Objective. To identify factors that predict failure to diagnose congenital heart disease in newborns.
Design. All fatal cases in the Baltimore-Washington Infant Study were compiled. The Baltimore-Washington Infant Study includes 4390 cases of infants with congenital cardiovascular malformations identified in a population-based study between 1981 and 1989 in the Baltimore-Washington metropolitan area. Death occurred in 800 such infants in the first year of life. In 76 of these infants, death occurred before diagnosis of heart disease. These cases were identified by community search of autopsy records. Their characteristics are compared with those of infants who died after a cardiac diagnosis was made.
Results. Infant characteristics (birth weight, gestational age, intrauterine growth retardation, and chromosomal anomaly) are associated with death of infants with congenital cardiovascular malformations and with death of such infants before diagnosis. Diagnoses of coarctation of the aorta, Ebstein's anomaly, atrial septal defect, and truncus arteriosus are overrepresented in infants found by community search, particularly in those infants without associated malformations. Paternal education is associated with failure to diagnose congenital heart disease in life but other sociodemographic characteristics of the infant's family are not.
Conclusions. Diagnosis of congenital cardiovascular malformations requires close observation in the neonatal period. Analysis of age at death of infants with undiagnosed congenital cardiovascular malformation suggests that such infants may be at risk if discharged within the first 2 days of life.
- CCVM =
- congenital cardiovascular malformations •
- BWIS =
- Baltimore-Washington Infant Study •
- IUGR =
- intrauterine growth retardation
Physiologic changes in the circulation after birth mask the presentation of some congenital cardiovascular malformations (CCVM) for a variable period of early infancy. The average in-hospital period of observation of apparently healthy infants has been reduced from 2.9 days in 1981 to 2.0 days in 19901 ,2 raising concern about possible discharge of infants before diagnosis of CCVM. During the period 1981 through 1989, a population-based study of congenital heart disease in Maryland, the District of Columbia, and northern Virginia enrolled 4390 cases of CCVMs. This study offers a unique opportunity to examine the presentation and diagnosis of CCVM. Analysis of the cases of CCVMs that were identified only by search of autopsy records and death certificates (the community search deaths) was undertaken. The first objective of this study was to characterize the fatal cases for which diagnosis in life did not occur. The second objective was to determine if infant characteristics and family sociodemographic variables predict lack of diagnosis of CCVMs in life.
Case ascertainment by the Baltimore-Washington Infant Study (BWIS) has been previously described.3 ,4 Cases were enrolled by participating pediatric cardiologists at five university and one community hospital program, encompassing all pediatric cardiology services in a region with a population of six million. Approximately 52 of 53 regional hospitals participated in searches of pathology records for cases in which congenital heart disease was identified by autopsy findings. Death certificates in Washington, DC and Maryland were reviewed for all infants who died under 1 year of age. The logbook of the Medical Examiner was also reviewed for such cases in the State of Maryland. Home visits were made by trained interviewers to obtain demographic data, information about risk factors, and exposures. Race of the infant was designated as reported by the mother. Ninety percent of the families of eligible infant cases participated in interviews and 98.1% of these interviews were with the infant's mother. For infants diagnosed before death, follow-up visits and confirmation of diagnosis until 1 year of age were used to determine the age at death and whether or not surgery had been performed before death. No ascertainment of cause of death was made. Chromosomal studies were performed as clinically indicated.
Subgroups of cases were defined by survival and by diagnosis before and after death. Case group differences in mean birth weight and gestational age were analyzed using the Student's t test. χ2 tests were used to compare groups on the distributions of all other covariates. Subjects with intrauterine growth retardation (IUGR) were identified by classifying infants with gestational age and gender-specific birth weights below the tenth percentile of a population standard.5 Logistic regression was used to evaluate multiple potential risk factors for death in the first year of life.
Mortality for the 4390 infants with CCVMs registered in the BWIS was determined until age 1 year. A total of 800 cases (18%) died leaving 3590 survivors. Of the fatal cases, 10% (n = 76) were not diagnosed until after death, ie, were identified by community search. Of the remaining 724 deaths in the study group, 344 patients had undergone surgery.
