Analyses of all outcomes were conducted at the ICU level, followed
the intention-to-treat approach, and accounted for the
matched-pair design. All tests were 2-sidedwith 5%type I error.
For acquisition of either MRSA or VRE and other outcomeswith
baseline period data,weighted paired t tests compared
changes inrates from baseline (prerandomization) to the
end of the study between the intervention and control ICUs.16
For outcomes without baseline period data, weighted paired
t tests compared study period rates or means between intervention
and control ICUs.16 Weighting accounted for differences in cluster sizes (eg, patient-time at risk) between ICUs
with each pair weighted according to the inverse variance of
the estimated effect size.33 Testing and estimation were performed
on the log scale to account for different ICU sizes34; estimated
rates and 95%confidence intervalswere obtained by
exponentiating. A prespecified secondary analysis of the primary
outcome and key secondary outcomes was performed,
adjusting for ICU admission prevalence of MRSA or VRE. For
each pair, the weight was the inverse variance of the estimated
effect after adjusting for admission prevalence ofMRSA
or VRE. All weighted paired t tests had 9 degrees of
freedom. The statistical plan is in the Supplement.
Analyses of all outcomes were conducted at the ICU level, followed
the intention-to-treat approach, and accounted for the
matched-pair design. All tests were 2-sidedwith 5%type I error.
For acquisition of either MRSA or VRE and other outcomeswith
baseline period data,weighted paired t tests compared
changes inrates from baseline (prerandomization) to the
end of the study between the intervention and control ICUs.16
For outcomes without baseline period data, weighted paired
t tests compared study period rates or means between intervention
and control ICUs.16 Weighting accounted for differences in cluster sizes (eg, patient-time at risk) between ICUs
with each pair weighted according to the inverse variance of
the estimated effect size.33 Testing and estimation were performed
on the log scale to account for different ICU sizes34; estimated
rates and 95%confidence intervalswere obtained by
exponentiating. A prespecified secondary analysis of the primary
outcome and key secondary outcomes was performed,
adjusting for ICU admission prevalence of MRSA or VRE. For
each pair, the weight was the inverse variance of the estimated
effect after adjusting for admission prevalence ofMRSA
or VRE. All weighted paired t tests had 9 degrees of
freedom. The statistical plan is in the Supplement.
การแปล กรุณารอสักครู่..