Trends in Preterm Delivery among Singleton Gestations with Critical Congenital Heart Disease

Abstract

Objective

To examine state-wide population trends in preterm delivery of children with critical congenital heart disease (CHD)over an 18-year period. We hypothesized that, coincident with early advancements in prenatal diagnosis, preterm delivery initially increased compared with the general population, and more recently has declined.

Study design

Data from the Texas Public Use Data File 1999–2016 was used to evaluate annual percent preterm delivery (<37 weeks) in critical CHD (diagnoses requiring intervention <1 year of age). We first evaluated for pattern change over time using joinpoint segmented regression. Trends in preterm delivery were then compared with all Texas livebirths. We then compared trends examining sociodemographic covariates including race/ethnicity, sex, and neighborhood poverty levels.

Results

Of 7,146 births with critical CHD, 1,339 (18.7%) were delivered preterm. The rate of preterm birth increased from 1999–2004 (mean increase 1.69% per year) then decreased between 2005–2016 (mean decline −0.41% per year). This represented a faster increase and then a similar decline to that noted in the general population. While the highest proportion of preterm births occurred in newborns of Hispanic ethnicity and Non-Hispanic Black race, newborns with higher neighborhood poverty level had the most rapidly increasing rate of preterm delivery in the first era, and only a plateau rather than decrease in the latter era.

Conclusion

Rates of preterm birth for newborns with critical CHD in Texas first were rising rapidly, then have been declining since 2005.

Prenatal diagnosis of congenital heart disease (CHD) has been associated with improved pre-operative clinical status,^1–4 neurodevelopmental outcomes,⁵ and reduced mortality.^1,6–8 This likely results from coordination of resources and in utero transport of the fetus with CHD to appropriate care facilities with rapid delivery of postnatal therapy.⁹ While benefits of prenatal diagnosis of CHD have been reported, many studies show no benefit^2,4,10–15 and some studies suggest that infants with a prenatal diagnosis have increased mortality.^9,16–18

One possible contributor to the apparent lack of benefit of prenatal diagnosis is that infants with a prenatal diagnosis of CHD are more likely to be delivered at earlier gestational ages than those without a prenatal diagnosis.^1,5,19–21 Earlier delivery may result from institutions performing induction of labor or planned cesarean section before natural labor. This may reduce the beneficial impact of prenatal diagnosis, because preterm infants with CHD have worse outcomes than those born term.^22–26, a study in the state of Utah demonstrated that even early term birth (37–38 weeks) is associated with increased mortality compared with later term birth.²⁷ A single center study identified that preterm and early term birth in infants with critical CHD (CCHD-lesions that require intervention early in life) were associated with worse outcomes than later term birth.²⁸ Since then, multicenter studies have supported these findings.^24,29,30 These studies emphasize that that every week in utero is important and advocate for efforts to extend the duration of pregnancy in situations where a prenatal diagnosis of CCHD is made with no clear indication for early delivery.³¹ Such considerations hold true for the general population as the American College of Obstetricians and Gynecologists has advocated against elective deliveries under 39 weeks of gestation.^32,33

We sought to determine if the proportion of preterm births in children diagnosed with CCHD changed between 1999–2016 in a large, diverse population-based cohort. We hypothesized that the proportion of preterm CCHD births would increase initially, coinciding with improving rates of prenatal diagnosis in the 1990s and 2000s, but would then decrease, due to changing medical management after publications about risks of preterm and early term birth in CCHD.^34,35 We also investigated potential influence of race/ethnicity, sex, and neighborhood poverty level on trends.

Methods:

We conducted a retrospective, population-based study using the 1999–2016 Texas Public Use Inpatient Data File (TPUIDF). The TPUIDF is a non-sampled, population-based, administrative database that includes all state licensed hospitals (except for a small proportion that are statutorily exempt from the reporting requirement before 2015, accounting for an estimated <3% of state birth hospitalizations).³⁶ The Texas Department of State Health Services validates the data through automated auditing and verification. Up to 25 discharge diagnoses and 25 procedures are coded with the use of the International Classification of Diseases (ICD), Ninth Revision, Clinical Modification (ICD-9-CM) through September 30, 2015 and ICD-10-CM codes from October 1, 2015 onwards. The Institutional Review Board of the Baylor College of Medicine determined that this activity did not constitute human subjects research, and therefore was exempt from review.

