We conducted a retrospective cohort study to evaluate if increased travel time between the birth location and a CSC was associated with increased neonatal mortality among infants with HLHS. This study used data from the TBDR, which has conducted active, statewide surveillance since 1999. Registry staff review records of all delivery units and pediatric hospitals throughout Texas. They abstract detailed demographic and diagnostic information for infants with a registry-monitored birth defect diagnosed within 1 year of delivery, to Texas-resident mothers. TBDR data are routinely linked to Texas vital records.

The TBDR classifies congenital anomalies using 6-digit birth defect codes that are based on the British Pediatric Association (BPA) Classification of Diseases (1979) and the International Classification of Diseases, ninth revision, clinical modification (1979). We included live births during January 1, 1999 -December 31, 2007 with the codes for HLHS (746.700), aortic atresia (746.480), or hypoplastic left ventricle (746.881). To ensure accurate diagnosis, a pediatric cardiologist (S.A.M.) reviewed available TBDR clinical data. A diagnosis of HLHS required at least one of the following:

We excluded infants with alternative primary cardiac diagnoses, including atrioventricular septal defect and double outlet or double inlet ventricle. We also excluded twins and infants with gestational age <23 weeks, birthweight <400 grams, or genetic disorders. Patients with either major extracardiac birth defects or more than 3 minor extracardiac birth defects were also excluded. Extracardiac birth defects are carefully catalogued by the TBDR. A major extracardiac birth defects was defined by the National Birth Defects Prevention Network as a congenital abnormality that requires medical or surgical treatment, has a serious adverse effect on health and development, or has significant cosmetic impact.¹⁴

The study was approved by the Internal Review Boards of Baylor College of Medicine, the Texas Department of State Health Services, and the University of South Florida.

Maternal socioeconomic status (SES) was estimated from two widely used, validated proxies of SES: education level and percent of population living in poverty in the maternal census tract, according to the 2000 U.S. census.^16,17 When a post office box or a non-mappable rural route was the residential address listed, the street or city was used. For categorical analysis of poverty, a cutoff of ≥20% was used in accordance with the U.S. Census Bureau.¹⁸.

Preterm birth (PTB) was defined as <37 weeks, low birthweight (LBW) as <2500 grams, and small for gestational age as <10% of expected weight using published criteria¹⁹. Birth era was divided into 1999-2002 and 2003-2007. Presence of total anomalous pulmonary venous return and highly restrictive atrial septal defect/intact atrial septum (RAS) also were noted. As RAS information was not available in approximately 25% of infants, it was not included in primary analyses. For secondary analyses, RAS was coded as “present,” “absent,” or “unknown”. The “present” RAS group included those with an atrial septum described as “severely restrictive,” “intact,” or who underwent catheter-based atrial septal intervention. Those with “absent” RAS were those with an atrial septum described as “unrestrictive,” other similar language, or who had no atrial septal descriptors in the echocardiography report but were described throughout the record as clinically stable with no catheter-based atrial septal intervention and Stage 1 palliation (S1P) >3 days of life. All others were included in the “unknown” group, including those with records using only the term “restrictive” atrial septum on the echocardiography report, given wide variation in the use of “restrictive.” CSC HLHS case volume was calculated from the total number of neonates with HLHS presenting to a given CSC, as actual surgical volume was not reliably available in the registry.

Neonatal mortality (death before age 28 days) was chosen as the primary outcome to minimize the competing effects of SES and access to healthcare on post-discharge mortality. Infants were classified as deceased based on the TBDR and Texas vital records. Neonatal death was categorized according to whether it occurred before transfer to a CSC; after arrival at a CSC but before S1P; or after S1P. CSC HLHS volume was used as a variable in the latter two analyses, so only infants with an identifiable CSC were included.

To also investigate the relationship between time to CSC and mortality, three intermediate outcomes were studied: (1) age at diagnosis, using the first abstracted postnatal confirmation of HLHS (typically an echocardiogram); (2) pre-surgical decompensation, determined from review of abstracted data for key phrases (Appendix); and (3) age at S1P.

Univariable comparisons were performed using the Wilcoxon rank sum or Kruskal-Wallis analysis for continuous variables, and Chi-square, Fisher exact, or Chi-square trend analyses for categorical data. For associations with categorical variables with more than two classes, multiple comparisons were adjusted using Bonferroni’s correction. Correlations were calculated using Spearman’s rho.

