To evaluate the extent and determinants of missed prenatal detection of congenital heart disease (CHD) in a population-based setting.
To evaluate the extent and determinants of missed prenatal detection of congenital heart disease (CHD) in a population-based setting.
This was a retrospective cohort study of cases with CHD, excluding minor defects, identified between 1997 and 2007 by a statewide surveillance program. We examined a comprehensive list of potential risk factors for which data were available in the surveillance database from abstracted medical charts. We analyzed the association of fetal, maternal and encounter factors with 1) whether a prenatal ultrasound was performed and 2) prenatal detection of CHD.
CHD was detected prenatally in only 39% of 1474 cases, with no improvement in detection rate over the 10-year period. Among the 97% (n = 1431) of mothers who underwent one or more ultrasound examinations, 35% were interpreted as abnormal; fetal echocardiography was performed in 27% of the entire cohort. Maternal and encounter factors increasing the adjusted odds of prenatal detection included: family history of CHD (OR, 4.3 (95% CI, 1.9–9.9)), presence of extracardiac defects (OR, 2.7 (95% CI, 1.9–3.9)) and ultrasound location i.e. high risk clinic vs clinic (OR, 2.1 (95% CI, 1.3–3.1)). Defects that would be expected to have an abnormal outflow-tract view were missed more often (64%) than were those that would be expected to have an abnormal four-chamber view (42%).
The majority of CHD cases over the 10-year study period were missed prenatally and detection rates did not increase materially during that time. The failure to detect CHD prenatally was related to encounter characteristics, specifically involving screening ultrasound examinations, which may be targeted for improvement. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.
Congenital heart disease (CHD) is one of the most common and lethal birth defects1. Approximately 1% of liveborn infants have CHD. Of these, 18% die within a year2 and 40% require some type of intervention3. CHD diagnosis prior to delivery allows for early parental counseling. Although data regarding the impact of prenatal diagnosis of CHD on mortality are conflicting4–8, it is widely accepted that for prenatally diagnosed infants requiring intervention, planned delivery and appropriate postnatal care improve preoperative hemodynamic stability, decreasing perioperative morbidity5, 9.
Efficient screening for fetal CHD is challenging, requiring a population-based approach, as most cases occur in mothers without known risk factors10, 11. Consensus recommendations in the USA advocate CHD screening during standard second-trimester ultrasound examination, using a four-chamber view of the fetal heart, plus, if ‘technically feasible’, an outflow-tract view12, 13. Published studies report detection rates for CHD as high as 55–65% with the four-chamber view alone and 80–84% with the addition of the outflow-tract view14, 15. However, current screening practices in most developed countries detect only 30–50% of CHD cases2, 11, 16, 17. While low rates of prenatal detection are well documented16–18, the reasons for failed detection have not been well studied.
Several studies cite low rates of prenatal CHD detection even when > 90% of women in the population undergo fetal ultrasound examination14, 16, 17, 19–21. Therefore, factors other than low use of prenatal ultrasound, including gestational age at the time of ultrasound, maternal habitus, technical ability to obtain appropriate views, CHD diagnosis and the ultrasound operator's and reader's experience, likely play a greater role in CHD detection. Studies of predictors of failed CHD detection have been performed in select cohorts20, 22. However, we are unaware of any systematic population-based study of the potentially modifiable factors related to failed detection.
We used population data from the Utah Birth Defect Network (UBDN) to: 1) determine the rate of failed prenatal detection of CHD, 2) determine when during pregnancy the opportunity to detect CHD is missed, and 3) identify maternal and encounter-related risk factors for failed prenatal detection.