We included all birth hospitalizations with any diagnosis code indicative of CCHD. A birth hospitalization was defined as a discharge with either a diagnostic code for live-born infant or a source of admission coded as a newborn from inside the hospital. We defined CCHD as a CHD likely to require a cardiac intervention within the first year of life (Table 1; available at www.jpeds.com). Given the heterogeneity in presentation and diagnosis of aortic arch obstruction, only infants with codes for both arch obstruction and arch intervention during birth hospitalization were included (Table 2; available at www.jpeds.com). Exclusion criteria included a diagnosis of multiple gestation birth, unknown sex, genetic syndrome or extracardiac birth defects (Table 3; available at www.jpeds.com).

Table 1.

List of included diagnoses by ICD-9-CM and ICD-10-CM codes.

Diagnosis	ICD-9-CM	ICD-10-CM
Atrioventricular septal defect	745.60, 745.61, 745.69	Q21.2
Coarctation/other arch obstruction	747.1, 747.10, 747.11, 747.21, 747.22	Q25.1, Q25.21, Q25.29, Q25.41, Q25.42
Congenital aortic stenosis	746.3	Q23.0, I35.0*
Double outlet left ventricle		Q20.2
Double outlet right ventricle	745.11	Q20.1
Ebstein anomaly	746.2	Q22.5
Hypoplastic left heart syndrome	746.7	Q23.4
Mitral stenosis	746.5	Q23.2, I34.2
Other single ventricle	745.3, 745.7	Q20.4, Q22.6
Pulmonary atresia	746.01	Q22.0, Q25.5
Tetralogy of fallot	745.2	Q21.3
Total anomalous venous return	747.41	Q26.2
Transposition of the great arteries	745.10, 745.12	Q20.3, Q20.5
Tricuspid atresia	746.1	Q22.4
Truncus arteriosus	745.0	Q20.0

^*Despite being a code for adult-onset aortic stenosis, I35.0 was included to account for common miscoding of this congenital lesion and no adults in this cohort.

Table 2.

Procedure codes used to define arch interventions

Surgical

38.04, 38.05, 38.34, 38.35, 38.44, 38.45, 38.64, 38.65, 39.56, 39.57, 39.58, 39.59, 39.73, 02BX0ZX, 02BX0ZZ, 02QX0ZZ, 02QW0ZZ, 02RX0*, 02UW0*, 02UX0*, 02RW0*, 027X0*, 027W0*, 02BW0*, 02BX0*, 021W08B, 021W08D, 021W09B, 021W09D, 021W0AB, 021W0AD, 021W0JB, 021W0JD, 021W0KB, 021W0KD, 021W0ZB, 021W0ZD, 021X08B, 021X08D, 021X09B, 021X09D, 021X0AB, 021X0AD, 021X0JB, 021X0JD, 021X0KB, 021X0KD, 021X0ZB, 021X0ZD

Catheter-based

395.0, 399.0, 027W3*, 027W4*, 027X3*, 027X4*

^*indicates all billable codes under this stem.

Table 3.

List of excluded genetic syndromes and birth defects

ICD-9-CM	ICD-10-CM
275.1, 279.11, 553.3, 747.81, 270.*, 271.*, 272.*, 277.*, 740.*, 741.*, 742.*, 743.*, 744.*, 748.*, 749.*, 750.1–750.9, 751.*, 752.*, 753.*, 754.*, 755.*, 756.*, 757.*, 758.*, 759.*	D82.1, K44.9, Q28.2, Q28.3, E70*, E71.*, E72.*, E74.*, E75.*, E76.*, E77.*, E78.*, E79.*, E84.*, E80.0, E80.1, E80.20, E80.21, E80.29, E80.3, E80.5, E83.01, Q0.*, Q1.*, Q3.*, Q4.*, Q5.*, Q6.*, Q7.*, Q8.*, Q9.*

^*indicates all billable codes under this stem.

The primary outcome was preterm birth, defined as birth less than 37 weeks gestation using ICD-9-CM and ICD-10-CM codes (Table 4; available at www.jpeds.com). ICD-9-CM codes do not distinguish between early and late term births, so this could not be investigated. We examined covariates to define the population and evaluate for confounders, including sex, race/ethnicity, and insurance. Public insurance included federal programs apart from military insurance. To account for socioeconomic status, percent of the population living in poverty in the discharge home zip code was determined using publicly available United States (US) Census data (2000 Census data for 1999–2006, 2010 Census data for 2007–2011, American Community Survey for 2012–2016).³⁷ Poverty levels were dichotomized between <20% of a zip code living in poverty and ≥ 20%, as <20% defines a poverty area by the US Census.³⁸

Table 4.

