The Performance of Critical Congenital Heart Disease Screening in Rural Versus Urban Locations in the Northwest United States

Abstract

Objectives

Pulse oximetry screening of newborns detects critical congenital heart disease (CCHD). Rural birth location is known to affect timing and management of when infants with CHD undergo surgery, but its association with CCHD screening is unknown. We assess the relationship between rural location and postnatal CCHD diagnosis and describe lesion-specific modes of diagnosis.

Study design

Data were abstracted from medical records at 2 cardiac surgery centers in Washington state. Infants with CCHD, defined as CHD requiring either cardiac intervention or resulting in death at <1 month of age, born between July 2015 and June 2020, were included and classified by method of identification. Patient home ZIP codes were used to determine rural location.

Results

We included 561 newborns with CCHD; 35% were diagnosed postnatally. Predominant postnatal mechanisms of identification (n = 194) included symptomatic before CCHD screening period (56%), CCHD screening (22%), and symptomatic after false-negative screen (15%). Postnatal diagnosis rate increased with degree of ruralness (48% in small rural/isolated regions vs 32% in urban regions; P = .01) and incidence of undiagnosed CCHD at birth increased with lower nursery levels (5.5/10 000 live births in nursery level 1 vs 2.1/10 000 live births in level 4).

Conclusions

CCHD screening identifies 22% of postnatally diagnosed CCHD and 7% of cases overall in our region. Postnatal diagnosis is more common in rural regions. The incidence of undiagnosed CCHD at birth increases with decreasing nursery levels. This study supports the value of CCHD screening in routine newborn care, especially in rural areas and hospitals with lower nursery levels.

Universal newborn screening for critical congenital heart disease (CCHD) using pulse oximetry is efficacious and cost-effective and was recommended nationally in 2011.1, 2, 3, 4 An analysis of the National Center for Health Statistics data concluded that states that adopted mandatory screening demonstrated a 33% decline in infant deaths from CCHD. Single-center studies have shown that CCHD screening is useful in rural settings, but no studies in the US have compared the difference in postnatal diagnosis rates for CCHD between rural and urban settings.5, 6, 7, 8 A recent Canadian publication demonstrated a decreased rate of prenatal diagnosis in rural regions; however, it is unclear if these findings are generalizable to the US.⁹

Critical forms of CHD encompass a wide variety of lesions that share the common feature of poor adaptation from fetal to neonatal life within the first days to weeks after birth. Without timely diagnosis and management, these anomalies can result in shock, extreme cyanosis, severe heart failure, or death. Each lesion varies in birth prevalence, likelihood of prenatal diagnosis, sensitivity and specificity of CCHD screening, and timing of clinical presentation in the absence of prenatal diagnosis. These factors interact to affect postnatal screening's contribution to improving the timely identification, and thus outcome, of children with a given form of CCHD. Literature describing these lesion-specific profiles is scarce; most studies are relatively old and reflect a low prenatal diagnosis rate.10, 11, 12, 13, 14, 15, 16 One recent publication from Sweden, a country whose population and universal healthcare system differ significantly from those of the US, describes these lesion-specific profiles.¹⁷

Given the marked advances in prenatal diagnosis of CCHD over the last 20 years and the possibility for decreased prenatal diagnosis in rural regions, we aimed to assess the performance of CCHD screening in rural vs urban regions in the Northwest US.

Washington state has 2 congenital cardiac surgery programs that serve a large catchment area in the Northwest US, including most of Washington and Alaska, as well as portions of Montana and Idaho. Most of the land mass is rural, although the population is largely urban.¹⁸ All states in our study use the algorithm created by the American Academy of Pediatrics and Centers for Disease Control and Prevention (CDC), which recommends screening between 24 and 48 hours of life. Using contemporary data from this region, we aim to: (1) compare postnatal diagnosis rates in rural vs urban areas and in lower vs higher nursery levels, (2) compare lesion-specific modes of presentation of CCHD with prior estimates, and (3) define lesion-specific contribution of CCHD screening to the identification of affected newborns.

