Physician Barriers and Facilitators for Screening for Congenital Heart Defects on Routine Obstetric Ultrasound: A National U.S. Survey

Abstract

Objectives

Prenatal detection of congenital heart disease (CHD) with obstetrical screening remains <50% in most population studies, far from is thought to be achievable. We sought to identify barriers/facilitators to screening from the perspective of interpreting physicians and to understand how these barriers/facilitators may be associated with interpretation of screening images.

Methods

Our mixed-methods studies included 4 focus groups in centers across the U.S. with obstetric, maternal-fetal medicine, and radiology providers who interpret obstetric ultrasound. Themes around barriers/facilitators to fetal heart screening were coded from transcripts. A national web-based survey was then conducted which 1) quantitatively measured reported barriers/facilitators and 2) measured physician ability to interpret fetal heart screening images. Multivariable generalized linear random effect models assessed the association between barriers/facilitators and accuracy of image interpretation at the image level.

Results

Three main themes were identified in focus groups: intrinsic barriers (i.e. comfort with screening), external barriers (i.e. lack of feedback), and organizational barriers (i.e. study volumes). Among 180 physician respondents, 104 interpreted ultrasounds. Perception of barriers varied by practice setting with non-tertiary providers having lower self-efficacy and perceived usefulness around cardiac screening. Facilitators associated with odds of accurate interpretation of screening images were knowledge (OR 2.54, p=0.002) and volume of scans per week (OR 1.01 for every additional scan, p=0.04).

Conclusion

Some of the main barriers to cardiac screening identified and prioritized by physicians across the United States were knowledge around screening and minimal volumes of scans. Targeting these barriers will aid in improving prenatal detection of CHD.

Introduction

Since >80% of congenital heart disease (CHD) occurs in mothers who have no identifiable risk factors for delivering an affected child, prenatal detection is challenging and relies on population-level screening.¹ The sensitivity of obstetric ultrasound can be as high as 85% in efficacy studies when a four chamber (4C) and outflow tract views are included in routine sonographic screening. However, though >90% of pregnant women in developed countries have an obstetric ultrasound, population-based studies show that only 30–50% of CHD cases are detected in actual day-to-day clinical practice settings.^2–4

Regional and institutional variation in the procedure for and interpretation of obstetrical ultrasound screening for CHD may be a critical determinant of lower detection rates. For example, higher sensitivity of obstetric ultrasound screening has been achieved when both 4C and outflow tract views are obtained, although both views may not be effectively performed in all settings, largely because outflow views are more challenging.⁵ Interventions aimed at improving prenatal CHD detection have not had widespread impact on detection rates. Since most interventions were studied at tertiary or academic centers ⁶, where detection rates are already higher ^3,4,7, the impact on detection may be improved by looking at barriers/ facilitators in community settings where >75% of prenatal screening for CHD actually occurs.^3,4

The factors that influence health professional practice and behaviors are complex.^8,9 Because of this, most experts agree that multifaceted interventions targeting identified barriers/facilitators are more likely to be effective.¹⁰ A well–established framework for understanding physician behaviors is that knowledge influences attitudes which then influences behavior.¹¹ Several studies have identified core beliefs, perceived usefulness, organizational barriers, self-efficacy, and inertia of current practice as important for understanding physician implementation of guidelines and use of screening tests in all areas of medicine.^8,12,13

We sought to 1) explicitly engage physicians who interpret fetal cardiac screening images in identifying determinants of successful CHD detection, 2) explore how these determinants vary among practice settings so future interventions can be effectively targeted, and 3) engage physicians in prioritizing potential targets for modifying knowledge, skills, practice standards, organizational structures, and policy.

Materials and Methods

This mixed-methods study employed focus groups to inform the development and implementation of a quantitative cross-sectional national survey. In the first part of the study we conducted focus groups with local providers who interpret obstetric ultrasounds at four geographic sites across the United States. For the second part of the study we conducted a national survey targeting the same type of providers. Institutional Review Board approval was obtained at all participating institutions for the focus group part of the study and at the University of Utah for the survey study.

