A parallel group cohort to determine the measurement properties of an early inflammatory arthritis detection tool

Abstract

To determine the comprehensibility, internal consistency, test-retest reliability and discriminative properties of an early inflammatory arthritis (IA) detection tool. Four groups were recruited from outpatient clinics at two tertiary care hospitals: early IA, established IA, non-IA musculoskeletal conditions (MSK) and non-MSK. Participants attended a study visit where they completed the 11-item tool with binary yes/no response options. Comprehensibility was assessed for each tool item on a 5-point Likert scale. For test-retest assessment, the tool was mailed to participants following a 2-week recall washout interval. Two items were randomly selected to test internal consistency. Discriminative properties compared tool item responses with blinded rheumatologist clinical assessments. A previously developed rheumatology triage algorithm was externally validated. A total of 170 participants were enrolled in the study. Comprehensibility approached unity for all tool items. The internal consistency Kuder-Richardson Formula 20 was 0.985 (P < 0.0001). The mean test-retest reliability kappa (S.D.) was 0.81 (0.02). The intraclass correlation coefficient (ICC) (95% CI) for summed yes responses between test and retest phases was 0.94 (0.93, 0.95) and for algorithm scores was 0.97 (0.96, 0.98). Patient responses were significantly associated with the corresponding clinical evaluations (P < 0.0001, respectively). The sum of yes responses and rheumatology triage algorithm scores both differentiated early IA from each of the other three groups (P < 0.004, respectively). The scoring algorithm receiver operating characteristic plot area under the curve (S.E.) was 0.829 (0.003). The tool has favourable measurement and discriminative properties. The discriminative properties of the rheumatology triage scoring algorithm were externally validated.