1
ORIGINAL ARTICLE Temporal Variation of Birth Prevalence of Congenital Heart Disease in the United States Alexander Egbe, MD, MPH,* Santosh Uppu, MD,* Simon Lee, MD,* Annemarie Stroustrup, MD, MPH,*† Deborah Ho, MD,* and Shubhika Srivastava, MBBS* *Division of Pediatric Cardiology, Mount Sinai Medical Center, New York, NY, USA; †Department of Preventative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA ABSTRACT
Background. This is a longitudinal analysis of the largest and most comprehensive inpatient care database in the United States to determine temporal variation of birth prevalence of congenital heart disease (CHD) diagnosis, adjusting for potentially confounding factors. Methods. We compared all entries of CHD diagnoses in the Nationwide Inpatient Sample (NIS) database in 1998 and 2008 to determine differences in birth prevalence of overall CHD and specific CHD phenotypes stratified by race, gender, socioeconomic status, and geographical location. Results. CHD prevalence was 10.2/1000 in 1998 and 10.8/1000 live births in 2008, without significant changes in prevalence (P = .09). Prevalence of isolated patent ductus arteriosus increased from 1.9 to 2.8 per 1000 (P < .001), and this temporal increase remained statistically significant after stratification by race, income status, and geographic location. Prevalence of mild CHD increased from 8.0 to 9.1 per 1000 (P = .01), with most of this increase occurring among Caucasians and the upper socioeconomic class. There was a decrease in prevalence of severe CHD from 1.5 to 0.9 (P = .03), while prevalence of moderate CHD and all other specific CHD phenotypes remained unchanged in both cohorts. Conclusions. We report increased prevalence of isolated patent ductus arteriosus and mild CHD; decreased prevalence of severe CHD; and unchanged prevalence of overall CHD and all other specific CHD phenotype. We speculate that increased prevalence of mild CHD was due to increased case detection because of improvement in echocardiography. Decrease in diagnosis of severe CHD could be due to the impact of pregnancy termination. Key Words. Congenital Heart Disease; Pediatric; Epidemiology
Introduction
C
ongenital heart disease (CHD) encompasses any gross structural abnormality of the heart or intrathoracic great vessels present at birth that is potentially of functional significance.1 CHD is the most common type of congenital malformation and contributes significantly to infant mortality and morbidity.2–10 Comparison of published studies suggest temporal changes in prevalence of overall CHD and some specific CHD phenotypes both in the United States and worldwide since the 1950s.3–7,11,12 Differences in study design, inclusion criteria, and disease classification systems used in © 2014 Wiley Periodicals, Inc.
these different studies limit reliable comparison of estimates derived from these studies. Improvement and widespread application of fetal ultrasound have resulted in prenatal diagnosis of congenital malformations very early in gestation, allowing parents the choice of terminating affected pregnancies.13–17 There has also been a concordant growth in perinatology and genetics in identifying risk factors associated with CHD, making it possible to counsel high-risk couples to make informed decisions regarding future pregnancies especially in situations when the risk of CHD recurrence is high. In addition, sensitivity of transthoracic echocardiography has also improved Congenit Heart Dis. 2014;••:••–••
2 significantly, increasing the odds of diagnosing simple and asymptomatic cardiac defects in the postnatal period. We hypothesize that these changes impact the birth prevalence of overall CHD and specific CHD phenotypes in the United States. To test this hypothesis, we reviewed the Nationwide Inpatient Sample (NIS) database, which is the largest available database of hospitalization information in the United States. We compared birth prevalence of overall CHD and specific CHD phenotypes in 1998 and 2008 to determine temporal variation in CHD prevalence. Methods
All data were derived from the NIS, Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality.18 The 1998 and 2008 NIS is an all-payer administrative database reporting clinical and resource use information representative of hospitalizations in 22 and 42 states, respectively. We chose the NIS database instead of other databases such the Kids’ Inpatient Database because NIS is the largest available inpatient care database in the United States; containing approximately 8 million hospital stays each year from about 1000 hospitals sampled to approximate a 20% stratified sample of the US community hospitals. NIS large sample size makes it ideal for analysis of rare conditions such as specific CHD phenotypes. We chose 2008 dataset because it was the latest available NIS dataset at the inception of this study, and we chose 1998 because it was the earliest NIS dataset covering hospitalization in more than 20 states. We reviewed all birth entries in the NIS database from January to December 1998 and January to