Developing Best Practice for Analyzing Individual Patient Utility Data in Health Technology Assessment: A Modified...

Accurate health utilities are critical for health economic models because quality-adjusted life-years combine health utilities and life-years. Individual patient utility data are an important source of these inputs because they are used to calculate health-state utility values; yet, no dedicated best practice guidance exists for analyzing patient-level utility data. This research aimed to develop guidance for analyzing such data and identify gaps.

We conducted a scoping survey, literature and National Institute for Health and Care Excellence technology appraisal reviews, and a modified Delphi process to generate consensus-based recommendations. The research examined key analysis considerations including utility distributions (eg, upper bounded), data characteristics (eg, missing data, repeated collection), and associated analytical methods (eg, regression models). It also examined post-hoc adjustments required to address specific data properties. The derivation of individual patient utility data was out of scope.

Thirty-one participants-predominantly health economists, alongside statisticians-from academia, industry, consultancies, and National Institute for Health and Care Excellence contributed to the Delphi process. Thirteen topics were discussed, covering 3 stages of analysis: core methods, potential adjustments, and sensitivity analyses. Although no new recommendations emerged, the panel refined wording, added clarifications, and expanded recommendations. Broad consensus across all recommendations was achieved by the end of round 3.

The research informed the development of best practice guidance for analyzing individual patient utility data. Important evidence gaps remain in the published literature, particularly regarding optimal methods for responding to data characteristics. These findings support health technology assessment agencies in providing clearer, more consistent instructions for robust and transparent analysis of these data.