Infant characteristics of the population are shown in Table 1 comparing survivors, deaths of infants after diagnosis, and community search deaths. The case groups do not differ by racial distribution, but all other infant characteristics show a differential distribution. The infant characteristics of: lower mean birth weight, low birth weight for gestational age (IUGR), associated noncardiac malformations, chromosomal anomalies, prematurity (gestational age <38 weeks), and male sex are more common in infants with CCVM who die before 1 year of age compared with such infants who survive to at least 1 year of age (Table 1). Community search cases are more likely to have associated malformations, chromosomal anomalies, low birth weight, prematurity, and IUGR than are the cases who died after diagnosis. Male sex is not associated with failure of diagnosis before death; but male sex is correlated with death in other infants with cardiovascular malformations. Mean birth weight in the community search cases is lower than other cases; this parameter is not normally distributed (Fig 1). Of the cases identified at community search, about 25% have birth weights <1500 g compared with <5% of all infants with CCVMs. Of these very small infants, 12 of the 15 (80%) had trisomy or major noncardiac malformations. In contrast, only 12 of 35 (34%) infants with birth weights >2000 g had these severe anomalies.
Age at Death
Over 80% of all community search deaths occur within the first week of life. Figure 2 shows the distribution of ages at death in the first 8 days of life for those cases identified at community search compared with the distribution of deaths of cases with CCVM identified before death. In the first week of life, 82% of all community search deaths occur compared with 23% of all deaths of infants with CCVM and 32% of deaths of infants with antemortem diagnosis of CCVM. Community search deaths essentially all occurred before 12 weeks of life with a median age at death of 2 days when the day of birth is defined as day 1. The distribution of age at death of the community search deaths versus deaths after diagnosis differs with P < .01; (χ2). Age at death varied by diagnosis among the community search cases. Coarctation of the aorta and hypoplastic left heart syndrome cases were more likely to die after day 2 of life than before that age, with odds ratios of 5.6 (95% confidence interval [CI] .3–11.6) and 9.3 (95% CI 3.5–15.1) respectively. All cases of Ebstein's anomaly and of transposition of the great arteries identified at community search died before day 2 of life in this series.
Patients with truncus arteriosus, Ebstein's anomaly, coarctation of the aorta, and atrial septal defect are overrepresented in the community search cases compared with the proportion of these diagnoses among all deaths (Table 2). By χ2 analysis the significance of the overrepresentation of coarctation is P < .045; of Ebstein's anomaly,P < .008; of atrial septal defect, P< .002; and of truncus arteriosus, P < .09. The association of endocardial cushion defect with the clinically obvious trisomy Down syndrome may be a reason that endocardial cushion defect is underrepresented (P < .025; χ2) in the cases of CCVM dying before diagnosis.
Infants with multiple major or lethal malformations were less likely to be in the community search death group than infants without such malformations. The odds ratio of being identified at community search death versus death after diagnosis was 36.6 (95% CI = 32.2–41.0) for coarctation of the aorta in all cases but was 46.6 (95% CI = 41.8–51.3) for infants with coarctation of the aorta who had no major malformations or trisomy. For the diagnosis of truncus arteriosus, the corresponding odds ratios for community search death were 2.4 (95% CI = -.7–5.5) for all cases and 3.7 (95% CI = 0.5–7.1) for cases of truncus arteriosus without associated malformations or trisomy. In cases of atrial septal defect, the same relationship held with the odds ratio for being community search death increased from 5.09 (95% CI = 1.8–8.4) to 7.7 (95% CI = 4.8–10.5) when cases with major malformations or trisomy were excluded. These data suggest that the presence of external malformations may have provoked more comprehensive diagnostic assessment of such infants and improved the likelihood of premortem diagnosis of cardiovascular malformations.
Family Sociodemographic Characteristics
The characteristics of maternal age, maternal education, family income, paternal education, surgical intervention, and infant race were entered as independent variables in a stepwise logistic regression model with death versus survival as the dependent variable. A separate model was analyzed including the same family characteristics and using community search death as the dependent variable but excluding surgical intervention as an independent variable. Paternal education is predictive of death of infants with CCVM and specifically of community search death (P = .0011). Paternal education less than high school increases the probability of the child's death 62% compared with children of fathers with college education (95% CI = 20–110%) Adjustment for type of cardiac defect does not alter the magnitude or significance of this association.
