Playground injury rates (injuries per 1000 children per month)
were compared at the intervention schools and at the nonintervention
schools before equipment removal and after equipment
replacement. Person-time denominators used the average yearly
enrollment in each school multiplied by the number of months
that the school was in session. The same 10-month calendar periods
were selected before and after the intervention to avoid bias
related to seasonal variation in injury. The pre-intervention period
was from September 1999 to June 2000, and the postintervention
period was from January 2002 to December 2002
(excluding July and August). To adjust for clustering within
schools, injury rates and relative risks (RRs), along with their confidence
confidence
intervals (CIs), were estimated using random-effect metaanalytic
methods as proposed by Thompson and colleagues.13
Expected post-intervention injury rates in the intervention
schools were calculated by multiplying the pre-intervention rates
in intervention schools by the ratio of post-intervention to preintervention
rates in the nonintervention schools. The proportion
of injuries prevented (etiological fractions)14 were calculated by
subtracting observed from expected rates and expressing the result
as a percentage of the expected rate. The analysis was repeated for
the subset of injuries flagged as equipment related.