Prenatal diagnosis mediates the relationship between disease complexity and intervention timing for congenital heart defects

Abstract

Introduction

Previous studies comparing timing of postnatal intervention among patients with prenatal versus postnatal diagnosis of a congenital heart defect (CHD) have yielded conflicting results due to challenges in controlling for CHD complexity. The objective of this analysis is to take an alternative approach: to determine whether prenatal diagnosis mediates, or explains, the relationship between CHD complexity and age at intervention.

Methods

A retrospective cohort study including infants who received intervention for CHD at < 60 days of life, 2015–2021, at a high‐volume CHD surgical center. The exposure was CHD complexity (low, moderate, and high). The mediator was prenatal diagnosis (yes, no). The outcome was age at intervention (cardiac surgery or interventional catheterization, whichever came first) in days. Covariates included sociodemographic and clinical characteristics of the pregnant individual and infant. Mediation analysis used linear probability regressions with a COVID‐19‐era fixed effect and robust standard errors.

Results

Of 429 infants, 28% had high‐, 63% had moderate‐, and 8% had low‐complexity CHD; 68% had a prenatal diagnosis. The median age at intervention was 6 (IQR 4–14) days. In adjusted models, prenatal diagnosis mediated 27%–29% of the association between CHD complexity and age at intervention (indirect effects: moderate complexity vs. low complexity: 3.7, 95%CI 1.4–6.1 days; high complexity vs. low complexity: 5.5, 95%CI 2.2–8.8 days).

Conclusion

Complex CHDs are intervened upon earlier, partly because they are likelier to be prenatally detected, than low‐complexity defects. Prenatal diagnosis plays a mechanistic role in improving access to care for infants with complex CHD.

1. INTRODUCTION

Congenital heart defects (CHDs) are the most common and one of the most fatal birth defects [1]. Prenatal CHD diagnosis allows families to plan their pregnancy [2] and aids clinicians in optimizing postnatal management [3, 4]. However, quantitative studies evaluating the benefits of prenatal diagnosis are unclear [5, 6, 7]. Some have shown that for hypoplastic left heart syndrome and D‐transposition of the great arteries, prenatal diagnosis is associated with higher probability of delivery at a congenital heart center, and thus improved preoperative outcomes [8]. However, others have shown no difference in, or even worse access to care, for infants with prenatal versus postnatal diagnoses of severe CHDs [9, 10].

Many researchers have since noted that these conflicting findings are because more complex CHDs are typically more easily diagnosed with prenatal ultrasound and have worse outcomes than less complex CHDs, i.e., CHD complexity confounds the relationship between prenatal diagnosis and outcomes. Therefore, if disease complexity is insufficiently controlled for, prenatal diagnosis will be falsely associated with worse outcomes [5, 6, 11, 12, 13]. Theoretically, randomization could eliminate the issue of confounding from disease complexity [14], but it is not possible to randomize receipt of prenatal diagnosis.

In this analysis, we take an alternative approach from previous studies that compared outcomes by prenatal diagnosis status. Instead, our objective was to quantify the impact of prenatal diagnosis on access to care, by showing how prenatal diagnosis mediates—or explains the relationship between [15, 16]—disease complexity and timing to postnatal intervention. We hypothesized that those with higher‐complexity CHDs receive intervention sooner than those with the lowest‐complexity CHDs, in part, due to higher prenatal diagnosis rates for more complex defects.

2. METHODS

2.1. Data source

Data were obtained from the locally maintained Society of Thoracic Surgeons Congenital Heart Surgery Database (STS‐CHSD) at Ann & Robert H. Lurie Children's Hospital of Chicago (Lurie Children's). STS‐CHSD is the largest congenital heart surgery registry in the world. It captures highly granular information on patient demographics, hospital, surgeon, 110 anatomically‐specific cardiac diagnoses, 187 procedures and procedural subtypes, 22 comorbid conditions and preoperative risk factors, elective/urgent/emergent case status, bypass, cross‐clamp, and circulatory arrest times [17]. At Lurie Children's, specialists dedicated to the STS‐CHSD collect data prospectively.

