Accuracy of Linking VR-12 and PROMIS Global Health Scores in Clinical Practice.

OBJECTIVES: To examine the accuracy of general health cross-walk tables in a clinical sample of patients with spine disorders. Published tables (Schalet BD, Rothrock NE, Hays RD, et al. Linking physical and mental health summary scores from the Veterans RAND 12-Item Health Survey (VR-12) to the PROMIS(®) Global Health Scale. J Gen Intern Med 2015;30:1524-30) link scores from the Veterans RAND 12-Item Health Survey (VR-12) to the 10-Item Patient-Reported Outcome Measurement Information System (PROMIS), a global health scale metric for both mental (GMH) and physical (GPH) summary scores. METHODS: We assessed the accuracy of administered PROMIS and VR-12 scores with scores predicted by cross-walks in 4606 adult patients seen in a spine clinic from October 2015 to 2016. Accuracy of linking scores was evaluated using Pearson correlation, intraclass correlation coefficients, and mean and SD of score differences. Bland-Altman plots were used to graphically assess the levels of agreement. The consistency in scores' discrimination across levels of pain severity, depression, and other patient characteristics was assessed. Bootstrap methods estimated linking precision across varying sample sizes. RESULTS: Actual and cross-walked PROMIS scores showed moderate correlation (ICC(3,1): GMH 0.73; GPH 0.81), with Bland-Altman plots suggesting smaller differences between scores in patients with lower and higher general health. Significant discrimination between patient subgroups was demonstrated reliably by both actual and estimated scores. Bootstrapped resamples indicated adequate precision for 200 patients (95% confidence interval for mean difference: GMH -1.38 to 0.60; GPH 0.39 to 1.93). CONCLUSIONS: VR-12 and PROMIS global health scores can be accurately linked within a sample of patients with spine disorders; nevertheless, bias is high and precision is low for linking on the patient level. Linked scores at the group level for more than 200 patients can be used in comparative effectiveness research and for comparing results across studies.