The infants were also divided into groups of diagnoses where sufficient cases (n > 100) existed to allow multiple regression. The dependent variable death was assessed against the variables denoting family characteristics. The results show that surgical intervention is not correlated with death within cases of the same type. However, paternal education less than high school remains strongly correlated with death of the infant within cases that have the same diagnosis.
Race is associated with mortality in this population of infants with CCVMs only among infants of low birth weight, not among normal birth weight infants. White low birth weight infants are 1.7 times more likely to die than non-white low birth weight infants are (95% CI = 1.2–2.4) in this study.
Infants with CCVM who die before diagnosis of the CCVM differ from survivors and from cases diagnosed before death with respect to birth weight, prematurity, IUGR, presence of associated malformations, and trisomy. The same infant characteristics associated with death before diagnosis (low birth weight, prematurity, IUGR, associated malformations, and trisomy) might be expected to mitigate against early discharge. These characteristics may also limit diagnostic and other interventions. Review of the autopsy summaries available in the BWIS data were made. From the available information, the community search cases were determined to have been deaths that were anticipated or unanticipated. Information to make this judgment was available for 69 of 76 such cases. In 45 cases death was not anticipated. Anticipated death as a dependent variable is strongly associated with chromosomal abnormality when cases of trisomy were excluded (P = .0045), or with trisomy if trisomies were included (P = .0026).
Previous reports of congenital heart disease identified only at autopsy in infants have described a spectrum of malformations unlike this report. Coronary artery anomalies,4–7 total anomalous pulmonary venous drainage,8 type I truncus arteriosus,9 asplenia syndrome,10 and subarterial ventricular septal defect11 are described in these case reports. In our series the diagnoses that are overrepresented in proportion to their frequency in the population of infants dying with CCVM are coarctation of the aorta, atrial septal defect, truncus arteriosus, and Ebstein's anomaly. Over 60% of the infants with cardiac malformations diagnosed at community search in our series weighed >2000 g, a weight sufficient for surgical interventions. Community search deaths represent 10% of all mortality in this population of infants with congenital heart disease. These deaths may represent only part of the morbidity attributable to failure to diagnose cardiovascular malformations in infancy; diagnosis of infants after onset of life-threatening symptoms may affect long-term outcomes in infants who survive.
Sociodemographic factors have been reported to correlate with increased mortality and morbidity in children in a number of studies.12–17 Both neonatal and postneonatal death rates are increased in poor children. Mortality rates are increased in children from low-income families or families with less maternal education. Mortality at age 1 year of infants with cardiovascular malformations related to two demographic variables in this study. Paternal education is a risk factor for increased infant mortality of infants, a similar effect to that noted for all cause mortality.18 Low levels of paternal education were associated with increased mortality of infants with CCVM whether diagnosis is made or not made in life. This suggests that the effect of paternal education was not in obtaining initial access to tertiary medical care. During the era 1981–1989 in the Baltimore-Washington region, Children's Specialty Services programs and Medicaid were accepted at all tertiary care cardiology programs in the region. Our finding that death before diagnosis did not correlate with lower income, less maternal education, or lower socioeconomic status scores implies that access for tertiary care was provided for children from families with these characteristics.
Length of Hospital Stay
Length of stay for uncomplicated vaginal delivery in the United States was significantly reduced over the period of this study. The in-hospital period of observation of apparently healthy infants has been reduced from 2.9 days in 1981 to 2.0 days in 19901and has been reduced to 24 hours or less in some communities now. A recent national telephone survey of mothers of newborn infants demonstrated that 51% of mothers discharged in the first 5 months of 1996 were discharged when their infant was <24 hours old.19
Beebe et al studied all cause mortality to age 28 days of term infants with birth weight of 2500 g or more as a function of length of hospital stays. The neonatal death rate was 115 of 109 256 eligible births and no significant relationship was found between length of stay and neonatal mortality. However, of 17 cases that were well at discharge and subsequently died, 7 infants (41%) died at ages 2 to 27 days from CCVMs. Five of these 7 infants had cardiac malformations of left heart obstruction.20 In the BWIS study, congenital heart disease was diagnosed on day 1 of life (the day of birth) in 436 cases; on day 2 in 409 cases, on day 3 in 247 cases; day 4 in 178 cases, and day 5 in 147 cases. Of all CCVMs identified by 1 year of age, a third of the diagnoses were apparent within the first 5 days of life. These data were collected in an era when in-hospital neonatal observation averaged 2.1 to 2.9 days.