2.2. Study design and population

This is a retrospective cohort study of all patients who received their initial intervention at Lurie Children's at < 60 days of life, between January 1, 2015, and December 31, 2021. We used a cutoff time of 60 days [18, 19, 20, 21] to eliminate confounding from those defects with more elective intervention timing, such as balanced atrioventricular septal defects. We chose these seven study years in order to: (1) mitigate secular trends in CHD management, such as changes in surgeons, and (2) eliminate confounding from universal statewide pulse oximetry screening, which was instituted in August 2013 [22]. Patients were excluded if the birthing parent had no prenatal care, cardiac transplantation was their first surgery, they did not require surgical or catheterization‐based intervention, or their fundamental diagnosis was acquired disease (e.g., endocarditis). Patients with CHDs that are rarely diagnosed prenatally due to the technical limitations of prenatal ultrasound—e.g., secundum‐type atrial septal defects [23], bicuspid aortic valve, coronary anomalies, and partial anomalous pulmonary venous return [24]—were also excluded.

2.3. Primary outcome and mediator

The primary outcome measure was age (in days) at postnatal intervention (cardiac catheterization or surgery, whichever occurred first). We chose age at intervention, rather than mortality, as the outcome because: (1) mortalities were rare in our cohort, (2) age at postnatal intervention is a proxy for access to postnatal care which is an important benefit related to prenatal diagnosis [25, 26], (3) using age at intervention eliminates statistical bias from procedural and post‐procedural factors, and (4) earlier intervention is associated with improved long‐term outcomes for many types of critical CHD [27, 28, 29].

A mediator is defined as a variable that explains the process or mechanism by which the exposure variable influences the outcome. It must therefore occur sequentially after the exposure variable, but before the outcome variable [15]. We hypothesized that prenatal diagnosis, analyzed as a binary variable (yes/no), was a mediator of disease complexity and age at postnatal intervention. There were no missing data for the outcome or mediator variables.

2.4. Exposure variable

The primary exposure variable was disease complexity, a three‐category ordinal variable defined by how readily the patient's CHD could be detected postnatally based on anatomic and physiologic factors. The three categories (low, moderate, and high complexities) were defined based on tertiles of postnatal diagnosis rates in the EPICARD study, which quantified the relationship between CHD anatomy and timing of diagnosis [30]. High‐complexity CHD presents with more significant signs and symptoms at birth and is thus detected sooner (e.g., hypoplastic left heart syndrome), whereas low‐complexity CHD could present without signs or symptoms and is thus detected later (e.g., ventricular septal defects). The CHDs assigned to each disease complexity category are listed in Table S1.

Existing measures of CHD complexity such as the STAT, RACHS‐2, and Aristotle scores are defined by the cardiac surgery that the patient received, not by cardiac anatomy. We could not use these surgical scores to define disease complexity, since mediation analysis requires that the exposure must occur before the mediator; surgery occurs after prenatal diagnosis. Therefore, disease complexity had to be classified by cardiac anatomy, which is established by 10 weeks’ gestation [31].

2.5. Additional explanatory variables

Sociodemographic characteristics included infant sex assigned at birth, payor status at the time of surgery (public vs. private insurance), and infant race/ethnicity, extracted by STS‐CHSD data collectors from the patient's electronic medical record, or directly collected from clinical providers. A combined race/ethnicity variable was included as a covariate given its socially defined nature, which results in different outcomes due to structural racism. We grouped race and ethnicity into five categories: Hispanic, non‐Hispanic Black, non‐Hispanic Asian or native Pacific Islander, other (multiple races, races other than those in defined categories, or no race or ethnicity reported), and non‐Hispanic White. Hispanic was the reference group due to previous data demonstrating disparate outcomes among Hispanic and non‐Hispanic Black patients, as compared to non‐Hispanic White patients [32, 33].

