Estimating measurement error of the Oswestry Disability Index with missing data

BACKGROUND CONTEXT: The Oswestry Disability Index (ODI) is the most commonly used patient-reported outcome measure for low back pain. Incomplete ODI questionnaires may induce measurement error, causing misclassification of disability level. PURPOSE: We determined which scoring method—alternative scoring (AS) or multiple imputation (MI)—produced the best estimate of measurement error with missing items; and assessed the precision of AS and MI when scoring the ODI and classifying disability level with increasing numbers of missing items. STUDY DESIGN/SETTING: Quantitative study PATIENT SAMPLE: We included 995 patients who underwent lumbar spine surgery from 2014 to 2020 and who had complete ODI data. OUTCOME MEASURES: Measurement error. METHODS: We (1) simulated 1000 datasets for each number of ODI items (1–9) for which data could be missing by randomly setting items to missing; (2) calculated simulated scores using AS or MI; (3) calculated absolute percentage error (APE) of simulated vs. true scores; (4) classified disability level using the simulated score; and (5) compared APE between scoring methods to assess precision and misclassification rates. RESULTS: For 1, 5, and 9 missing items using AS, APE was 0.6%, 3.1%, and 12%, respectively, and misclassification rates were 0.6%, 4.6%, and 13%. For 1, 5, and 9 missing items using MI, APE was 1.3%, 7.9%, and 56%, respectively, and misclassification rates were 4.6%, 9.7%, and 58%, respectively. CONCLUSION: With increasing numbers of missing ODI items, MI introduced more measurement error than AS did. Clinicians should consider APE when interpreting ODI scores with missing data and classifying disability level. Inaccurate interpretations could negatively affect assessments and treatment plans.