Overall, this study of cardiac care in Haiti demonstrates significant associations between ecological independent variables and clinical outcomes. The young median age in the patient database (5.87 years) supports existing evidence that cardiac diseases of childhood, including CHD and RHD, are a notable source of CVD burden in Haiti [6,7,8,9, 22]. However, because significant CHD and RHD are often symptomatic, they may rise to clinical attention earlier and more frequently than hypertensive and atherosclerotic forms of CVD. By natural course, hypertension and ASCVD remain asymptomatic for decades; thus, for identification they require routine population-level screening that is largely unavailable in Haiti [23]. The lack of ASCVD patients in the HCA database is unsurprising given the limitations of the primary healthcare system in Haiti to detect these pathologies. Therefore, this study does not provide insight into the comparative burden of hypertension and ASCVD in Haiti relative to CHD and RHD.

These data further revealed that the West department was overrepresented in the HCA patient database relative to the population share of West in Haiti at large. Meanwhile, the more rural South, South-East, North, North-East, Grande d’Anse, and Artibonite departments were relatively underrepresented. Beyond the fact that three of the five clinical sites are located in West, this patient distribution likely reflects differences in referral patterns due to a higher degree of urbanization in West [13], which contains the entire Port-au-Prince metropolitan area. Urbanization may increase HCA referrals because patients are closer to the referring facilities where their cases can be identified and closer to HCA clinics where referrals can be completed. Also, urban patients are more likely to have higher socioeconomic status, which may facilitate care access.

Indeed, the relative paucity of rural patients in the HCA database result aligns with existing literature on rural access barriers to specialty care in low- and middle- income countries. For instance, a 2021 study in rural Madagascar found that even in the presence of referral programs strengthened by the health system, geographic barriers leading to increased referral travel times were a primary driver of diminished access to specialty care in rural Madagascar [24]. Of note, a large portion of HCA patients present for cardiac complications of RHD, which begins with an infectious etiology and is less likely to be adequately treated and prevented in rural settings. Overall, these findings suggest that more robust, decentralized networks of HCA referrals and evaluations are needed to bring cardiac care closer to the rural poor, as exemplified by the World Health Organization PEN-Plus strategy [25].

While most reflective of a similar rural–urban access disparity within Haiti, the urban-heavy distribution of HCA patients may also be compatible with other epidemiologic trends. In a 2018 review, Bickler et al. note that in Sub-Saharan Africa, rates of non-communicable diseases like CHD often rise fastest in the urban regions of a country, as compared to rural and less industrialized areas; this may be partially attributable to documented genetic and epigenetic modifications that accompany urbanization [26], although no research to date explores similar biomolecular changes in a Haitian context.

Additionally, it is critical to note that the sociopolitical circumstances in Haiti during the timeframe of HCA data collection (April 2012–December 2020) may have masked some intra-departmental disparities in access to care. Due to the COVID-19 pandemic and the Threat Level 4 travel advisory issued by the US State Department in response to kidnapping and uprising threats in Haiti, its national border was effectively closed for large parts of 2020 [14]. While HCA continued to operate medically, this translated into a universal halt in surgical therapy (which, due to these security concerns, is now exclusively conducted at HCA’s international partner sites given the absence of cardiac surgical capability in Haiti) because eligible patients could not be transited abroad. This uniform reduction in treatment availability may have diminished some of the regional differences normally present in HCA care outcomes.

Furthermore, patient movement across administrative departments over time was not a significant analytical concern during the time frame of this study; while some patients emigrate abroad after coming into contact with HCA, this also comprised only a small minority of the analyzed dataset. After the conclusion of the study period, however, Haitian population movement increased substantially due to a catastrophic earthquake and widespread sociopolitical unrest following the assassination of former president Jovenel Moïse in 2021. Therefore, future studies of Haitian and LMIC cohorts should also consider the potential of ongoing internal displacement and international migration flows to impact clinical outcomes of the studied population.

