OBJECTIVES: The domains of the MOS 36-ltem Short-Form Health Survey (SF-36) can be aggregated to physical (PCS) and mental (MCS) component summary scores, which are widely used measures of patient-reported health. PCS and MCS were originally derived using an uncorrelated factor model, potentially leading to problems with interpretation of results. Consequently, modified scoring algorithms for correlated SF-36 summary scores (PCS_c and MCS_c) have been suggested. The PROMIS-29 v2.0 is a newer generic health measure which is increasingly used as an alternative to the SF-36. PROMIS-29 physical and mental summary scores can be derived from individual PROMIS-29 domains. The objective of this study was to establish algorithms to predict SF-36 component summary scores from PROMIS-29 scores.

METHODS: Baseline data from n=713 participants of the Berlin Longterm Observation of Vascular Events (BeLOVE) study were used. We estimated seperate linear regression models, with either PROMIS-29 domain scores or PROMIS-29 physical/mental summary scores as predictors and SF-36 physical (PCS and PCS_c) and mental (MCS and MCS_c) summary scores as dependent variables. Follow-up data from n=194 participants were used to validate these regression models. Pearson correlation coefficients (r) were calculated to determine the agreement between empirical and predicted SF-36 summary scores.

RESULTS: Individual PROMIS-29 domains as well as PROMIS-29 summary scores showed high predictive value for PCS, PCS_c, and MCS_c (R²≥70%), and moderate predictive value for MCS (R²=58% and R²=41%, respectively). The agreement between empirical and predicted SF-36 summary scores in the validation sample was high for PCS, PCS_c, and MCS_c (r>0.80), but considerably lower for MCS (r=0.64 and r=0.71).

CONCLUSIONS: This study provides regression coefficients which can be used to predict SF-36 physical and mental summary scores based on either individual PROMIS-29 domains or PROMIS-29 summary scores. The prediction of SF-36 mental health summary scores appeared to be more precise for the correlated factor model.