This retrospective cohort study included all cases of major CHD identified by the UBDN from 1997 to 2007 for all live births, stillbirths and terminations at > 20 weeks' gestation. We excluded cases with only isolated septal defects (except for inlet-type ventricular septal defects) or mild valve abnormalities (isolated stenosis or regurgitation without associated ventricular chamber hypoplasia). Inlet ventricular septal defects were included as most can be seen on an appropriate four-chamber screening view at the level of the atrioventricular valves (while outflow tract or perimembranous defects may be missed) and most will require postnatal intervention. Cases of severe valvar stenosis with associated ventricular hypoplasia were included as these again should be seen on a four-chamber screening view and in these cases intervention is almost always required. Cases were reviewed and then coded using the Center for Disease Control recommended modified ICD-9-DM codes23. If the case had multiple CHD codes, it was assigned a primary diagnosis based on the most significant defect. For each case we determined which, if any, ultrasound view would be expected to be abnormal at screening, according to the particular defects present. If a case had multiple defects, all defects and their expected abnormalities on screening images were used to designate them as either an ‘expected abnormal four-chamber screening view’, ‘expected abnormal outflow-tract view’, ‘expected abnormal both views’ or ‘expected abnormal neither view’.
The UBDN is a well-established, robust population-based statewide surveillance system that meets the requirements of the Centers for Birth Defects Research and Prevention methodology and participates in the National Birth Defects Prevention Study. The UBDN, under the auspices of the Utah Department of Health, prospectively monitors all births (live births, stillbirths and pregnancy terminations) of mothers who reside in Utah to identify major birth defects. Age at first diagnosis is up to 24 months. The UBDN has over 100 data sources, resulting in a high level of case ascertainment. Potential cases are reviewed by three medical geneticists (including one who is also board-certified in maternal–fetal medicine (MFM)). Most CHD cases are also reviewed by a pediatric cardiologist. The UBDN began collecting CHD data in 1997 for conotruncal and left-sided obstructive lesions. In 1999, ascertainment expanded to include all heart defects with the exception of isolated ventricular septal defects, which were included from 2003. The database includes detailed information regarding maternal characteristics, prenatal care and imaging and postnatal diagnosis and imaging.
Maternal and encounter characteristics were collected from the UBDN database. A positive family history was defined as a history of CHD in a first-degree relative. Ultrasound reader was defined in a hierarchical fashion in the order in which referrals would typically be made. Thus, cases in which multiple ultrasound examinations had been performed and interpreted by obstetricians and/or radiologists and MFM specialists were coded as read by a MFM; those in which ultrasound examinations had been interpreted by obstetricians and radiologists were coded as read by a radiologist; those in which they were interpreted only by obstetricians were coded as read by an obstetrician. Location of ultrasound examination was treated in a similar hierarchical fashion, with high-risk clinics, followed by hospitals and then general clinics. We defined a screening ultrasound as the first ultrasound examination performed between 16 and 24 weeks' gestation, as this is when anomaly screening is performed. Cases delivered in 2003–2007 were reviewed for available prenatal ultrasound reports. Though data from ultrasound reports, including timing, location, reader and diagnoses, had been abstracted for all cases, paper reports were not retained prior to 2003. Reports were reviewed solely for detailed documentation regarding the cardiac screening views obtained and whether they were read as normal or abnormal. Paper reports were not used as a source for other study variables.
Additional socioeconomic variables and measures of distance were obtained from the 2000 census data using the University of Utah's Department of Geography's Digitally Integrated Geographic Information Technologies (DIGIT) lab. Using the maternal address at delivery, the DIGIT lab provided census-tract level measures of socioeconomic status, including education, median income and population below the poverty level (defined by the Census bureau for family size and number of dependents24). Census-tract rural-urban commuting areas were used to define residence as ‘urban’ (codes 1–3) or ‘rural’ (codes 4–10)25. Travel time to the nearest pediatric hospital with a fetal cardiology program was calculated using distance and road speed data, with a maximum speed of 55 mph.
The cohort was described using frequencies and proportions. Odds ratios were used to examine the association of fetal, maternal and encounter factors with documentation of having undergone a prenatal ultrasound examination and prenatal diagnosis of CHD. Logistic regression was used to model risk factors for undergoing a prenatal ultrasound examination and detection of CHD. Covariates were included in the model if on univariate analysis P < 0.2. Models were examined for collinearity and, if found, the variable with the strongest association was retained. Log likelihood ratios were used to backwards eliminate covariates. All analyses were conducted using Stata 11.0 (StataCorp, College Station, TX, USA).
The study was approved by the institutional review boards of the University of Utah and the Utah Department of Health.
