Identifying Key Psychosocial Predictors of Hospital Readmission Among Older Adults: Application of Textual Analytics in Electronic Health Records

Abstract

One in 5 older adults aged 65 years and older hospitalized in Singapore public hospitals are readmitted within 30 days of discharge. Hospital readmissions are distressing to patients, burdensome to their families, and costly for the health care system. Their predictors include prior health care utilization, multimorbidity, and hospitalization severity.

In addition, psychosocial factors that influence physical health outcomes through a psychological mechanism, account for non-trivial proportions of the total risk among older adults. While information on typical predictors may be extracted from structured data in electronic health records (EHR), psychosocial risk factors are largely embedded as unstructured data which are challenging to extract. An alternative method using textual analytics to curate data found in EHR clinical notes. To date, only a few textual analytic studies have examined psychosocial information in EHR data. Thus, the overarching goal of this study is to obtain proof-of-concept for the text analytic approach to identify key psychosocial factors that are predictors of 30-day readmission, namely depressive symptoms, anxiety, poor social support, financial strain, living alone, and caregiver stress.

The specific aims are to assemble a lexicon of EHR-available search terms that define these psychosocial factors, to develop lexicon-based case detection algorithms for their identification, and to ascertain whether they add to existing prediction models for 30-day hospital readmission based on clinical and administrative data of older adults.

To this end, natural language processing (NLP) text-mining will be used to codify psychosocial factors through machine learning. Training, testing, and validation samples will comprise older adults hospitalized at Ng Teng Fong General Hospital and who are at higher risk for readmission. We will test our hypothesis that key psychosocial factors identified by textual analysis improve prediction of 30-day hospital readmission in older adults beyond that achieved by clinical and administrative variables alone.

Principal Investigator

Goh Kim Huat

Prof Goh Kim Huat

Nanyang Business School

Kim Huat is the Associate Dean (Graduate Studies) and Senior Editor for the Journal of the Association for Information Systems. He received his Ph.D. in business administration with a specialization in economics and information systems from the Carls ...

Appointments:
Associate Dean (Graduate Studies), College of Business (Nanyang Business School) Professor, College of Business (Nanyang Business School) - Division of Information Technology & Operations Management

Keywords: Business and Management | Economics | Info-Communication Technology | Internet & Communications | Science, Technology and Society

Lydia AU (Dr)
NG TENG FONG GENERAL HOSPITAL