Focus Groups

Focus Group Study Population and Recruitment

Focus groups were conducted at the University of Utah, Stanford University, Northwestern University and Duke University from January to May of 2014. These four sites were selected to obtain representation from across the country and because they offered collaboration with established investigators from both the Pediatric Heart Network and the Maternal Fetal Medicine Units Network. The target population for the focus groups were obstetricians [including maternal fetal medicine (MFM) specialists] and radiologists who interpret second trimester obstetric ultrasound (fetal anomaly scans). We mailed letters to potential participants using addresses obtained from the National Provider Identifier (NPI) file. We also sent out fliers through American College of Obstetrics and Gynecology state chapters to member and local institutional email listservs to recruit participants from each state.

Focus Group Methods

Using recommended techniques,¹⁴ focus groups were designed and conducted using a scripted moderator guide and face-to-face meetings while also allowing for participation by teleconference. Participants were asked open-ended questions regarding the importance of cardiac screening, their perception of its difficulty, barriers/facilitators to screening, and suggestions for improvement. Participants were then asked the relevance of candidate factors (compiled from our sonographer study¹⁵ and a literature review of key determinants of physician practice ^8,9,11,16,17) that might influence cardiac screening. They were also encouraged to contribute factors not previously identified. Participants reviewed the draft survey instrument and provided suggestions for refinement. All sessions were audio recorded and transcribed. Observational notes captured non-verbal nuances. Informed consent was obtained from all participants.

Focus Group Analysis

Transcripts were analyzed using a five-stage method: 1) familiarization with data, 2) identifying thematic framework, 3) indexing data, 4) charting, and 5) mapping and interpretation.¹⁸ Transcripts were first reviewed to determine themes and investigators met to discuss face validity and reach consensus regarding the thematic designation. The content of the audio files was coded using TAMS Analyzer software (Version 4.14b1h).

National Survey

The second part of this study consisted of a cross-sectional national survey of providers involved in ultrasound interpretation for pregnant women. The web-based survey was designed and refined using input from the focus groups discussed above.

Survey Study Population and Recruitment

The target population for the survey was primarily physician providers who interpret general obstetrical ultrasound studies. The comprehensive NPI file for taxonomy codes specific for Obstetrics (207V00000X) and Diagnostic Ultrasound in Radiology (2085U001X) was used as the sampling frame. Estimates of obstetricians and radiologists practicing in different regions of the United States were determined using the Area Resource File of the Health Resources and Services Administration¹⁹ and the 2010 geocoded NPI file from the North American Association of Central Cancer Registries.²⁰ We employed probability sampling stratified by 1) disproportionate sampling by practitioner (2:1 obstetrician to radiologist, since 10–45% of obstetric ultrasounds are interpreted by radiologists),^3,4 2) proportionate sampling by census region, and 3) disproportionate sampling of 1:1 rural and urban addresses utilizing Rural Urban Commuting Area (RUCA) codes^21,22 to create a sample of 2000 physician mailing addresses. These addresses were then matched to the American Medical Association resource email file to create a final list of 1300 physicians that could be reached by mail and email. To enhance response rates, we included a $5 pre-incentive gift card mailed with the invitation letter, since advance participant support of even small monetary amounts has proven to be more successful than larger sums promised upon survey completion with professional groups such as physicians.^23,24 To further increase participation, we used a multi-phase mixed mode recruitment strategy including a flyer, pre-notification letter, letter with a survey link, and two additional electronic reminders.²⁵

Survey Methods

The draft survey was reviewed by survey experts from the University of Utah Center for Clinical and Translational Science’s Biostatistics and Study Design Survey Core for instrument design, question format, question order and wording. It was programmed online using Checkbox v 6.0 and pretested on physicians who had participated in the focus groups. Refinements were made before was the survey was launched.

Initial survey questions determined if physicians qualified (provided prenatal care and/or interpreted 2^nd trimester obstetric ultrasound in the last six months). The web survey had two main parts. Part 1 assessed barriers and facilitators to prenatal cardiac screening using validated scales obtained from the literature on common behavioral domains used to understand physician behavior and their adoption of evidence-based recommendations ^8–11 (Table 1). The main behavioral determinants assessed included 1) knowledge, 2) perceived usefulness of fetal ultrasound cardiac screening, 3) physician self-efficacy or confidence in his/her ability to interpret fetal cardiac images, 4) perceived ease of use or thoughts on ease of interpreting cardiac screening, 5) professional expectations or beliefs about whether cardiac screening is expected by peers, supervising physicians or guidelines, 6) feedback which measured importance of some assessment of cardiac screening performance and how much feedback was received, and 7) organizational barriers/facilitating conditions of the infrastructure support to perform fetal cardiac screening. Respondents were also asked to assess their own ability to interpret cardiac screening images and estimate the time to perform interpretation.