There were 1474 cases of CHD ascertained by the UBDN in 1997–2007 that met our study inclusion criteria; their characteristics are given in Table S1. The number of cases of CHD was lower in 1997–1998, when only conotruncal and left-sided obstructive heart lesions were collected by the UBDN, but stable through the rest of the study period. Most mothers were white and had an education at high-school level or lower. A family history of CHD was reported in 3% of cases. Extracardiac malformations were present in 38% of cases and 1% had heterotaxy.
|US exams received||Prenatal detection of CHD|
|Characteristic||n (%)*||OR (95% CI)||n (%)†||OR (95% CI)|
|≥ 35 years||197 (96)||1||96 (49)||1|
|21–34 years||1096 (97)||1.61 (0.76–3.44)||430 (39)||0.51 (0.32–0.80)|
|< 21 years||138 (98)||2.10 (0.56–7.90)||45 (33)||0.68 (0.50–0.92)|
|Plurality||Collinear (—)||1.45 (0.93–2.41)|
|Singleton gestation||1360 (97)||536 (39)|
|Multiple gestation||71 (100)||35 (49)|
|Initiation of prenatal care|
|First trimester||1400 (98)||1||488 (35)||1|
|Second trimester||135 (96)||0.35 (0.13–0.98)||55 (41)||1.04 (0.73–1.49)|
|Third trimester||38 (90)||0.12 (0.04–0.39)||11 (29)||1.03 (0.63–1.69)|
|Per additional pregnancy||0.89 (0.79–1.00)||1.10 (1.04–1.15)|
|Maternal BMI at first visit|
|< 25 kg/m2||838 (97)||1||340 (41)||1|
|≥ 25 kg/m2||320 (98)||0.95 (0.73–1.24)||126 (39)||0.95 (0.73–1.23)|
|≥ 30 kg/m2||273 (96)||0.91 (0.69–1.20)||105 (38)||0.90 (0.69–1.19)|
|≤ 16 kg (c. 35 lb) (normal)||419 (42)||1|
|> 16 kg (c. 35 lb) (excessive)||152 (35)||0.72 (0.58–0.91)|
|College graduate||169 (95)||1||87 (51)||1|
|High school||633 (98)||2.14 (0.88–5.19)||249 (39)||0.56 (0.40–0.78)|
|< High school||629 (97)||1.41 (0.62–3.28)||235 (37)||0.62 (0.44–0.86)|
|Family history||49 (98)||1.49 (0.20–11.04)||28 (57)||2.06 (1.16–3.66)|
|White||1191 (97)||1||464 (39)||1|
|Non-white||234 (97)||0.93 (0.40–2.12)||103 (44)||1.23 (0.93–1.64)|
|Census-tract level % of adults ≥ 25 years with < high school education|
|0.0–14.9%||952 (97)||1||387 (41)||1|
|15.0–24.9%||314 (97)||0.86 (0.41–1.79)||111 (35)||0.79 (0.61–1.02)|
|25.0–39.9%||99 (96)||0.68 (0.23–1.98)||41 (41)||1.00 (0.66–1.51)|
|40.0–100.0%||24 (96)||0.66 (0.09–5.03)||13 (54)||1.64 (0.74–3.63)|
|Census-tract level % of adults ≥ 25 years with college degree|
|40.0–100.0%||176 (98)||1||74 (42)||1|
|25.0–39.9%||387 (96)||0.41 (0.12–1.43)||171 (44)||1.02 (0.72–1.46)|
|15.0–24.9%||484 (98)||0.69 (0.19–2.46)||173 (36)||0.75 (0.53–1.06)|
|0.0–14.9%||342 (97)||0.58 (0.16–2.14)||134 (39)||0.85 (0.59–1.23)|
|Census-tract level % below poverty level‡|
|0.0–4.9%||516 (98)||1||211 (41)||1|
|5.0–9.9%||437 (98)||0.93 (0.39–2.21)||169 (39)||0.90 (0.69–1.17)|
|10.0–19.9%||277 (94)||0.32 (0.15–0.70)||112 (40)||0.92 (0.69–1.23)|
|20.0–100.0%||159 (99)||1.7 (0.37–7.72)||60 (38)||0.88 (0.61–1.27)|
|Census-tract level rural residence§|
|Urban||1247 (97)||1||506 (41)||1|
|Rural||142 (96)||0.66 (0.27–1.61)||46 (32)||0.69 (0.78–0.99)|
The majority of mothers (87%) had their first prenatal visit in the first trimester. About half (53%) of the cohort had prenatal ultrasound examinations performed only in a clinic (family practice or obstetric), 32% had one or more ultrasound examinations performed in a hospital and 15% had one or more performed in a MFM clinic. The interpreting physician's specialty could be identified in 690 (47%) cases. For screening ultrasound examinations, 62% were read by an obstetrician, 12% by a radiologist and 25% by a MFM.
