Model Comparison for Hip Fracture Patients Hospital Readmission Prediction

• Activity Number: 160 - SPEED: Biometrics
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Qingqing Dai* and Zhuqi Miao and lan zhu and Scott Shepherd and William Paiva
• Companies: Oklahoma State University and Oklahoma State University and oklahoma state university and Oklahoma State University and Oklahoma State University
• Keywords: readmission prediction ; logistic regression ; data mining

Abstract

Readmission is common in hip fracture patients and often leads to adverse consequences, especially for geriatric population. To reduce this life threatening risk, the analytical model, e.g. logistic regression model is often applied to identify risk factors for readmission of hip fracture patients, but with low accuracy due to the unknown factors associated with readmission. In this paper, we apply various data mining algorithms, such as random forest algorithm, decision tree and support vector machine, to identify the risk factors and perform predictions on 30-day all-cause inpatient readmissions of hip fracture patients using Cerner Health Facts database. 47,233 elective index admissions of patients aged 50 years and older who undertook hip fracture surgery are included in this study. The prediction accuracy and risk factors identified from data mining approaches are compared with that from logistic regression model through cross-validation process.