Drivers of Stroke Independence During Inpatient Rehabilitation Care: Causal Forest Modeling

Research Objectives: - to understand the medical patterns driving independence among stroke survivors during inpatient rehabilitation care
- to understand the influence of therapy type (e.g., PT, OT, SLP), medication, mental health services, and illness on independence among stroke survivors' independence upon discharge from inpatient rehabilitation care

Design: Machine learning analysis via causal forest modeling to examine electronic medical record data of stroke survivors from their inpatient rehabilitation care.

Setting: Shepherd Center Inpatient Stroke Treatment

Participants: 662 stroke survivors who were inpatient, treated at Shepherd Center between 2020-2023.

Interventions: No intervention, analysis of electronic medical record data for Shepherd Center.

Main Outcome Measures: Stroke survivors independence which is defined via discharge location, individual's ability to navigate their surroundings, individual's agency in personal care routines, supervision level the patient needs at discharge (SRS), and patient goals met (NOMS).

In our analysis, in addition to standard metrics of disability and demographics, we focused on four distinct treatment types, to assess the impact of each unique treatment on patient independence: therapy participation, medication with focus on Lasix and Botox, psychiatric services, and prevalence of urinary tract infections.

Results: Discharge Disposition Findings: Stroke patients who receive Botox treatments had lower discharge disposition score. Lasix treatment showed minimal impact on stroke patients.

Mobility and Supervision Requirements Findings: Botox treatment for stroke patients was associated with decreased levels of assistance needed, indicating a significant improvement in independence. Conversely, Lasix treatment increased the need for assistance, suggesting reduced mobility or increased care needs of patients needing Lasix treatment.

Urinary Tract Infection Findings: For TBI patients, UTI treatment showed significant improvements in mobility and bowel management outcomes, indicating enhanced mobility and improved bowel management post-treatment. However, UTIs were related to an increase in the duration of hospital stays.

These findings highlight the complexity of treatment responses across different patient populations and underscore the importance of personalized treatment approaches.

Conclusions: This study's utilization of causal inference methods, particularly causal forests, aimed to deepen understanding of factors influencing independence among stroke survivors during their inpatient rehabilitation treatment. Through the exploration of treatments such as Botox, Latex, and Urinary Tract Infection (UTI) treatments, we sought to provide valuable insights that could inform clinical decision-making and enhance patient care strategies in the rehabilitation setting. Our analysis revealed significant differences in treatment effects on patient independence by patient subgroup, highlighting the importance of tailored interventions based on individual demographic and medical characteristics.

Author(s) Disclosures: Authors have no disclosures.