OBJECTIVES: Over 200 mapping algorithms have been published that allow for prediction of EQ-5D utilities. A database of usable algorithms is needed to allow researchers and HTA agencies to identify quickly, the quality and suitability of algorithms used. We present a freely available online database of such algorithms across all disease areas, with a classification based on a ‘traffic light’ system of red (warning), amber (use with caution) and green (usable).

METHODS: We searched the available databases (including PubMed, Cochrane Library) to identify all published mapping algorithms and classified them by disease area. We report important metrics including the sample sizes used, the source and quality of data and the reported predictive properties. We classify these using simulation methods across properties and advocate a new classification system of the form:

Pr [(Ω_j > 0) |∧, A_i] ≥ Δ , for ∧ ∈ {measures such as R² , % predicted}

Pr [(Ω_j < 0) |∧, A_j] ≥ Δ , for ∧ ∈ {measures such as MSE, MAE}

Where, Ω_j = ( θ*._j - μ), θ*._j is an overall average performance metric and μ is an overall performance measure across algorithms and A_i is each algorithm.

RESULTS: We identified over 200 mapping algorithms over 40 disease areas. The most common disease area was Oncology; 60% of algorithms were classified as red (warning); 20% as amber (use with caution) and 20% as green (usable). Identified mapping algorithms included those that predicted EQ-5D utility scores and EQ-5D dimension levels.

CONCLUSIONS: Mapping algorithms can be classified and although the classification approach may require further work, the database offers a rich source of structured information on the use of mapping algorithms consistent with the guidance provided in NICE DSU TSD 22 (June 2023). This will be a valuable resource freely available to academics and pharmaceutical companies for economic evaluation purposes.