Methods

Patient Identification

Since July 2015, at the request of the Washington State Department of Health, the 2 congenital cardiac surgery programs in Washington have gathered data on all infants cared for at their institutions who either required cardiac intervention (surgical or transcatheter) or died from CHD at <1 month of age. Patients who required intervention at >1 month of age were excluded. The 2 programs are the Providence Sacred Heart Medical Center program (PSHMC, Spokane, WA) and the Seattle Children's Heart Center program, which has 2 sites of practice at Seattle Children's Hospital (SCH, Seattle, WA) and Mary Bridge Children's Hospital (MBCH, Tacoma, WA). All cardiac admissions at each institution were screened for neonates meeting the inclusion criteria. Data were obtained by review of medical records and query of each institution's Society of Thoracic Surgeons (STS) Congenital Heart Surgery Database, a clinical registry that includes all patients undergoing congenital heart surgery at the institution. The Seattle Children's STS registry covers both sites of practice. Data elements included cardiac diagnosis, gestational age, date and time of birth, location of birth (hospital, birth center, or home), state of residence, patient ZIP code, mechanism of identification, screening data, evidence of complications at diagnosis, and a brief description of the clinical case.

Diagnoses were classified and assigned by using both the CDC's list of CCHD and the STS diagnosis short list.^19,20 If an infant had multiple cardiac diagnoses, the diagnosis that necessitated intervention in the first month of life was assigned. For example, an infant with total anomalous pulmonary venous return (TAPVR) and balanced complete atrioventricular canal who required intervention for the TAPVR in the first month of life would be classified as TAPVR. Cardiac diagnoses were finalized in consultation with an attending pediatric cardiologist at each site.

For this study, we included infants who met the criteria for reporting to the Washington State Department of Health with birthdates between July 1, 2015, and June 30, 2020. Infants cared for at these institutions were included regardless of their state of residence or state of birth. Babies born in Washington with CCHD are typically treated at 1 of the 2 centers participating in this study, except for a limited number of infants in southwest Washington who are in greater proximity to Portland, Oregon. No congenital cardiac surgery programs exist in Alaska, Montana, or Idaho, but referrals for CCHD care are not exclusively to the Washington centers.

The study protocol was reviewed and approved by the Institutional Review Boards at SCH and PSHMC.

Classifications

Newborns were classified by mechanism of identification as listed in Table I. Rural-urban commuting area (RUCA) codes were used to classify patients’ home ZIP codes.^21,22 RUCA codes are a national census tract-based classification scheme that combines standard Bureau of Census Urbanized Area and Urban Cluster definitions with work commuting information. According to standard methods, the RUCA codes were aggregated into 1 of 4 clusters: urban, large rural, small rural, and isolated. Small rural and isolated were combined in this analysis owing to small numbers in each category. Nursery levels were classified from 1 to 4, in accordance with the American Academy of Pediatrics Policy Statement on Levels of Neonatal Care and as reported in publicly available documents from state health departments.²³ Increasing nursery level corresponds with greater capabilities for the care of sick newborns.

Table I.

Definitions of methods to identify CCHD

Mechanism	Definition
Prenatal	Prenatal diagnosis of CHD
Symptomatic before the screening period	Any symptoms prompting evaluation for CCHD from birth to 48 hours of age, as long as the infant had not yet been screened
Screening	Identified by failed CCHD screen between 24 and 48 hours of life per American Academy of Pediatrics guidelines
Symptomatic after false negative screen	Identified owing to symptoms after passing CCHD screening
Not screened	CCHD screening not performed by 48 hours of life and baby subsequently developed symptoms
Other	Diagnosis prompted by reason other than symptoms or failed screen, typically echocardiogram ordered for other anomalies or dysmorphisms in the absence of recognized symptoms

Statistical Methods

Descriptive statistics were used to analyze these data. To maximize the ability to detect effects of rural location, the entire catchment area of the 2 cardiac surgical centers was included. Lesion-specific prevalence at birth and estimated prevalence at birth of undiagnosed CCHD by nursery level were limited to newborns who were Washington residents, because there was near-complete capture of births in Washington. Publicly available data sources were consulted to determine the number of live births by hospital and the nursery level of each hospital in Washington, which are reported by calendar year.²⁴ Because our study period was July 2015 through June 2020, the closest reported 5-year period of January 2015 through December 2019 was a used as a reasonable estimate of total in-hospital births (416 422) for the study period.

