Development of a patient-reported measure of treatment burden after stroke.

Problem

Treatment burden is the workload of healthcare and wellbeing impact for people with long-term conditions. Stroke rehabilitation engenders treatment burdens not captured in patient-reported measures (PRMs) developed for general or multimorbid populations with no index condition. We aimed to adapt a PRM of treatment burden in multimorbidity, the PETS (Patient Experience with Treatment and Self-Management v2.0), to create a stroke-specific measure, PETS-Stroke. We aimed to examine factor structure, content validity, construct validity, reliability and feasibility in a stroke survivor population.

Approach

First, we adapted the 60-item PETS into the PETS-Stroke (34-items) using a taxonomy of treatment burden developed in previous qualitative work. We then conducted content validity testing through cognitive interviews (n=15) to explore the importance, relevance, and clarity of each item. Evaluation of psychometric properties was conducted through analysis of data from stroke survivors recruited via postal survey (n=386). Demographic information was self-reported. Factor structure was tested with Confirmatory Factor Analysis and internal consistency were indexed with coefficient’s alpha and omega. Construct validity was examined against: The Stroke Southampton Self-Management Questionnaire; The Satisfaction with Stroke Care Measure; and The Shortened Stroke Impact Scale. Test-retest reliability was examined through within-subject correlations of PETS-Stroke scores at two assessment points 3 weeks apart. Acceptability and feasibility was explored via missing data rates and telephone interviews with 30 participants.

Findings

In total 386 stroke survivors completed the baseline survey and 300 completed the follow-up. Mean age was 68 (SD:11.2), 56% male, 98.4% white, 37% in the two most deprived socioeconomic quintiles (SIMD), 20% had speech difficulties and 26% lived alone. CFA demonstrated the best fit was for a higher order model reflecting a global treatment burden score and nine sub-factors. Cronbach's alpha was >0.69 and McDonald’s omega was >0.88, both suggesting good internal consistency. For test-retest reliability, the intra class correlation co-efficient (ICC) was >0.7 for all domains. Convergent validity demonstrated that all domains of the PETS-stroke were moderately correlated with higher stroke burden, lower readiness to self-manage and lower satisfaction with care (Spearman’s ρ ranged 0.226-0.671). Missing data varied for each domain between 3% and 77%, with the majority of missingness being due participants choosing the ‘not applicable’ answer rather than omitting answers. Content validity testing suggested that items with content irrelevant to some participants e.g. obtaining walking aids, were still important to include. No major issues arose with feasibility in the interviews.

Consequences

We have created a PRM of treatment burden after stroke (PETS-stroke) that shows promise as a population-specific measure for use in stroke rehabilitation clinical trials. This will be an important outcome measure alongside measures of efficacy, to ascertain if treatments are manageable and implementable into the lives of stroke survivors. The lack of diversity in ethnicity in the sample was a limitation.