Our univariate logistic regression analysis showed a negative association between departmental childhood growth retardation and active care representation in HCA. Conversely, higher rates of childhood growth retardation were positively associated with loss to follow-up in HCA. Given the high prevalence of severe malnutrition in Haiti, childhood growth retardation has been identified as a key public health target [27]. In addition, the association between growth retardation and the percentage of patients receiving any form of healthcare is supported by existing literature. Analyses of Haitian national survey data in the aftermath of the 2010 earthquake suggested that rates of under-nutrition among children below five years declined in tandem with increased antenatal care attendance and associated “baby tents” designed to promote infant health [28].

At the department level, multivariable regression models trended towards a positive association of active care representation with adult employment rate (although this association was not statistically significant in our sample). Moreover, there was a statistically significant, negative correlation between departmental economic index and rates of preoperative death in HCA. The relationship of adult employment and income to healthcare access is well-established in LMICs given the increased ability of employed individuals to afford out-of-pocket expenses in areas without widespread health insurance [18].

In multivariable models, travel time to the nearest healthcare facility also trended towards a negative association with the proportion of patients in active care without reaching the threshold of statistical significance. Similarly, there was a nearly-significant, positive correlation between travel time to emergency healthcare and the departmental rate of HCA patients lost to follow-up. Because increased travel time is often interlinked with other social determinants of health such as financial barriers and the cultural acceptability of seeking professional medical services, multivariable analysis here was necessary to parse the individualized contribution of travel time [29]. Travel time has also been identified as a driver of healthcare disparities between urban and rural areas in LMICs [30].

In this study, however, it is worth considering that referral biases may cause overrepresentation of patients with shorter travel times to HCA clinics from within the broader departmental population. For instance, we found that the rate of HCA patients deceased preoperatively by department was negatively correlated with travel time to emergency healthcare facilities in multivariable regression. While this unexpected finding requires further investigation, it raises the possibility that a greater portion of symptomatic patients—like those with decompensated CHD or advanced RHD—in departments with accessible, well-dispersed healthcare facilities are more likely to be referred to HCA for advanced disease, whereas symptomatic patients in departments where healthcare facilities are less readily accessible to all may be more likely to die before even making contact with a specialty cardiac provider like HCA.

Notably, the effect sizes for each covariate within these multivariable regressions appear small, as may be reasonably expected when correlating population-level indicators with the clinical outcomes of a specialty cardiac patient group. Importantly, however, the collective set of covariates in each model provides adequate predictive power, as indicated by high R2 values. This finding, in turn, demonstrates how aggregate, widely available population health and sociodemographic data predicts clinical outcomes disparities within the particular cardiac patient cohort of Haiti Cardiac Alliance. For this reason, our study suggests that readily available population-level data may be a useful starting point for identifying healthcare inequities confronted by specific public health interventions and policies, specifically within Haiti and possibly in other resource-constrained, data-limited settings as well.

Counterintuitively, active care representation was negatively associated with access to qualified prenatal care and positively associated with lost to follow-up representation when controlling for other covariates. While a precise mechanism for this finding is unclear, departments with greater availability of qualified prenatal care are also more likely to have other established primary and secondary healthcare providers. It is plausible that patients in these areas may view HCA as a pathway to cardiac surgery rather than as a long-term cardiac care provider, and may transfer their postoperative follow-up care and long-term medical management from HCA to local providers.

This hypothesis is supported by the intuitive finding that prenatal care access was negatively correlated with pre-operative death rates. In other words, the prenatal care variable in EMMUS-VI may serve as an indicator of overall healthcare availability. Departments with weaker regional health systems would presumably have higher preoperative mortality in HCA due to both: a) greater risk of patients facing other unmet health needs and/or medical emergencies while awaiting cardiac surgery; and b) delayed referral to secondary healthcare services like HCA, until surgery is no longer clinically feasible and mortality is imminent. At the same time, departments with stronger regional health systems might plausibly be less dependent on HCA for long-term medical management of postoperative patients, who may transfer their care to regional providers postoperatively and lose contact with HCA over time.

This study has numerous strengths. It is the first analysis of Haiti’s largest known cardiac patient database, which has significant statistical power owing to a large sample size. This rare source of patient-based data from Haiti allows us to offer one of the first academic approaches to characterizing regional health disparities within Haiti, as defined by clinical outcomes. Because the study uses data derived from zero-cost cardiac consultations in public and aid-based clinics, it is broadly inclusive and likely to represent the general Haitian population with a high degree of external validity. Drawing from multiple clinic sites across Haiti, the study includes data from all 10 administrative departments and thus provides a national-level picture of the cardiac care landscape.

