Early Unplanned Readmissions and Mortality After Induction Chemotherapy in AML

Area of Science:

• Hematologic Oncology
• Clinical Informatics
• Predictive Analytics

Background:

• Acute myeloid leukemia (AML) is an aggressive cancer.
• Induction chemotherapy for AML carries a significant risk of complications, including unplanned readmissions and early mortality.
• Identifying high-risk patients is crucial for improving outcomes.

Purpose of the Study:

• To determine the primary reasons for unplanned readmissions in AML patients post-induction chemotherapy.
• To develop predictive models for unplanned readmissions or early death using electronic health record (EHR) data.
• To identify patients at highest risk for adverse events.

Main Methods:

• Retrospective analysis of 1,111 inpatient encounters from 305 adult AML patients.
• Inclusion of patients undergoing induction chemotherapy; exclusion of specific leukemia types and transplant recipients.
• Utilized logistic regression and least absolute shrinkage and selection operator for risk factor identification and variable selection.

Main Results:

• 22% of patients experienced unplanned readmissions within 30 days post-discharge, primarily due to fever/infection (53.5%).
• 30-day mortality was 22%, leading to a combined 39.7% adverse event rate.
• Predictive model performance improved (AUC 0.67 to 0.74) with the inclusion of comorbidities and symptom frequency from clinical notes.

Conclusions:

• A significant proportion of AML patients experience adverse events within 30 days of discharge.
• Predictive models integrating structured and unstructured EHR data can identify at-risk patients.
• Risk stratification tools can guide timely interventions to reduce hospitalizations and improve survival.