OBJECTIVES

: This study aims to build a predictive model for hospital readmission among diabetes patients using artificial neural network and compare its performance to logistic regression model.

METHODS

: A public database was used in this study. All the participants who were eligible were randomly assigned into 2 groups: training sample and testing sample. Two models were built using training sample: artificial neural network and logistic regression. We used these models to predict the risk of Hospital Readmission among Diabetes Patients in the testing sample. Receiver operating characteristic (ROC) were calculated and compared for these two models for their discrimination capability and a curve using predicted probability versus observed probability were plotted to demonstrate the calibration measure for these two models.

RESULTS

: According to this neural network, the top 5 most important predictors were number of ER visits, referral from other facility, discharged to outpatient office, age 30 or younger, and number of outpatient visits. The cumulative weights of these 5 accounted for around 80% of all weights.

For training sample, ROC was 0.66 for the Logistic regression and 0.68 for the artificial neural network. In testing sample, ROC was 0.66 for the Logistic regression and 0.68 for the artificial neural network. Artificial neural network had similar performance with Logistic regression.

CONCLUSIONS

: In this study, we identified several important predictors for hospital readmission among diabetes patients. When compared to artificial neural network model, logistic regression had a similar discriminating capability and calibration between predicted probability and observed probability.