Table 1.

Selected Survey Measures

	Definition	Scale/Item	Cronbach’s Alpha
Self-Assessment
Time	Average time to interpret fetal heart images	1-item	NA
Ability- 4-chamber	Ability to interpret four chamber views	1-item	NA
Ability - outflow	Ability to interpret outflow tract views	1-item	NA
Determinants of Practice
Knowledge	Familiarity with screening guidelines, efficacy of fetal heart screening and current detection rates	5-item	TBD
Perceived Usefulness	Usefulness of obstetric cardiac screening	2-item	0.66
Self-efficacy	Belief about own ability to interpret screening images	6-item	0.82
Social/Professional Expectancies	Belief about whether screening is expected by peers, superiors or professional guidelines	3-item	0.82
Perceived Ease of Use	Degree to which screening is free of effort	4-item	0.89
Organizational Barriers	Infrastructure to allow for screening	3-item	0.70

Part 2 of the survey provided 20 video clips of typical fetal heart screening views. The first set asked physicians to interpret the technical adequacy of 4C fetal heart images provided for screening purposes. The second set of images asked them to interpret the 4C view images provided as either “normal” or “abnormal”. The remaining two sets repeated the same for outflow tract images. To attempt to replicate real-life situations, images included those affected by maternal obesity, suboptimal fetal position, and early versus late gestational ages. Images were independently reviewed by two fetal cardiologists and a fetal cardiac sonographer and only those with 100% inter-rater reliability were included. To decrease the impact of the image itself on physician performance, there were 3 sets of 20 images and the image set (test number) was randomly selected for each participant.

Survey analysis

We reported descriptive statistics as mean ± standard deviation for continuous variables or as frequency (percentage) for categorical variables. We compared characteristics and outcomes between tertiary/academic or high-risk clinics versus low-risk or community clinics by using a t-test and chi-squared test as appropriate. Image items were classified as correct or incorrect and a total score was calculated for all images. Raw and mean scores with standard deviations were assessed for construct scales used to measure the factors identified as perceived barriers/facilitators. In addition to the validated scales, practitioner type (radiologist, obstetrician or MFM), age, experience, and gender were assessed. We analyzed the association of correctness and potential barriers/facilitators in univariate and multivariable analysis using general and generalized linear mixed effects models. The mixed effect model allowed identification of factors that influence imaging ability in a manner similar to logistic regression, but had the added advantage of allowing us to consider correlation of responses by physician (that each physician will perform similarly on questions). Variables were included in initial models if p <0.1 in univariate models and their Pearson correlation coefficients with other variables was <0.7. Final multivariable model covariates were selected using a backward stepwise technique.

We repeated subgroup analyses for physicians who practice in tertiary/academic or high-risk clinics and physicians who practice in low-risk or community clinics using the analytic strategy described above. These subgroup analyses allowed us to assess differences in the manner in which candidate factors were associated with physician screening abilities. We tested for interaction between subgroups (academic versus community practice subgroup) and the association of potential barriers with accuracy of image interpretation. All statistical tests were two-sided with a significance level of 0.05, and all analyses were conducted using Stata/MP 15 (College Station, TX).

RESULTS

Focus Groups

Four focus groups (n=43 participants) were held across the U.S. in Salt Lake City, UT (1/2014 with three MFMs, two obstetricians and one radiologist), Durham, NC (2/2014 with six MFMs, two obstetricians and one radiologist), Palo Alto, CA (3/2014 with five MFMs and one obstetrician) and Chicago, IL (5/2014 with four MFMs and six obstetricians). Three main themes around barriers/facilitators to ultrasound screening for CHD emerged: 1) individual level barriers including comfort/confidence with cardiac imaging, training, and experience, 2) external barriers such as lack of feedback and incongruent patient expectations of anomaly scans; and 3) system-level barriers including varying volume, access to expertise, and the lack of formal standards/quality control. Exemplar quotes for themes are shown in Table 2. Provider knowledge about current detection rates of CHD, attitudes regarding the usefulness of prenatal detection, and current screening protocols (i.e. recommended screening images) varied among participants. Additional common threads in the discussions were 1) variation in these barriers by practice setting and 2) tension between increased expertise/volume versus conserving resources and ensuring access. When coding by practitioner type, MFMs and obstetricians more frequently discussed system-level and individual level barriers such as ability and training (Figure 1). Among the few participating radiologists, more discussion focused on the usefulness of ultrasound prenatal screening for heart defects.