Clinical characteristics included birth weight; presence of a genetic syndrome, non‐chromosomal syndrome, or extra‐cardiac structural abnormality; and birth hospital level of perinatal care. Gestational age was highly correlated with birth weight and was excluded from the model. Birth hospital levels of care are defined by the American Academy of Pediatrics guidelines:[34] Levels I–II hospitals have nursery‐level care where pediatric cardiac services (e.g., echocardiography) may not be available on‐site. Level III hospitals must have a formal neonatal intensive care unit and level IV hospitals (e.g., CHD surgical centers) must have pediatric surgical subspecialty services available.

2.6. Model specifications

Standard descriptive statistics were used to analyze each explanatory variable. Cells with 10 or fewer patients were combined or suppressed to protect patient privacy. Bivariate comparisons were performed using omnibus Kruskal–Wallis, Chi‐squared, or t‐tests.

We used linear probability regression models with robust standard errors to estimate the associations between the exposure and the outcome, and the exposure and the mediator. All models included a binary fixed effect for infants born during the COVID‐19 pandemic (March 2020–December 2021), as the pandemic may have impacted both prenatal care utilization and postnatal timing of surgery. We analyzed infant congenital comorbidities (i.e., chromosomal abnormality, syndrome, or noncardiac anatomic abnormality) as binary, in accordance with the STS‐CHSD standards [33, 35]. The final model specifications were:

$$\begin{array}{ccc}& & \mathrm{prenatal}\mathrm{diagnosis}={\beta }_0+{\beta }_1\ast \mathrm{disease}\mathrm{complexity}\hfill \\ \hfill & & +{\beta }_2\ast \mathrm{race}/\mathrm{ethnicity}\hfill \\ \hfill & & +{\beta }_3\ast \mathrm{noncardiac}/\mathrm{genetic}/\mathrm{chromosomal}\mathrm{abnormality}\hfill \\ \hfill & & +{\beta }_4\ast \mathrm{payor}+{\beta }_5\ast \mathrm{COVID}+{\epsilon }_1\hfill \end{array}$$ (1)

$$\begin{array}{ccc}& & \mathrm{age}\mathrm{at}\mathrm{intervention}={\beta }_0+{\beta }_1\ast \mathrm{prenatal}\mathrm{diagnosis}\hfill \\ & & +{\beta }_2\ast \mathrm{disease}\mathrm{complexity}+{\beta }_3\ast \mathrm{sex}+{\beta }_4\ast \mathrm{birth}\mathrm{weight}\hfill \\ & & +{\beta }_5\ast \mathrm{race}/\mathrm{ethnicty}\hfill \\ & & +{\beta }_6\ast \mathrm{noncardiac}/\mathrm{genetic}/\mathrm{chromosomal}\mathrm{abnormality}\hfill \\ & & +{\beta }_7\ast \mathrm{hospital}\mathrm{level}\mathrm{of}\mathrm{care}+{\beta }_8\ast \mathrm{COVID}+{\epsilon }_1\hfill \end{array}$$ (2)

We utilized linear rather than logistic regression for Equation 1, because odds ratios obtained from logistic regression are neither additive nor multiplicative with linear regression coefficients (Equation 2) to yield interpretable mediation effects [36, 37]. Infant sex and birth weight were not included in Equation (1) due to the lack of association with prenatal diagnosis in bivariate comparisons. While birth hospital level of care was associated with prenatal diagnosis in bivariate comparisons, it was not included in Equation (1) because it did not contribute to model fit after including disease complexity and noncardiac/genetic/chromosomal abnormality. Payor was excluded from Equation (2) due to the lack of association with age at intervention in bivariate comparisons.

