Congenital diaphragmatic hernia (CDH) is a rare congenital anomaly (2.5 per 10,000 live births) involving a defect in the diaphragm and displacement of abdominal organs into the thoracic cavity, which impedes lung development. It is associated with approximately 30–40% mortality during infancy as well as long-term neurodevelopmental impairment among survivors, and it is costly to treat.¹² Specific genetic abnormalities and environmental exposures have been proposed to contribute to the risk of CDH, but its etiology remains largely uncertain.³ Infants with CDH often have pulmonary hypoplasia and pulmonary hypertension, thereby necessitating complex management approaches immediately at birth to maximize survival and minimize short and long-term morbidity.

The care of infants born with CDH is complex, and definitive management may include advanced respiratory treatment, extracorporeal life support, and surgical repair. It has been suggested that delivery of infants with CDH at hospitals with high-level neonatal intensive care units (NICUs) that are capable of this definitive management is preferable to post-delivery transport, so that access to optimized care is immediate.² However, the extent to which infants with CDH are inborn versus transported to higher-level care NICUs and the impact of these practices on health outcomes is not well understood. Prior studies suggest that transport may be associated with higher mortality,⁴⁵ but further study is needed, especially in populations that more completely capture population-based referral patterns. The extent to which individual-level clinical or sociodemographic factors are associated with where infants with CDH are born, whether they are transferred, and whether they survive also merits further study, so that at-risk infants can be better identified.

To address these gaps in understanding, this paper describes level of care of the birth hospital, inter-hospital transfer before repair, and survival to one year of age among infants born with CDH, based on data from a cohort of over three million births in California from 2006–2011. Secondly, it examines the association of clinical and sociodemographic factors with these outcomes. To do so, we used linked data from vital records, hospital discharge records, and the California Perinatal Quality Care Collaborative (CPQCC), which collects information from NICUs.

Overall, 52% of the mothers of infants born with CDH were Hispanic, 46% had greater than a high school education, 50% had public insurance at delivery, and 98% began prenatal care in the first or second trimester, and 14% of the infants with CDH had an accompanying cardiac or non-cardiac birth defect (Table 1). In total, 25% of the infants were born at level I-II hospitals and 75% at level III-IV hospitals; 62% of the infants (n=357) were transferred for repair. Almost a third of the infants died (31%, 180/577). Half (50%) of the deaths occurred during the first week of life (early neonatal mortality, n=90), 29% from 7–28 days (late neonatal mortality, n=52), and 21% later during infancy (i.e., >28 days, n=38). Among the 180 infants who died, 33% died after repair surgery (59/180); 67% died before surgery (121/180, 39 of whom died after transfer but before surgery); and 43% (77/180) died after discharge from their delivery hospitalization.

Transfer varied by level of care of the birth hospital (Table 2). Very few infants born at level I (3%) or level II (5%) hospitals were not transferred, whereas 30% born at level III and 88% born at level IV hospitals were not transferred (Table 2). Most infants who were transferred were transferred to level IV hospitals. Overall, 80% of CDH repairs occurred at level IV hospitals and 20% at level III hospitals; only three occurred at a level II hospital and none at a level I hospital (data not shown).

Among infants born at level I-II hospitals, 96% were transferred, 24% of those who were transferred died, and 100% of those who were not transferred died (Table 3). Among infants born at level III-IV hospitals, 50% were transferred, 20% of those who were transferred died, and 45% of those who were not transferred died. Among infants who were transferred, 61% (217/357) were born at level III-IV hospitals, whereas among infants who were not transferred, 97% (214/220) were born at level III-IV hospitals.

The analysis of level of care of birth hospital yielded minimal evidence of association with the study variables, especially after adjustment for covariates (Table 4). As one example, after adjustment, maternal education less than high school was associated with a 1.3-fold increased risk of delivering at a level I-II, versus level III-IV, hospital (95% CI 0.9–1.9), but maternal education greater than high school was also associated with increased risk (adjusted RR 1.3, 95% CI 0.9–2.0). Late or no prenatal care was associated with a 1.9-fold increased risk of delivering at a level I-II hospital (95% CI 1.0–3.6). Relative risks for variables related to case complexity were all less than one, but all confidence intervals included 1.0.

Adjusted relative risks for the association of covariates with transfer (any vs. none) ranged from 0.9–1.1, with one exception; infants born at level I-II hospitals were 20% more likely to be transferred than infants born at level III-IV hospitals (RR 1.2, 95% CI 1.1–1.4) (Table 4). The unadjusted and adjusted relative risks tended to be similar, with some notable differences; i.e., the unadjusted relative risk for transfer among infants born at a level I-II hospital was 1.9 (versus 1.2 after adjustment), and unadjusted relative risks ranged from 0.6–0.8 for the clinical factors (versus 0.9–1.0 after adjustment). After excluding infants who died on their first day of life, RRs were similar (e.g., the adjusted RR for transfer was 1.2, 95% CI 0.9–1.5).

