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

Demographic Health Analysis by Incorporation of 
Census Data with Patient Records
Christopher R. Cuellar* and Wayne Loschen

Johns Hopkins University Applied Physics Laboratory, Columbia, MD, USA

Objective
The objective of this project is to enable a deeper analysis of 

patient health by correlating patient health records with the census 
demographic data. Based upon these correlations, the ESSENCE 
system will be enhanced with new query filtering capabilities.

Introduction
Electronic disease surveillance canonically represents analysis 

performed on health records with respect to their syndromes, 
complaints, lab data, etc. This data can tell the story of a patient’s 
current status but does not provide a holistic look at the where the 
patient is from. By incorporating census data, a deeper examination of 
the patient’s area can be performed which may result in discovery of 
risk factors associated with race, economic status, and culture.

Methods
Data was gathered from census surveys conducted by the US 

government in a comma delimited file [1]. Datasets used included 
household income, poverty status in the past 10 years, and race. All 
data within each file was associated with a census ZCTA block. 
ZCTA’s, while similar to zip codes, do not share a one-to-one 
correlation with zip codes, and subsequently, need to be mapped 
appropriately to use them with zip code-based health records. The 
most common approach to this mapping is to calculate the centroid 
of the ZCTA region and project that onto the zip code region. We 
used a dataset that performed such a mapping for our zip code to 
ZCTA correlation [2]. Finally, through collaboration with the Florida 
Department of Health, we developed several groupings for the data. 
Partitions of health records were then built based upon factors such as 
percentage of race or ethnicity in a specific ZCTA, median household 
income, and predominant race for the resident ZCTA block. These 
groupings were then incorporated into the Query Portal feature of 
the system.

Results
Users can now filter for all health data by a demographic factor, 

such as finding all patients from a predominantly Hispanic zip code, 
or all fever cases from zip codes with a median income level of 
$50,000 or more. Primary complications were due to the usage of 
ZCTA for census data and the usage of zip code for patient records. 
Since there is no exact one-to-one mapping, it is impossible to assure 
that patients reside within a specific ZCTA without a more granular 
location being specified. ZCTA data is also subject to the number 
of people responding to surveys within a given area and can be 
unreliable for areas of low participation. The binning for census data 
is variable depending on the location which it is being applied, there is 
no perfect separation that can be applied especially when considering 
median household income.

Conclusions
Application of census data can be burdensome, especially as zip 

codes have a tendency to change in shape; however, the usefulness of 
census data in determining societal risk factors can be immeasurable. 
Future work includes developing the capability to automatically 

determine changes in ZCTA and zip code representations. 
Development of appropriate binning for census data is a problem 
of localization as features such as median household income can be 
overall higher or lower within a state compared to other states. Finally, 
inclusion of a metric comparing cost of living should be added to get 
a better idea of the worth of specific median household incomes. This 
talk will detail the technical approaches, including complications, 
along with the types of data incorporated and the resulting forms of 
analysis possible. The datasets themselves will be discussed in terms 
of their granularity and what information they can provide.

Keywords
Census Data; System Development; Query Improvement

References

1) U.S. Census Bureau; using American FactFinder; http://factfinder2.
census.gov;(May 2014)

2) UDS Mapper; http://www.udsmapper.org

*Christopher R. Cuellar
E-mail: Christopher.Cuellar@jhuapl.edu    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e16, 201