A comprehensive literature search was conducted to identify studies published in the English language between January 1980 and June 2019 that reported on mortality and survival from CHD, separated by SDoH category. The database searches used a combination of Medical Subject Headings (MeSH) and keywords for social determinants of health, mortality and congenital heart disease in four electronic databases: OVID Medline, OVID Embase, Web of Science (Core Collection) and CINAHL (Supplementary Section 1). Iteration of reference lists from included articles was performed to identify studies that were not returned in the database search.

Title and abstract screening was performed using the Rayyan web application (10), followed by full text screening by two reviewers (RT and RF). Data were abstracted for individual study by the primary reviewer (RT) and checked for accuracy by the second reviewer (RF). Conflicts were resolved by consensus and discussion with a third reviewer (AK). Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines were followed (11).

Randomized and non-randomized clinical trials, observational and case-control studies on mortality and survival from CHD separated by SDoH group were included in the study. Case reports, case series with <10 patients, systematic/narrative reviews, abstracts only, editorials, letters to editors, commentaries, studies reporting prenatal outcomes and focused solely on gender, nutrition, and patent ductus arteriosus were excluded.

Data were abstracted for individual study by the primary reviewer (RT) and checked for accuracy by the second reviewer (RF).

Relevant social, economic, environmental, cultural, and behavioral SDoH were included in the analysis. We noted that standard and/or similar definitions of SDoH have not been used across all the included studies. However, in order to gain a comprehensive understanding of the SDoH and their impact on mortality in CHD, we chose to include and analyze studies providing data on most SDoH and categorized patients whenever possible and if needed within standard definitions available. For example, deprivation was classed in different studies either on the basis of household income or relative poverty/deprivation level. For this study, we analyzed patients by categorizing them as the most and least deprived. Another such example is for hospital status, based on case volume or teaching hospital. In this instance, we chose to keep the original study definitions of high volume or teaching hospital versus low volume or non-teaching hospital. Similarly, geographical barriers for access to care were explored by using a combined category of urban versus rural hospital or residence, and distance to hospital, without altering the original definitions used in the included studies. Details of definitions of all SDoH used in the included studies and their categorizations are provided in Supplementary Section 2.

Summary statistics were used to describe extracted data. The primary outcome of interest was mortality; no secondary outcomes were studied. Meta-analysis was carried out to determine a pooled estimate of the direction of effect and degree of association between a SDoH and CHD mortality only if three or more studies met the inclusion criteria. To facilitate comparison between patient groups, data were stratified by timing of mortality (inpatient mortality, neonatal mortality, mortality in the first year and long-term mortality after 1 year) and risk category. All ductal dependent CHD or those with significant operative intervention or postoperative care were classified as critical CHD (Table 1). When it was not possible to perform an analysis on a particular SDoH using studies from the same risk category or mortality timing group (due to lack of clarity within the studies), analyses were carried out to understand a possible overall effect of this SDoH on CHD mortality. If two or more studies from the same cohort or database and timeframe qualified, both studies were included in the overall results, but only the study with the greatest number of participants was included in the MA. If more than one CHD type from the same risk category was present in a study and it was not possible to combine data (for example, when patients had two or more coexistent forms of CHD), data for the CHD type that best facilitated comparison with other studies was extracted and used in MA.

A random effects model was used to calculate odds ratios. The true effect size was assumed to vary between studies. The chi-square test for homogeneity was utilized to assess heterogeneity when combining studies with similar exposure and outcome measures. Statistical significance was defined at p <0.1 rather than p <0.05 level, given the low power of this test in a meta-analysis when there are few included studies or studies with small sample sizes (12). An I² statistic was calculated to quantify heterogeneity in the MA. An I² ≤ 30% was considered to have low heterogeneity, 30–50% moderate heterogeneity and >50% was likely to represent substantial or considerable heterogeneity (12). All analyses were performed using STATA Version 16.1, College Station, TX, United States.

All included studies were assessed for risk of bias (ROB) by the primary reviewer (RT) and checked for accuracy by the secondary reviewer (RF). All studies were observational, non-randomized; therefore, the Cochrane Risk of Bias in Non-Randomized Studies (ROBINS-I) tool was used. ROB was classed as low, moderate, serious or critical (13). A funnel plot was generated for studies that included data on race/ethnicity (the variable with the largest number of included studies) to assess publication bias.

This project was judged by the Research Governance & Integrity Office at the London School of Hygiene and Tropical Medicine as exempt from ethical review when RT was affiliated with the institution in 2019. Individual patient health data were not accessed for this study.

Of the 65 studies included in the review (2, 14–77), 63 were retrospective, observational studies; two were prospective. Data from regional, national, or international databases or registries was published in 53 studies. Forty-eight studies (74%) contained data exclusively from the United States. The basic characteristics of the 65 included studies are described in Supplementary Table 1.

Among the total 546,981 patients with CHD, 34,080 died. The age range of the patients ranged from 0 days to 91 years, and seven studies included data on adult CHD (ACHD) and/or followed patients up for more than 20 years (22, 45, 67, 69–72). Although adult data were described in the individual studies, sufficient SDoH variables were not available on these patients to allow for MA in the adult age group. Not all patients underwent surgical interventions, and thus unrepaired ACHD patients were included in the analysis [Verheught et al. (71)]. Mortality according to both SDoH and CHD risk categories could be determined in 40 studies. The number of studies that published data on each SDoH were as follows: race/ethnicity (n = 52), deprivation status (n = 16), insurance status (n = 15), maternal age (n = 12), maternal education (n = 9), single/multiple pregnancy (n = 5), urban/rural hospital/residence or distance to specialist care (n = 7), hospital case volume (n = 11), surgery at teaching/non-teaching hospital (n = 3), parental occupation (n = 2), receiving public assistance (n = 1), and number of adult care givers (n = 1). Only relevant positive and negative results are discussed due to the large number of variables analyzed.

