Factors Influencing Health Behavior Adherence in Young and Middle-Aged Stroke Patients

BACKGROUND: Health behavior adherence (HBA) is important for effective rehabilitation and secondary prevention after stroke. Young and middle-aged patients may face particular challenges in maintaining recommended health behaviors during hospitalization because of functional impairment, cognitive deficits, pain, and competing social or occupational responsibilities. However, factors associated with HBA in this population remain insufficiently characterized, and practical tools for the early identification of patients at risk of suboptimal adherence are limited. To identify factors associated with HBA in young and middle-aged patients with stroke during hospitalization and to develop a combined predictive model for the early identification of patients with non-high adherence. METHODS: A single-center retrospective observational study was conducted, including 298 young and middle-aged stroke patients hospitalized between January 2021 and June 2024. Patients were categorized into a high-adherence group and a non-high-adherence group based on the HBA Index. Differences in demographic and functional characteristics between the groups were compared. Correlation analysis and multivariate logistic regression were performed to identify independent factors affecting HBA. Receiver operating characteristic (ROC) curves were constructed to evaluate the predictive performance of individual indicators and the combined model. Adherence was positively correlated with BI, MMSE, FMA, and BBS scores, and negatively correlated with Numeric Rating Scale (NRS) pain scores. In multivariate logistic regression analysis, MMSE, NRS, educational level, and first stroke status were independently associated with non-high adherence. In the final parsimonious multivariable logistic regression model, MMSE score, NRS score, educational level, and first-stroke status were independently associated with non-high adherence. The final four-predictor model demonstrated good discrimination. CONCLUSIONS: A multidimensional combined predictive model demonstrated good discrimination performance and may help to identify patients at higher risk of non-high adherence.