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.

206

ISDS 2014 Conference Abstracts

Using Ambulatory Syndromic Surveillance Data for 
Chronic Disease: A BMI Case Study
Andrew Walsh*

Health Monitoring Systems, Inc, Pittsburgh, PA, USA

Objective
To demonstrate the utility of ambulatory syndromic surveillance 

data to public health domains beyond communicable diseases

Introduction
Syndromic surveillance is one of the meaningful use public health 

menu set objectives for eligible professionals. The value of this data 
for syndromic surveillance as an adjunct to the more widely adopted 
emergency department registrations has not been studied extensively. 
It may be that it would improve the sensitivity or timeliness of 
detecting certain communicable disease events, or it may just 
contain signals comparable to what is available via other syndromic 
surveillance data streams. The value of making the effort to collect 
this data is considered contingent on the answer to that question.

Public health is concerned with more than just communicable 
diseases, however. Chronic diseases and their underlying causes are 
also a significant public health concern. Obesity alone is estimated to 
be responsible for 2.5% of the global disease burden, and represents 
a higher fraction in many developed nations. Since chronic diseases 
are not associated with singular events of brief duration, they are 
difficult to track with traditional surveillance methods. They are also 
not typically managed via emergency departments, so syndromic 
surveillance does not capture them well either.

Chronic diseases are often treated by physicians at ambulatory 
practices. Thus data from eligible professionals may provide a means 
for monitoring chronic diseases, or metrics associated with chronic 
diseases, that would not otherwise be as feasible. As a proof of 
concept, this study seeks to determine if body mass index (BMI), 
the standard measure of obesity, can be obtained from ambulatory 
syndromic surveillance messages.

Methods
Syndromic surveillance messages were collected from 

approximately 100 eligible professionals at 21 practices in the greater 
Philadelphia metropolitan area over a period of 1 month. These were 
received and processed by the EpiCenter syndromic surveillance 
system. Height and weight were extracted from OBX segments and 
used to calculate BMI.

The study population was restricted to adults, as determined by visit 
date relative to the patient’s reported date of birth. The data were also 
filtered to exclude outlier height and weight values, as determined by 
the distribution of these values in the dataset. These extreme values 
appeared to represent data entry errors. The mean height and weight 
for each patient were then calculated from all remaining observations, 
and those means were used to calculate the patient BMI.

Results
A total of 194,523 messages were received from 21 practices over 

the 1 month study period. These represented 24,076 unique visits from 
18,858 patients, as determined by the identifiers sent by the practices. 
Of these, 18,047 were adults (95.7%). Of those, 18,022 (99.9%) had 
height and weight measurements determined to be within the limits 
determined from the distributions.

BMI was calculated for those 18,022 patients. It ranged from 10.5 
to 122.7. The median BMI was 27.1 and the mean was 28.1. The full 
distribution of BMI is shown in Figure 1. 270 patients (1.5%) had an 
underweight BMI, 6,194 patients (34.4%) had an overweight BMI, 
and 5,679 (31.5%) had an obese BMI.

Conclusions
This study successfully demonstrated that BMI can be extracted 

from ambulatory syndromic surveillance messages. Minimal data 
processing was necessary to achieve high data quality. Expanding 
to include children would make it more difficult to set lower limits 
for filtering errant height and weight values, but in this setting that 
filtering did not substantially affect the results. Taking the mode 
of values from a given visit rather than the mean might provide an 
alternative method to exclude some errant values.

A longer, longitudinal study would be necessary to determine 
whether monitoring patient BMI in this fashion could actually be 
used to detect meaningful trends of interest to public health, or used 
to improve significant public health outcomes.

Keywords
Ambulatory data; Chronic disease surveillance; BMI; Syndromic 
surveillance

Acknowledgments

We wish to thank the Pennsylvania Department 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):e58, 201