New Dizziness Impact Measures of Positional, Functional, and Emotional Status Were Supported for Reliability, Validity, and Efficiency

Abstract

ObjectiveTo calibrate the 25 items from the Dizziness Handicap Inventory (DHI) patient-reported outcome measure (PROM), using item response theory (IRT), into 1 or more item banks, and assess reliability, validity, and administration efficiency of scores derived from computerized adaptive test (CAT) or short form (SF) administration modes. DesignRetrospective cohort study. SettingOutpatient rehabilitation clinics. ParticipantsPatients (N=28,815; women=69%; mean age [SD]=60 [18]) included in a large national dataset and assessed for dizziness-related conditions who responded to all DHI items at intake. InterventionsNot applicable. Main Outcome MeasuresIRT model assumptions of unidimensionality, local item independence, item fit, and presence of differential item functioning (DIF) were evaluated. Generated scores were assessed for reliability, validity, and administration efficiency. ResultsPatients were treated in 976 clinics from 49 US states for either vestibular-, brain injury-, or neck-related impairments. Three unidimensional item banks were calibrated, creating 3 distinct PROMs for Dizziness Functional Status (DFS, 13 items), Dizziness Positional Status (DPS, 4 items), and Dizziness Emotional Status (DES, 6 items). Two items did not fit into any domain. A DFS-CAT and a DFS 7-item SF were developed. Except for 2 items by age groups and 1 item by main impairment, no items were flagged for DIF; DIF impact was negligible. Median reliability estimates were 0.91, 0.72, and 0.79 for the DFS, DPS, and DES, respectively. Scores discriminated between patient groups in clinically logical ways and had a large effect size (>0.8), with acceptable floor and ceiling effects (<15%), except for a floor effect for DPS (20.4%). DFS-CAT scores were generated using a median of 8 items; they correlated highly with full-bank scores (r=0.99). ConclusionThe 3 dizziness impact PROMs demonstrated moderate to high reliability, were valid, and highly responsive to change; thus, they are suitable for research and routine clinical administration.