Achintya Dey

Practical comparison of aberration detection algorithms for biosurveillance systems

National syndromic surveillance systems require optimal anomaly detection methods. For method performance comparison, we injected multi-day signals stochastically drawn from lognormal distributions into time series of aggregated daily visit counts from the U.S. Centers for Disease Control and Preventions BioSense syndromic surveillance system. The time series corresponded to three different syndrome groups: rash, upper respiratory infection, and gastrointestinal illness. We included a sample of facilities with data reported every day and with median daily syndromic counts ⩾1 over the entire study period. We compared anomaly detection methods of five control chart adaptations, a linear regression model and a Poisson regression model. We assessed sensitivity and timeliness of these methods for detection of multi-day signals. At a daily background alert rate of 1% and 2%, the sensitivities and timeliness ranged from 24 to 77% and 3.3 to 6.1days, respectively. The overall sensitivity and timeliness increased substantially after stratification by weekday versus weekend and holiday. Adjusting the baseline syndromic count by the total number of facility visits gave consistently improved sensitivity and timeliness without stratification, but it provided better performance when combined with stratification. The daily syndrome/total-visit proportion method did not improve the performance. In general, alerting based on linear regression outperformed control chart based methods. A Poisson regression model obtained the best sensitivity in the series with high-count data.

Automated monitoring of clusters of falls associated with severe winter weather using the BioSense system.

Objectives To identify and characterise clusters of emergency department (ED) visits for fall injuries during the 2007–2008 winter season. Methods Hospital ED chief complaints and diagnoses from hospitals reporting to the Centers for Disease Control and Prevention BioSense system were analysed. The authors performed descriptive analyses, used time series charts on data aggregated by metropolitan statistical areas (MSAs), and used SaTScan to find spatial–temporal clusters of visits from falls. Results In 2007–2008, 17 clusters of falls in 13 MSAs were found; the median number of excess ED visits for falls was 71 per day. SaTScan identified 11 clusters of falls, of which seven corresponded to MSA clusters found by time series and five included more than one state/district. Most clusters coincided with known periods of snowfall or freezing rain. Conclusion The results show the role that a national automated system can play in tracking widespread injuries. Such a system could be harnessed to assist with prevention strategies.

Coding of Electronic Laboratory Reports for Biosurveillance, Selected United States Hospitals, 2011

Objective Electronic laboratory reporting has been promoted as a public health priority. The Office of the U.S. National Coordinator for Health Information Technology has endorsed two coding systems: Logical Observation Identifiers Names and Codes (LOINC) for laboratory test orders and Systemized Nomenclature of Medicine-Clinical Terms (SNOMED CT) for test results. Materials and Methods We examined LOINC and SNOMED CT code use in electronic laboratory data reported in 2011 by 63 non-federal hospitals to BioSense electronic syndromic surveillance system. We analyzed the frequencies, characteristics, and code concepts of test orders and results. Results A total of 14,028,774 laboratory test orders or results were reported. No test orders used SNOMED CT codes. To describe test orders, 77% used a LOINC code, 17% had no value, and 6% had a non-informative value, “OTH”. Thirty-three percent (33%) of test results had missing or non-informative codes. For test results with at least one informative value, 91.8% had only LOINC codes, 0.7% had only SNOMED codes, and 7.4% had both. Of 108 SNOMED CT codes reported without LOINC codes, 45% could be matched to at least one LOINC code. Conclusion Missing or non-informative codes comprised almost a quarter of laboratory test orders and a third of test results reported to BioSense by non-federal hospitals. Use of LOINC codes for laboratory test results was more common than use of SNOMED CT. Complete and standardized coding could improve the usefulness of laboratory data for public health surveillance and response.

