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

Using Emergency Department Data for Detection of a 
Synthetic Marijuana Outbreak
Lourdes W. Yun*1, Robert Beum1, Michele Askenazi2, Yushiuan Chen2, Dean McEwen1, 
Moises Maravi1 and Arthur Davidson1

1Denver Health and Hospitals, Denver, CO, USA; 2Tri-County Health Department, Greenwood Village, CO, USA

Objective
The aims of this presentation is to use ED chief complaint data, 

to test BioSense 2.0 for detection of a novel public health event 
(i.e., serious adverse events related to synthetic marijuana use) not 
currently categorized in the BioSense syndromic surveillance library.

Introduction
Timely access to Emergency Department (ED) Chief Complaint 

(CC) data, before the definitive diagnosis is established, allows 
for early outbreak detection and prompt response by public health 
officials.BioSense 2.0 is a cloud-based application that securely 
collects, tracks, and shares ED data from participating hospitals 
around the country. Denver Health (DH) is one of several Colorado 
hospitals contributing ED Chief Complaint data to BioSense 2.0. In 
August 2013, ED clinicians reported an increase in patients presenting 
with excited delirium, possibly related to synthetic marijuana (SM). 
We used this event to test the use of CC field of ED data for detection 
of a novel public health event (i.e., serious adverse events related to 
synthetic marijuana use) not currently categorized in the BioSense 
syndromic surveillance library.

Methods
Daily routines retrieved DH hospital data from BioSense 2.0 to 

local public health servers: 1. ‘R’ programs access BioSense servers 
that contain the Raw, Binned and Exceptions syndromic data; 2. 
UNIX CRON jobs were created to schedule these ‘R’ jobs to run on 
a regular daily basis and pull the results down to the DH Servers; 3. 
Scheduled SAS batch job that APPENDS the retrieved data into one 
SAS dataset. Using SAS INDEX functions, we searched for specific 
keywords likely related to SM-usage in the CC field. For example, “if 
(index(cc, “blueberry”) > 0 or index(cc, “boogie”) > 0 or index(cc, 
“cannabinoid”) > 0 or index(cc, “cannabis”) > 0 or ….The ‘IF’ 
logic grew quickly as the list of possible keywords/exclusions and 
the criteria became unmaintainable. To develop a maintainable code 
solution, a custom built library containing 89 SM-related keywords 
and exclusion words was developed. SM-related keywords were 
provided by ED clinicians. Exclusion words were words in the CC 
field that came back as “hits” but were not SM-related. For example, 
“spice” is a keyword but “hospice” is an exclusion word. Not all 89 
keywords were found in the CC data; only 20 keywords provided 
significant numbers of hits for possible SM-use. Because of the sheer 
volume of records returned and lack of specificity of some keywords, 
15 of the 20 keywords were further excluded.

To assess the sensitivity and specificity of this method, we 
compared the patients seen at DH ED with a known SM-related 
diagnosis (gold standard) during the event period (September 4-19, 
2013) with the patients with a CC where a SM-related keyword was 
observed.

Results
Between August 2013-April 2014, using the 5 keywords (mamba, 

marijuana, mj, spice, thc), 106 SM- related records were identified in 
the analysis. During the outbreak period (September 4-19, 2013), 21 

cases of SM-related illness presented to the DH emergency department 
and a total of 27 of SM-syndrome associated records were identified. 
Sensitivity and specificity analysis was based on the sample of these 
27 records. Using this method to detect SM-related events was highly 
specific (99.6%) but only moderately sensitive (44%).

Conclusions
Keywords and exclusions were used to identify a novel synthetic 

marijuana outbreak with adverse health effects through analysis of ED 
record CC field. Keyword data were difficult to analyze due to many 
acronyms, abbreviations, misspellings, and truncation of words. As 
triage personnel may be unfamiliar with SM-related terms and enter 
the target keyword efforts may be required to educate ED clinicians 
during the event to improve sensitivity.

Keywords
syndromic surveillance; synthetic marijuana; emergency department

*Lourdes W. Yun
E-mail: lourdes.yun@dhha.org    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e178, 201