Adjusted hazard ratios for mortality indicated that infants who were low birthweight, had other accompanying cardiac or non-cardiac birth defects, had a low 5-minute Apgar score, and whose mothers started prenatal care in the third trimester or not at all had approximately two- to four-fold higher risk of mortality than their respective comparison groups; 95% CIs for these estimates excluded 1.0 (Table 5). Among infants with other birth defects who survived, 42% (10/24) had cardiac defects, of whom 60% (6/10) also had non-cardiac defects; among infants with other birth defects who died, 63% (35/56) had cardiac defects, of whom 54% (19/35) also had non-cardiac defects. Adjusted hazard ratios for the sociodemographic variables were closer to 1.0 and their CIs did include 1.0. Transfer from birth hospital to another hospital for repair was associated with 70% lower risk of mortality (adjusted HR 0.3, 95% CI 0.2–0.5). After excluding infants who died on the first day of life from the model, the adjusted HR for transfer was 0.5 (95% CI 0.3–0.7); the pattern of results for other variables did not change substantially (data not shown).

Among infants born with CDH in California, 25% were born at hospitals with level I-II neonatal care capacity, 62% were transferred to another hospital for repair, and 31% did not survive infancy. Most sociodemographic and clinical variables did not substantially influence these outcomes, with the exception that infants who had clinical risk factors (e.g., low birthweight, other major birth defects) were at higher risk of dying, and infants whose mothers had late or no prenatal care initiation were at higher risk of being born at level I-II hospitals and of dying.

We are unaware of studies reporting level of care of birth hospital within other contemporary, large cohorts of infants with CDH. It is typically thought that delivery at a Level I-II hospital should be avoided for infants with CDH, given that infants with CDH are likely to need services offered only at higher-level hospitals. However, 25% of infants in this cohort were born at Level I-II hospitals. Late or no prenatal care was associated with 1.9-fold increased risk of this outcome, suggesting that lack of prenatal diagnosis, or complications leading to rapid delivery at an unanticipated location may have contributed.

We also examined hospital transfer; 62% of infants were transferred. Transfer is an important part of care, but ideally, it should be pre-empted when possible, especially for infants with potentially life-threatening conditions like CDH. Transfer can incur unnecessary medical risk, results in at least short-term separation of mothers from their newborns, and is expensive.⁹¹⁰ Some prior data on transfers among infants with CDH are available but tend to be based on highly selected study populations (e.g., from single or few institutions) and focus on mortality among transferred infants rather than predictors of transfer itself. A study by Aly et al. of over 2,000 cases cared for in selected U.S. hospitals from 1997 to 2004 reported that 48% of infants were transported before surgery, and that infants who were transported were 1.5 times more likely to die after surgery than non-transported infants.⁴ A smaller study of 140 prenatally diagnosed cases born from 2005–2008 and treated within a network of Canadian pediatric surgical centers reported that infants who were transported to the centers for care were more likely to die.⁵ In contrast, our study found that infants who were transferred were more likely to survive. Decision-making around where to deliver and whether to transfer an infant is complex, and many factors could contribute to the differences in findings. The prior studies noted above were conducted among more restricted subgroups of infants, rather than being population-based, which could affect their results. Given that most (98%) of women in our study received prenatal care by the second trimester, we expect that at least the most severe CDH cases could have been prenatally diagnosed and advised to be delivered at level III-IV hospitals. On the other hand, less severe cases may not have been detected on ultrasound; although the diaphragm is formed by the seventh week of gestation, visceral herniation may not occur until later, up to the time of delivery, if the extent of abnormal musculature is limited and the hernia is small.¹¹ It is thus possible that the cases who were transferred in our study (at least those transferred from I-II to III-IV level hospitals) had less severe hernias and this contributed to their better survival. In addition, clinical risk factors were less common among infants who were transferred, which also suggests that they were inherently a lower-risk group. At the other extreme, transfer may not occur for some very severe or complex cases with multiple anomalies, because the family had prospectively made the decision for comfort care. It is also possible that some births may occur intentionally at a level II or level III center and then transfer to a level IV center that does not have a delivery service. We were unable to distinguish this type of transfer from a transfer in which the place of birth was not intentionally planned based on the diagnosis of CDH. Unfortunately, we do not have information on prenatal diagnosis or severity of the hernia to enable us to directly assess the contribution of these factors to our findings. In addition, CPQCC reporting would miss cases that were born at non-CPQCC hospitals who died before transfer to a CPQCC hospital. However, because CPQCC includes data from >90% of NICUs in California, and >90% of non-CPQCC hospitals are Level I, we do not expect this source of under-reporting to be a major contributor to our findings.