2.7. Mediation

To evaluate whether prenatal diagnosis explained any of the relationship between CHD complexity and age at intervention, and decompose the effects of prenatal diagnosis from other factors which explain why moderate‐ or high‐complexity CHDs are intervened upon sooner than low‐complexity CHDs, we first ensured that conditions [38, 39] for mediation were met: (1) disease complexity was associated with age at intervention, (2) disease complexity was associated with prenatal diagnosis, and (3) the association between disease complexity and age at intervention weakened with the addition of prenatal diagnosis to the model. We used the mediate package in Stata18 to estimate indirect effects (the effect of disease complexity on age at intervention that occurs through prenatal diagnosis), direct effects (the effect of disease complexity on age at intervention that does not occur through prenatal diagnosis), total effects (the sum of the direct and indirect effects), and the percent mediated (indirect effect/total effect).

2.8. Sensitivity analyses and ethical oversight

We performed three sensitivity analyses. First, to check whether the cohort appropriately excluded elective cases, we repeated analyses excluding patients admitted from home prior to surgery. Second, to check our disease complexity specification, we repeated analyses using a three‐level ordinal variable for complexity based on the probability of prenatal diagnosis rather than postnatal diagnosis. Like the disease complexity variable in the primary analysis, the disease complexity variable in this sensitivity analysis was determined from existing literature [40, 41, 42] (Table S1). Third, to check our outcome specification, we repeated the analyses using the outcome of age at surgery, rather than age at intervention.

All analyses were conducted using Stata18 (StataCorp). The study was deemed exempt from human subjects review by the Institutional Review Board at Lurie Children's (IRB 2022‐5367), as data had already been extracted from electronic medical records and de‐identified when obtained from the locally held STS‐CHSD registry.

3. RESULTS

3.1. Cohort characteristics

In total, 429 infants met the inclusion criteria. Of these, 122 (28%) had high‐, 271 (63%) had moderate‐, and 36 (8%) had low‐complexity disease; 293 (68%) had prenatal diagnosis; and 66 (15%) had a cardiac catheterization as their initial intervention. The median age at initial intervention was 6 days (IQR 4–14 days). The most common diagnoses were aortic arch anomalies including coarctation (n = 96, 22%), D‐ or L‐transposition of the great arteries (n = 88, 20%), and hypoplastic left heart syndrome (n = 72, 17%). Table 1 describes cohort characteristics stratified by prenatal diagnosis status.

TABLE 1.

Cohort characteristics stratified by prenatal diagnosis status (N = 429).

	N (%) or median (IQR) or mean (SD)	N (%) or median (IQR) or mean (SD)
	Postnatal Diagnosis (n = 136)	Prenatal diagnosis (n = 293)	p value
Age at intervention (days)	11 (6–29)	6 (4–9)	<0.001
Disease complexity a			<0.001
Low	26 (6.1)	10 (2.3)
Moderate	91 (2.1)	180 (42.0)
High	19 (4.4)	103 (24.0)
Infant sex			0.330
Male	82 (19.1)	162 (37.8)
Female	54 (12.6)	131 (30.5)
Birth weight (kg)	3.07 (0.7)	3.03 (0.6)	0.643
Race and ethnicity			0.172
Hispanic	41 (9.6)	78 (18.2)
Non‐Hispanic White	73 (17.5)	149 (34.7)
Non‐Hispanic Black	15 (3.5)	34 (7.9)
Non‐Hispanic Asian, other or multiple race b	7 (1.6)	32 (7.5)
Other abnormality in infant c			0.067
No	102 (23.8)	194 (45.2)
Yes	34 (7.9)	99 (23.1)
Payor			0.779
Private	91 (21.2)	192 (44.8)
Public	45 (10.5)	101 (23.5)
Birth hospital level of care d			<0.001
I, II	55 (12.8)	6 (1.4)
III	53 (12.4)	31 (7.2)
IV (CHD surgical center)	28 (6.5)	256 (59.7)
Fundamental diagnosis e			<0.001
Aortic arch anomalies including coarctation	48 (11.2)	48 (11.2)
D‐ or L‐transposition of the great arteries	18 (4.2)	70 (16.3)
Hypoplastic left heart syndrome	10 (2.3)	62 (14.5)
Tetralogy of Fallot	6 (1.4)	23 (5.4)
Atrioventricular or ventricular septal defect	17 (4.0)	21 (4.9)
Tricuspid atresia, Ebstein anomaly, or double inlet left ventricle	6 (1.4)	21 (4.9)
Pulmonary atresia or discontinuous pulmonary arteries	8 (1.9)	18 (4.2)
Total anomalous pulmonary venous return	13 (3.0)	7 (1.6)
Semilunar valve stenosis	5 (1.2)	14 (3.3)
Truncus arteriosus	5 (1.2)	9 (2.1)