Insufficient data were available to perform MA for some SDoH. Taylor et al. noted that patients with a single adult caregiver were more likely to have increased interstage mortality following the Norwood procedure (66). This study did not find a relationship between employment status or public assistance status and mortality. Kucik et al. presented data on employment and found no significant relationship with neonatal or infant mortality for seven different types of CHD (55).

A number of individual studies performed multivariable analyses, attempting to understand the modifying as well as confounding effects of different SDoH on CHD mortality. Kucik, et al. presented adjusted hazard ratios (HR) for mortality by race/ethnicity category and concluded that higher infant mortality among the Black patients persisted after adjustment of potential confounding SDoH factors (54). Similarly, Nembhard et al. presented HRs for mortality by racial/ethnic category in Texas, United States (46). They noted that Black children had a significantly increased likelihood of death in the post-neonatal period compared to White children after adjustment for factors such as maternal age, education and residence, infant sex, gestational age and birth weight.

Using the ROBINS-I tool, 14 studies had moderate ROB, 48 serious and three with critical ROB (Supplementary Table 1).

A funnel plot was generated to assess publication bias in studies presenting White versus non-White race/ethnicity as this comprised the largest number of studies in this review (Supplementary Section 3). This funnel plot shows relatively large studies with small standard errors at the apex. There is asymmetry suggesting a publication bias, with smaller studies tending to show a larger effect of non-White race/ethnicity as a risk for CHD mortality.

This SR-MA provides important perspective into the ages and types of CHD that are most likely to be affected by SDoH domains and therefore, a role for focused interventions during those stages of CHD care. A number of interventions and practices could be introduced to achieve equity in CHD outcomes. For example, by increasing research and data collection in routine clinical care, we could further improve our understanding around the impacts of SDoH on numerous other CHD outcomes such as hospitalizations and quality of life. There is an urgent need to implement a more robust and comprehensive patient assessment with tools to identify SDoH and manage those at highest risk of poor outcomes in CHD. Frameworks to create such tools exist (94) and can be expanded to account for a wide array of SDoH.

Tailored services could be developed to improve CHD outcomes among individuals with specific SDoH. The delivery of health services by community health workers (CHW) and medical home model to offer centralized access to CHW resources, care coordination, home visiting and interagency collaboration are attractive options that require further research (95–97). It is likely that many families may have multiple SDoH which put them at higher risk of poor health outcomes, and therefore larger health policy changes would be required to allow for adequate resource allocation and development of the community-level SDoH interventions. Overcoming SDoH barriers in patients with chronic needs will require large health system level innovations to engage communities and multi-pronged interventions at the patient, provider, health system and local and state government levels to address these barriers. There is limited research on interventions for these marginalized communities that can inform widespread policy efforts to improve access to timely, high quality healthcare for CHD patients in order to narrow the racial/ethnic disparities in CHD outcomes highlighted in our review. There is also a great need to provide healthcare workers with the training required to address health disparities and literacy barriers. Finally, systematic research strategies using experimental and quasi-experimental designs will need to be employed to evaluate the impact of SDoH interventions in achieving a change in patient level outcomes.

There are several strengths to this analysis. The study followed recommendations on best practice for systematic reviews by registering in PROSPERO, adhering to a priori determined project protocol and a methodological robustness in search, extraction and analysis as per the PRISMA Guidelines.

It is unknown if studies had reliable documentation from patients of their race/ethnicity. While this review attempted to categorize SDoH such as race/ethnicity by utilizing clear definitions which best reflected standardized terms recommended in the literature (Table 1 and Supplementary Section 2), it is possible that some studies may have used variations in definitions (98). Many of the SDoH studied lack standardized definitions, posing challenges to analysis. However, this study attempted to create broad definitions to enable analyses and provide an indication of the impact of these SDoH on CHD mortality. It was also not possible to adjust for confounding factors, such as prematurity and low birth weight due to lack of data within studies. Double-counting may have occurred if patients were included in more than one database or registry. However, few analyses contained more than one study from a database/registry and specific efforts were made to avoid inclusion of studies from the same geographical area and timeframe in the same analysis. Only studies that presented crude data were included. It is likely that certain SDoH which are experienced solely as an adult may affect these outcomes differently, but the majority of studies included in this review described pediatric outcomes. The majority of included studies were from the United States. It is likely that SDoH and their associated inequities may be different in other countries. Many studies included in this SR-MA were deemed to have a high RoB, explained by observational study design and that they did not primarily aim for assessment of the SDoH. These studies, therefore, had a higher likelihood of confounding factors that may have influenced mortality outcomes. As such, with appropriate categorization of studies and acknowledgment of RoB, MA was felt to provide the best indication possible of SDoH impacts on CHD mortality and a foundation for future research on this important issue. Careful research examining the direct impact of SDoH on CHD care is necessary to collect and publish the fullest possible data on all potential SDoH variables that could impact mortality. Numerous factors such as food insecurity and housing could not be evaluated due to a lack of literature that directly assessed these factors in patients with CHD.