C + I − O. N is the eligible population, U is uptake, C is cost per participant, I is implementation cost, and O is assumed cost offsets. The exact scenarios are produced by src/health_economics.py and written to outputs/budget_impact_scenarios.csv. They explore assumptions. They are not observed expenditure and not a forecast.
How to read the ICER
The ratio is ΔC / ΔQ, and only when incremental QALYs are positive and the ratio is interpretable. A low illustrative ICER is not evidence that a real digital mental health intervention is cost-effective. The QALYs here are approximate demonstration values, not a validated patient-level QALY analysis.
Reproducibility
Analysis logic is separated from generated outputs. The repository includes unit tests and a documented way to run the workflow again. Before the same functions were applied to real research data, the analyst would still have to check missingness, duplicates, units, cost ranges, outcome definitions, follow-up, and the provenance of every input.
Code and tests: digital-mental-health-health-economics.