To derive CCHD lesion-specific prevalence at birth, the number of identified cases in Washington residents was divided by the total in-hospital live births in Washington for the years 2015-2019. To estimate the incidence of postnatally diagnosed newborns with CCHD by nursery level in Washington, for each nursery level, the number of postnatally diagnosed infants with CCHD was divided by the total number of births in hospitals of that nursery level. The analysis was repeated excluding births at 3 hospitals in southwest Washington that typically refer to Portland, Oregon, with no cases of CCHD in our dataset.

The rate of prenatal diagnosis for lesions with a grossly abnormal 4 chamber view (Ebstein anomaly and single ventricle [SV] lesions including tricuspid atresia, hypoplastic left heart syndrome, anatomical SV, and pulmonary atresia), was compared with the rate of prenatal diagnosis for other lesions using a χ² test. The rate of postnatal diagnosis for urban RUCA codes was compared with the rate of postnatal diagnosis for small rural/isolated RUCA codes using a χ² test. We also compared Washington and non-Washington residents by whether their RUCA codes were urban or small rural/isolated using a χ² test.

Results

561 newborns with CCHD were born during the study period from July 2015 through June 2020. Descriptive features of the cohort are shown in Table II. Seventy-eight percent of infants were cared for at SCH, 22% at PSHMC, and <1% at MBCH. Neonates hospitalized at MBCH requiring surgery with cardiopulmonary bypass are typically transferred to SCH; the 2 cases performed at MBCH were patients with isolated coarctation of the aorta (CoA) repaired through a left thoracotomy approach. Infants were predominantly residents of Washington (77%), with most of the remainder having residency in Alaska (8%), Montana (6%), or Idaho (4%).

Table II.

Description of the cohort of infants with CCHD

Descriptor	No. (%)
Surgical center
SCH	436 (77.7)
MBCH	2 (0.4)
PSHMC	123 (21.9)
State of residence
WA	431 (76.8)
AK	43 (7.7)
MT	36 (6.4)
ID	25 (4.5)
OR	3 (0.5)
Other or missing	23 (4.1)
Location of birth
Hospital	547 (97.5)
Non-hospital birth center	5 (0.9)
Home	7 (1.2)
Other or missing	2 (0.4)
Gestational age
Term (≥37 weeks estimated gestational age)	510 (90.9)
Preterm (<37 weeks estimated gestational age)	51 (9.1)
Mechanism of identification
Prenatal	367 (65.4)
Symptoms before screening	109 (19.4)
Screening	42 (7.5)
Symptoms after false negative screen	30 (5.3)
Not screened	7 (1.2)
Other	6 (1.1)
Cardiac diagnosis (CDC screening targets)
Ebstein anomaly	1 (0.2)
Tricuspid atresia	3 (0.5)
Hypoplastic left heart syndrome	77 (13.7)
SV	44 (7.8)
Pulmonary atresia	33 (5.9)
Truncus arteriosus	17 (3.0)
TOF	32 (5.7)
DORV	16 (2.9)
d-TGA	92 (16.4)
CoA	120 (21.4)
Interrupted aortic arch	27 (4.8)
TAPVR	44 (7.8)
Other	55 (9.8)
RUCA code for home zip code
Urban	437 (77.9)
Large rural	55 (9.8)
Small rural	30 (5.3)
Isolated	39 (6.9)
Nursery level (data are available for Washington births only)
1	30 (6.1)
2	44 (9.0)
3	83 (17.0)
4	332 (67.9)
Other (home or non-hospital birth)	0 (0.0)

Of the 431 Washington state residents, 372 (86%) had urban RUCA codes and 25 (6%) had small rural/isolated RUCA codes; of the 114 non-Washington state residents, 54 (47%) had urban RUCA codes and 40 (35%) had small rural/isolated RUCA codes (P < .0005). Figure 1 demonstrates that as patients’ ZIP codes became more rural, as defined by RUCA code, postnatal diagnosis rate increased from 32% to 48%; the difference between urban and small rural/isolated was statistically significant (P = .01). Figure 2 demonstrates that the estimated incidence of postnatal CCHD diagnosis in Washington birth hospitals was inversely related to hospital nursery level, more than doubling from level 4 to level 1. When births from 3 hospitals that refer to Portland, Oregon, were excluded from the denominator (7% of overall births in Washington), the patterns observed did not change. Because these data were derived using counted prevalence data, no statistical analysis was performed.

