Acute Care Utilization Patterns During Chemotherapy and Predictive Model Development at a Rural Community Cancer...

Area of Science:

• Oncology
• Health Services Research
• Data Science

Background:

• Acute care use (ACU) in oncology patients is costly, prolonged, and disrupts treatment.
• Identifying patients at risk for ACU is crucial for proactive intervention.
• Rural oncology patients face unique barriers to care access.

Purpose of the Study:

• Develop predictive models for all-cause acute care use (A-ACU) and preventable acute care use (P-ACU) in oncology patients.
• Identify key predictors for A-ACU and P-ACU.
• Assess rural-specific barriers to acute care access.

Main Methods:

• Retrospective cohort study of adult oncology patients receiving intravenous chemotherapy (Oct 2021 - Apr 2024).
• Utilized electronic medical records and insurance claims data.
• Trained LASSO and Random Forest models to predict 30-, 90-, and 180-day risk of A-ACU and P-ACU.

Main Results:

• 45.3% of patients experienced A-ACU; 10.3% experienced P-ACU within 180 days.
• Predictors included prior inpatient stays, comorbidities, insurance type, age, and lab values.
• Models demonstrated strong predictive performance (AUC=0.73, F1=0.79).

Conclusions:

• Models effectively identify high-risk oncology patients for ACU using routine data.
• Validated known risk factors in a rural oncology population.
• Future integration into practice and addressing rural challenges can reduce ACU.