ABSTRACT Objectives When health outcomes relevant for economic evaluations are unavailable, algorithms can be developed to map utilities using available clinical outcome measures. This study aims to develop two mapping algorithms estimating EuroQol-5 dimension-3 level (EQ-5D-3 L) utilities using the clinician-rated Health of the Nation Outcome Scores (HoNOS) and Positive and Negative Syndrome Scale (PANNS). Methods A dataset with 2,029 observations of patients with psychotic disorders included EQ-5D-3 L, HoNOS, PANSS item scores, and demographics. Correlations between instruments were evaluated. Least Absolute Shrinkage and Selection Operator (LASSO) regression and random forest (RF) algorithms with various predictor variable sets were applied. Model performance was cross-validated using R-squared and Root Mean Square Error (RMSE). Results High ceiling effects were observed for EQ-5D-3 L, with weak to moderate negative correlations between EQ-5D and HoNOS (r = -0.34) and PANSS (r = -0.27). Overall, LASSO models outperformed RF models, with individual item models performing best for the HoNOS and PANSS, with the best observed RMSEs of 0.241 and 0.231, respectively. Conclusions The HoNOS and PANSS could be mapped onto EQ-5D-3 L utilities but lack accuracy for individual patient predictions. However, in the absence of alternatives, they could adequately predict population-based utility score differences for health economic evaluations.