The proportion of CHD cases detected prenatally in this cohort was 39% (574/1474), with no significant differences according to year of delivery (Figure 1, P = 0.10). The lowest detection rates (Figure 2) were for aortopulmonary windows (0%) and total anomalous pulmonary venous return (6%). Detection was also low for conotruncal or outflow tract anomalies, including truncus arteriosus (24%), tetralogy of Fallot with pulmonary stenosis (26%) and transposition of the great arteries (14%).
Almost all (97%) mothers of CHD cases underwent at least one prenatal ultrasound examination and 77% had an ultrasound exam between 16 and 24 weeks' gestation. However, 60% of CHD cases in which a prenatal ultrasound examination had been performed were missed. Fetal echocardiograms were performed in 27% of CHD cases. Family history of CHD was associated with having undergone a fetal echocardiogram (65% vs 47%, P = 0.02). However, 35% of mothers with a family history of CHD did not receive a fetal echocardiogram. Although most (89%) cases with an abnormal ultrasound were seen by a MFM, 42% of these cases never had a fetal echocardiogram. Of those with a fetal echocardiogram, 3% had a missed CHD diagnosis, predominantly coarctation of the aorta (n = 8), with one case each of double outlet right ventricle and double inlet left ventricle.
Factors associated with failure to receive a prenatal ultrasound examination included later initiation of prenatal care, higher number of previous pregnancies and maternal residence in a census tract in which 10–20% of the population were below the poverty level; they did not include maternal age, education or race (Table 1). In the multivariate model, only late initiation of prenatal care (in second trimester: odds ratio (OR), 0.35 (95% CI, 0.12–0.976) and in third trimester: OR, 0.1 (95% CI, 0.0–0.4)) was associated independently with failure to undergo prenatal ultrasound.
Among mothers who underwent ultrasound examination, maternal factors associated with lower prenatal detection of CHD included younger age, fewer years of education, excessive weight gain during pregnancy (> 16 kg (c. 35 lb)) and rural residence (Table 1). In contrast, a family history of CHD increased the odds of prenatal detection (OR, 2.1 (95% CI, 1.2–3.7)).
Encounter factors associated with lower prenatal detection of CHD included ultrasound examinations performed solely at general clinics compared with one or more performed at a hospital or MFM clinic (Table 2). Nevertheless, 67% of CHD cases which underwent ultrasound examination in a hospital and 25% in a MFM clinic were missed. Another factor related to lower detection was travel time to the nearest fetal cardiology program (per additional hour of travel: OR, 0.9 (95% CI, 0.78–0.97)). CHD cases without additional non-cardiac defects were less likely to be diagnosed prenatally (56% vs 29%, P < 0.001).