December 2008, and identified all cases of echocardiographically diagnosed CHD based on unique CHD ICD-9 code 745.0–747.42. For the purpose of our study, we excluded preterm neonates with isolated patent ductus arteriosus (PDA), ICD-9 code 747.0, because we considered this a normal variant. Based on these criteria, we identified 8171 and 13 249 cases of CHD diagnoses in 1998 and 2008, respectively. We then excluded 154 patients (1.9%) and 156 patients (1.2%) with CHD ICD-9 code from the 1998 and 2008 cohort, respectively, because they did not have echocardiography procedure code 93303. We did this to ensure that all cases of CHD diagnosis included in our study were confirmed by echocardiography. To avoid double counting, we included only CHD diagnosis made during birth hospitalization. We Congenit Heart Dis. 2014;••:••–••
Egbe et al. eliminated readmissions and interhospital transfers by including only hospitalizations with ICD-9 code for normal and complicated delivery (650.0– 669.0). Approval for this study was obtained from both HCUP and institutional review board. In patients with multiple CHD diagnoses, we coded one diagnosis per person based on hierarchical system described in the BaltimoreWashington Infant Study and the PAN study.8,19 We then grouped specific CHD phenotypes based on severity and anatomic location of lesions based on classification systems used in prior studies.6,19 See Supporting Information Table S1 and S2. For race, socioeconomic and geographic stratification, we adopted the classification system used in NIS database.18 As per HCUP data use agreement prohibiting reporting of cell size ≤10, CHD diagnoses with cell size ≤10 were excluded from sub-analysis and grouped together as “others” as shown in table legends. Standard descriptive statistic methods were used. Incidence of overall CHD and individual CHD phenotypes were expressed per 1000 and per 100 000 live births, respectively, for the general population and then stratified by race, gender, socioeconomic status, and hospital geographic location. We used chi-square test to assess difference between groups. There were multiple analyses and comparisons in this study, which increases the chance of type 1 error. To avoid type 1 error, we applied Bonferroni correction and set P value of
ORIGINAL ARTICLE Temporal Variation of Birth Prevalence of Congenital Heart Disease in the United States Alexander Egbe, MD, MPH,* Santosh Uppu, MD,* Simon Lee, MD,* Annemarie Stroustrup, MD, MPH,*† Deborah Ho, MD,* and Shubhika Srivastava, MBBS* *Division of Pediatric Cardiology, Mount Sinai Medical Center, New York, NY, USA; †Department of Preventative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA ABSTRACT
Background. This is a longitudinal analysis of the largest and most comprehensive inpatient care database in the United States to determine temporal variation of birth prevalence of congenital heart disease (CHD) diagnosis, adjusting for potentially confounding factors. Methods. We compared all entries of CHD diagnoses in the Nationwide Inpatient Sample (NIS) database in 1998 and 2008 to determine differences in birth prevalence of overall CHD and specific CHD phenotypes stratified by race, gender, socioeconomic status, and geographical location. Results. CHD prevalence was 10.2/1000 in 1998 and 10.8/1000 live births in 2008, without significant changes in prevalence (P = .09). Prevalence of isolated patent ductus arteriosus increased from 1.9 to 2.8 per 1000 (P < .001), and this temporal increase remained statistically significant after stratification by race, income status, and geographic location. Prevalence of mild CHD increased from 8.0 to 9.1 per 1000 (P = .01), with most of this increase occurring among Caucasians and the upper socioeconomic class. There was a decrease in prevalence of severe CHD from 1.5 to 0.9 (P = .03), while prevalence of moderate CHD and all other specific CHD phenotypes remained unchanged in both cohorts. Conclusions. We report increased prevalence of isolated patent ductus arteriosus and mild CHD; decreased prevalence of severe CHD; and unchanged prevalence of overall CHD and all other specific CHD phenotype. We speculate that increased prevalence of mild CHD was due to increased case detection because of improvement in echocardiography. Decrease in diagnosis of severe CHD could be due to the impact of pregnancy termination. Key Words. Congenital Heart Disease; Pediatric; Epidemiology
Introduction
C
ongenital heart disease (CHD) encompasses any gross structural abnormality of the heart or intrathoracic great vessels present at birth that is potentially of functional significance.1 CHD is the most common type of congenital malformation and contributes significantly to infant mortality and morbidity.2–10 Comparison of published studies suggest temporal changes in prevalence of overall CHD and some specific CHD phenotypes both in the United States and worldwide since the 1950s.3–7,11,12 Differences in study design, inclusion criteria, and disease classification systems used in © 2014 Wiley Periodicals, Inc.
