Author Response

We thank Dr Shitara and colleagues for their thoughtful comments on our article, “Congenital Heart Surgery on In-Hospital Mortality in Trisomy 13 and 18.” Their questions highlight important considerations we made during the design of the study as well as intrinsic limitations to the use of large administrative data sets such as the Pediatric Health Information Systems (PHIS).

Although the primary schema for inclusion and exclusion in our study was presented in Fig 1 of the original article, we appreciate the opportunity to provide further clarity here. As for inclusion, the PHIS database was queried for all discharges during the study period for which patients had an ICD-9 code corresponding to either trisomy 13 or 18. This query resulted in a list of encrypted medical record numbers of candidates for study inclusion. We further narrowed inclusion to those patients who had a first admission within the first 14 days of life. The goal for this criterion was to capture as many patients as possible because the mean length of survival for both trisomy 13 and 18 is <14 days. It also served to help create a cohort that had some baseline similarity, that being they all required hospitalizations in a pediatric hospital before 2 weeks of age. For similar reasons, we included patients in the congenital heart surgery group who underwent a congenital heart surgery during their first admission. With regard to study exclusion, in some cases, there were patients whose records included a diagnostic code for trisomy 21, and there were cases in which codes for both trisomy 13 and trisomy 18 were present. All of these cases were excluded because it was impossible to determine which diagnosis was correct. Any patient discharged during the first 14 days of life who was readmitted and subsequently underwent congenital heart surgery was included in the study but was grouped with the no surgery cases. There were 19 patients (1.1% of total) who were discharged and readmitted within the first 14 days of life, few of whom underwent congenital heart surgery.

Regarding the point on patients who might have been discharged and readmitted later for a congenital heart surgery, such patients would have been included in the no surgery group. The reason we decided to adhere strictly to group assignment based on surgery within the first admission was actually centered on decreasing bias in the surgical group. If a patient was discharged without surgery and returned for surgery on a later date, that given patient likely had a better intrinsic physical constitution and a different management plan than the majority of patients in the study. In truth, to include such patients would introduce a significant degree of the immortal bias Dr Shitara and colleagues have cautioned against. Additionally, the inclusion of patients who had surgery during a later admission and who possibly had better outcomes as a result would falsely elevate the survival rate in the no-surgery group. Therefore, the methodology we chose to employ would, if anything, lessen the differences in survival between the study groups rather than increasing them.

The question of immortal bias in the study is one of interest and importance. Above, we have touched on our attempt to decrease immortal bias through the chosen study methodology. Regarding Dr Shitara and colleagues’ specific concerns over those in whom surgery was planned but who died before surgery and what seems to be an implication that such cases should have been included in the surgery group, there are 2 major considerations. The first is that the study data are derived from a large administrative database without any data on clinical decision-making. As a result, it is impossible to determine who may have been considered for surgery but did not survive to have surgery. Second, in the study, we sought to determine the impact that surgery had on survival as opposed to the impact that the decision for surgery had on survival. We believe there is an important distinction between the two. From our perspective, to determine if surgery impacts a patient’s outcome, the patient must undergo the actual surgical procedure in question. Additionally, regarding the issue of cause of death, the PHIS database does not have any data that could be reliably construed as indicating the cause of death. In addition, there are no outpatient data in the PHIS to allow for determination of the cause of death for those who died outside of the hospital. For these reasons, the analysis recommended by Dr Shitara and colleagues is not possible. Regarding the number of survivors in each survival period, it should be noted that those numbers were already included in Fig 3 of the original article.

The patients in our study were younger at the time of surgery than previous studies.¹ We disagree with the suggestion that this represents an effect of prenatal diagnosis and familial socioeconomic status. The age of our patient cohort is a reflection of our study methodology; we were looking at surgeries that occurred during admissions that began before 14 days of age. If we had examined all congenital heart surgeries in the 3273 patients during any of their 9761 hospitalizations, the mean age would have been considerably higher. Additionally, socioeconomic data are limited in the PHIS database, and there is no way to determine cases wherein a prenatal diagnosis was made.

Given the complexity of clinical decision-making in the care of patients with trisomy 13 and 18 and given the limitations of available data sets for robust study of these complex questions highlighted by the present comments of Dr Shitara and colleagues, it seems that a large national and/or international registry to track and study these patients would be beneficial. Certainly, such an endeavor would allow for higher resolution understanding of the complex variables that impact outcomes in these patients. At present, the use of large administrative databases such as PHIS allow us to develop a better understanding, although we must remain cognizant of their inherent limitations.

• Copyright © 2018 by the American Academy of Pediatrics