Abstract / Summary
Abstract Background After stroke, families take long-term responsibility for rehabilitation and care, and family resilience is an important measure of family adaptation. Predicting later family resilience from information collected early during hospitalization may help assess family support needs and plan follow-up care. This study aimed to develop and conduct a preliminary external validation of a model for predicting family resilience 6 months after stroke in older adults. Methods We used a prospective two-cohort design. The development and independent external validation analyses included 310 and 102 older adults with stroke, respectively. Data were collected at baseline (T0), 1 month (T1), and 6 months after stroke (T2). Using 23 candidate predictors measured early during hospitalization (T0), we developed a least absolute shrinkage and selection operator (LASSO) linear regression model to predict the continuous score on the shortened Chinese version of the Family Resilience Assessment Scale (FRAS-C) at 6 months (T2). The model was internally validated using repeated nested 10-fold cross-validation and bootstrap resampling, and the final model was applied directly to the independent external cohort. Performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), and calibration measures and compared with the baseline FRAS-C-only model. Results The final model included five predictors: baseline family resilience, self-care ability, social support, caregiver sex, and caregiver burden. In repeated nested cross-validation, the model yielded an R² of 0.555 (95% CI 0.499–0.603) and an RMSE of 2.959. In the external validation cohort, the R² was 0.409 (95% CI 0.269–0.508), the RMSE was 3.516, and the calibration slope was 1.165 (95% CI 0.897–1.429). Compared with the baseline FRAS-C-only model, the full model increased R² by 0.238 and reduced RMSE by 0.649 points in external validation. Conclusion The model, which included baseline family resilience, patient functional status, and family caregiving characteristics, predicted family resilience at 6 months and added predictive information beyond baseline family resilience alone. It retained predictive performance in the independent external cohort and may be used to support family assessment early during hospitalization and follow-up planning after discharge.