A multivariable logistic regression model was computed for neonatal mortality outcomes. In the primary analysis, all variables were considered for inclusion if p<0.10 in association with mortality.

For subanalyses evaluating timing of death, the variables noted to be significant in association with overall neonatal mortality were included in the models. For the subanalyses of death prior to surgery (after arrival at a CSC) and death after surgery, which included CSC HLHS volumes in the models, generalized estimating equation models were computed to accommodate correlation between patients at the same center. P values <0.05 were considered statistically significant.

Time from birthplace to CSC ranged from 0-577 minutes: 210 newborns (45.4%), <10 minutes; 147 (31.7%), 10-90 minutes; and 106 (22.9%), >90 minutes. Mothers delivering farther from a CSC were less educated and more often lived in a poverty area (Table 1). Thirty percent of mothers delivering >90 minutes from a CSC were born in Mexico, compared to 17% of those delivering <90 minutes from a CSC. Very few black mothers delivered >90 minutes from a CSC.

HLHS was prenatally diagnosed in 39% of newborns, with a higher proportion diagnosed prenatally in the later birth era (49.0% versus 25.5%, p<0.001). Prenatally diagnosed infants were delivered significantly closer to a CSC, with median time to CSC 2.6 minutes (interquartile range, IQR 1.2,9.9 minutes) versus 28.8 minutes (IQR 10.0,136.3 minutes, p<0.001).

Death before age 28 days occurred in 123 infants (26.6%). Mortality was higher in the earlier birth era (Table 2). In crude analysis, increased time to a CSC was associated with increased mortality: <10 minutes, 21.0%; 10-90 minutes, 25.2%; and >90 minutes, 39.6% (p for trend <0.001, Table 2). This pattern was evident in both birth eras and in both prenatally and postnatally diagnosed infants (Figure 1). In infants diagnosed prenatally, mortality was 22.8% compared to postnatally, 29.0% (p=0.14) (Table 2).

In multivariable analysis, factors significantly associated with neonatal mortality were time to CSC, LBW, and earlier birth year (Table 3). PTB was not included in the multivariable model because of its high correlation with LBW.

In sub-analyses that adjusted for RAS, the results for time to CSC were similar: 10-90 minutes odds ratio (OR) 1.23 (95% confidence interval: CI 0.73-2.08); >90 minutes, OR 1.97 (95% CI 1.14-3.38). The OR for RAS “present” versus “absent” was 4.19 (95% CI 1.92-9.12).

To account for the possible effect of parental intention not to treat, we excluded six infants who were diagnosed prenatally, born ≥10 minutes from a CSC, and never transferred. The results were similar: 10-90 minutes, OR 1.16 (95%CI 0.69-1.95); >90 minutes, OR 1.85 (95% CI 1.07-3.20).

Thirty patients (6.5%) died before arriving at a CSC. Pre-transport mortality in infants delivered >90 minutes from a CSC was 18.9% compared to those born 10-90 minutes (2.7%, p<0.001) and <10 minutes (2.9%, p<0.001, Figure 2). In those prenatally diagnosed, delivery >90 minutes from a CSC was associated with significant pre-transport mortality (28.6%), compared to 7.1% in those born 10-90 minutes (p=0.06), and 2.2% in those born <10 minutes from a CSC (p<0.001).

In multivariable analysis, time to CSC, LBW, and birth year remained associated with pre-transport neonatal death (Table 3). When RAS was included, the results remained similar: 10-90 minutes, OR 0.94 (95% CI 0.26-3.47); >90 minutes, OR 6.26 (95% CI 2.35-16.70).