Table 2.

Themes/Sub-themes and Representative Quotes from Focus Group Participants

Theme	Exemplar Quotes
Intrinsic Barriers
Ability/Self-Efficacy	“to be perfectly honest, I never feel horribly comfortable with heart screening” “I feel reasonably comfortable, though I’m always in pursuit of learning more”
Training	“Radiologists that just come out of practice are not comfortable with fetal hearts.” “All you really do during residency is try biometry and assess how big is the baby..”
Attitude/Usefulness	“look at the huge debate over mammography …stating that screening really isn’t changing outcomes. Has anybody looked at screening in fetal echoes?” “It’s never been a lack of awareness that doing a screen on the heart was important” “it might be that it we are just stuck at the moment with a crummy screening test or one that has several significant deficits”
Extrinsic Barriers
Expectations	“a lot of patients think their ultrasound is to see their baby and get the gender. But they need to go to high-quality places, and they don’t …demand better quality.”
Feedback	“I’ve been doing this for a long time, but I have really no sense of how many I’ve missed. I could be..missing a lot” “I would also mention that getting the feedback from my referrals would increase my knowledge, … hearing back, really was valuable. So that communication between us… and those we refer to is really important”
Organizational Barriers
System-level	“with initial prenatal detection, ultrasound is done in thousands of different offices, by physicians doing low volume and low risk… “locations with much less resources you’re not going to have the access to the specialists, to have the resources to refer to them.” “I do not think you get the same ability to look at something that you want to see better … reading a scan from home.” “I did rely on the ultrasonographer, you know, I just looked at the pictures. She was the one that told me if something was weird”
Process Implementation	“So how do we get that implemented as a requirement? How do we get that implemented across the nation?” “..at the end of the day financial incentives and disincentives at the level of the organization will be what ultimately pushes change across the board”
Policy/Quality Control	“Standardize the terminology and make sure everybody uses the same thing to prevent this kind of hedgy-ness“ “We are taking on certification for nuchal translucencies and cervical length, but yet for cardiac images, a basic 4-chamber and outflow tracts views, there is practically no accreditation system”

Figure 1.

Frequency of discussion of themes in physician focus groups by physician specialty.

Survey Results

From 1300 survey invitations, 1267 were delivered. Initially 282 physicians responded to the web survey (response rate 22%) between June 2015 and May 2016. Among these only 190 met qualifying criteria. An additional, 86 did not interpret obstetric ultrasounds (or had not in the last six months), leaving 104 qualified respondents. As demographic information was not a required response, only 61 respondents gave their zip code. Among these, 28% resided in the Northeast, 31% in the South, 13% in the Midwest and 28% in the West regions of the United States.

Among 104 physicians who completed the survey, only 82 completed all the image questions. Their average age was 46 years with 22±10 years of experience and 37% were female (Table 3). The majority (58%) were MFMs, another quarter obstetricians, and the remainder radiologists; 61% practiced in a tertiary center. Table 3 shows the average reading volume and domain measures for the overall cohort and by primary practice setting. Compared to those practicing in tertiary centers, those practicing in non-tertiary settings were similar in experience and age, but differed in the volume of ultrasounds read per week. Average scores for all measured domains were higher for physicians in tertiary settings. For example, physicians in tertiary practices rated their self-efficacy in interpreting outflow images on average 0.8 points higher on a 5-point Likert response scale compared to those in non-tertiary settings (p=<0.001). The differences between average scores by setting was larger for measures of outflow tract barriers compared to 4C domains.

Table 3.

Characteristics of Participants and Univariate Analysis.

