Outcomes in pregnant patients with congenital heart disease by rurality

Study design and population

This is a retrospective population-based study, using the Healthcare Cost and Utilization Project-National Inpatient Sample (HCUP-NIS) database between 2016 and 2021. The HCUP-NIS is sponsored by the Agency for Healthcare Research and Quality and contains data about hospital utilization, outcomes, and costs at a national level. The HCUP-NIS is the largest inpatient database in the country and represents 96 % of the American population, with over 9 million admissions annually. The data is publicly available and anonymized. Hence, institutional review board approval was not required based on the guidelines of the Tri-Council Policy Statement (2018).

The NIS Core File was used to identify patient-level characteristics such as age at admission, median household income, primary insurance, and location of patient residence. For this analysis, we included hospitalized pregnant patients ≥18 years of age with a diagnosis of CHD during the study period. CHD was stratified by disease complexity based on the 2018 AHA/ACC Guideline for the Management of Adults With Congenital Heart Disease [3]. The list of ICD-10 codes included, stratified by disease complexity (complex, moderate and simple complexity) is shown in Supplementary Table S1. Patients who had a diagnosis of abortion or ectopic pregnancy or who were missing data on rural or urban residence were excluded. As per data available in HCUP-NIS, rural residence was defined as living in micropolitan counties with populations between 10,000 and 50,000 or in areas that are neither micropolitan nor metropolitan, based on the U.S. Department of Health and Human Services definition. All other patients were considered to reside in urban settings [9]. The rurality of the hospital at which care was accessed was abstracted separately.

Study outcomes

The primary outcomes for our study were three distinct measures: (1) major adverse cardiovascular events (MACE), (2) adverse pregnancy outcomes (APO), and (3) neonatal outcomes. MACE was defined as the occurrence of any event associated with an ICD-10 code for acute heart failure, arrhythmia, cerebrovascular accident, embolic events (cerebrovascular, pulmonary embolism), and unspecified cardiovascular complications of pregnancy and hospital death. APO was defined as the occurrence of any event with an ICD-10 code associated with gestational hypertension, preeclampsia, placental complications (including antepartum bleeding), fetal complications, preterm delivery, anesthetic complications, postpartum hemorrhage, complication with amniotic fluid, gestational diabetes, liver disorder in pregnancy, or antepartum/postpartum infection. Neonatal outcomes were defined as the occurrence of any event with an ICD-10 code associated with small for gestational age, intrauterine fetal demise, neonatal birth trauma, and congenital disorder. The complete list of ICD-10 codes can be found in Supplementary Table 1.

Statistical analysis

Descriptive statistics for the study population by rurality status were compared using counts and percentages for categorical variables, means and standard deviations for normally distributed continuous variables, and medians and interquartile ranges for non-normally distributed continuous variables. To compare differences between groups, Pearson’s χ² tests, two-sample t-tests, and Wilcoxon rank-sum tests were used, respectively.

Univariate logistic regression was applied to determine the odds of MACE, APOs, and neonatal outcomes by rurality. Multivariable logistic regression was applied to determine odds of MACE, APOs, and neonatal outcomes by rurality adjusting for age, race, income, insurance payer, smoking, alcohol use, drug use in pregnancy, diabetes, obesity, pulmonary disease, liver disease and uterine fibroids [10]. Each regression was stratified by CHD complexity, and a sensitivity analysis was conducted to where moderate and complex CHD were grouped together due to the small sample size in each of these two complexity categories. A univariate analysis was conducted to assess the interaction term between rurality and MACE, rurality and APO, as well as rurality and neonatal outcomes.

All analyses used survey procedures and SAS 9.4 to account for the complex design of HCUP-NIS database. All proportions and odds ratios were weighted to reflect national estimates. two-sided p values of <0.05 were considered statistically significant.

Baseline characteristics

A total of 4859 patients met inclusion criteria, representing 24,295 patients nationally. Of those, 691 patients (14 %, n=3455 patients represented nationally) were rural residents. Compared to patients residing in urban settings, residents of rural areas were slightly younger (27.1 vs. 29.5 years old, p<0.001), more likely to be of White race (84.1 % vs. 62.3 %, p<0.001), and tended to have lower income and public insurance. There was no difference in the rates of vaginal or c-section deliveries between rural and urban areas (vaginal delivery 45 % vs. 46 %, p=0.75 and, c-section 33 % vs. 32 %, p=0.53, for those residing in rural vs. urban areas respectively). Of note, the database was missing information on mode of delivery for 22 % of deliveries. Of the patients who accessed care at an urban hospital, 85.4 % were living in an urban location while 10.4 % were from a rural location (Table 1). Of all patients living in a rural setting in this cohort, only 26.8 % (n=185/691) accessed care at a rural hospital.

In terms of CHD complexity among rural patients, 48.1 % had simple CHD, while 54.2 % of urban patients were in this category (p=0.003). Similarly, 45.3 % of rural patients had moderate CHD complexity, compared to 40.1 % in urban areas (p=0.01). There was no significant difference between rural and urban patients in the prevalence of complex CHD (6.7 % vs. 5.7 %, respectively, p=0.29). Patients with CHD residing in rural areas were more likely to have a history of smoking (9.9 % vs. 5.1 %, p<0.001), but were less likely to have gynecologic disorders at baseline. Otherwise, chronic comorbidities were similar between the two groups (Table 1).