Lists of ICD-9-CM and ICD-10CM codes used to define preterm birth

ICD-9-CM	ICD-10-CM
362.20–362.27, 765.0x, 765.1x, 765.20–765.28	H35.1x, P07.2x, P07.3x, O.061x

x indicates all billable codes under this stem.

Statistical Analyses

For validation of the dataset, we first compared the number of annual live births in our dataset to those published for Texas by the National Center for Health Statistics (NCHS).³⁹ Of note, some difference was expected as the NCHS dataset includes information on home births. For further validation, we then compared the total live births in our dataset with those in the Texas Birth Defects Registry (TBDR) for a sample of lesions: hypoplastic left heart syndrome (HLHS), truncus arteriosus, and tetralogy of Fallot (TOF). Some difference was also expected, as our dataset only included hospitalizations during which a diagnosis was made before the end of the newborn hospitalization, whereas the TBDR captures diagnoses made through the first year of life. Then, after applying inclusion and exclusion criteria, discharge characteristics and diagnoses were compared between infants with and without preterm birth using generalized estimating equations (GEE) accounting for clustering by hospital using log-binomial regression to estimate relative risk.

To evaluate change in percent of preterm births over time, joinpoint segmented regression was performed to identify any temporal changes in the data. This delineated a significant change after the year 2004. First, crude rates and rates of change of preterm delivery before and after 2004 were reported in the group with CCHD and compared with the NCHS data for all Texas births using linear regression, stratified by the 2 eras. Then analyses were performed using GEE with log-binomial regression employing autoregression and hierarchical modeling accounting for clustering by hospital to calculate relative risk per year in the two eras. Models were created (Model 1) for the entire cohort, accounting only for time (using year, a binary division <2004 and >2004, and an interaction of the two to allow for separate slopes) and delivery hospital. Then multivariable analysis was performed accounting for time, hospital, and demographics (Model 2). The analysis was stratified by race/ethnicity and then by poverty in secondary models to allow better estimates of changes in preterm birth within subgroups (Models 3 and 4, respectively).

Sub-analysis was then performed limiting the study population to CCHD lesions likely to be prenatally diagnosed: tricuspid atresia, HLHS, pulmonary atresia, other single ventricle lesions, and combinations of these lesions.⁴⁰ We performed the same analyses with this sub-population as the entire cohort. Statistical analysis was performed on SAS version 9.4 (SAS Institute, Cary, NC).

Results:

Annual live births in the TPUIDF ranged from a minimum of 326,348 in 1999 to a maximum of 400,443 in 2015. There were 3.4% fewer live births in TPUIDF relative to live births reported by the NCHS (6,707,262 versus 6,945,023, respectively). When comparing live births with selected CCHD in the TPUIDF to that of the TBDR, the case counts were HLHS: 1248 versus 1336, truncus arteriosus: 454 versus 519, TOF: 2220 versus 2298; in the TPUIDF versus the TBDR, respectively.

A total of 7,146 total births fulfilled CCHD inclusion criteria (Figure 1; available at www.jpeds.com), of which 1,339 (18.7%) were born preterm (Table 5). Variables associated with higher percentage of preterm birth included Hispanic ethnicity, non-Hispanic Black race, living in a high poverty neighborhood, and public health insurance. We observed considerable variation in preterm prevalence by type of CCHD, from 11.2% to 31.6%. Transposition of the great arteries was used as the reference due to having the lowest defect-specific percentage of premature births and due to its historically poor rates of prenatal diagnosis.⁴¹

Figure 1, online. Study population.

Of the newborns in the Texas Public Use Inpatient Data File from 1999–2016, births of multiple gestation, those without CCHD, and those with birth defects or genetic syndromes were excluded. A total of 7,146 newborns were included in the primary analysis.

Table 5.