Factor	Overall		Non-tertiary		Tertiary		P value*
	N	Mean/N (STD/%)	N	Mean/N (STD/%)	N	Mean/N (STD/%)
Female	68	25 (36.76%)	24	7 (29.17%)	44	18 (40.91%)	0.34
Experience (years)	103	21.60 ± 10.27	32	21.03 ± 10.23	50	22.34 ± 9.84	0.56
Volume (obstetric scans/week)	102	35.27 ± 28.29	32	14.47 ± 10.37	50	49.84 ± 27.73	<0.001
Age (years)	67	55.94 ± 9.51	24	54.79 ± 8.30	43	56.58 ± 10.16	0.46
Additional training in cardiac	71	48 (67.61%)	25	13 (52.00%)	46	35 (76.09%)	0.038
Feedback	93	4.44 ± 0.83	32	4.13 ± 1.05	50	4.66 ± 0.58	0.005
Facilitating conditions	93	4.28 ± 0.88	32	3.81 ± 0.98	50	4.58 ± 0.68	<0.001
Self-efficacy 4Cr	94	3.98 ± 0.65	32	3.68 ± 0.67	50	4.18 ± 0.54	<0.001
Self-efficacy outflow	94	3.62 ± 0.90	32	3.15 ± 0.93	50	3.98 ± 0.71	<0.001
Outcome expectancy 4C	95	4.56 ± 0.76	32	4.36 ± 0.90	50	4.65 ± 0.71	0.11
Outcome expectancy outflow	95	4.59 ± 0.75	32	4.25 ± 0.88	50	4.73 ± 0.68	0.007
Perceived ease of use 4C	95	4.39 ± 0.86	32	4.00 ± 0.91	50	4.63 ± 0.78	0.001
Perceived ease of use outflow	95	4.20 ± 0.83	32	3.78 ± 0.90	50	4.48 ± 0.73	<0.001
Knowledge	71	0.50 ± 0.26	25	0.37 ± 0.24	46	0.58 ± 0.25	<0.001
Outcomes
Time to interpret 4C images (min)	82	2.78 ± 2.89	32	3.25 ± 3.65	50	2.48 ± 2.27	0.24
Time to interpret outflow images (min)	81	3.57 ± 3.27	31	3.74 ± 3.90	50	3.47 ± 2.86	0.72
Overall accuracy (%)	82	0.67 ± 0.14	32	0.62 ± 0.13	50	0.70 ± 0.15	0.02
Accuracy 4C (%)	82	0.71 ± 0.18	32	0.66 ± 0.16	50	0.75 ± 0.19	0.02
Accuracy outflow (%)	72	0.61 ± 0.15	26	0.59 ± 0.17	46	0.63 ± 0.14	0.32
Accuracy adequate (%)	82	0.59 ± 0.16	32	0.55 ± 0.15	50	0.61 ± 0.17	0.12
Accuracy normal (%)	80	0.76 ± 0.18	30	0.70 ± 0.19	50	0.79 ± 0.17	0.04

^*Data are presented as mean ± SD for continuous measures and frequency (percentage) for categorical measures. Comparisons between non-tertiary and tertiary use Chi-square test for the categorical variables and T-test for continues variables 4C – four chamber

A quarter of physician respondents indicated that they spoke with the imaging sonographer “less than half of the time” or “never” when reading obstetric ultrasounds. About half of physician respondents reported having access to a pediatric cardiologist. About 60% of obstetricians and radiologists had access to an MFM specialist. A small but notable number of physicians (4%) reported that cardiac imaging was not a required part of a 2^nd trimester obstetric ultrasound. An additional 10% reported that imaging of the outflow tracts was not required. The majority interpreted both cine and still clips of the heart. About half (53%) of reading physicians indicated their training did not adequately prepare them to interpret fetal cardiac images and the majority (67%) had completed additional training in cardiac screening.

Image interpretation

The overall accuracy in interpretation of the 1420 fetal cardiac screening images was 67±14% (Table 3). Accuracy for outflows was lower (61%) than for 4C views (71%) and lower for adequacy of the screening image than for interpretation of normal versus abnormal images (59% and 76% accuracy respectively). Most measured factors were significantly associated with odds of correct interpretation of all fetal heart screening images as well as subsets of image types (4C, outflow, adequacy and normal vs abnormal) on univariate analysis (Table 4).

Table 4.

