An analysis of data found specific patient, hospital, and community factors can predict which hospitalized patients with cancer are at greater risk for readmission 30 days after discharge. These findings were published in JCO Clinical Cancer Informatics.
Researchers analyzed electronic health record data from a retrospective cohort of 2460 patients and a comprehensive machine learning analysis pipeline to train and test prediction models for 30-day readmission rates. They sought to determine the effects of their best model compared with existing models in the Epic electronic health record (EHR).
Data from 2 time points of a patient’s inpatient encounter — the first 48 hours and the entire encounter — were used to determine which factors were more likely to influence readmissions.
Patient, hospital, and community variables obtained within the first 48 hours that could predict hospital readmissions within 30 days of discharge in oncology patients included:
- Patient Weight change over 365 days, depressive symptoms, laboratory test results, and cancer type
- Hospital Winter discharge and hospital admission type
- Community Zip code, income, and marital status
These factors can help providers determine actions that could reduce the likelihood of readmission for affected patients. For example, significant weight changes over the last year or low albumin levels could indicate issues such as food scarcity, an inability to eat, or nausea and vomiting. “Providing patients with access to food programs/social workers, a nutritionist, or medications to control nausea and vomiting could mitigate 30-day readmissions with associated symptoms, respectively,” the researchers wrote in their report.
A combination of clinical decision support and emerging digital technologies may be able to help providers support their patients’ transitions and provide the continuous care needed to reduce hospital readmission rates.
This research had some limitations. First, the models were evaluated at only 1 hospital in our health system. A root cause analysis was not performed, nor did the researchers optimize the decision boundary for flagging patients at risk for readmission.
Hwang S, Urbanowicz R, Lynch S, et al. Toward predicting 30-day readmission among oncology patients: identifying timely and actionable risk factors. JCO Clin Cancer Inform. Published online February 21, 2023. doi:10.1200/CCI.22.00097