^a Fundamental diagnoses included in each complexity level are shown in Table S1.

^b Race/ethnicity are combined due to cell size < 5, though non‐Hispanic Asian and multiple/other race are separate categories in the multivariable regressions.

^c Denotes a noncardiac anatomic abnormality (e.g., gastroschisis), genetic syndrome (e.g., Alagille syndrome), or chromosomal abnormality (e.g., Trisomy 21) present in infant.

^d Birth hospital levels of care as defined by the American Academy of Pediatrics. Levels I and II entail general nursery care, level III entails neonatology care, and level IV entails the availability of CHD surgical services.

^e Fundamental diagnoses are combined if cell size < 5.

3.2. Associations of disease complexity, prenatal diagnosis, and age at intervention

To estimate the mediation effects of prenatal diagnosis on the associations between disease complexity and age at intervention, three conditions must be met. First, disease complexity must be associated with age at intervention. In unadjusted models, on average, those with moderate‐complexity disease received intervention 13 days sooner (95% CI 6–20 days), and those with high‐complexity disease received intervention 18 days sooner (95% CI 11–25 days), than those with low‐complexity disease. Second, disease complexity must be associated with prenatal diagnosis. In unadjusted models, on average, those with moderate‐complexity disease were 38.6 percentage points (95% CI 22.9–55.4pp) more likely, and those with high‐complexity disease were 56.6 percentage points (95% 40.6–72.7pp) more likely, to receive prenatal diagnosis than those with low‐complexity disease. Finally, the association between disease complexity and age at intervention must decrease after adjusting for prenatal diagnosis. In the fully adjusted model which included prenatal diagnosis, those with moderate‐complexity disease received intervention 10 days sooner (95% CI 3–17 days) and those with high‐complexity disease received intervention 13 days sooner (95% CI 6–20 days) than those with low‐complexity disease (Table 2).

TABLE 2.

Adjusted associations between disease complexity, prenatal diagnosis, and age at intervention.

	Difference in probability of prenatal diagnosis	Average difference in age at intervention, in days
Disease complexity a
Low	Ref	Ref
Moderate	0.40 (p < 0.001)	−9.9 (p = 0.005)
High	0.59 (p < 0.001)	−13.2 (p < 0.001)
Race/ethnicity
Hispanic	Ref	Ref
Non‐Hispanic White	0.05 (p = 0.415)	1.3 (p = 0.349)
Non‐Hispanic Black	0.04 (p = 0.550)	1.2 (p = 0.553)
Non‐Hispanic Asian or Native Pacific Islander	0.31 (p < 0.001)	3.0 (p = 0.328)
Other b	−0.07 (p = 0.617)	−4.1 (p = 0.166)
Other abnormality in infant c
No	Ref	Ref
Yes	0.08 (p = 0.070)	4.6 (p = 0.003)
Payor		Not in model
Private	Ref
Public	0.02 (p = 0.632)
Sex	Not in model
Male		Ref
Female		0.4 (p = 0.729)
Birth weight (kg)		−3.8 (p = 0.001)
Birth hospital level of perinatal care d
Levels I and II		Ref
Level III		3.8 (p = 0.164)
Level IV (CHD surgical center)		3.7 (p = 0.207)

Note: Bolded values indicate statistical significance (p < 0.05).

