Unveiling the Impact of Occupational Therapy on Acute Care Outcomes: A Machine Learning Approach

Comprehensive Summary

Lee et al. conducted a retrospective cohort study to characterize optimal occupational therapy (OT) service patterns associated with improved health outcomes in orthopedic patients, using decision tree models. Eligible patients met all of the following criteria: 1) admission to and discharge from one of eleven Cleveland Clinic hospitals between January 1, 2017 and December 1, 2021; 2) receipt of at least one OT session; 3) length of stay (LOS) of less than 30 days; and 4) an orthopedic surgeon listed as their primary provider. OT service patterns were defined by 1) total number of OT sessions; 2) total OT session time; and 3) the proportion of hospital days that included an OT session. Three outcome variables were assessed: 1) 30-day readmission to a Cleveland Clinic hospital; 2) discharge to a home setting; and 3) minimal detectable change in daily activities at discharge, as measured by the Activity Measure for Post-Acute Care (AM-PAC) 6-Clicks daily activity short form, scored by occupational therapists at each visit. The study comprised a total of 36,300 orthopedic patients. Prediction models demonstrated strong performance (AUC: 0.71-0.94; specificity: 0.66-0.91; sensitivity: 0.67-0.92; accuracy: 0.69-0.91). The pruned decision tree model for 30-day readmission identified the following risk patterns: 1) patients with LOS less than 3 days, a Charlson Comorbidity Index (CCI) of 4 or greater, and OT coverage on less than 67% of hospital days were at increased readmission risk; 2) patients with LOS less than 3 days, CCI below 4, and OT coverage below 39% were similarly at elevated risk; and 3) patients with LOS of 3 days or more and CCI below 2 exhibited complex multifactorial risk patterns. Thresholds associated with minimal detectable change in daily activities included: 1) three or more OT sessions; 2) OT sessions on more than 82% of hospital days; and 3) a cumulative OT session time of 39 minutes or more.

Outcomes and Implications

This study underscores the importance of OT session frequency and duration for acute orthopedic patients. A higher proportion of hospital days with an OT session was a significant predictor of both 30-day non-readmission and minimal detectable change in daily activities. Notably, elevated readmission risk was paradoxically associated with greater total OT session time, an association the authors attribute to triage bias, as higher-acuity patients typically require longer OT sessions yet remain at an inherently greater readmission risk. None of the OT service variables emerged as significant predictors of home discharge. Key limitations include a restricted number of OT practitioners, resulting in 70.8% of the cohort receiving only one session and 71.7% not receiving follow-up evaluations, as well as the inability to track readmissions occurring outside of the Cleveland Clinic hospital system.