Univariate Associations of Factors with Accuracy of Fetal Heart Image Interpretation

Barriers	Overall (N=1420)	4C (N=710)	Outflow (N-710)	Adequacy (N=710)	Normality (N=710)
	Odds Ratio (95% CI)	Odds Ratio (95% CI)	Odds Ratio (95% CI	Odds Ratio (95% CI)	Odds Ratio (95% CI)
Self-efficacy 4C	1.30 (1.07, 1.58)	1.45 (1.11, 1.91)	1.17 (0.88, 1.56)	1.14 (0.86, 1.50)	1.56 (1.17, 2.08)
Self-efficacy outflow	1.20 (1.05, 1.37)	1.26 (1.05, 1.52)	1.15 (0.95, 1.40)	1.14 (0.94, 1.37)	1.30 (1.07, 1.58)
Outcome expectancy 4C	1.28 (1.11, 1.48)	1.35 (1.10, 1.66)	1.22 (0.98, 1.51)	1.26 (1.02, 1.56)	1.33 (1.08, 1.64)
Outcome expectancy outflow	1.28 (1.11, 1.48)	1.40 (1.15, 1.72)	1.17 (0.95, 1.46)	1.19 (0.96, 1.46)	1.43 (1.15, 1.76)
Perceived ease of use 4C	1.20 (1.05, 1.37)	1.24 (1.03, 1.49)	1.17 (0.96, 1.42)	1.15 (0.95, 1.39)	1.27 (1.05, 1.55)
Perceived ease of use outflow	1.32 (1.16, 1.52)	1.43 (1.18, 1.72)	1.24 (1.01, 1.51)	1.27 (1.04, 1.54)	1.42 (1.17, 1.73)
Time to interpret 4C images (min)	1.00 (0.95, 1.04)	1.01 (0.95, 1.07)	0.97 (0.90, 1.06)	1.03 (0.96, 1.10)	0.97 (0.90, 1.03)
Time to interpret outflow images (min)	1.01 (0.99, 1.03)	1.02 (0.99, 1.05)	1.00 (0.98, 1.03)	1.01 (0.99, 1.04)	1.01 (0.98, 1.04)
Experience (years)	1.00 (0.99, 1.01)	1.01 (0.99, 1.02)	1.00 (0.98, 1.02)	1.01 (0.99, 1.03)	0.99 (0.98, 1.01)
Volume (obstetric scans/week)	1.01 (1.00, 1.01)	1.01 (1.00, 1.02)	1.01 (1.00, 1.01)	1.01 (1.00, 1.01)	1.01 (1.00, 1.02)
Feedback	1.32 (1.15, 1.52)	1.49 (1.22, 1.82)	1.17 (0.95, 1.44)	1.26 (1.02, 1.54)	1.44 (1.17, 1.77)
Knowledge	3.10 (1.86, 5.17)	4.63 (2.20, 9.76)	1.16 (0.79, 1.69)	3.57 (1.73, 7.37)	2.94 (1.39, 6.26)
Facilitating conditions	1.29 (1.13, 1.48)	1.33 (1.10, 1.60)	1.27 (1.04, 1.55)	1.19 (0.98, 1.44)	1.46 (1.20, 1.78)
Tertiary practice	1.37 (1.06, 1.76)	1.59 (1.13, 2.24)	7.06 (2.23, 22.38)	1.35 (0.94, 1.92)	1.43 (0.98, 2.08)
Defect Image	5.25 (2.86, 9.65)	3.77 (1.84, 7.73)	1.17 (0.88, 1.56)	5.23 (1.90, 14.41)	1.46 (0.60, 3.55)
Specialty
MFM	Ref	Ref	Ref	Ref	Ref
Ob/Gyn	0.60 (0.45,0.80)	0.51 (0.35, 0.75)	0.73 (0.47,1.13)	0.73 (0.48, 1.09)	0.47 (0.31, 0.72)
Radiology	0.83 (0.59, 1.17)	0.76 (0.47, 1.23)	0.90 (0.54, 1.50)	0.81 (0.50, 1.31)	0.84 (0.50, 1.41)

Odds ratios are for each 1 score increasing in above barriers or each 10% increasing in Knowledge, Tertiary practice vs. Non-tertiary practice, and Abnormal or Inadequate image vs. Normal or Adequate image.

4C – four chamber, MFM – maternal fetal medicine, Ob/Gyn – obstetrics and gynecology

Multivariate analysis

In multivariable analysis, overall accuracy in image interpretation was associated with physician knowledge (increasing odds of accurate interpretation by 2.2 if the knowledge score increased to 100%, 95% CI 1.2–3.8), practice setting (non-tertiary OR 0.7, CI 0.4–0.99) and volume of ultrasounds read per week (OR 1.01 for every additional scan, p=0.04, Table 5). While volume remained associated with accurate interpretation of outflow tract screening images, it was not associated with accuracy for 4C screening images. Perceived feedback was independently associated with accuracy for both outflow and 4C imaging (OR 1.5, CI 1.1–2.1). The interactions between measured barriers and type of screening image (4C versus outflow) were not significant (Supplementary Table 1), but interactions between barriers and practice setting (tertiary versus non-tertiary) were significant.