As opposed to current literature that primarily describes healthcare access in specific settings of Haiti [17, 22], this study is also among the first to identify factors associated with current disparities in cardiac care access using readily available department-level sociodemographic data. Using objective data from the EMMUS-VI national survey, the study proposes a starting point to triangulate “access to care,” a concept that is difficult to measure directly. This difficulty is particularly pronounced in Haiti, where experiences of poverty and resource scarcity are quite heterogeneous between rural and urban settings. In these ways, our study draws from department-level data on both HCA clinical outcomes and national sociodemographic and population health indices to advance the literature on social determinants of health in Haiti.

This study is not without limitations, which were often imposed by data availability. First, defining clinical outcomes in three broad categories (active care, deceased preoperatively, or lost to follow-up) does not account for the full spectrum of clinical experiences over time. For instance, departments in which a high proportion of patients have been in consistent active care for years may differ from those where most patients have been in active care for weeks to months, yet our classification system does not reflect these differences. It also does not differentiate patients with sustained active care from those who were lost to follow-up for some time and subsequently returned to care (possibly after interim disease progression). Moreover, while risks of patient misclassification are not uncommon in data analyses given the possibility of primary data source errors, Haiti Cardiac Alliance mitigates these risks through fieldwork protocols that involve frequent updates and routine quality controls.

Second, we were unable to analyze patients by type or severity of cardiac pathology, which may substantially impact clinical outcomes. For example, medically-treated patients for whom surgery is recommended are especially vulnerable because they face higher mortality rates without surgical intervention [31, 32], despite optimal medical management; our study merges this group with patients who lack operative indications. On the other hand, many patients with surgical cardiac pathologies present to HCA at such an advanced stage in the natural course of their disease that surgery is prohibitively high-risk. Because we could not explicitly label or know if patients' primary reason for death was structural unavailability of surgery, there is potential for lead-time bias as patients screened or diagnosed early on in a severe or unalterable course of disease would be classified in the active care group as opposed to the deceased preoperatively group. Nevertheless, it is important to note that this study does not serve to evaluate effectiveness of HCA operations or therapeutic interventions, but rather baseline care access across Haiti. Therefore, we believe that including mortalities during treatment in the active care group still provides a reliable marker of patients’ access to care, in line with the study aims. As such, we do not believe these biases would have negated the impact shown by our statistical analysis of departmental sociodemographic factors given the opportunities to access and remain in care during our project timeframe.

Relatedly, the ability to segment patients by more granular characteristics of clinical status or disease could have improved interpretation by allowing for sensitivity analyses and quality controls in which illness severity could have been incorporated as a covariate. While disparities in the types of cardiac pathologies seen across different administrative departments of Haiti would translate to interregional differences in clinical outcome status, the current analyses are based on the underlying assumption that different pathologies are approximately evenly distributed across Haiti’s departments. RHD is a post-infectious condition where prevalence may vary based on the robustness of public health infrastructure in different regions, but overall, this assumption of homogeneity is a reasonable approximation given that a large share of the observed presentations is secondary to spontaneous congenital anomalies.

Third, the HCA database includes only patients for whom information about treatment outcome and departmental location was available; excluded patients might have altered aggregate outcomes in meaningful ways. Finally, given the lack of available sociodemographic data at finer administrative levels like the commune or section, our analysis cannot address cardiac care access barriers within a given department.

Importantly, all findings of this study should be interpreted as hypothesis-generating. Future studies should aim to model cardiac care disparities in Haiti with greater precision by using patient-level data (home address, income or other indicators of economic status, travel time to cardiac clinic, employment status, surgical vs non-surgical cardiac pathology, and comorbidities) to predict clinical patient outcomes. To increase internal validity and better inform policymaking, regional-level studies of HCA data should be conducted at the neighborhood level. In this way, precise estimates of on-the-ground travel time and travel distance to the nearest cardiac clinic may be incorporated.