Univariable Analysis of Characteristics Associated with Preterm Delivery Among Infants with CCHD

Characteristic	All n	Preterm n (%)	RR (95% CI)	p-value
CCHD	7146	1339 (18.7)
Sex				0.077
Male	4188	756 (18.1)	reference
Female	2958	583 (19.7)	1.09 (0.99, 1.20)
Race/Ethnicity
Non-Hispanic White	2568	434 (16.9)	reference
Hispanic	3072	593 (19.3)	1.14 (1.02, 1.28)	0.020
Non-Hispanic Black	667	162 (24.3)	1.44 (1.23, 1.68)	<0.001
Other/Missing	839	150 (17.9)	1.06 (0.89, 1.25)	0.513
Poverty Level of zip code
< 20%	4668	821 (17.6)	reference
≥ 20%	2233	475 (21.3)	1.21 (1.09, 1.34)	<0.001
Insurance
Private	2685	477 (17.8)	reference
Medicare/other federal	3534	699 (19.8)	1.11 (1.00, 1.24)	0.045
Self-pay	284	46 (16.2)	0.91 (0.69, 1.20)	0.513
VA/CHAMPUS	114	13 (11.4)	0.64 (0.38, 1.08)	0.094
Other	99	21 (21.2)	1.19 (0.81, 1.76)	0.371
CCHD Lesion
Transposition of the great arteries	961	108 (11.2)	reference
Arch obstruction and VSD	164	21 (12.8)	1.14 (0.74, 1.76)	0.558
Hypoplastic left heart syndrome	899	121 (13.5)	1.20 (0.94, 1.53)	0.146
Total anomalous venous return	380	58 (15.3)	1.36 (1.01, 1.83)	0.043
Combination	540	87 (16.1)	1.43 (1.10, 1.86)	0.007
Other single ventricle	143	24 (16.8)	1.49 (1.00, 2.24)	0.053
CCTGA	29	5 (17.2)	1.53 (0.68, 3.47)	0.305
Arch obstruction, isolated	174	31 (17.8)	1.59 (1.10, 2.28)	0.013
Tricuspid atresia	428	81 (18.9)	1.68 (1.29, 2.19)	<0.001
Double outlet right ventricle	382	75 (19.6)	1.75 (1.33, 2.29)	<0.001
Pulmonary atresia-VSD	175	36 (20.6)	1.83 (1.30, 2.57)	<0.001
Congenital aortic stenosis	487	103 (21.1)	1.88 (1.47, 2.41)	<0.001
Ebstein anomaly	303	66 (21.8)	1.94 (1.47, 2.56)	<0.001
Tetralogy of Fallot	1268	298 (23.5)	2.09 (1.71, 2.56)	<0.001
Pulmonary atresia-IVS	286	69 (24.1)	2.15 (1.64, 2.82)	<0.001
Truncus arteriosus	233	63 (27.0)	2.41 (1.83, 3.17)	<0.001
Atrioventricular septal defect	294	93 (31.6)	2.81 (2.20, 3.59)	<0.001

For CCHD lesions, categories are mutually exclusive. When >1 critical CHD was present, lesion is listed as “Combination.” RR: Relative risk, CCHD: critical congenital heart disease, VA: U.S. Department of Veterans Affairs, CHAMPUS: Civilian Health and Medical Program of the Uniformed Services, VSD: ventricular septal defect, IVS: intact ventricular septum, CCTGA: congenitally corrected transposition of the great arteries.

Based on the yearly percentage of preterm birth, joinpoint segmented regression models identified a change in trend after 2003 for infants overall in Texas, and after 2004 for infants with CCHD (Figure 2). In both populations the annual percent of preterm births initially increased, with mean change of +1.69 percent/year (95%CI +0.95 to +2.44) for the CCHD cohort and +0.31 percent/year (95%CI +0.25 to +0.38) for the total population (CCHD versus population change p<0.001). Following 2004, the annual percent of preterm births for infants with CCHD and the general population decreased, with a mean change of −0.41 percent/year (95%CI −0.73 to −0.09) in CCHD, and −0.16 percent/year (96%CI −0.18 to −0.14) in the general population (CCHD versus population P = .123).

Figure 2. Annual percentage of infants born preterm in Texas with CCHD, 1999–2016.

The dotted line represents joinpoint segmented regression for newborns born with CCHD and the solid line represents the percentage preterm births for all newborns. A point in which there is a change in the slope of the data for the population with CCHD is noted in 2004, with the slope of the earlier era being +1.69 percent/year and the slope of the later era being −0.31 percent/year.

