Abstract

Despite several improvements in various domains of healthcare systems, the inability to reduce patients’ readmission rates is still a major problem faced by healthcare providers. It is extremely important to reduce the probability of readmissions because these not only increase the burden of healthcare costs on the patient but also expose them to prolonged psychological stress (e.g., trauma, pain, and discomfort due to altered physical functions) and healthcare-associated infections. Readmissions basically lead to the use of healthcare resources by the same person twice instead of being utilized by another patient. Furthermore, the readmission rate is used as a potential measure of healthcare quality. The high readmission rates may be due to a poor quality of care provided and could tarnish the reputation of the healthcare facilities. It could also reduce hospitals’ reimbursements from the insurance companies. The goal of this research is to quantify the risk of hospital readmissions by analyzing significant factors in patients undergoing skin procedures and to also identify the best time-frame and the corresponding predictors that could be used for predicting future readmissions related to skin procedures. Specifically, data analysis and predictive modeling approach will be adopted to identify the predictors of readmissions using a dataset of over 22,000 hospitalizations. The proposed the methodology will concentrate on patients’ demographics such as their age and gender along with the type of service, place of service, and others in order to predict if readmissions could be explained by these factors. Our study will analyze the significance of the above-mentioned factors for readmissions occurring over six different time intervals. The time-intervals being considered under this study are within 7 days, 15 days, 30 days, 45 days, 90 days and 1 year of initial admission. After analyzing the predictors over different time intervals, we found that although the significant factors differ for different time intervals, readmissions for the selected group of patients can be correlated to specific predicting variables for each of the time-interval. One of the predictors that seemed to be consistent over five different time intervals is the patient’s age. The care provider is another predictor, which was identified as statistically significant for more than one of the time intervals. Utilizing the training, validation, and testing data split, we were able to predict a probable outcome of whether or not it would be a case of readmission. By employing the confusion matrix to compare this predicted outcome against the actual outcome, the study checked the authenticity and accuracy of the models developed for each of the time intervals. Based on the performance measures developed using the confusion matrix, the best time-intervals to predict the readmissions are 7 days and 90 days with the F-1 score of 0.53, whereas the worst-time interval is 15 days.

Library of Congress Subject Headings

Hospital patients--Rehabilitation--Data processing; Hospital care--Quality control--Data processing; Medical care--Data processing

Publication Date

5-4-2021

Document Type

Thesis

Student Type

Graduate

Degree Name

Industrial and Systems Engineering (MS)

Department, Program, or Center

Industrial and Systems Engineering (KGCOE)

Advisor

Nasibeh Azadeh-Fard

Advisor/Committee Member

Katie McConky

Campus

RIT – Main Campus

Plan Codes

ISEE-MS

Download

Share

COinS