Figure 1.

Percent of postnatal diagnosis of CCHD by RUCA code.

Figure 2.

Estimated incidence of undiagnosed CCHD per 10 000 live births.

The postnatal diagnosis rate was 35% overall and varied little over the 5 years of the study. The postnatal mechanisms of identification included: symptomatic before the CCHD screening period (56%), CCHD screening (22%), symptomatic after false-negative screen (15%), not screened (4%), or other (3%) (Figure 3). Of the 7 neonates not screened, 2 did not qualify for CCHD screening owing to neonatal intensive care unit admission and 5 were missed. The most common CCHD diagnoses that were identified by symptoms before the screening period were TAPVR (26%), dextro-transposition of the great arteries (d-TGA; 26%), CoA (15%), interrupted aortic arch (6%), and other CCHDs (15%). Screening accounted for 25% of all TAPVR diagnoses. CCHD screening was also important in the diagnosis of interrupted aortic arch (19%), tetralogy of Fallot (TOF; 16%), and double outlet right ventricle (DORV; 13%). CoA represented 73% of false-negative screens. TAPVR has a very low prenatal diagnosis rate, but nearly complete diagnosis by the end of the CCHD screening period.

Figure 3.

Mechanism of identification of CCHD by percentage of patients.

Table III shows estimated live birth prevalence per 10 000 live births of different CCHD diagnoses requiring intervention in the first month of life in Washington. Figure 4 shows mechanisms of identification for each CCHD lesion. Lesions with SV physiology plus Ebstein anomaly, which typically have a grossly abnormal 4-chamber view, had very high rates of prenatal diagnosis. The overall rate of prenatal diagnosis for these lesions was 93% (147/158) vs 55% for all other lesions (220/403; P < .001). The remaining SV physiology diagnoses in the cohort were diagnosed owing to symptoms before screening and with screening.^2,9

Table III.

Estimated prevalence of CCHD lesions per 10 000 live births

CCHD lesion	Estimated prevalence per 10 000 live births
Ebstein anomaly	0.02
Tricuspid atresia	0.05
Truncus arteriosus	0.29
DORV	0.29
Interrupted aortic arch	0.53
TOF	0.60
Pulmonary atresia	0.62
TAPVR	0.74
SV	0.79
Other critical CHDs	1.06
HLHS	1.46
d-TGA	1.61
CoA	2.28

Figure 4.

Mechanism of identification of CCHD by lesion. Each bar represents one lesion and the total for each lesion is listed at the top of the bar.

Discussion

Pulse oximetry screening for CCHD has become the standard of care worldwide. However, data describing the importance of screening and the effects of rural location on postnatal CCHD diagnosis, as well as lesion-specific data on the mechanisms of diagnosis in the era of universal CCHD screening, have not been available previously.

Our data indicate two important relationships: that prenatal diagnosis occurs less commonly in more rural areas throughout the catchment region and in hospitals with a lower nursery level in Washington state. This observation may explain why in other studies only a small number of CCHD cases are identified by CCHD screening at large academic institutions, which are typically in urban areas and have high levels of care, despite national data indicating a much larger impact of screening leading to a mortality benefit.10, 11, 12, 13, 14, 15, 16 In our region, urban hospitals’ nursery levels ranged from 1 to 4, whereas rural settings have mostly lower nursery levels; thus, decreased prenatal diagnosis rates in this regional cohort may be influenced by increasing degrees of ruralness, lower nursery levels, or both. These data can inform quality improvement efforts for both prenatal detection and postnatal screening by targeting rural hospitals and those with lower nursery levels.