Many of the studied factors did not predict which CDH infants were likely to deliver at level I-II (versus III-IV) hospitals, be transferred, or die. Notably, race-ethnicity and variables related to socioeconomic status were not associated with the outcomes. Potential predictors that were not available in this study but that would be useful to examine in the future include, for example, prenatal diagnosis, planned delivery hospital, distance from a mother’s home to the birth hospital versus repair hospital, and more detail regarding case severity, especially information that could be detected prenatally (e.g., size and sidedness of the hernia, observed-to-predicted estimated fetal lung volume, position of the liver). Also, detailed information about the specialty services and expertise available at some level III-IV sites was not available in this dataset. These factors may be particularly important to understanding why so many infants were delivered at a level III-IV hospital but transferred for repair, and why non-transferred infants had higher mortality. At this point, further study is needed in order to understand these complexities and what changes in practice could improve outcomes among these vulnerable infants.

Mortality estimates among liveborn infants with CDH have varied widely across studies. Some contributing factors are years of study (approaches to care may have changed over time), extent of follow-up (e.g., some studies follow infants only through hospital discharge, whereas others follow them through infancy), exclusions (e.g., some studies exclude cases who die before repair, or complex cases such as those with underlying genetic abnormalities), and study design (varying from single-center studies, to those that are population-based). More recent, population-based studies tend to estimate infant mortality to be approximately 30%, which is similar to our study.^12–14 Improvements in survival in recent decades may be attributable to improvements in care. Increased termination of pregnancy, especially of more severe cases, may also contribute.¹⁴ In our study, clinical factors were associated with increased risk of dying, which agrees with prior studies.²¹²¹⁵ We examined a more extensive set of sociodemographic variables than most prior studies (maternal race-ethnicity, education, payer and census tract poverty), but these factors were not predictive of mortality. The study by Aly et al. reported that babies born to black mothers were 2.4 times more likely to die, after adjustment for factors such as newborn co-morbidities, age at repair, and transport.⁴ Stevens et al. also reported worse survival among black infants after adjustment for other factors, among infants born from 1995–2004.¹⁶ Both of these studies were limited to infants born at or transferred to participating hospitals, which may have influenced the findings.¹⁷ Sola et al., in a study of the Kids Inpatient Database from a sample of U.S. hospitals, considered race/ethnicity, type of insurance, and zip code median income; they reported black race-ethnicity as the only significant predictor after adjustment for the other factors.¹⁸ However, the study was based on ICD-9 codes for diagnosis, only 40% of cases (1,095/2,774) had evidence of repair, and 30% were missing data on race-ethnicity. A recent state-wide study of 1,025 infants with CDH born in Florida from 1998–2012, however, did not find that maternal race-ethnicity or education was associated with infant survival.¹²

Important strengths of the current study include its state-wide data capture, large and diverse study population, and use of linked data from vital records, administrative data, and CPQCC. Many studies of care and mortality among infants with CDH are limited to data from either the delivery hospitalization or the repair hospitalization, and they are unable to determine mortality after discharge.¹⁴ In our study, over half of the infants were transferred for repair, 43% of deaths occurred after discharge from the delivery hospitalization, and almost one-fourth of all deaths occurred after 28 days, thus illustrating the usefulness of the linked data resources. We relied on diagnosis data from CPQCC, which has advantages. For example, some prior studies that use administrative data have relied on ICD-9 codes to identify cases,¹²¹⁸ but the ICD-9 code for CDH (756.6) does not differentiate eventration of the diaphragm from CDH. The percent of infants with other major birth defects (14%) was lower than other prior studies;³¹³¹⁴ we primarily relied on CPQCC codes for specific birth defects, and this system likely errs on the side of including more severe birth defects than, for example, reliance on ICD-9 codes. Although our study had access to a wide range of variables, it was also lacking some factors which will be important for future studies, e.g., prenatal diagnosis and severity of the hernia (as noted above), planned delivery hospital, specialty services available at each hospital such as availability of extracorporeal membrane oxygenation (ECMO), and immediate complications in the newborn. We were unable to identify cases with known syndromes or chromosomal abnormalities with certainty; however, we expect them to comprise a relatively small proportion of all cases (5–15%).³¹³¹⁴ The diversity and size of the study population are advantages. As this was a population-based study conducted in the state of California, we recognize that our findings may not be generalizable to other states that have varying geographical and healthcare system patterns. However, California uses level of care classification criteria specified by the American Academy of Pediatrics, which is a nationally recognized organization, and regional systems and networks across the U.S. may consider our findings applicable to their practice. Additionally, our results are applicable to a substantial proportion of U.S births, since one in eight U.S. births occurs in California.

In conclusion, this California state-wide study of infants with CDH found that a substantial proportion were born at level I-II hospitals or underwent hospital transfer before repair. Sociodemographic factors (e.g., race-ethnicity and markers of socioeconomic status) were not predictive of delivery hospital level of care, hospital transfer, or mortality. Factors related to case complexity (low birthweight, other accompanying birth defects, and low Apgar score) were not predictive of level of care or transfer, but they were predictive of mortality. These data support the importance of prenatal care, diagnosis, and delivery planning into level III-IV centers wherein the care and outcomes for newborns with CDH can be optimized.