Of 433 newborns arriving at a CSC, 94 (21.7%) died, 15 prior to undergoing S1P, and 52 after S1P; 27 had insufficiently detailed hospital procedure information abstracted. For pre-surgical mortality analysis, 384 infants had sufficient information for classification. In those born <10 minutes from a CSC, 3/190 (1.6%) patients died before surgery, compared to 8/130 (6.2%) in those born 10-90 minutes, and 4/64 (6.3%) in those born >90 minutes (Figure 2). CSC HLHS volume was inversely correlated with CSC pre-Stage 1 mortality (Spearman rho=−0.498, p<0.001, Figure 3A). In the final multivariable model, time to CSC 10-90 minutes was significantly associated with increased pre-Stage I mortality, while there was a trend towards increased mortality in patients delivered >90 minutes from a CSC and those cared for in lower volume CSCs (Table 3). When RAS was included, the results were similar: 10-90 minutes, OR 3.43 (95% CI 0.96-12.33); >90 minutes, OR 4.44 (95% CI 1.12-17.57).

Three infants never underwent S1P but were alive at 28 days and ultimately died, one underwent S1P in another state, two were referred out of state for a primary cardiac transplant, and one had no detailed hospital information abstracted and died at 40 days of life.

Evidence of S1P at an identifiable CSC existed for 369 infants. CSC HLHS volume was inversely correlated with S1P mortality (Spearman rho=−0.707, p<0.001, Figure 3B). Only low HLHS volume and earlier birth year remained associated with higher S1P mortality in multivariable regression (Table 3), even when RAS was included.

We assessed intermediate outcomes including age at diagnosis (in most cases, time to first echocardiogram), presurgical decompensation, and age at surgery. Age at diagnosis was available for 96% of infants. Although infants born further from a CSC were older at diagnosis (Table 1), older age at diagnosis was not associated with increased mortality (Table 2), even after restricting the analysis to postnatally diagnosed infants (p=0.50) or to no RAS (p=0.74).

Thirty-five infants had evidence of pre-surgical decompensation (Appendix). Infants born further away from a CSC had greater frequency of decompensation (Table 1), which was associated with increased mortality (Table 2) even after restricting to no RAS (p=0.03). Of note, almost two-thirds (n=22) of infants with decompensation were diagnosed at ≤2 days of life. Of these 22 patients, 2 of 8 (25%) born <10 minutes from a CSC died, while 5 of 5 (100%) patients born 10-90 minutes from a CSC died, and 8 of 9 (89%) patients born >90 minutes from a CSC died.

Age at surgery was available for 302 of 384 infants undergoing S1P. Infants born further from a CSC were older at surgery (Table 1), but this factor was not associated with mortality (Table 2), even after restricting to no RAS (p=0.84).

This study was limited by the use of abstracted registry data. However, this registry source is exceptional for its active surveillance for HLHS in all Texas delivery hospitals and pediatric hospitals and extensive data collection. A study of similar content and size cannot be prospectively accomplished.

The use of registry data did not allow us to distinguish families who chose nonintervention, which would result in death. If prenatally-diagnosed mothers chose more often to deliver >10 minutes from a CSC, those with longer times to CSCs would have increased mortality, as our data demonstrate. However, only 42 mothers with a prenatal diagnosis delivered >10 minutes from a CSC. Thirty-six of these infants were transferred to a CSC, suggesting that the families desired care. When excluding the other six children, the results were unchanged, confirming that this factor was not responsible for our results. Although families of infants diagnosed postnatally and born far from a CSC may have chosen non-intervention at a higher rate, that factor would not negate our findings, as differential counseling could serve as an additional mechanism of increased mortality with increased distance from a CSC.

Data on termination of pregnancy (TOP) was not included in this study. The TBDR does collect information on TOP in the facilities in which data are collected for births. Although this does not likely include all TOPs, only 0.71% of all 1999-2007 cases of HLHS in the TBDR, as defined by birth defect code 746.700, were induced terminations.²⁵

Whereas the primary study variables, time to CSC and neonatal mortality, were universally available, some variables of interest were not, the most important of which was the presence of RAS. Hence, RAS was incorporated into the analyses secondarily and did not significantly alter the effect estimates of time to CSC. Medication use was not available in the birth defects registry, so time to prostaglandin (PGE) initiation was not evaluated in this study.

The findings of this study may not be applicable in other states or countries with less variation in geography or access to healthcare. However, most states in the United States have even a greater proportion of the population considered rural than Texas. Per the 2010 US Census, 35 states ranked above Texas in proportion of households living rurally.²⁶. Therefore, we believe a significant proportion of infants with HLHS would benefit from prenatal diagnosis and relocating close to a high volume CSC.