LITERATURE REVIEW FOR PREDICTING 30-DAY HOSPITAL READMISSION

Author

Sachithra Perera, Michigan Technological UniversityFollow

Date of Award

2018

Document Type

Campus Access Master's Report

Degree Name

Master of Science in Mathematical Sciences (MS)

Administrative Home Department

Department of Mathematical Sciences

Advisor 1

Qiuying Sha

Committee Member 1

Nilufer Onder

Committee Member 2

Kui Zhang

Abstract

The hospital readmission of patients within 30-days of discharge is crucial to healthcare quality and has involved with high costs in Medicare expenditures. Reducing hospital readmissions has earned great importance in recent relevant healthcare studies. Thus, it is important to identify the factors which cause readmissions and methods to optimally predict the risk of 30-day readmissions for readmit patient populations. This study summarizes the findings of a literature review to answer above mentioned questions. Literature reviews on social studies identified the following high impact factors on readmissions: sociodemographic, socioeconomic, social environment, behavioral, sociocognitive, neighborhood, activities of daily living, geography and hospital size, and hospital admission and discharge information. Also, the studies emphasized the need of adjusting the readmission rate calculation by adding social determinants that have not taken into account. Almost all statistical and machine learning models have been proven to provide promising predictive performance when compared to standardize readmission tool, LACE index used by hospitals. Neural networks, penalized model, GBM, and SVM have outperformed other machine learning models when predicting risk of 30-day readmissions.

Recommended Citation

Perera, Sachithra, "LITERATURE REVIEW FOR PREDICTING 30-DAY HOSPITAL READMISSION", Campus Access Master's Report, Michigan Technological University, 2018.

https://doi.org/10.37099/mtu.dc.etdr/676