Management of Hypoplastic Left Heart Syndrome

To the Editor.—We read with interest the article by Chang et al1 on the management of hypoplastic left heart syndrome (HLHS) in newborn infants. Chang et al used data from the Nationwide Inpatient Sample (NIS) covering the years 1988–1997 to examine trends in the use of the Norwood procedure, heart transplantation, and mortality over time. They also examined patient and hospital characteristics to assess whether they had any influence on management decisions. Chang et al claimed the in-hospital mortality rate fell from 54.4% in 1988 to 38.1% in 1997 and use of the Norwood procedure increased from 8% to 34% over the same time period. They also found that hospitals in the South were more likely to use the Norwood procedure relative to other regions of the country—a finding that “cannot be easily explained.” In this letter, we address these findings.Central to the Chang et al analysis is the development of the sample from the NIS. Chang et al claim to have identified patients in the NIS with multiple hospitalizations. This is not possible because the NIS does not contain unique individual identifiers (personal communication with intramural staff at the Agency for Healthcare Research and Quality). This error led Chang et al to include patients in their sample who were transferred to an acute care hospital “not included in the NIS data.” The result of this sample selection procedure was to artificially increase the sample size by the inclusion of over 500 patients who were coded as “transfers.” Their inclusion lowered both the estimated mortality rate and the rate of patients receiving a Norwood.A second concern with the Chang et al analysis involves the identification of patients undergoing the Norwood procedure. Because there is no specific procedure code for the Norwood procedure, Chang et al identified patients based on an algorithm where cardiopulmonary bypass was present (International Classification of Diseases, Ninth Revision, Clinical Modification code 39.61) and 1 or more of the following codes: surgical creation of a septal defect (which was incorrectly listed as 34.42 rather than 35.42), repair of heart or pericardium (37.4), incision excision, or occlusion of aorta (38.14), and systemic to pulmonary shunt (39.0). Reliance on such an algorithm may be too strict for administrative databases. It may be more accurate to rely on the inclusion of any evidence that a procedure occurred. A patient may have a code for cardiopulmonary bypass but not have any of the other 4 codes in the Chang et al algorithm. Use of an algorithm based on “or” criteria without evidence of transplantation appears more appropriate.Chang et al also incorrectly listed the actual procedure codes they used in identifying Norwood patients. Our replication of the Chang et al analysis found 284 Norwood patients using the codes provided compared with 346 reported by Chang et al. The code 37.4 was found in 50% of patients received a Norwood in 1997 (33% over all of the study years) and the mortality rate approached 50%. We also found no significant difference in Norwood procedure rates by region of the country.The different sample selection methods and algorithms to identify Norwood patients lead to different results. Our findings are consistent with findings from a consortium of university hospitals.2 Chang et al assert an overrepresentation of aggressive surgical strategies in the consortium study. We believe the difference in findings results from their sample selection methods and algorithm to identify Norwood patients.In Reply.—Tilford and colleagues provide thoughtful comments regarding our study on the clinical management of hypoplastic left heart syndrome (HLHS).1 We applaud their efforts to offer an alternative approach to the analysis of the Nationwide Inpatient Sample (NIS) data using different assumptions.Tilford et al correctly pointed out that the NIS database identifies each hospitalization, not each patient. Using the unique sequence numbers in the NIS database helped us to eliminate duplicate records, but not patients with multiple hospitalizations. Additionally, a typographical error was made in the International Classification of Diseases, Ninth Revision code for surgical creation of septal defect, which should be 35.42. We apologize for this misrepresentation of the International Classification of Diseases, Ninth Revision code reported in our paper; however, the correct code was used in our analysis.We do not agree with Tilford et al that including patients who were transferred to another hospital in our analysis artificially increased the sample size and decreased the measured prevalence of the Norwood procedure. Patients who were transferred from other hospitals to a NIS hospital would also affect the sample size and Norwood procedure prevalence. Following the argument of Tilford et al, an analysis that excludes all patients who were transferred out should also exclude patients who were transferred in. In the “admission source” of the 1986 patients in our study, 740 patients (37.3%) were transferred from other hospitals. This number is comparable to the number of patients who were transferred out (N = 520). The patients whose admission source was “transferred from other hospitals” had a Norwood procedure rate of 26.6%, which was significantly higher than patients who were not transferred from other hospitals (12.0%). As we have stated in the “Discussion” section of our article, the effects of patients who were transferred out are likely to be offset by patients who were transferred in to NIS hospitals. However, we agree that it may be difficult to determine the magnitude of the “offset effect” of patients transferred in and patients transferred out.The determination of a Norwood procedure in an administrative database can be controversial. Whether one should use the strict algorithm that was used in our study or a more inclusive algorithm as suggested by Tilford et al remains a topic for further debate. We are concerned that using cardiopulmonary bypass as the only selection criterion for Norwood procedure may 1) increase the chance of selecting miscoded cases, and 2) select some milder forms of HLHS that undergo procedures such as aortic valvotomy. Moreover, when we assigned patients in the Norwood procedure group, we examined all of the first 4 procedure code fields, which may explain the higher number of cases which we found, as compared with the analysis performed by Tilford et al.It is not surprising that by excluding the patients in the NIS dataset who were transferred out Tilford et al would obtain results similar to the study by Gutgesell and Massaro,2 which used data from a consortium of university hospitals. Our data indicate that patients who were transferred between hospitals were more likely to undergo aggressive management. Given the nature of tertiary referrals to university hospitals, it is likely that these hospitals receive more transfers in and a lower proportion of transfers out, thus raising the overall prevalence of the Norwood procedure in these hospitals.As we acknowledge in our article,1 a major limitation of our study is the use of data from an administrative database, rather than a clinical database. Because of many known deficiencies in administrative data, various assumptions such as criteria for identification of cases and the determination of Norwood procedure were required to conduct our analysis. We recognize that our study can be repeated using different assumptions, and that this process may lead to somewhat different findings.