TY  - JOUR
T1  - RApid response team implementation in a children’s hospital—reply
AU  - Joffe AR, Anton NR, Burkholder SC
Y1  - 2011/12/01
N1  - 10.1001/archpedi.165.12.1139-b
JO  - Archives of Pediatrics & Adolescent Medicine
SP  - 1139
EP  - 1140
VL  - 165
IS  - 12
N2  - 
First, the statistical technique of ARIMA time-series analysis is superior to simple before-and-after comparisons, as it can model the effects of random error over time. We do not believe that ARIMA can “eliminate the biases of a simple before-and-after design,” completely account for unmeasured confounding factors that are temporally coincident (or even temporally related by delayed effects) to the intervention, nor solve the problem of systematic bias noted in studies involving historical controls.1- 3 With only 19 postimplementation event rates in their ARIMA model,4 there is risk of type I error.1 A recent publication graphs 8 additional postimplementation event rates: hospital mortality increased again, and the ARIMA model parameter estimate for PMET intervention was no longer visually or statistically significant (Figure 1 and Table 3).5 Moreover, the hospital had several quality interventions (infection and adverse drug event reduction strategies and creation of a hospitalist program) implemented over the period of the PMET study that may be confounders.5 Of note, ARIMA was not used for analysis of ward code rates.4 More than 40% of deaths were in “newborns and other neonates with conditions originating in the perinatal period”; improvements may have been driven by perinatal care and less by care in the general pediatric population.4

SN  - 1072-4710
M3  - doi: 10.1001/archpedi.165.12.1139-b
UR  - http://dx.doi.org/10.1001/archpedi.165.12.1139-b
ER  -