^a Fundamental diagnoses included in each complexity level are shown in Table S1.

^b Denotes other, multiple, or missing race/ethnicity.

^c Denotes a noncardiac anatomic abnormality (e.g., gastroschisis), genetic syndrome (e.g., Alagille syndrome), or chromosomal abnormality (e.g., Trisomy 21) present in infant.

^d Birth hospital levels of care as defined by the American Academy of Pediatrics (PMID 22926177). Levels I and II entail general nursery care, level III entails neonatology care, and level IV entails the availability of CHD surgical services.

3.3. Mediation analysis

The mediation effect of prenatal diagnosis was statistically significant for moderate‐ versus low‐complexity disease (indirect effect 3.7 days, 95% CI 1.4–6.1 days) and for high‐ versus low‐complexity disease (indirect effect 5.5 days, 95% CI 2.2–8.8 days). This resulted in a 27% mediation effect for moderate versus low‐complexity disease, and a 29% mediation effect for high‐ versus low‐complexity disease. In other words, slightly more than one‐quarter of the relationship between disease complexity and age at intervention for moderate‐ and high‐ versus low‐complexity disease is attributable to prenatal diagnosis having occurred. The direct and total effects were also statistically significant in the mediation model (Figure 1).

FIGURE 1.

Directed acyclic graph of mediation analysis. A directed acyclic graph demonstrating the mediating effect of prenatal diagnosis on congenital heart defect (CHD) complexity and age at intervention. The indirect effect represents the effect of CHD complexity on age at intervention that occurs through prenatal diagnosis, the direct effect represents the effect of CHD complexity on age at intervention that does not occur through prenatal diagnosis, the total effect is the sum of the direct and indirect effects, and the percent mediated is the ratio of indirect effect/total effect.

3.4. Sensitivity analyses

Results were similar in the first two sensitivity analyses. In the third sensitivity analysis, which replaced age at intervention with age at surgery, the association between disease complexity and age at surgery was not statistically significant (Tables S2–S4).

4. DISCUSSION

In this study, we found that those with high‐ and moderate‐complexity CHD are intervened upon sooner than those with low‐complexity CHD, in part, because they are likelier to be prenatally detected. Our mediation models demonstrate statistically significant direct and total effects, suggesting that prenatal diagnosis is not the only factor that explains the relationship between CHD complexity and intervention timing. This is consistent with clinical experience; clinical factors such as a restrictive atrial septum requiring urgent balloon atrial septostomy also likely mediate the relationship between CHD complexity and intervention timing. Nevertheless, we demonstrate how prenatal diagnosis has a mechanistic role in improving access to care for those infants with moderate‐ or high‐complexity CHD.

There are two advantages in evaluating prenatal diagnosis as a mediator, rather than an exposure variable (i.e., comparing outcomes by prenatal diagnosis status). First, a mediation model has more explanatory power because it estimates the mechanism by which prenatal diagnosis impacts postnatal events, whereas evaluating prenatal diagnosis as an exposure variable can only estimate associations (i.e., relationships without directionality) of prenatal diagnosis with postnatal events [15]. Further, associations between prenatal diagnosis and outcomes can yield paradoxical results if there is insufficient controlling for CHD complexity:[5, 6, 11, 12, 13] for instance, a study of infants who received critical CHD surgery between 2004 and 2008 at high‐volume center showed no association between prenatal diagnosis and surgical timing [9]. Another study from five tertiary centers in California, 2011–2013, showed that prenatal diagnosis was associated with later surgery compared to prenatal diagnosis [10].