In the regional cohort reported here, screening accounted for 7% of all CCHD diagnoses and 22% of all postnatal CCHD diagnoses (Figure 3). Postnatally diagnosed CCHD was otherwise primarily diagnosed by symptoms before screening (56%) or symptoms after false-negative screening (15%). Only 3% of the qualifying newborns in the postnatally diagnosed cohort (<1% overall) missed being screened. The only other contemporary cohort which provides lesion-specific profiles of mode of diagnosis in the setting of universal CCHD screening, used a similar definition of CCHD (CHD requiring surgery or catheter intervention, or resulting in death without such treatment within 28 days of birth), but the prenatal diagnosis rate in our cohort was higher (65% vs 42%).¹⁷ The proportion diagnosed by symptoms before screening was similar, although the proportions diagnosed owing to CCHD screening or symptoms after screening were both lower in our cohort. Examining lesion-specific modes of diagnosis, the rate of prenatal diagnosis was higher in our cohort for nearly all forms of CCHD, except Ebstein anomaly (100% prenatal diagnosis in both cohorts) and TAPVR (<10% in both cohorts). In both studies, CoA was the diagnosis most frequently missed both by prenatal detection and postnatal screening.

Overall, our estimates of prevalence at birth for each CDC screening target are similar to prior reports, with notable exceptions being DORV, Ebstein anomaly, and TOF which were much less prevalent in our cohort.¹⁰ In both a 2014 study from Massachusetts and the national estimates generated from Ailes et al. in 2015, the prevalence of TOF was estimated around 4 per 10 000 live births, whereas in this dataset it was less than 1 per 10 000 live births.^10,13 We hypothesize that because we limited our definition to infants requiring cardiac intervention or surgery within the first month of life, we captured severe forms of TOF, but not less severe cases who undergo elective repair at a few months of age. This hypothesis is supported by similar estimates of birth prevalence of these 3 lesions in the Swedish study, which used a definition of CCHD similar to ours.¹⁷ Additionally, in our study the prenatal detection rate of TOF was 75%, compared with 52% in the Massachusetts study and 26% in the Ailes et al study.¹⁰ It is unclear whether this is finding attributable to improved overall prenatal detection of TOF over time or inclusion of only the severe forms in our definition. This argument can be extended to DORV and Ebstein anomaly as well.

Further analysis of the CCHD screening targets by diagnosis, lesions with an abnormal 4-chamber view (Ebstein anomaly, tricuspid atresia, hypoplastic left heart syndrome, anatomical SV, pulmonary atresia) are much more readily detected prenatally than all other lesions (93% vs 55%). In general, the primary mechanisms of identification differed for each form of CCHD.

The relatively high rates of postnatal diagnosis of TAPVR and screening failures for CoA demonstrate challenges in detection of CCHD and illustrate how lesion-specific modes of diagnosis differ markedly. TAPVR had the lowest prenatal diagnosis rate, but nearly all cases were diagnosed by the end of the screening period. Additionally, TAPVR was the most common diagnosis among neonates with failed CCHD screens. In contrast, CoA had a moderate prenatal diagnosis rate, but 25% were undiagnosed by the end of the screening window. Furthermore, 73% of all false-negative screens were in patients with CoA. CoA and TAPVR have a normal 4-chamber view with normal outflow tracts and are therefore likely to have a falsely normal 20-week anatomy scan. However, postnatally, TAPVR is likely to present with cyanosis or tachypnea before the screening period or to be identified on CCHD screening. In contrast, CoA is often asymptomatic before closure of the ductus arteriosus and only has a saturation differential detectable by CCHD screening in the setting of pulmonary hypertension and right-to-left shunting across the ductus. CoA is the diagnosis most likely to be missed by all modes of identification and these analyses argue that we must continue to devote efforts to improving the diagnosis of CoA.