Table 5.

Multivariable Regression Model of Factors Associated with Accuracy in Fetal Heart Image Interpretation

Barrier	Overall (n=1420)		4C (N=710)		Outflow (N=710)		Adequacy (N=710)		Normality (N=710)
	Odds Ratio (CI)	P value	Odds Ratio (CI)	P value	Odds Ratio (CI)	P value	Odds Ratio (CI)	P value	Odds Ratio (CI)	P value
Self-efficacy outflow	1.02 (08,1.3)	0.840	0.96 (0.7,1.3)	0.783	1.08 (0.8, 1.4)	0.586	1.06 (0.8, 1.4)	0.678	0.97 (0.7,1.3)	0.841
Outcome expectancy 4C	1.08 (0.8, 1.5)	0.648	0.89 (0.6,1.4)	0.61	1.29 (0.8, 2.0)	0.254	1.28 (0.8, 1.9)	0.269	0.88 (0.5,1.4)	0.593
Outcome expectancy outflow	0.99 (0.7, 1.5)	0.945	1.34 (0.8, 2.3)	0.307	0.73 (0.4, 1.3)	0.262	0.77 (0.4, 1.3)	0.357	1.29 0.7,2.3)	0.384
Perceived ease of use 4C	0.89 (0.7, 1.1)	0.328	0.79 (0.6, 1.1)	0.182	1.00 (0.7, 1.4)	0.992	0.99 (0.7, 1.4)	0.972	0.77 (0.5,1.1)	0.149
Volume (obstetric scans/week)	1.01 (1.0, 1.01)	0.038	1.00 (0.99,1.01)	0.349	1.01 (1.0, 1.02)	0.043	1.00 (0.99,1.01)	0.412	1.01 (1.0,1.02)	0.028
Feedback	1.15 (0.9, 1.5)	0.233	1.48 (1.1, 2.1)	0.023	0.89 (0.6,1.3)	0.521	1.21 (0.9, 1.7)	0.256	1.11 (0.8,4.5)	0.536
Facilitating conditions	1.19 (0.9, 1.5)	0.175	1.06 (0.7, 1.5)	0.768	1.35 (0.9, 1.9)	0.102	0.97 (0.7, 1.4)	0.865	1.52 (1.04,2.2)	0.031
Knowledge	2.54 (1.4, 4.6)	0.002	3.66 (1.6,8.6)	0.003	1.83 (0.8, 4.2)	0.151	3.35 (1.4, 7.6)	0.004	1.90 (0.8,4.5)	0.149
Non-tertiary practice	0.67 (0.5, 0.99)	0.045	0.71 (0.4, 1.3)	0.248	0.61 (0.3, 1.1)	0.086	0.73 (0.4,1.3)	0.269	0.60 (0.3,1.1)	0.082
Image type	6.87 (3.5,13.5)	<0.001	5.51 (2.4, 12.9)	<0.001	7.56 (2.3,24.6)	0.001	6.51 (2.3,18.7)	0.001	1.75 (0.6,4.9)	0.29

^*Multivariable models were also adjusted for image test version and view (outflow vs. 4C). Odds ratios are for each 1 score increasing in above barriers, each additional ultrasound per week for volume, or increase in knowledge to 100%, and abnormal/inadequate image vs. normal/adequate image

In subgroup analysis of physicians practicing in tertiary or high-risk settings, knowledge was the only factor associated with accuracy of image interpretation (OR 2.9, CI 1.3–6.7, Figure 2a). For physicians who practiced in non-tertiary settings, facilitating conditions were associated with improved accuracy (OR 1.5, CI 1.02–2.3, Figure 2b).

Figure 2.

Multivariable models of odds of accurate image interpretation by practice setting. A. Tertiary Practice. B. Non-tertiary practice. Factors with p>0.05 on univariate analysis or correlation ≥0.7 were excluded from multivariable models.

