Development and validation of the General Rehabilitation Adherence Scale (GRAS) in patients attending physical therapy clinics for musculoskeletal disorders

BACKGROUND: Non-adherence to physical therapy ranges from 14 to 70%. This could adversely affect physical functioning and requires careful monitoring. Studies that describe designing and validation of adherence measuring scales are scant. There is a growing need to formulate adherence measures for this population. The aim was to develop and validate a novel tool named as the General Rehabilitation Adherence Scale (GRAS) to measure adherence to physical therapy treatment in Pakistani patients attending rehabilitation clinics for musculoskeletal disorders. METHODS: A month-long study was conducted in patients attending physical therapy sessions at clinics in two tertiary care hospitals in Karachi, Pakistan. It was done using block randomization technique. Sample size was calculated based on item-to-respondent ratio of 1:20. The GRAS was developed and validated using content validity, factor analyses, known group validity, and sensitivity analysis. Receiver operator curve analysis was used to determine cut-off value. Reliability and internal consistency were measured using test-retest method. Data was analyzed through IBM SPSS version 23. The study was ethically approved (IRB-NOV:15). RESULTS: A total of 300 responses were gathered. The response rate was 92%. The final version of GRAS contained 8 items and had a content validity index of 0.89. Sampling adequacy was satisfactory, (KMO 0.7, Bartlett’s test p-value< 0.01). Exploratory factor analysis revealed a 3-factor model that was fixed and confirmed at a 2-factor model. Incremental fit indices, i.e., normed fit index, comparative fit index and Tucker Lewis index, were reported > 0.95 while absolute fit index of root mean square of error of approximation was < 0.03. These values indicated a good model fit. The value for Cronbach (α) was 0.63 while it was 0.77 for McDonald’s (ω), i.e., acceptable. Test-retest reliability coefficient was 0.88, p < 0.01. Education level was observed to affect adherence (p < 0.01). A cut-off value of 12 was identified. The sensitivity and accuracy of the scale was 95%, and its specificity was 91%. CONCLUSION: The scale was validated in this study with satisfactory results. The availability of this tool would enhance monitoring for adherence as well as help clinicians and therapists address potential areas that may act as determinants of non-adherence.