JMIR Publications

ABSTRACT

Background:

Older adults account for the largest percentage of the long-term care (LTC) and total medical spending. Prior studies have identified several data sources and many predictors for nursing home admission among the elderly [5]-[10]; however, such studies have often used data sources with limited availability across large populations and methods with widely disparate sample populations. Insurance claims and electronic health records (EHR) are widely available across large populations of older adults. Predictors extracted from insurance claims or EHR data can be extremely useful for predicting nursing home admission among older adults on a population scale.

Objective:

This study is to synthesize previous findings and to identify advances and gaps in the recent literature on predictors of nursing home admission that are collected from insurance claims or EHRs.

Methods:

To report the study selection, PRISMA extension for scoping reviews (PRISMA-ScR) guidelines were used.

Results:

A total of 27 papers were selected for final inclusion in this review. In most studies, mean or average age of total sample ranged from 70 to 79 years (n=15, 56%). In addition to nursing home admission, all-cause mortality (n=12; 32%) and hospitalization/ rehospitalization (n=7; 19%) were frequently used as outcome measures. In the selected studies, 33.0%(n=9) used Cox proportional hazard models while 30.0%(n=8) used logistic regression models. Comorbidity counts, frailty, cognitive function, medication, medical and social service utilization were used as predictors of nursing home admissions using EHR and insurance claims data.

Conclusions:

Prediction tools based on EHRs or claims data may assist clinicians and patients to make informed decision. Future studies should develop discussion on how to adapt the predictive models of nursing admissions in healthcare systems.