Objective: To examine temporal trends and geographic variations and predict inpatient rehabilitation (IPR) length of stay (LOS) and home discharge for stroke patients.
Methods: Patients aged ≥18 years who were admitted to an IPR facility in Alberta, Canada, between 04/2014 and 03/2018 (years 2014-2017) were included. Predictors of LOS and home discharge were examined using 2014-2016 data and validated using 2017 data. Multivariable linear regression (MLR), multivariable negative binomial (MNB), and multivariable quantile regressions (MQR) were used to examine LOS, and logistic regression was used for home discharge.
Results: We included 2686 rehabilitation admissions between 2014 and 2017. The mean LOS decreased (2014: 71 days; 2017: 62.1 days; p = 0.003) during the study period and was shortest in Edmonton (59.1 days) compared to Calgary (66 days) or other localities (70.8 days; p < 0.001). Three-quarters of patients were discharged home and this proportion remained unchanged between 2014 and 2017. Calgary patients were more likely to be discharged home than those in Edmonton (OR = 0.62; p = 0.019) or other localities (OR = 0.39; p = 0.011). The MLR and MNB models provided accurate prediction for the mean LOS (predicted = 59.9 and 60.8 days, respectively, vs. actual = 62.1 days; both p > 0.5), while the MQR model did so for the median LOS (predicted = 44.3 days vs. actual = 44 days; p = 0.09). The logistic regression resulted in 82.4% of correct prediction, a sensitivity of 91.6%, and a specificity of 50.7% for home discharge.
Conclusions: Rehabilitation LOS decreased while the proportion of home discharge remained unchanged during the study period. Both varied across health zones. Identifiable statistical models provided accurate prediction with a separate patient cohort.
Keywords: Stroke; Stroke rehabilitation.