In a recent article in Pediatrics, “Reducing Absenteeism From
Gastrointestinal and Respiratory Illness in Elementary School Students: A
Randomized, Controlled Trial of an Infection-Control Intervention”,
Sandora and colleagues stated that “a multifactorial intervention
including hand sanitizer and surface disinfection reduced absenteeism
caused by gastrointestinal illness in elementary school students” (1).
While we applaud all efforts to improve hygiene in schools, we question
both the validity and importance of their findings.
To begin with, the authors used a clustered randomized design with
teams as the unit of assignment. This is also commonly called a group
randomized trial (GRT). We agree with the authors that a general
estimating equation with teams as a random effect can be used to analyze
this data, but it has been noted that in situations where there are less
than 40 groups the sandwich estimator used in GEE is biased downward (2,
3). This means that without proper correction of the variance you will
get an incorrectly low estimate of your standard error. It is unclear if
the authors accounted for this, and in their later estimates of sample
size it appears they didn’t. The authors used the total number of
students in their power calculations without consideration that the actual
unit of randomization was the team. There was no mention in the power
calculation of adjusting for the intraclass correlation, or the variance
inflation factor which is needed when randomization is at the group level
(4). This casts some doubt over the p-values of the intervention effect
that they found.
However, even if we accept the validity of their effect measures, the
difference in absenteeism rates between the two groups was 3.7 absences
per 1,000 student-days of observation. This translates into a predicted
22 fewer absences due to gastrointestinal illnesses in the treatment group
over the study period. How important is this? In public schools,
increased absenteeism reduces state subsidies. Based on state allocations
in Minnesota, the value of these absences would be around $1,200.
However, the costs of 205 containers of sanitizing wipes and 70 bottles of
hand sanitizers appear to be in the same ballpark.
Finally, the results of environmental sampling do not provide
compelling evidence for a clear and consistent effect from the use of
these products in the classroom. During week 1 the treated classrooms
clearly had fewer samples positive for norovirus and lower heterotrophic
plate counts. It would be nice to credit these results with the lowered
rates of gastrointestinal illness in the treatment classes. However,
doing so is not that straightforward. During week 2 the control rooms
were cleaner than the intervention rooms had been during week 1. During
week 4, both control and intervention rooms resembled the control rooms
from week 1. No explanation is offered for the instability of the
environmental sampling results, but they suggest that the use of hand
sanitizers and surface disinfectants was not the primary determinant of
the hygiene levels in the classroom.
Working out the fine details of how the use of hand washing and
personal hygiene measures can reduce absenteeism in schools is a difficult
task. Sandora and colleagues have demonstrated an approach that could be
useful on a larger scale with better linkage between environmental
monitoring and observed patterns of absenteeism.
1. Sandora TJ, Shih MC, Goldmann DA. Reducing Absenteeism From
Gastrointestinal and Respiratory Illness in Elementary School Students: A
Randomized, Controlled Trial of an Infection-Control Intervention.
Pediatrics 2008;121(6):e1555-e1562.
2. Murray DM, Varnell SP, Blitstein JL. Design and analysis of group-
randomized trials: a review of recent methodological developments. Am J
Public Health 2004;94(3):423-32.
3. Murray DM. Design and analysis of group-randomized trials. New York:
Oxford University Press; 1998.
4. Hsieh FY. Sample size formulae for intervention studies with the
cluster as unit of randomization. Stat Med 1988;7(11):1195-201.
Conflict of Interest:
None declared
![[Read eLetters]](/icons/misc/sectionDOWN.gif){kind=icon}
