Psychometric properties and reference values of nine PROMIS parent-proxy measures for children aged 5 to 7 years in the Netherlands

The Patient-Reported Outcomes Measurement Information System (PROMIS®) provides standardized, generic measures to assess self-reported aspects of physical, mental, and social health. While PROMIS parent-proxy measures exist in multiple languages, evaluations outside the U.S. are limited. This study evaluated psychometric properties and provided Dutch reference values for PROMIS parent-proxy measures for children aged 5–7 years. Parents (n = 529) completed PROMIS parent-proxy measures (Global Health, Anxiety, Depressive Symptoms, Anger, Peer Relationships, Mobility, Fatigue, Sleep Disturbance, Pain Intensity) and Pediatric Quality of Life Inventory (PedsQL) questionnaires as legacy measures. Structural validity was assessed using a Graded Response Model (GRM) with item fit statistics (S-X2, p < 0.001:misfit). Construct validity was evaluated by comparing PROMIS T-scores with PedsQL subscales. Reliability was estimated based on participants with a standard error (SE) ≤0.32 (reliability ≥0.90). Additionally, Test Information Functions (TIFs), and theoretical and empirical marginal reliability coefficients (>0.70:acceptable) were examined. Relative efficiency ((1-SE(θ)2)/nitems) compared performance across PROMIS full form, short form, CAT, and PedsQL measures. Measurement invariance was assessed through Differential Item Functioning (DIF) analyses (McFadden’s pseudo R2 > 0.02:DIF), with impact evaluated by re-estimating GRMs without DIF item(s) and comparing T-scores, reliability indices, and Test Characteristic Curves (TCCs). Dutch reference values (mean T-score (M), standard deviation (SD)) were reported for PROMIS v1.0/3.0 and v2.0. All PROMIS measures demonstrated acceptable psychometric properties. Five items showed misfit (Global Health:1; Fatigue:1; Sleep Disturbance:3), and three items exhibited DIF (Depressive Symptoms:1; Mobility:2), all with negligible impact. Construct validity confirmed alignment between PROMIS and PedsQL subscales. Reliability was high near the sample mean and extended into clinically relevant ranges. TIFs and marginal reliability coefficients confirmed good/excellent precision (0.85–0.94) for all item banks except for Mobility (0.56), with theoretical and empirical coefficients closely aligned. PROMIS CAT was most efficient, providing more information per item than the full form, short form, and PedsQL. Dutch reference values were obtained for v1.0/3.0 (M:[44.8–54.1], SD:[7.4–11.8]) and v2.0 (M:[44.5–50.7], SD:[9.1–12.5]). PROMIS parent-proxy Global Health, Anxiety, Depressive Symptoms, Anger, Peer Relationships, Mobility, Fatigue, Sleep Disturbance, and Pain Intensity measures showed acceptable psychometric properties in Dutch children aged 5–7 years. This study provides reference values for this population.