|Characteristic||CHD detected (n (%))||Odds ratio (95% CI)|
|Hospital||65 (31)||1.60 (1.14–2.25)|
|MFM/high-risk clinic||341 (74)||10.00 (7.64–13.10)|
|Radiologist||10 (16)||2.77 (1.21–6.32)|
|MFM||268 (79)||56.41 (32.7–97.2)|
|Suspected abnormality||74 (75)||24.38 (14.75–40.27)|
|Abnormal||398 (86)||52.03 (36.96–73.24)|
|Screening US location*|
|Hospital||59 (38)||1.60 (1.14–2.25)|
|MFM/high-risk clinic||211 (71)||10.00 (7.64–13.10)|
|Screening US interpreter*|
|Radiologist||36 (50)||2.77 (1.21–6.32)|
|MFM||110 (75)||56.41 (32.7–97.2)|
|Number of US exams||1.79 (1.65–1.93)|
|(per additional exam)|
|Presence of additional congenital defects||311 (57)||3.18 (2.55–3.98)|
|Presence of heterotaxy||12 (60)||2.29 (0.92–5.63)|
|Travel time to fetal cardiology program||0.90 (0.78–0.97)|
|(per additional hour)|
Defects were categorized based on the screening view(s) expected to be abnormal (Figure 2). If a case had multiple defects, all expected abnormal views (in addition to their primary diagnoses) were considered. Compared with defects for which neither view would be expected to be abnormal, those with an expected abnormal four-chamber view had the highest chance of being detected prenatally (OR, 4.6 (95% CI, 3.6–5.7)), while those with an isolated expected abnormal outflow-tract view had a slightly increased likelihood of detection (OR, 1.8 (95% CI, 1.4–2.4)). However, 42% of cases with an expected abnormal four-chamber view, 64% with an expected abnormal outflow-tract view and 30% with both views expected to be abnormal were not detected prenatally.
On multivariate analysis, after adjusting for maternal race and rural residence, prenatal detection was related independently to several encounter factors, including total number of fetal ultrasound examinations, location of ultrasound examination and an abnormal screening ultrasound result (Table 3). Cases with a family history of CHD and those with an additional non-cardiac congenital defect had higher odds of prenatal detection. Maternal age, education and weight gain during pregnancy were not retained in the final model.
|Odds ratio (95% CI)||P|
|Family history||4.3 (1.9–9.9)||< 0.01|
|Maternal rural residence||0.6 (0.3–1.2)||0.14|
|Each additional US exam||1.6 (1.4–1.7)||< 0.01|
|Suspected abnormality on US||17.3 (9.8–30.5)||< 0.01|
|Abnormal US||31.3 (20.7–47.6)||< 0.01|
|Additional congenital defect||2.7 (1.9–3.9)||< 0.01|
|US performed at high-risk clinic||2.1 (1.3–3.1)||< 0.01|
|US performed at hospital||0.8 (0.4–1.3)||0.31|
Of the 705 cases delivered after 2003, 297 (42%) had ultrasound reports available for review. Compared with the whole cohort (n = 1474), this subset of patients had a higher rate of prenatal detection (79%). The review of their reports showed that 95% had documented cardiac screening, including 65% with specific documentation of a four-chamber view and 57% with documentation of both four-chamber and outflow-tract views. A fetal echocardiogram was never performed in 10% of cases that documented an abnormal cardiac screen.
This study is the first in 15 years to provide data on longitudinal trends in prenatal detection of CHD in the USA. It also provides novel population-based findings on potential predictors of missed prenatal detection of CHD. Utilizing a statewide birth defect surveillance system, we examined a comprehensive list of potential risk factors not readily available in previous population studies of prenatal CHD detection. We found that the majority of CHD cases were missed prenatally and that detection rates did not increase materially over the 10-year study period. Although discouraging, the findings also suggest missed opportunities in the screening process that could be targeted to improve CHD detection.
The low rate of prenatal detection (39%) in our cohort of approximately 1500 patients is consistent with previous national and European publications14, 16, 17, 19–21. The finding that CHD detection rates did not improve significantly over the 10-year study period (1997–2007) is in contrast to an earlier population study in Atlanta26. However, in that study the initial rate of detection was very low (2.6%), and the study period overlapped with the period of rapid evolution of fetal ultrasound technology in the 1990s. Nevertheless, the lack of improvement in prenatal detection rates in our more recent time period is concerning and emphasizes the importance of identifying modifiable factors that, if appropriately targeted, could increase the sensitivity of current screening approaches.
Similar to previous studies14, 16, 17, 19–21, we found that fetal ultrasound use was nearly universal. Thus, increasing the rates of screening alone is unlikely to improve CHD detection. The major risk factor for missed CHD detection was failure to detect a cardiac abnormality on routine ultrasound, particularly for defects expected to have only an abnormal outflow-tract view. Potentially, interventions aimed at improving the skills of those performing and reviewing prenatal screening ultrasound examinations could increase the detection of serious cardiac anomalies, including many conotruncal malformations.