When accounting for hospital clustering, we observed an increasing relative risk of preterm births with CCHD from 1999–2004 (era 1, RR per year=1.10, Table 6, Model 1) followed by a decreasing relative risk in the period after 2004 (era 2, RR=0.98, p-value comparing both eras=<0.001). These findings were consistent after additional adjustment for sociodemographic variables (Table 6, Model 2). In the third model (Table 6, Model 3), to account for observed interactions between race/ethnicity and with time, multivariable analysis was performed while stratifying by race/ethnicity. The overall pattern of an increase in the first era and decrease in the second era were present across all race/ethnicities. When comparing race/ethnicity within each era, there were no significant differences in rate of change of preterm delivery. In the 4^th model (Table 6, Model 4 and Figure 3), to account for observed interactions between neighborhood poverty level and time, multivariable analysis was performed while stratifying by poverty level. This demonstrated that those living in a high poverty area had a faster increase in risk of preterm delivery in the first era and a plateau in the second era.

Table 6.

Multivariable Models of Relative Risk per year of Preterm Delivery Among Infants with CCHD in Era 1 and Era 2

Model	RR per year (95%CI), 1999–2004 (Era 1)	RR per year (95%CI), 2005–2016 (Era 2)	Interaction p-value*
Model 1: Adjusted only for hospital	1.10 (1.04–1.16)	0.98 (0.96–0.99)	<0.001
Model 2: Adjusting for hospital, poverty, race/ethnicity, and sex	1.10 (1.04–1.16)	0.98 (0.96–0.99)	<0.001
Model 3: Adjusting for hospital, poverty, race/ethnicity, and significant interactions, stratified by race/ethnicity**
Non-Hispanic White	1.07 (0.98–1.16)	0.98 (0.94–1.01)	0.070
Hispanic	1.09 (1.01–1.19)	0.98 (0.96–1.01)	0.009
Non-Hispanic Black	1.14 (0.99–1.31)	0.96 (0.93–1.00)	0.026
Other	1.15 (0.93–1.42)	0.98 (0.92–1.04)	0.125
Model 4: Adjusting for hospital, poverty, race/ethnicity, and significant interactions, stratified by poverty**
Low poverty	1.06 (0.99–1.13)†	0.97 (0.95–0.99)	0.018
High poverty	1.23 (1.08–1.39)	1.00 (0.93–1.53)	<0.001

^*Compares the RR per year of era 1 and era 2.

^**Interaction noted between race/ethnicity and era and poverty and era.

^†Denotes significant differences within the era.

Figure 3. Annual percentage of infants born preterm in Texas with CCHD stratified by poverty level, 1999–2016.

The dashed line represents joinpoint segmented regression for infants born with CCHD in low poverty areas. The solid line represents joinpoint segmented regression for infants born with CCHD in high poverty areas. Those living in a low poverty area had an increase in preterm delivery in the first era and decrease in the second. Those living in a high poverty area had a faster increase in preterm delivery in the first era, and only a plateau, but not a decline in the second era.

There were 2,349 live births and 401 (17.1%) preterm births in the sub-population of CCHD most commonly prenatally diagnosed (Table 7; available at www.jpeds.com). Variables associated with higher percentage of preterm birth included Hispanic ethnicity, non-Hispanic Black race, and living in a high poverty neighborhood. The percent of preterm births over time for this population is shown in Figure 4 (available at www.jpeds.com) and compared with overall live births. Joinpoint segmented regression identified a change in trend after 2004. The annual percent of preterm births initially increased with a mean change of +1.95 percent/year (95%CI 0.71 to 3.19), which was similar to the overall change rate in CHD (p=0.729), and faster than the general population (p=0.010). Following 2004, the annual percent of preterm births had a similar change to that noted in the overall CHD and population groups (p=0.845 and p=0.535 respectively), with a mean change of −0.48 percent/year (95% CI −1.00 to 0.03).

Table VII.

Univariable analysis of characteristics associated with preterm delivery among the subgroup of infants with CCHD most likely to be prenatally diagnosed

