Predictors for non-response to PROMs across three areas of inpatient rehabilitation - a nationwide cross-sectional study in Switzerland.

Abstract

Background: Patient-reported outcome measures (PROMs) are valuable tools to integrate the patient perspective and improve quality of rehabilitation services. However, there is a risk of non-response bias. Identifying characteristics associated with non-response is important to reduce barriers to PROM completion and to ensure quality care for all patients undergoing rehabilitation.

Methodology: In this cross-sectional observational study, we used routine data from the Swiss national quality assurance program for inpatient rehabilitation, in which PROMs are employed at admission and discharge for all patients in cardiac, psychosomatic and pulmonary rehabilitation. We included all patients aged 18 years or older, who completed cardiac, pulmonary or psychosomatic rehabilitation between 01.01.2016 and 31.12.2022. We used mixed-effects logistic regression to identify predictors for non-response to PROMs across the three areas of rehabilitation. Non-response was defined as recorded refusal to complete the PROMs (cardiac: MacNew Heart, psychosomatic: Hospital Anxiety and Depression Scale (HADS) and Patient-Health Questionnaire (PHQ)-15, pulmonary: Chronic Respiratory Disease Questionnaire CRQ) at admission and/or discharge. Several socio-demographic and medical characteristics, as well as clinic and year were included as covariates in the regression model.

Results: We included 39,630 patients completing the MacNew Heart, 12751 HADS, 13195 PHQ and 16193 completing the CRQ. Non-response rates were highest for the MacNew Heart (28.6%) and CRQ (25.0%) and lower for the HADS (14.4%) and PHQ (17.5%). Across all three PROMs we observed strong significant associations of a nationality-based proxy for being a non-native speaker, admission from or discharge into a nursing home/rehabilitation and discharge into acute care after rehabilitation with non-response. Having no supplementary health insurance increased the odds for non-response on the MacNew Heart, HADS and CRQ. Additionally, higher comorbidity Cumulative Illness Rating Scale (CIRS) score and age were also significantly associated with non-response. Between-clinic differences explained a substantial proportion of variance in non-response, with adjusted ICCs ranging from 0.34 in psychosomatic to 0.62 in cardiac rehabilitation.

Conclusion and relevance: Non-response to PROMs in rehabilitation is predicted by impaired health, older age, and a proxy for non-native language skills. This should be addressed by reducing barriers for patients with these characteristics to reduce bias and to ensure health equity.