Secondary Logo

Detecting the Onset of Infectious Disease Outbreaks Using School Sign-out Logs

Weiss, Zoe

doi: 10.1097/EDE.0000000000000969
Letters
Free
SDC

Georgia Institute of Technology, Lakeside High School, Atlanta, GA, zoeweiss@gatech.edu

The authors report no conflicts of interest.

Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com).

Data: The manuscript contains all analyzed data. The raw data (the attendance sheets) are available from the author. All analysis performed with standard Excel functions.

IRB Protocols: All data were provided to the author in redacted form with no names, grades, or other identifying information of any study subjects. Thus, no IRB approval was necessary. See Supplemental Materials for data in the received form.

Back to Top | Article Outline

To the Editor:

We report on an investigation of a novel method to detect and track an outbreak of influenza or other infectious disease in real time. The method employs sign-out logs in school attendance offices. This is distinct from standard attendance data, which are limited to a count of absences and does not include the reason for the absence.1,2 The new method overcomes limitations of current methodologies, including multiweek lag time, coverage gaps, and cost.3,4

The method reported herein details a pilot study conducted in a large middle school (approximately 1500 students) in Atlanta, Georgia. The goal was to detect the onset of an influenza outbreak using school sign-out log data and to compare the prediction with the onset obtained from influenza-like illness data collected by the Georgia Department of Public Health through their statewide surveillance network of healthcare providers. This study addressed two key aspects of detection using school sign-out logs: accuracy and advance detection. Middle-school sign-out logs spanning the dates of 14 August 2013 through 21 February 2014 were obtained (see eFigure 1 for data format; http://links.lww.com/EDE/B462). For each day, we tabulated the number of students who were signed out by their parent or guardian with the written reason of sign-out as “sick”, “doctor”, or “other”. We then compared these data (eTable 1; http://links.lww.com/EDE/B460) to the Georgia Weekly Influenza Report for weeks 40 to 52 for 2013 and weeks 1 to 9 for 2014, spanning 29 September 2013 through 1 March 2014.5 It is important to note that the middle-school data were available 2 weeks earlier than the influenza-like illness data during this study.

We evaluated the agreement between the weekly number of school sign-outs and the weekly influenza-like illness data (received after a 2-week lag) from the same week. A linear regression yielded a correlation coefficient of 0.64. We also used a first-order autoregressive model with Poisson noise6 to estimate the expected number of statewide influenza-like illness cases given the sign-out data. A lag time of 0 weeks in the model corresponds to comparing sign-outs and statewide surveillance influenza-like illness visits during the same week although the sign-out data were available 2 weeks before the influenza-like illness data. The Figure shows a comparison of the number of observed statewide influenza-like illness cases, the number of sign-outs, and the number of expected cases based on the first-order autoregressive model for each of the 25 weeks, using a lag time of 0 weeks. As a sensitivity analysis, multiple lag times were explored, resulting in the lowest Akaike Information Criterion value for the model with a 0-week time lag (eTable 2; http://links.lww.com/EDE/B461). The surveillance data collected from school sign-out logs were highly correlated with the influenza-like illness data, strongly suggesting the possibility of detecting the onset of an outbreak sooner than current methodologies.7

FIGURE

FIGURE

One limitation of this method is that it necessarily misses a few weeks of sign-out data during the Thanksgiving and winter breaks. The missing data were filled in by standard averaging.8 This method will detect the outbreak of any infectious disease, not just influenza. Thus, detecting an influenza outbreak requires this method be combined with another method. The volatility of the sign-out data can be reduced by including additional schools, a future research project. One benefit of an inexpensive, rapidly assimilable, advanced indicator of the onset of an outbreak using this method is the fact that the middle-school data were available 2 weeks in advance of the publication of the statewide influenza-like illness data collected during this study.

Back to Top | Article Outline

ACKNOWLEDGMENTS

I acknowledge the following individuals: Wendy Smith, Epidemiology Preparedness Director at the Georgia Department of Public Health, who provided the Georgia Weekly Influenza Reports; Tiffany Brown, Middle School Assistant Principal, who provided the redacted middle-school attendance office data; and Vicki Hertzberg, Professor of Nursing at Emory University, who assisted with analyzing missing data.

Zoe Weiss

Georgia Institute of Technology

Lakeside High School

Atlanta, GA

zoeweiss@gatech.edu

Back to Top | Article Outline

REFERENCES

1. Egger JR, Hoen AG, Brownstein JS, Buckeridge DL, Olson DR, Konty KJ. Usefulness of school absenteeism data for predicting influenza outbreaks, United States. Emerg Infect Dis. 2012;18:1375–1377.
2. Nsoesie EO, Brownstein JS, Ramakrishnan N, Marathe MV. A systematic review of studies on forecasting the dynamics of influenza outbreaks. Influenza Other Respir Viruses. 2014;8:309–316.
3. Ferguson NM, Cummings DA, Fraser C, Cajka JC, Cooley PC, Burke DS. Strategies for mitigating an influenza pandemic. Nature. 2006;442:448–452.
4. Webby RJ, Webster RG. Are we ready for pandemic influenza? Science. 2003;302:1519–1522.
5. Georgia Department of Public Health, Georgia Weekly Influenza Report (2014). MMWR Weeks 40 – 42. 2014 and Weeks 1–9, 2014.
6. Brandt PT, Williams JT. A linear Poisson autoregressive model: the Poisson AR (p) model. Political Analysis, 2001;9:164–184.
7. Wilson K, Brownstein JS. Early detection of disease outbreaks using the Internet. CMAJ. 2009;180:829–831.
8. Little RJ, Rubin DB. Statistical Analysis With Missing Data (2014.Vol 333). New York, NY: John Wiley & Sons.

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.