Mobile Messaging as Surveillance Tool during Pandemic (H1N1) 2009, Mexico

Pandemic (H1N1) 2009 highlighted challenges faced by disease surveillance systems. New approaches to complement traditional surveillance are needed, and new technologies provide new opportunities. We evaluated cell phone technology for surveillance of influenza outbreaks during the outbreak of pandemic (H1N1) 2009 in Mexico.

symptoms. Mean number of other persons with ILI in the household was 1.6 among respondents reporting severe disease and 0.3 among those with nonsevere disease (p<0.0001, by t test).
Epidemic curves for disease onset for confi rmed and suspected cases of pandemic (H1N1) 2009 from SI-NAVE and daily proportion of severe cases from the cell phone survey are shown in the Figure. Daily counts of ILI were clustered around multiples of 5, and no distinct pattern was observed (online Technical Appendix Figure 2). Use of the daily proportion of severe cases may partially correct for clustering and artifactual peaks by standardizing by total number of cases. The proportion of severe cases increased throughout the month beginning on April 1 (36.4%) and peaking on April 26 (57.9%). Two distinct decreases in severity of disease coincided with Semana Santa school vacation and school closures on April 24. These decreases are consistent with the decrease in the SINAVE epidemic curve.
The pattern of change in the proportion of severe ILI may be consistent with a decrease in transmission after control measures were implemented. The low response rate (5.8%) made it likely that respondents were not representative of the total population. Therefore, we did not make estimates of disease incidence. We were unable to determine whether a pathogen for which susceptibility was higher was responsible for the difference in number of ILI cases within the household of those reporting severe disease or whether respondents in households with several affected persons were more likely to report severe disease (online Technical Appendix Table). We observed unexpected peaks and a clustering of date of fever onset. However, the peak on April 1 may refl ect disease at the end of March, and the decrease in daily proportion of severe cases may indicate lower incidence of ILI after school closures. Comparison of these data with epidemic curves for pandemic (H1N1) 2009 showed less variability than expected; no geographic variation was detected.
Our study was limited by potential selection bias, recall bias, and in-  clusion of mostly young persons from urban areas. Comparisons between reported cases and noncases are invalid because of the low response rate. However, comparisons within cases may be less prone to bias if they are more likely to respond. Persons had diffi culty remembering the exact date of fever onset. In 2 telephone surveys in New York City outside the infl uenza season (March and October-November 2003), a total of 20.8% and 19.6% of respondents, respectively, reported ILI within the past month (4), which were more than the rate of 12% during the peak of pandemic (H1N1) 2009 in New York (5). Use of daily proportion of severe cases may have partially corrected for this recall error. Also, persons may be more likely to report ILI if the date of onset was closer to the date of the survey. Nevertheless, a lower number of cases by the end of the month indicates that more accurate recall for recent dates may not be a serious problem. Generalizability of these results is of concern. However, the age group that was captured was most affected in the early stages of this outbreak (6).
Effi cient estimation of extent of disease caused by a novel infectious agent may be costly and logistically diffi cult. When carefully deployed, unstructured supplementary service data surveys may be a practical, low-cost, and timely complement to traditional surveillance. Further refi nements of this tool are required to improve its validity. To limit recall errors and increase response rate, repeated surveys at short intervals and specifi c strategies to improve response rate should be considered.