Discussion

Our study assessing barriers/facilitators for detection of CHD using fetal ultrasound is among the first to identity physician cited determinants of performance. These may serve as intervention targets for improving the effectiveness of prenatal screening. Physicians who interpret these ultrasounds identified several barriers including not only knowledge, training and ability as expected, but also lack of feedback, quality control and issues with consistent volumes. Of these determinants, knowledge about screening and volume of studies appeared most pertinent to measured ability to interpret fetal heart screening images on our survey. Feedback received on screening and facilitating conditions also influenced accuracy of interpretation in certain aspects.

The gap between what ultrasound screening for heart defects can achieve and what is actually accomplished in practice has been the subject of many studies and editorials.^26–29 While training and experience are the targets of many interventions,^6,30–34 they are largely focused on sonographers. Expanding interventions to include physicians who interpret these images may be warranted since lack of training and familiarity with cardiac imaging were cited as key barriers. Even physicians practicing in high risk settings have little knowledge about fetal cardiac screening guidelines ³⁵ and sensitivity of screening views. Since physicians who interpret ultrasounds represent multiple specialty and subspecialty fields with variation in training in cardiac imaging in the U.S., the level of familiarity likely depends on individual interest.³⁶ If the expected baseline of knowledge for image interpretation is standardized with well-accepted guidelines for training and maintenance of expertise in obstetric ultrasound, detection should increase.³⁷ As some have suggested, providing focused training for the cardiac portion of the exam, thought to be the most complex part of the fetal ultrasound, may be effective.¹⁵

Systematic feedback has been useful for improving detection of heart defects.³⁸ While providers in tertiary settings typically have ready access to expertise, feedback may not be formalized. Feedback is even more challenging for the lone radiologist or obstetrician practicing in low-risk community settings. Despite this, single centers have developed supportive data structures with formalized feedback.³⁹ Participants in the survey also cited certification similar to that required for nuchal translucency and cervical length assessments as a high priority to improve screening. Indeed, regular audits for maintenance of certification or quality control have been successful ^38,40–42 and could be implemented on a wider scale by either a governing body such as American College of Obstetrics and Gynecology or the American Institute of Ultrasound Medicine.

The impact of perceived facilitating conditions on image interpretation in community settings where detection rates are low is intriguing, ^3,4,7,43 since equipment, age at referral and maternal body habitus have been previously cited as common barriers.⁷ The lack of supportive structures and equipment in these settings may influence ability and skill compared to tertiary settings with sufficient technology and assistance. Stakeholders (MFMs and pediatric cardiologists) should identify the minimal level of infrastructure needed to effectively perform screening ultrasounds and advocate for inclusion of these elements into health systems.

Physicians cited volume as an important factor in accurately interpreting cardiac images consistent with studies in other areas.⁴⁴ The guidelines for maintenance of practice and accreditation/certification services already recognize that a minimum number of obstetric ultrasounds is needed to maintain skills/expertise.³⁷ Possible ways to address volume issues include restricting reimbursements to centers that read a minimum number of studies, but maybe more realistically, providing support within referring regions for telemedicine services, consults and over reads. Measures that improve practice in even low volume settings have the advantage of allowing continued local access for pregnant mothers.^39,45

National screening programs in countries with more standardized protocols for training, feedback, and minimum study volumes have led to increased detection of CHD. ⁴⁶ Determining how to implement similar multifaceted programs in the highly variable screening settings in the U.S. will be crucial to improving care for CHD families.

Limitations

This study is limited by the possibility of selection bias in the focus group participants and survey respondents who may have been more interested in the topic of cardiac imaging than non-respondents. The results may not be generalizable to all geographic regions particularly for the Northeast which was not represented in our focus groups and the Midwest which had lower representation in the survey. Additionally, like all surveys, there is a risk of nonresponse bias. This risk may be accentuated by our small sample size. While response bias is also a possibility, the survey instrument and the individual self-administration were designed to minimize this bias. While we examined differences in barriers to screening by practice setting, there are likely additional variabilities in settings and providers which were not represented. Though we analyzed barriers to screening within a behavioral framework, our measure of image interpretation on a web survey theory was a surrogate outcome for true detection of CHD on obstetric ultrasound. Given the anonymity of the survey, we were unable to relate barriers/facilitators measured to true detection rates of individual physicians or practices.

Conclusion

In this national survey of physicians, we identified and prioritized barriers/facilitators to effective cardiac screening that are affecting physicians in practice across the United States. These novel results can be combined with the previously identified barriers for sonographers¹⁵ to inform targeted interventions based on the perspectives of those who are at the forefront of population screening prenatally for CHD.