A MULTI-LEVEL ASSOCIATION EXTRACTION RECOMMENDATION SYSTEM FOR SURVEYS

Author

Mohammad Reza Khoie, Southern Illinois University CarbondaleFollow

Date of Award

5-1-2017

Degree Name

Master of Science

Department

Computer Science

First Advisor

Rahimi, Shahram

Abstract

Surveys are used by hospitals to evaluate patient satisfaction and to improve general hospital operations. Collected satisfaction data is usually represented to the hospital administration using statistical charts and graphs. Although such visualization is helpful, typically no deeper data analysis is performed to identify important factors which contribute to patient satisfaction. This work presents an unsupervised data-driven methodology for analyzing patient satisfaction survey data. The goal of the proposed exploratory data analysis is to identify patient communities with similar satisfaction levels and the major factors which contribute to their satisfaction. This type of data analysis helps hospitals pinpoint the prevalence of certain satisfaction factors in specific patient communities or clusters of individuals and implement more proactive measures to improve patient experience and care. To this end, two layers of data analysis is performed. In the first layer, patients are clustered based on their responses to the survey questions. Each cluster is then labeled according to its salient features. In the second layer, the clusters of first layer are divided into sub-clusters based on patient demographic data. Associations are derived between the salient features of each cluster and its sub-clusters. Such associations are ranked and validated using statistical validation tools through a different project entitled “A Validation Methodology for Extracted Associations from HCAHPS Dataset,” and valid extracted associations are turned into comments and recommendations for healthcare providers and patients.