MACE

Rural patients tended to have a higher rate of MACE compared to urban patients, although this was not statistically significant (rural: 44.3 %, urban: 40.3 %, p=0.05) (Table 1). Similar results were seen after adjusting for relevant covariates (adjusted-OR 1.17; 95 %-CI 0.98–1.40; p=0.09) (Table 2, Figure 1). When stratified by complexity of CHD, no statistically significant difference in odds of MACE was observed (P_interaction=0.66) (Table 2). In terms of specific adverse cardiac events, patients residing in rural areas were more likely to be diagnosed with atrial arrhythmias (adjusted-OR 1.26; 95 %-CI 1.05–1.52; p=0.01), but otherwise, there were no other significant differences between groups (Table 3). Among rural residents with CHD, those that accessed care at a rural hospital had lower odds of MACE (adjusted-OR 0.54; 95 %-CI 0.38–0.79; p=0.001); however, nearly half (47 %) of those receiving care at rural hospitals had simple CHD, suggesting a potential referral bias of more complex cases to urban centers (Table 4).

Figure 1:

Forest plot of odds of major cardiovascular event, adverse pregnancy outcomes, and neonatal outcomes. APO, adverse pregnancy outcomes; MACE, major adverse cardiovascular events.

APO

Rural patients had lower rate of APO compared to urban patients (rural: 46.5 %, urban: 51.5 %, p=0.02) (Table 1). These findings were consistent after adjusting for relevant covariates (adjusted-OR 0.76; 95 %-CI 0.63–0.91; p=0.003) (Table 2, Figure 1) and remained consistent in a sensitivity analysis restricted to rural residents comparing outcomes by hospital location (Table 4). When stratified by CHD complexity, rural patients with simple disease had lower odds of APO (adjusted OR 0.70, 95 % CI 0.54–0.91, p=0.008), whereas this association was not statistically significant in patients with moderate or complex CHD (Table 2). However, there was no significant effect modification by CHD complexity (P_interaction=0.60). Regarding specific APOs, compared to patients residing in urban settings, residents of rural areas were less likely to have infection complications (adjusted-OR 0.58; 95 %-CI 0.40–0.86; p=0.006) (Table 3).

Neonatal outcomes

There was no difference in adverse neonatal outcomes between CHD patients residing in rural compared to urban areas (10.9 % and 12.5 %, respectively, p=0.25) (Table 1). Similar results were observed after adjusting for relevant covariates (adjusted-OR 0.78; 95 %-CI 0.59–1.03; p=0.082) (Table 2, Figure 1). There was no evidence of effect modification by CHD complexity (P_interaction=0.75) and no difference in specific adverse neonatal outcomes by rurality (Table 3). However, in a sensitivity analysis restricted to rural residents, those who accessed care at a rural hospital had lower odds of adverse neonatal outcomes (adjusted-OR 0.31; 95 %-CI 0.14–0.67; p=0.003; Table 4).

Sensitivity analysis

A sensitivity analysis in which moderate and complex CHD was grouped together also showed no statistically significant difference in odds of MACE, APO and neonatal outcomes amongst these patients. The adjusted odds of MACE in patients with either moderate or complex CHD was 1.148 (95 %-Cl 0.894–1.473; p=0.28). The adjusted odds of APO in patients with either moderate or complex CHD was 0.782 (95 %-Cl 0.610–1.002; p=0.051). The adjusted odds of neonatal in patients with either moderate or complex CHD was 0.854 (95 %-Cl 0.591–1.235; p=0.40).

Strengths and limitations

The use of the HCUP-NIS database imposes certain constraints on our study. Its focus on inpatient care does not allow for the incorporation of outpatient settings, post-discharge complications or longitudinal data. The lack of detailed clinical information in the database may lead to challenges in controlling for confounding variables, potentially affecting the validity of the conclusions drawn from the analysis. Limitations in the clinical information available in the database such as lack of granular anatomic and physiological details, also prevent the use of other cardiac risk categories (such as mWHO or CARPREG) in the pregnant population. Additionally, the absence of information on patient outcomes or mortality post-discharge restricts our ability to investigate antenatally or post-partum outcomes. Also, the inability of the NIS to link maternal and infant records may limit the accuracy of capturing neonatal outcomes, although stillbirth is expected to be reliably coded on the maternal claim. Furthermore, ICD-10 diagnostic codes do not differentiate by disease severity or repaired status of congenital heart disease, and some of the patients may have been misclassified into the wrong complexity categories. Also, we are unable to stratify by state, assumption of independence of each datapoint may be questionable as we don’t have information on parity. The changes in coding practices and data collection over time may hinder the validity of trend analyses when using HCUP-NIS data across multiple years. Finally, dichotomizing rurality may not fully capture the complexity of the rural-urban spectrum. County-based measures of rurality, as used in this analysis, can be limited by the varying geographic sizes and distribution of infrastructure and resources across counties.