An additional key difference in our analysis compared to previous studies is that the outcome was defined as age at intervention, rather than age at surgery. When we replaced age at intervention with age at surgery, there was no significant association between CHD complexity and age at surgery. Therefore, using age at surgery as the primary outcome, we could not have estimated the true mediating effect of prenatal diagnosis on postnatal access to care. This is consistent with the increasing prevalence of catheterization‐based neonatal interventions in the modern era. Catheterization interventions can delay surgery, as in balloon valvuloplasty for critical pulmonary stenosis, and in some cases, substitute for major neonatal cardiac surgeries such as the Norwood operation [43, 44]. We therefore recommend future analyses to consider the contemporary role of catheterization‐based therapies.

At the population level, this this analysis serves as a reminder of how disparities in prenatal diagnosis may lead to disparities in access to care. Prenatal diagnosis rates are lower in the publicly insured and non‐urban populations [42, 45]. The reasons for these disparities remain unclear, but may be related to differences in ultrasound quality [46] and socioeconomic differences in prenatal care utilization [47]. For instance, our team has found that insurance‐based disparities in prenatal diagnosis may be mitigated by urban residence [48], and that the receipt of second trimester obstetric ultrasound—and not fetal echocardiogram–is a key driver for mitigating insurance‐related disparities in prenatal CHD diagnosis [49]. The latter finding reflects the need for multidisciplinary collaboration between obstetrics, maternal‐fetal medicine, neonatology, and pediatric cardiology at the patient level to improve prenatal diagnosis rates. Further, this mediation analysis reinforces that patient level efforts to overcome barriers to obtianing pernatal diagnosis can have downstream benefits on postnatal outcomes [50].

With the knowledge that prenatal diagnosis leads to earlier intervention to quantify the long‐term impact of prenatal diagnosis, one must consider how earlier intervention impacts surgical outcomes. Our clinical experience tells us that optimal surgical timing is unique to each patient's circumstances. Yet on average, timelier intervention is associated with improved short and long‐term outcomes for the three most common CHDs in our cohort: for patients with critical aortic arch anomalies, later surgery may predict worse vascular reactivity and a higher likelihood of systemic hypertension [28, 51]. For patients with D‐transposition of the great arteries, the odds of major morbidity increase up to 47% with each day later that surgery is performed [27]. For patients with hypoplastic left heart syndrome, delayed surgery after the first week of life is associated with increased major morbidity [52] and white matter injury [53].

Our analysis has some limitations. First, our data are from a single, albeit high‐volume, CHD surgical center, excluding pregnancy terminations, infants who died before surgery, or infants who received surgery elsewhere. While our prenatal diagnosis rate and sociodemographic distribution are similar to the only other CHD surgical center in the Chicago metropolitan area [48], our findings may not extend to other geographic regions of the United States where prenatal diagnosis rates are more/less disparate. However, a benefit of single‐center data is the elimination of between‐center practice variation that might confound the relationship between prenatal diagnosis and age at postnatal intervention, and selection bias due to missing data. Second, our analysis focuses on timing to intervention, a measure of access to care, rather than morbidity and mortality, which are rare in this population. While the long‐term benefits of timelier surgery have been described for the three most common lesions in our cohort [27, 28, 51, 52, 53, 54], these benefits may not extend to the entirety of our cohort [55]. Third, the outcome variable, intervention timing, does not capture other impacts of prenatal diagnosis that have been previously described, such as improved delivery planning or familial functioning [5].

5. CONCLUSION

Prenatal diagnosis plays a consequential role in the timing of postnatal intervention for neonates with moderate or high‐complexity CHD. Future analyses studying the impacts of prenatal diagnosis on outcomes should consider alternatives to evaluating prenatal diagnosis as an exposure variable. Research and advocacy efforts to improve prenatal diagnosis rates hold promise for improving access to care for this clinically high‐risk group.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supporting information

DATA AVAILABILITY STATEMENT

Due to the presence of protected health information, supporting data cannot be shared.