Characteristics	All (n)	Preterm, n (%)	RR (95% CI)	P value
CCHD subgroup	2349	401 (17.1)
Sex				.123
Female	985	182 (18.5)	Reference
Male	1364	219 (16.1)	1.15 (0.96–1.38)
Race/ethnicity
Non-Hispanic white	801	110 (13.7)	Reference
Hispanic	1027	190 (18.5)	1.35 (1.09–1.67)	.007
Non-Hispanic black	254	56 (22.0)	1.61 (1.20–2.14)	.001
Other	267	45 (16.9)	1.23 (0.89–1.69)	.207
Poverty level of zip code
<20%	1470	215 (14.6)	Reference
≥20%	767	168 (21.9)	1.50 (1.25–1.80)	<.001
Insurance
Private	822	125 (15.2)	Reference
Medicare/other federal	1199	214 (17.8)	1.17 (0.96–1.44)	.120
Self-pay	93	18 (19.4)	1.27 (0.82–1.99)	.288
VA/CHAMPUS	100	64 (10.0)	0.66 (0.26–1.69)	.384
Other	40	10 (25.0)	1.64 (0.94–2.88)	.082
CCHD lesion
HLHS	899	121 (13.5)
Other single ventricle	143	24 (16.8)
Complex combination	418	70 (16.7)
Tricuspid atresia	428	81 (18.9)
Pulmonary atresia-VSD	175	36 (20.6)
Pulmonary atresia-IVS	286	69 (24.1)

CHAMPUS, Civilian Health and Medical Program of the Uniformed Services; IVS, intact ventricular septum; VA, US Department of Veterans Affairs; VSD, ventricular septal defect.

For CCHD lesions, categories are mutually exclusive. When >1 CCHD was present, lesion is listed as a complex combination.

Figure 4, online. Annual percentage of infants born preterm in Texas with CCHD most likely to be prenatally diagnosed, 1999–2016.

The dotted line represents joinpoint segmented regression for infants born with CCHD most likely to be prenatally diagnosed and the solid line represents the percentage preterm births for all infants. A point in which there is a change in the slope of the data for this subgroup of CCHD is noted in 2004, with the slope of the earlier era being +1.95 percent/year and the slope of the later era being −0.48 percent/year.

When accounting for hospital clustering, there was an increase in proportion of preterm births with CCHD most likely to be diagnosed from 1999–2004 (RR per year=1.14, Table 8, Model 1 (available at www.jpeds.com)) followed by a plateau in the period after 2004 (RR per year=0.97, p-value comparing the RR of both eras=0.002). This pattern was consistent in the second multivariable model that adjusted for demographics (Table 8, Model 2). When stratifying by race/ethnicity (Table 8, Model 3), the overall patterns of an increase in the first era and plateau in the second era were similar between race/ethnicities but no significant difference in the relative risks per year between eras noted for any race/ethnicity. Stratifying by poverty level (Table 8, Model 4), demonstrated that those living in a high poverty area only had a plateau in the later era while those in a lower poverty area noted a decline in preterm delivery rates.

Table 8.

Multivariable Models of Characteristics Associated with Preterm Delivery Among the Subgroup of Infants with CCHD Most Likely to be Prenatally Diagnosed

Model	RR per year (95%CI), 1999–2004 (Era 1)	RR per year (95%CI), 2005–2016 (Era 2)	Interaction p-value*
Model 1: Adjusted only for hospital	1.14 (1.04–1.16)	0.97 (0.94–1.01)	0.002
Model 2: Adjusting for hospital, poverty, race/ethnicity, and sex	1.12 (1.01–1.25)	0.98 (0.95–1.01)	0.012
Model 3: Adjusting for poverty, race/ethnicity, and significant interactions, stratified by race/ethnicity**
Non-Hispanic White	1.08 (0.93–1.26)	0.95 (0.88–1.03)	0.136
Hispanic	1.15 (0.96–1.38)	0.99 (0.94–1.04)	0.088
Non-Hispanic Black	1.10 (0.84–1.44)	0.99 (0.95–1.05)	0.498
Others	1.22 (0.79–1.90)	0.98 (0.89–1.08)	0.337
Model 4: Adjusting for hospital, poverty, race/ethnicity, and significant interactions, stratified by poverty**
Low poverty	1.10 (0.99–1.21)	0.95 (0.91–0.99)	0.021
High poverty	1.24 (0.94–1.63)	1.01 (0.97–1.05)	0.098

^*Compares the RR per year of era 1 and era 2.

^**Interaction noted between race/ethnicity and era

Discussion:

Although prenatal diagnosis of CCHD is associated with improvements in pre-operative clinical status and neurodevelopmental outcomes, it has also been associated with earlier gestational age at birth, which is associated with worse outcomes.^{1,2,5,19–23,28–30} The dilemma involves the interplay between the benefits of prenatal diagnosis and the earlier gestational age at birth. Although we cannot directly evaluate the reasons for the change of trend in the rate of preterm births, we hypothesize that the earlier increase was due to increasing prenatal CCHD diagnosis.⁹ Later, a change in practice may have occurred in response to the growing body of literature associating earlier gestational age at birth with worse outcomes.^{22,23,28–30} Other factors that may have influenced the gestational age at birth during the time period of our study include literature describing that progestogens in select populations will decrease the rate of preterm delivery in at-risk women.^43–46 Adoption of this agent over the time period could have further influenced gestational age at birth.

