Prediction of 30-Day Hospital Readmissions for All-Cause Dental Conditions using Machine Learning

Area of Science:

• Healthcare Analytics
• Dental Public Health
• Predictive Modeling

Background:

• The risk and contributing factors for hospital readmission among patients admitted for all-cause dental conditions (ACDC) remain largely unknown.
• Understanding these factors is crucial for developing targeted interventions to reduce readmission rates.

Purpose of the Study:

• To determine the prevalence of 30-day hospital readmissions for ACDC patients.
• To identify key risk factors associated with these readmissions.
• To develop and evaluate artificial intelligence (AI) driven machine learning (ML) models for predicting 30-day readmission risk.

Main Methods:

• Utilized data from the 2013 Nationwide Readmissions Database (NRD), analyzing 11,341 cases of all-cause dental admissions.
• Employed descriptive statistics to characterize patient demographics and clinical factors.
• Applied five distinct ML techniques to construct predictive models and identify significant risk factors, evaluating performance using AUC, accuracy, sensitivity, specificity, and precision.

Main Results:

• Eleven percent of patients admitted for ACDC experienced a 30-day hospital readmission.
• Significant predictors of readmission included total charges, number of diagnoses, patient age, chronic conditions, length of stay, number of procedures, and insurance type (Medicare/Medicaid).
• The artificial neural network model achieved the highest predictive performance with an AUC of 0.739.

Conclusions:

• Readmission following ACDC hospitalization is a notable issue, with potential for substantial cost savings if reduced.
• AI-powered ML algorithms demonstrate promise in identifying high-risk dental patients, facilitating targeted interventions to decrease readmissions and improve patient care.
• Further validation of these ML models is recommended to support clinical decision-making and enhance patient-centered care strategies.