Minimal Detectable Changes of the Health Assessment Questionnaire–Disability Index, Patient‐Reported Outcomes Measurement Information System‐29 Profile Version 2.0 Domains, and Patient Health Questionnaire‐8 in People With Systemic Sclerosis: A Scleroderma Patient‐Centered Intervention Network Cohort Cross‐Sectional Study

Abstract

ObjectiveSystemic sclerosis (SSc) is a rare, chronic autoimmune disorder associated with disability, diminished physical function, fatigue, pain, and mental health concerns. We assessed minimal detectable changes (MDCs) of the Health Assessment Questionnaire–Disability Index (HAQ‐DI), Patient‐Reported Outcomes Measurement Information System‐29 Profile version 2.0 (PROMIS‐29v2.0) domains, and Patient Health Questionnaire (PHQ)‐8 in people with SSc.MethodsScleroderma Patient‐Centered Intervention Network Cohort participants completed the HAQ‐DI, PROMIS‐29v2.0 domains, and PHQ‐8 at baseline assessments from April 2014 until August 2023. We estimated MDC95 (smallest change that can be detected with 95% certainty) and MDC90 (smallest change that can be detected with 90% certainty) with 95% confidence intervals (CIs) generated via the percentile bootstrapping method resampling 1,000 times. We compared MDC estimates by age, sex, and SSc subtype.ResultsA total of 2,571 participants were included. Most were female (n = 2,241; 87%), and 38% (n = 976) had diffuse SSc. Mean (±SD) age was 54.9 (±12.7) years and duration since onset of first non‐Raynaud phenomenon symptom was 10.8 (±8.7) years. MDC95 estimate was 0.41 points (95% CI 0.40–0.42) for the HAQ‐DI, between 4.88 points (95% CI 4.72–5.05) and 9.02 points (95% CI 8.80–9.23) for the seven PROMIS‐29v2.0 domains, and 5.16 points (95% CI 5.06–5.26) for the PHQ‐8. MDC95 estimates were not materially different across subgroups.ConclusionMDC95 and MDC90 estimates were precise and similar across age, sex, and SSc subtype groups. HAQ‐DI MDC95 and MDC90 were substantially larger than previous estimates of HAQ‐DI minimal important difference from several small studies. Minimally important differences of all measures should be evaluated in large studies using anchor‐based methods.