Surgery and neurodevelopmental impairment-reply
In Reply We appreciate the careful reading of our article by Weisz et al recently published in JAMA Pediatrics, who point out that our propensity score calculations omit consideration of the timing of surgery, an important variable in adjusting for selection bias.1,2 While we have dates of occurrence for some of the morbidities that we adjusted for, the Neonatal Research Network Generic Database does not contain information on the timing of surgery for the birth years 1998 to 2009. Consequently, there is essentially no precise way to know whether surgery predated some of the morbidities and we cannot do the type of time-aware propensity analysis that Weisz et al suggest. However, if anything, we may have adjusted for more covariates than we should have (ie, for morbidities that followed surgery); therefore, our results are more likely to be biased toward the null because of overadjustment. Nonetheless, we still found a sizeable and statistically significant association between surgery and outcome. This attests to the robustness of the effect and makes it more likely that we observed a true association, although we may have underestimated its magnitude because of overadjustment.
Das, A., Saha, S., & Morriss, F. H. (2014). Surgery and neurodevelopmental impairment-reply. JAMA Pediatrics, 168(12), 1169. DOI: 10.1001/jamapediatrics.2014.1787