Title: Predicting re-admission to hospital for diabetes treatment: a machine learning solution
Authors: Satish M. Srinivasan; Yok-Fong Paat; Philmore Halls; Ruth Kalule; Thomas E. Harvey
Addresses: School of Graduate Professional Studies, Penn State Great Valley, Malvern, PA 19355, USA ' Department of Social Work, The University of Texas at El Paso, El Paso, TX 79968, USA ' School of Graduate Professional Studies, Penn State Great Valley, Malvern, PA 19355, USA ' School of Graduate Professional Studies, Penn State Great Valley, Malvern, PA 19355, USA ' School of Graduate Professional Studies, Penn State Great Valley, Malvern, PA 19355, USA
Abstract: Predictive analytics embrace an extensive range of techniques for identifying patterns within data to predict future outcomes and trends. The objective of this study is to design and implement a predictive analytics system that can be used to forecast the likelihood that a diabetic patient will be readmitted to the hospital. Using the Diabetes 130-US hospitals dataset we modelled the relationship between the patient re-admission (predictor) and the response variable using the Random Forest classifier. We obtained a maximum AUC of 0.684 and an F1 Score of 52.07%. Our study reveals that attributes such as number of inpatient visits, discharge disposition, admission type, and number of laboratory tests are strong predictors for the re-admission of patients. Findings from this study can help hospitals design suitable protocols to ensure that patients with a higher probability of re-admission are recovering well and possibly reduce the risk of future re-admission.
Keywords: RF; random forest; hyperparameter tuning; data cleaning; predictive analytics; data cleaning; re-admission; hospitals; diabetes; AUC; recall.
DOI: 10.1504/IJCBDD.2020.113823
International Journal of Computational Biology and Drug Design, 2020 Vol.13 No.5/6, pp.539 - 554
Received: 04 Nov 2019
Accepted: 18 May 2020
Published online: 31 Mar 2021 *