Dengue Surveillance in Veterans Affairs Healthcare Facilities, 2007–2010

Background Although dengue is endemic in Puerto Rico (PR), 2007 and 2010 were recognized as epidemic years. In the continental United States (US), outside of the Texas-Mexico border, there had not been a dengue outbreak since 1946 until dengue re-emerged in Key West, Florida (FL), in 2009–2010. The objective of this study was to use electronic and manual surveillance systems to identify dengue cases in Veterans Affairs (VA) healthcare facilities and then to clinically compare dengue cases in Veterans presenting for care in PR and in FL. Methodology Outpatient encounters from 1/2007–12/2010 and inpatient admissions (only available from 10/2009–12/2010) with dengue diagnostic codes at all VA facilities were identified using VAs Electronic Surveillance System for Early Notification of Community-based Epidemics (ESSENCE). Additional case sources included VA data from Centers for Disease Control and Prevention BioSense and VA infection preventionists. Case reviews were performed. Categorical data was compared using Mantel-Haenszel or Fisher Exact tests and continuous variables using t-tests. Dengue case residence was mapped. Findings Two hundred eighty-eight and 21 PR and FL dengue cases respectively were identified. Of 21 FL cases, 12 were exposed in Key West and 9 were imported. During epidemic years, FL cases had significantly increased dengue testing and intensive care admissions, but lower hospitalization rates and headache or eye pain symptoms compared to PR cases. There were no significant differences in clinical symptoms, laboratory abnormalities or outcomes between epidemic and non-epidemic year cases in FL and PR. Confirmed/probable cases were significantly more likely to be hospitalized and have thrombocytopenia or leukopenia compared to suspected cases. Conclusions Dengue re-introduction in the continental US warrants increased dengue surveillance and education in VA. Throughout VA, under-testing of suspected cases highlights the need to emphasize use of diagnostic testing to better understand the magnitude of dengue among Veterans.

National and regional representativeness of hospital emergency department visit data in the National Syndromic Surveillance Program, United States, 2014

OBJECTIVE We examined the representativeness of the nonfederal hospital emergency department (ED) visit data in the National Syndromic Surveillance Program (NSSP). METHODS We used the 2012 American Hospital Association Annual Survey Database, other databases, and information from state and local health departments participating in the NSSP about which hospitals submitted data to the NSSP in October 2014. We compared ED visits for hospitals submitting data with all ED visits in all 50 states and Washington, DC. RESULTS Approximately 60.4 million of 134.6 million ED visits nationwide (~45%) were reported to have been submitted to the NSSP. ED visits in 5 of 10 regions and the majority of the states were substantially underrepresented in the NSSP. The NSSP ED visits were similar to national ED visits in terms of many of the characteristics of hospitals and their service areas. However, visits in hospitals with the fewest annual ED visits, in rural trauma centers, and in hospitals serving populations with high percentages of Hispanics and Asians were underrepresented. CONCLUSIONS NSSP nonfederal hospital ED visit data were representative for many hospital characteristics and in some geographic areas but were not very representative nationally and in many locations. Representativeness could be improved by increasing participation in more states and among specific types of hospitals. (Disaster Med Public Health Preparedness. 2016;10:562-569).

Use of Syndromic Data for Enhanced Surveillance: MERS Like-Syndrome

The goal is to identify and monitor MERS like syndrome cases in the syndromic surveillance system. In consultation with the state and local jurisdictions, five case definitions were developed to monitor MERS like syndromes. From May through July, 2014 fifteen reporting jurisdictions participated in MERS enhanced surveillance. . During this enhanced surveillance time period 171 probable MERS cases were identified and all of them were ruled out. The MERS collaborative efforts between BioSense programs, CDC subject matter experts and jurisdictions will help develop more comprehensive definitions to conduct enhanced surveillance at the national level using multiple syndromic surveillance systems.

Overcoming Operational Differences to Attain a National Picture for Novel Threats

Soon after discovery of a MERS case in Indiana, CDC through its BioSense Syndromic Surveillance (SyS) Program joined with certain public health jurisdictions to improve the national-level MERS surveillance picture. Activities were undertaken to bolster local surveillance efforts, despite jurisdictions use of differing SyS tools. This resulted in the ability to generate periodic reports of aggregated MERS-like surveillance data. Many seem to see this initiative to enhance the national MERS surveillance picture as a model to build upon, and a success that can help improve trust and generate hope for creating a meaningful national SyS picture.

Mental Illness and Medical Co-morbidity Using Automated Surveillance Data: BioSense 2008 - 2011

Background: Recent national surveys indicate that 5% of ambulatory care visits involved patients with mental disorder diagnosis. Objective: The objective of this study is to demonstrate the use of automated surveillance data for describing the burden of co-morbidity among patients with mental illness. Methods: We used Emergency Department (ED) visits data from over 650 non-federal hospitals that participated in BioSense from 2008-2011. The variables used in this descriptive analysis are age, gender, and syndromes as defined by BioSense program. The study included only ED visits from people of ≥ 18 years old and with the discharge diagnosis ICD-9-CM codes of mental illness (290 – 312). Co-morbidity was defined broadly as the co-occurrence of other medical condition