2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

ISDS 2014 Conference Abstracts

Impact of Demographics on Healthcare Utilization
Andrew Walsh*

Health Monitoring Systems, Inc, Pittsburgh, PA, USA

Objective
To determine if data collected for syndromic surveillance can 

inform policy questions related to emergency department utilization 
and inpatient readmission

Introduction
The Affordable Care Act (ACA) was promoted with two goals: 

expanding health insurance coverage and reducing healthcare costs. 
Expanded coverage is expected to partially reduce costs. Emergency 
department (ED) visits are costlier than comparable primary care 
physician visits. If uninsured patients use the local ED more often 
than insured patients with comparable conditions, insuring them may 
change usage and lower costs.

Some reports in the literature do not fit this model of ED usage. 
In one study, nonurgent ED visits were mainly the result of patient 
uncertainty about the severity of their condition. While trained 
medical personnel distinguished urgent and nonurgent cases after the 
fact, initial presentations were similar. In Oregon, an expansion of 
Medicaid increased health insurance coverage; ED usage increased 
rather than decreased. Thus, the motivating narrative about insurance 
coverage and ED usage informing the ACA may not be the complete 
story.

Reduction of hospital readmissions is also expected to cut costs 
under the ACA. Hospital process improvements are expected to 
realize this reduction. Recently it was reported that up to 60% of 
hospital readmissions are predicted by patient demographics, raising 
questions about how much control a hospital has over its readmission 
rate. This research will examine whether data collected via syndromic 
surveillance can corroborate these findings.

Methods
US census data was obtained for New Jersey (NJ), Ohio (OH), 

and Pennsylvania (PA) and aggregated to zip code tabulation area 
(ZCTA) for comparison with hospital data which was only available 
at that granularity.

ED registration counts were obtained from 312 hospitals in NJ 
(48), OH (167), and PA (97) for 2011-2013. ED usage rates for the 
ZCTAs were calculated using the maximum of total population values 
for that ZCTA and adjusted for days reporting. Readmission rates 
were obtained from a subset of these hospitals which send inpatient 
data via their syndromic surveillance feed.

Regression analysis used R v3.1.1, with packages lme4 v1.1-7 and 
MASS v7.3-33.

Results
ED registrations were tabulated for 3,537 ZCTAs in NJ, OH, and 

PA. A total of 34,412,761 registrations were observed for a usage rate 
of 0.337 registrations per person per year.

Figure 1 shows the percentage of the ZCTA population with 
public insurance coverage as an indicator of ED usage. ZCTAs 
were assigned to quartile groups based on public insurance coverage 
rate. The more people in a ZCTA with public insurance, the more 
ED visits. Regression confirmed these differences as statistically 
significant. The best fitting multivariate regression model of ED visits 
included several insurance covariates. Demographic data accounted 
for 59% of ED usage variation between ZCTAs.

Inpatient admissions and readmissions for 9 months in 2013 and 
2014 were tabulated for each ZCTA in NJ, OH, and PA; 1,736 ZCTAs 
were represented in the dataset. A total of 425,291 admissions and 
60,501 readmissions were observed for a readmission rate of 0.142 
readmissions per admission.

Regression models were also created for readmission rates. Several 
insurance covariates were also determined to be significant for these 
models. Demographic data accounted for 53% of the variation in 
readmission rates between hospitals.

Conclusions
The consistent association between insurance coverage within 

a ZCTA and healthcare utilization of residents from that ZCTA 
warrants further investigation. Increasing coverage may not have 
the expected effect on healthcare usage and costs. The influence of 
demographic factors also indicates that hospitals may have limited 
ability to improve the targeted outcomes.

As with any study of aggregate data, there is the possibility 
of conflating group observations with individual effects. Before 
implementing policy changes, it is necessary to replicate these 
correlations at an individual level.

Keywords
Affordable Care Act; Census Data; Readmission; Medicaid

Acknowledgments

We thank the New Jersey, Ohio, and Pennsylvania Deparments of Health 
for funding support and data for this work.

*Andrew Walsh
E-mail: andy.walsh@hmsinc.com    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e57, 201