When in-hospital observation of neonates is <24 to 48 hours, some cardiovascular malformations in infants will be diagnosed from an outpatient status. In our study, left heart obstructions were identified after day 2 of life. One can speculate that morbidity and mortality attributable to such malformations may increase. Congenital malformations are responsible for 40% of neonatal mortality in term infants. Strategies to ensure diagnosis of congenital malformations after hospital discharge are important in limiting neonatal mortality and morbidity.
One in 10 infants with cardiovascular malformations who die in the first year of life did not have a diagnosis made of the malformation before death. Of all infants with cardiovascular malformations who died in the first week of life, 1 in 4 did not have the cardiovascular malformation identified before death. These data are representative of an era when length of stay for uncomplicated vaginal delivery ranged from 2.9 to 2.1 days, compared with a length of stay of <24 hours for more than half of mothers delivering in the first 5 months of 1996. Low levels of paternal educational were a significant risk factor for infant mortality from cardiac disease; socioeconomic status is therefore a relevant consideration in determining length of hospital stay after childbirth.
Low birth weight, prematurity, chromosomal anomalies, noncardiac malformations, and IUGR were associated with mortality in children with congenital heart disease and are thus signals for evaluation of heart disease in such infants. However, CCVMs were even more frequently undiagnosed in infants without chromosomal anomalies, noncardiac malformations, or IUGR. Coarctation of the aorta, truncus arteriosus, and Ebstein's anomaly were overrepresented in the group of infants who died without diagnosis in life. Strategies to improve the identification of infants with treatable cardiac defects may include ensuring appropriate length of stay for mothers and infants after childbirth; frequent clinician visits in the first week of life; and education of clinicians to identify cardiac disease in the absence of murmurs. Legislative attempts to define an appropriate length of stay do not address the variable physiology of lethal neonatal disease.
This study was funded by NHLBI Grant HL-37–25629. Invaluable assistance was provided by the BWIS investigators who originally participated in the study and registered these cases. These co-investigators who supported this research are Joel I. Brenner, MD, Catherine A. Neill, MD, Lowell W. Perry, MD, Gerard R. Martin, MD, John W. Downing, MD, Seymour I. Hepner, MD, Mohamed Mardini, MD, Judith D. Rubin, MD, MPH, Joann A. Boughman, PhD, Robert J. McCarter, ScD, and P. David Wilson, PhD.
- Received April 17, 1997.
- Accepted September 29, 1998.
Reprint requests to (K.S.K.) Children's National Medical Center, 111 Michigan Ave, NW, Washington, DC 20010.
- ↵Commission on Professional and Hospital Activities. Length of Stay by Diagnosis, North Central Region 1981–1985. Ann Arbor, MI: Commission on Professional and Hospital Activities; 1982–1985
- ↵Commission on Professional and Hospital Activities. Length of Stay By Diagnosis And Operation, North Central Region 1986–1990. Ann Arbor, MI: Commission on Professional and Hospital Activities; 1986–1991
- ↵Ferencz C, Rubin JD, Loffredo CA, Magee CA, eds. Perspectives in Pediatric Cardiology, IV. Epidemiology of Congenital Heart Disease. Mt Kisco, NY: Futura; 1993
- Egbuonu L,
- Starfield B
- Newacheck PW,
- Stoddard JJ,
- McManus M
- ↵Young KT, Davis K, Schoen C. The Commonwealth Fund Survey of Parents With Young Children. August 1996
- Beebe SA,
- Britton JR,
- Britton HL,
- Fan P,
- Jepson B
- Copyright © 1999 American Academy of Pediatrics