Limitations

An accurate denominator (ie, total number of births) was required to calculate the prevalence of CCHD and the estimated incidence of undiagnosed CCHD at birth; as such, these analyses were limited to Washington state residents. For the analysis of the effects of RUCA code on postnatal diagnosis, all patients within the catchment area of the 2 major surgical centers in Washington were included to maximize our ability to assess for differences by rural location. Additionally, owing to small group sizes small rural and isolated were combined which limits the ability to differentiate these 2 groups. Data from southwest Washington residents cared for in Portland, Oregon, were unable to be included; however, this subset is small compared with the number of infants treated at the 3 reported institutions. Furthermore, analysis of the nursery levels did not change when these infants’ birth hospitals were removed from the dataset. Although it is possible that the noted differences related to the degree of ruralness or nursery level will vary over time, the present study only had the ability to assess the change in overall postnatal diagnosis rate over time. Only live newborns were considered because the goal of the study was to evaluate postnatal CCHD screening. Infants who died outside of the referral centers are not included in this study, although this number is thought to be small. An age-based definition, rather than a diagnosis-based definition, was used to simplify data collection and focus on the most vulnerable children with CCHD, specifically neonates. However, three infants who would be classically deemed to have CCHD (one with anatomical SV and two with d-TGA) were excluded because they presented at >30 days of age, and thus did not meet our definition of CCHD. Similarly, infants requiring intervention at >30 days (eg, those with less severe forms of TOF, DORV, or Ebstein anomaly) were also not captured. Additionally, although this is a large cohort of newborns, the total number of the rarer CCHD lesions is small. ZIP codes were abstracted from the medical record; thus, if a patient moved after birth and was readmitted to the hospital, the abstracted ZIP code would reflect the new home location thus we were unable to screen for patients who may have moved closer to a higher level of nursery owing to prenatal concerns. To decrease this error, ZIP codes that did not match birth state of residence were excluded and this represented <5% of the overall data. Race and ethnicity data were obtained from the STS database, but a high proportion of missing data did not allow meaningful analysis of race or ethnicity. Whether patients lived on a reservation or identified as Native American or Alaskan Native was not included in the dataset. Although it is known that socioeconomic status affects outcomes in CHD, ZIP codes were not used as a marker of socioeconomic status owing to the possibility of significant socioeconomic variability within the same ZIP code.²⁵ Moreover, although it is possible that there were differences in prenatal care between mothers living in rural vs urban areas, maternal prenatal information could not be captured from our data sources.

Conclusions

CCHD screening identifies 22% of postnatally diagnosed CCHD and 7% of cases overall. The rate of postnatal diagnosis of CCHD is related to rural location, with an increase from 32% in urban regions to 48% in small rural/isolated regions. Additionally, there is a higher incidence of undiagnosed CCHD at birth in lower nursery levels than higher nursery levels. This study also highlights that CCHD comprises a heterogeneous group of lesions for which prenatal detection and CCHD screening are variably effective. For hospital-based newborn providers, these data show that the most common diagnoses that present with symptoms before CCHD screening are TAPVR, d-TGA, and CoA, and reassure that 92% of CCHD has been diagnosed by the end of the CCHD screening period. Additionally, these data inform all providers who take care of newborns that CoA represents 73% of false-negative CCHD screens and serve as a useful reminder that careful palpation of femoral pulses, especially during the first month of life, is critical to an effective cardiovascular examination. Screening remains crucial in the diagnosis of CCHD and is especially important in rural areas and hospitals with lower nursery levels.

CRediT authorship contribution statement

Brian S. Marcus: Writing – review & editing, Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Plicy Perez-Kersey: Writing – review & editing, Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Ann B. Lee: Writing – review & editing, Project administration, Methodology, Investigation, Data curation. Richard A. Jensen: Writing – review & editing, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Beth S. Dullanty: Writing – review & editing, Project administration, Methodology, Investigation, Data curation. Patrick R. Parrish: Writing – review & editing, Project administration, Methodology, Investigation, Data curation. Matthew V. Park: Writing – review & editing, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. William Tressel: Writing – review & editing, Methodology, Investigation, Formal analysis, Conceptualization. Richard Kronmal: Writing – review & editing, Methodology, Investigation, Formal analysis, Conceptualization. Amy H. Schultz: Writing – review & editing, Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of Competing Interest

The authors declare no conflicts of interest.