Work

Assessing 3 Outbreak Detection Algorithms in an Electronic Syndromic Surveillance System in a Resource-Limited Setting

Year: 2020

Type: article

Abstract: S yndromic surveillance uses prediagnostic health- related data to signal probable outbreaks warranting public health response (1).Alerta and Vigila are internet-based syndromic surveillance systems successively implemented by Peru's navy (2-4).Among other disease syndromes, individual cases of acute diarrheal disease (ADD) are self-reported to healthcare workers at sites providing care for service members, dependents, and civilian employees.We assessed the performance of 3 ADD aberration detection algorithms in this resource-limited setting: X-bar chart, exponentially weighted moving average (EWMA), and Early Aberration Reporting System (EARS) C3 cumulative sums (CUSUM) models (5-9). The StudyWe defined ADD as >3 loose stools within 24 hours lasting <14 days, epidemic threshold as the incidence of cases in excess of normal for a given period, outbreak as the detection of ADD incidence above the epidemic threshold, true outbreak as an outbreak identified by the system with confirmation by trained field personnel (e.g., enteropathogen isolation), and epidemiological silence as a period during which no cases were reported.We performed descriptive analysis of ADD cases by using data from all reporting sites during 2007-2011, then conducted subsequent analyses by using data on nonbloody ADD.We calculated ADD incidence from weekly reports from Alerta and Vigila during 2007-2011 for naval bases with population denominator data available.We compared the number of site-weeks during which >1 ADD case was reported on shore-based sites with those of sea-based sites and nonbloody ADD incidence in summer with incidence in nonsummer months by using a Mann-Whitney test.Sites with <4 months of epidemiologic silence during 2007-2011 (60 months) were included for outbreak detection analysis.We aggregated nonbloody ADD case counts by epidemiologic week and performed a timeseries analysis by using X-bar chart, EWMA, and modified EARS C3 CUSUM aberration detection algorithms to flag potential outbreak weeks (8).Algorithm details are provided in the Appendix (https://wwwnc.cdc.gov/EID/article/26/9/19-1315-App1.pdf).To account for seasonal variability, nonbloody ADD cases for each week during 2009-2011 (36 months) were compared with an 8-week sliding historical baseline calculated from the current and previous 2 years (10,11).We excluded signals from weeks with <5 cases to minimize false signals associated with epidemiologic silence.Because of

Source: Emerging infectious diseases

DOI: 10.3201/eid2609.191315