ILLUSTRATIVE EXAMPLE
Cost-Effectiveness Analysis for a
Nurse-Led Hypertension Program
Patient-level costs and QALYs, a gamma cost model, a bootstrapped ICER and an acceptability curve for a trial of about 400 patients, reported to CHEERS 2022.
This example does not describe a real client or project. Every engagement is confidential, during the project and after it ends.
RESEARCH QUESTION
What the Evaluation Set Out to Answer
This cost-effectiveness analysis example follows a health services research team that has trial data on a nurse-led hypertension program compared with usual care. The team wants to know whether the program offers good value from the health system perspective over 12 months, expressed as the additional cost per quality-adjusted life year (QALY) gained.
A decision maker needs more than a single ratio, because the ratio can be unstable and hides how uncertain it is. The plan therefore fixes the perspective, the time horizon and the primary analysis before outcomes are examined, and it reports the uncertainty around both costs and effects together.
STUDY DESIGN AND DATA
Costs and Utilities Collected Alongside a Trial
A trial-based evaluation with about 400 patients randomized to the program or usual care and followed for 12 months. The dataset holds program delivery costs, clinic visits, medication and admissions from administrative records, and EQ-5D responses at baseline, six months and 12 months. Costs are right-skewed, and some patients leave before follow-up ends, so both costs and utilities are incomplete for some of them.
- Design: economic evaluation alongside a randomized trial, health system perspective
- Outcome: incremental cost and incremental QALYs over 12 months
- Software requested: Stata, with do-files and logs
METHOD AND ASSUMPTIONS
Why Resample Costs and QALYs Together
QALYs are calculated as the area under each patient's utility curve, adjusted for baseline utility, because patients start from different health states. Mean costs are modeled with a gamma model and a log link, which suits skewed positive data, and the result is checked against a two-part model where some cost categories contain many zeros. The comparison of arms stays tied to the randomization, so baseline differences are handled by adjustment and not by selecting patients.
The ratio of two uncertain differences has no simple standard error, so uncertainty is estimated by resampling patients with their costs and QALYs kept together, which preserves the correlation between them. Missing values are imputed within each resample, so the uncertainty from missing data is carried into the intervals, and complete-case results are shown as a check. The willingness-to-pay values used for decisions are set in the plan and are not chosen after the results are seen.
OUTPUT AND INTERPRETATION
How the Results Are Read
The output has incremental costs and incremental QALYs, each with a 95% confidence interval, the incremental cost-effectiveness ratio (ICER), and a cost-effectiveness plane showing the resampled estimates. The ICER is read with care: when the QALY difference is close to zero the ratio becomes unstable, so incremental net benefit across a range of willingness-to-pay values and an acceptability curve carry the decision message.
Sensitivity analyses vary the cost model, the imputation approach and the perspective where the data allow it. The write-up states the probability that the program is cost-effective at each threshold as a statement about uncertainty, not as a verdict, and notes that results from one trial setting may not transfer to another health system.
REPORTING
What the Researcher Received
A written report structured to CHEERS 2022 for economic evaluations, with statistical wording for the methods, matched to the output, for the team to check and incorporate, in the target journal's format. The decision context, interpretation of value and conclusions remain with the research team.