Opportunities for CHD detection may also have been missed because some mothers did not receive higher level imaging. Thirty-five percent of mothers with a family history of CHD and 10% of those with abnormal cardiac findings on screening ultrasound did not receive a fetal echocardiogram. While some patients may have been referred to MFM clinics first, evidence suggests that evaluation by a MFM specialist alone is insufficient when suspicion for CHD is increased by risk group or an abnormal screen27. In our study, 25% of patients scanned at a MFM office had a CHD that went undetected.
Potential risk factors for failed prenatal CHD detection include sociodemographic factors and factors that affect image quality, such as maternal body habitus. Studies examining sociodemographic factors show conflicting results20, 22. In our study, neither individual nor geographic measures of socioeconomic status were associated with prenatal CHD detection. This lack of association may be related to the current nearly universal use of ultrasound. We also did not observe an independent association of any maternal factor, such as body habitus, with CHD detection.
The primary risk factors for missed prenatal CHD detection were fetal ultrasound location, ultrasound interpreter and absence of extracardiac malformations. Most heart defects are isolated, with no extracardiac anomalies to increase suspicion. Additionally, most patients received their screening ultrasound in low-risk outpatient settings, where detection rates are lower compared to tertiary or university-based settings20, 22. The effects of location and interpreter are likely due to variations in experience, training, and equipment. This study was not designed to explore these factors. However, even in best-case scenarios, among mothers who had one of more ultrasound examinations performed at a MFM clinic and interpreted by a MFM, a quarter of significant CHD was missed prenatally.
Prenatal CHD detection was lower for those cases with an expected abnormal outflow-tract view than for those with an expected abnormal four-chamber view (of which 40% were still missed). This is consistent with other studies16, 18, 20, 22, 28, 29. The addition of an outflow view significantly enhances prenatal CHD detection, yet its successful visualization is inconsistent and not uniformly required30, 31. Currently, these views are not mandatory in recommendations for cardiac screening in the USA12, 13, although standardizing cardiac screening protocols and enhancing training could improve the ability to obtain this difficult view and improve detection rates28, 29, 32–35. In the UK and Canada, the routine assessment of outflow tracts is already part of the guidelines for screening36, 37.
This study had some limitations. It used data obtained for surveillance purposes; miscoding or data entry errors could have been present. Miscoding is unlikely, since all UBDN CHD cases undergo clinical review by a pediatric geneticist specializing in CHD or a pediatric cardiologist. Utah is predominately Caucasian, which decreases our study's generalizability, although our results confirm findings reported in other regions of the country. We were also unable to determine whether the lack of indicated fetal echocardiograms was due to a lack of referral or due to referred patients not actually having a fetal echocardiogram. The analysis of ultrasound reports must be interpreted with caution since only a minority of patients (those with a higher rate of prenatal detection) had reports available for review. Finally, in using secondary data, we were unable to review screening images directly. This will be important in future research to generate additional insights into prenatal detection.
In conclusion, in spite of nearly universal ultrasound screening, most (61%) significant CHD in this cohort was missed prenatally. Our study identified multiple points in the screening process amenable to improvement. The one likely to have the largest population effect is improvement to the initial screen done in the low-risk outpatient setting. The scarce improvement in detection rates over the recent decade suggests that initiatives to enhance the screening process have been neither effective nor widely disseminated. Since the primary risk factors for failed detection of CHD appear to be related to screening methods, targeted strategies amenable to widespread adoption in clinical practice may improve detection. The failure to detect these defects prenatally represents a missed opportunity to provide counseling and timely care for infants with CHD.
SUPPORTING INFORMATION ON THE INTERNET
The following supporting information may be found in the online version of this article:
Table S1 Demographics and descriptors of cases of congenital heart disease identified in Utah between and 2007
We would like to thank Professor Paula Woodward, MD, of the Department of Radiology, University of Utah School of Medicine and Professor Michael Varner, MD, of Maternal Fetal Medicine, Department of Obstetrics and Gynecology, University of Utah School of Medicine for their review and input on this paper. This study was supported in part by the Children's Health Research Center, University of Utah as well as by an NIH Institutional Career Enhancement Award for Dr Pinto (1KM1CA156723-01).