OBJECTIVES The aim of this study is to provide an overview of the quality of health economic evaluations (HEEs) of prediction models, the evidence used, and the challenges. METHODS The databases Medline, Embase, Econlit, and the NHS Economic Evaluations Database were systematically searched for HEEs of diagnostic and prognostic risk prediction models. The included HEEs were evaluated on their methodological quality using the Drummond checklist. Furthermore, an item list was developed incorporating descriptive items on the HEE, specific items on the HEE of prediction models, and statistical characteristics of the prediction model that could be incorporated into the evaluation. RESULTS The database search resulted in 791 unique papers, from which 653 were excluded based on abstract. After assessing full texts, 17 HEEs (all cost-utility studies) were included. A prediction model was compared to current practice in 11 HEEs and to an extended prediction model in 6 HEEs. On a 35-point scale the quality score ranged from 17 to 32 (median 25). In 7 papers there was no overlap between authors of the initial prediction model paper and those of the corresponding HEE. In 5 papers individuals were classified based on a single (set of) threshold(s); based on guidelines in 4 papers and once on expert opinion. In 8 papers the classification threshold was optimized in the CEA itself. A probabilistic sensitivity analysis was not included in 7 papers and uncertainty around predicted risks was only taken into account once. CONCLUSIONS In most papers limited (prediction model) details were available. Potentially due to this lack of evidence and a lack of specific guidelines on HEE of prediction models, a large variety in the quality and methodology was observed. This variation may complicate the validation and interpretation of HEE results and thereby the decision making on implementation of prediction models in practice.