Your child’s school was just closed due to an outbreak of flu. Instantly you wonder about the dangers of exposure–will you child show flu symptoms? Should you see your ped? Do you need the H1N1 vaccination? Is everyone over-reacting?
“Currently many U.S. schools don’t have specific or consistent algorithms for deciding whether to shut down,” says epidemiologist John Brownstein, PhD. “They don’t always use quantitative data, and it may be a political or fear-based decision rather than a data-based one.”
The Children’s Hospital Boston recently reported a study led by Brownstein and Anne Gatewood Hoen, PhD of the Children’s Hospital Boston Informatics Program, in collaboration Asami Sasaki of the University of Niigata Prefecture (Niigata, Japan), tapped a detailed set of Japanese data to help guide decision making by schools and government agencies. The analysis was published by the Centers for Disease Control and Prevention in the November issue of Emerging Infectious Diseases.
Sasaki, Hoen and Brownstein analyzed flu absenteeism data from a Japanese school district with 54 elementary schools. Japan makes a good model for studying influenza in schools because it closely monitors school absenteeism due to flu, requires testing for the flu virus in students who become ill, and has a track record of instituting partial or complete school closures during outbreaks. Tracking four consecutive flu seasons (2004-2008), they asked what pattern of flu absenteeism was best for detecting a true school outbreak — balanced against the practical need to keep schools open if possible.
“You’d want get a school closed before an epidemic peaks, to prevent transmission of the virus, but you also don’t want to close a school unnecessarily,” explains Brownstein. “We also wanted an algorithm that’s not too complex, that could be easily implemented by schools.”
A school outbreak was defined as a daily flu absentee rate of more than 10 percent of students. After comparing more than two dozen possible scenarios for closing a school, the analysis suggested three optimal scenarios:
- A single-day influenza-related absentee rate of 5 percent
- Absenteeism of 4 percent or more on two consecutive days
- Absenteeism of 3 percent or more on three consecutive days
The scenarios #2 and #3 performed similarly, with the greatest sensitivity and specificity for predicting a flu outbreak (i.e., the fewest missed predictions and the fewest “false positives.”) Both gave better results than the single-day scenario (#1). The researchers suggest that scenario #2 might be the preferred early warning trigger, balancing the need to prevent transmission with the need to minimize unnecessary closures.
“Our method would give school administrators or government agencies a basis for timely closure decisions, by allowing them to predict the escalation of an outbreak using past absenteeism data,” says Hoen. “It could be used with data from schools in other communities to provide predictions. It would leave decision-making in the hands of local officials, but provide them with a data-driven basis for making those decisions.”
Last spring, during the early days of the H1N1 influenza pandemic, the CDC recommended first a 7-day school closure, then a 14-day closure after appearance of the first suspected case. Later, as more became known about the extent of community spread and disease severity, the CDC changed the recommendation to advise against school closure unless absentee rates interfered with school function. CDC’s current guidelines (10/21/09) don’t provide a specific algorithm, but state that “the decision to selectively dismiss a school should be made locally,” in conjunction with local and state health officials, “and should balance the risks of keeping the students in school with the social disruption that school dismissal can cause.” When the decision is made to dismiss students, CDC recommends doing so for 5 to 7 calendar days.