Study on Prediction of hospitalization days in NICU and GCU using Birth Weight and Gestational day

Abstract

The purpose of this study is to construct a multiple regression equation that predicts the length of hospital stay for NICU and GCU using information obtained at the time of admission, and to verify the cross-validity of the prediction equation using test cases. The subjects were 1280 NICU and GCU inpatients and divided into 1149 model construction cases and 131 test cases by the Hold-Out method. The information obtained at the time of admission was used as an explanatory variable, and a hierarchial multiple regression analysis was using the number of days of hospitalization until discharge as the objective variable. The AIC value was obtained from the obtained model and the optimal statistical model was selected. The explanatory valiables of the selected model were weight at admission, gestational day, head circumference at admission, Apgar score 1 minutes, Apgar score of 5 minutes, presence of obstetric abnormalities, transvaginal delivery, cesarean delivery, primiparity and delivery. Comparison of the actual length of hospital stay with the predicted length of hospital stay in the test case showed no significant difference and showed a high correlation,supporting the cross-validity of the subjects at the facility.