these different studies limit reliable comparison of estimates derived from these studies. Improvement and widespread application of fetal ultrasound have resulted in prenatal diagnosis of congenital malformations very early in gestation, allowing parents the choice of terminating affected pregnancies.13–17 There has also been a concordant growth in perinatology and genetics in identifying risk factors associated with CHD, making it possible to counsel high-risk couples to make informed decisions regarding future pregnancies especially in situations when the risk of CHD recurrence is high. In addition, sensitivity of transthoracic echocardiography has also improved Congenit Heart Dis. 2014;••:••–••
2 significantly, increasing the odds of diagnosing simple and asymptomatic cardiac defects in the postnatal period. We hypothesize that these changes impact the birth prevalence of overall CHD and specific CHD phenotypes in the United States. To test this hypothesis, we reviewed the Nationwide Inpatient Sample (NIS) database, which is the largest available database of hospitalization information in the United States. We compared birth prevalence of overall CHD and specific CHD phenotypes in 1998 and 2008 to determine temporal variation in CHD prevalence. Methods
All data were derived from the NIS, Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality.18 The 1998 and 2008 NIS is an all-payer administrative database reporting clinical and resource use information representative of hospitalizations in 22 and 42 states, respectively. We chose the NIS database instead of other databases such the Kids’ Inpatient Database because NIS is the largest available inpatient care database in the United States; containing approximately 8 million hospital stays each year from about 1000 hospitals sampled to approximate a 20% stratified sample of the US community hospitals. NIS large sample size makes it ideal for analysis of rare conditions such as specific CHD phenotypes. We chose 2008 dataset because it was the latest available NIS dataset at the inception of this study, and we chose 1998 because it was the earliest NIS dataset covering hospitalization in more than 20 states. We reviewed all birth entries in the NIS database from January to December 1998 and January to December 2008, and identified all cases of echocardiographically diagnosed CHD based on unique CHD ICD-9 code 745.0–747.42. For the purpose of our study, we excluded preterm neonates with isolated patent ductus arteriosus (PDA), ICD-9 code 747.0, because we considered this a normal variant. Based on these criteria, we identified 8171 and 13 249 cases of CHD diagnoses in 1998 and 2008, respectively. We then excluded 154 patients (1.9%) and 156 patients (1.2%) with CHD ICD-9 code from the 1998 and 2008 cohort, respectively, because they did not have echocardiography procedure code 93303. We did this to ensure that all cases of CHD diagnosis included in our study were confirmed by echocardiography. To avoid double counting, we included only CHD diagnosis made during birth hospitalization. We Congenit Heart Dis. 2014;••:••–••
Egbe et al. eliminated readmissions and interhospital transfers by including only hospitalizations with ICD-9 code for normal and complicated delivery (650.0– 669.0). Approval for this study was obtained from both HCUP and institutional review board. In patients with multiple CHD diagnoses, we coded one diagnosis per person based on hierarchical system described in the BaltimoreWashington Infant Study and the PAN study.8,19 We then grouped specific CHD phenotypes based on severity and anatomic location of lesions based on classification systems used in prior studies.6,19 See Supporting Information Table S1 and S2. For race, socioeconomic and geographic stratification, we adopted the classification system used in NIS database.18 As per HCUP data use agreement prohibiting reporting of cell size ≤10, CHD diagnoses with cell size ≤10 were excluded from sub-analysis and grouped together as “others” as shown in table legends. Standard descriptive statistic methods were used. Incidence of overall CHD and individual CHD phenotypes were expressed per 1000 and per 100 000 live births, respectively, for the general population and then stratified by race, gender, socioeconomic status, and hospital geographic location. We used chi-square test to assess difference between groups. There were multiple analyses and comparisons in this study, which increases the chance of type 1 error. To avoid type 1 error, we applied Bonferroni correction and set P value of