This study noted that Hispanic ethnicity and non-Hispanic Black race were associated with a higher percentage of preterm birth of newborns with CCHD, a finding previously noted in the general population when comparing non-Hispanic blacks with non-Hispanic whites.^47–51 Though maternal risk factors were not part of our assessment, studies have noted racial disparities persist even when accounting for maternal risk factors for preterm birth.⁴⁷ Despite the association of Hispanic ethnicity and non-Hispanic black race with a higher percentage of preterm birth, the decreasing rate of premature birth in the recent era suggests improvement in preterm delivery in these populations.

Our ability to speculate on why trends in preterm delivery are modified by poverty level are limited. However, published data demonstrate that large city and urban areas showed a decrease in preterm births after 2005 whereas rural areas that were more socioeconomically isolated only exhibited a plateau in preterm delivery rates.⁵² It is possible that changes to public policy and clinical practice have had less impact on populations in poverty.

There were inherent limitations in our analysis. The specific type of CCHD was not included in our statistical models for two reasons. First, rates of prenatal diagnosis of CCHD varies by lesion.⁵³ As prenatal diagnosis has improved, the changes have been discrepant by lesion, with lesions that were more challenging to detect in the past now being recognized more often. Therefore, adjusting for lesion potentially would have adjusted for a factor in the mechanistic pathway. Second, our overall aim was to evaluate preterm delivery rates in the Texas population with CCHD. Accounting for individual types of CCHD and potential interactions would significantly limit statistical power and fail to answer the primary question.

Various factors likely influence the health outcomes of Hispanics in the US, including country of origin/cultural heritage and extent of acculturation to mainstream US culture.^54,55 Information regarding Hispanic ethnic subgroups and measures of acculturation were not available. In addition, the generalizability of this study may be limited. In 2016, Texas was the state with the lowest rate of first trimester initiation of prenatal care.⁵⁰ Additionally, Texas is a large state with many rural counties and the percentage of hospitals in rural counties with obstetrical services has decreased during the time period of our study.⁵⁶ Issues with access to care in rural counties is associated with increased preterm birth.⁵⁷

The database in this study involves passive surveillance using administrative codes without active case validation. Also, the analysis assumes that newborns without ICD codes for prematurity are not preterm at birth. Only a subset of infants born preterm are coded as preterm by ICD.⁵⁸ Previous results have demonstrated that the sensitivity of ICD-9-CM codes is low, particularly for 35 or 36 weeks’ gestation birth.^59,60 Although the implication is that the incomplete sensitivity of ICD codes for preterm births would cause our study to underestimate preterm birth, we believe this does not bias our study given that our primary outcome are changes in rates of preterm birth over time. Of note, when comparing 3 lesion-specific studies from the TBDR that include data on preterm delivery, the percent preterm in the TPUIDF for the same periods and lesions closely approximates those reported (Lupo et al reported 26% in Ebstein anomaly versus 24% in the TPUIDF; Morris et al reported 13% in non-syndromic HLHS versus 14% in the TPUIDF; Lara et al reported 12% in non-syndromic transposition of the great arteries versus 12% in the TPUIDF).^9,15,61 Additionally, the retrospective nature of the study does not allow the authors to draw conclusions on causality in the findings. Finally, we did not account for the clustering of births of the same mother (although multiples were excluded).

The rate of preterm birth in CCHD increased from 1999–2004 but decreased from 2005–2016. These patterns were similar in the population of CCHD most likely to be prenatally diagnosed. The change between these two periods may indicate a change in clinical practice in response to recommendations that infants with CCHD have worse outcomes when born preterm and early term.

List of abbreviations included >3 times:

CHD

Congenital heart disease

CCHD

Critical congenital heart disease

TPUIDF

Texas Public Use Inpatient Data File

ICD-CM

International Classification of Diseases, Clinical Modification

US

United States

TBDR

Texas Birth Defects Registry

HLHS

Hypoplastic left heart syndrome

TOF

Tetralogy of Fallot

GEE

Generalized estimating equation