Background
Breast cancer is the most frequently diagnosed malignancy and the leading cause of cancer death among females globally.1,2 In the United States, an estimated 300,590 cases will be diagnosed with breast cancer (15.3% of all new cancers) and 43,700 will die of the disease in 2023.3 The 5-year relative survival for female breast cancer is 90.3%, and it is estimated that by 2030 there will be 5 million breast cancer survivors in the United States.4,5 High survival expectancy and treatment improvements have increased concern about the concomitant increase in cardiovascular diseases (CVDs) associated with cancer treatment. In postmenopausal female breast cancer survivors, the risk of mortality attributable to CVD is higher than in those without a history of breast cancer, whereas CVD is also the leading cause of death in patients aged >50 years with active breast cancer.4,6,7 CVD and breast cancer share common risk factors (eg, obesity), and cancer treatments (eg, chemotherapy, hormone therapy, immunotherapy, and radiotherapy) are associated with an increased risk of cardiovascular toxicity, leading to increased CVD morbidity and mortality.4,8
Racial disparities have been reported for both breast cancer and CVD. Given the estimation that minority populations will exceed 50% of the US population by 2044, there is a concern for these disparate outcomes across racial and/or ethnic groups.4,9,10 Overall, the prevalence of CVD is higher in non-Hispanic Black (NHB) individuals relative to other racial subpopulations, and this disparity also exists for the primary risk factors for CVD, including hypertension and type 2 diabetes mellitus.11,12 For patients with breast cancer, mortality is higher among Black women compared with White women, with reports showing that Black women diagnosed with breast cancer are approximately 40% more likely to die compared with their White counterparts.13,14
The determinants of racial disparities in CVD outcomes have been categorized as a healthcare gap by the American Heart Association.4 Social determinants of health (SDOH) are known to be an important condition of peoples’ living environments that affect health outcomes, and are referenced in the Cancer Moonshot program.15,16 A large, recently published United Kingdom cancer registry–based analysis has shown that patients living in socioeconomically deprived areas have a higher rate of CVD.17 However, the individual-level impact of SDOH on racial disparities in cardiac events among female patients with breast cancer remains unknown.15,16 Thus, the primary objective of this study was to identify and quantify the role of SDOH in racial disparities in major adverse cardiovascular events (MACE).
Methods
Data Source
The study setting was the University Hospitals (UH) Seidman Cancer Center (Northeast Ohio), which is a tertiary care center that serves urban, suburban, and rural areas and is composed of an integrated network of 23 hospitals, >50 health centers and outpatient facilities, and >200 physicians offices in 16 counties throughout the region.18 Due to the inner-city population served, the UH patient population includes a higher percentage of Black individuals and a lower percentage of Hispanic and Asian individuals than the overall US population.
All patient data were obtained from the UH data repository based on the Caisis platform, which consists of an open-source, web-based, cancer data management system that integrates disparate sources of patient data (eg, Soarian, next-generation sequencing laboratories, Sunrise Clinical Manager, Tumor Registry, Via Oncology, OnCore, MOSAIQ, patient-reported outcome tools, and others).19–23 The information obtained was subsequently complemented with information from electronic medical records (EMRs) captured via EMERSE (Electronic Medical Record Search Engine) to obtain the most accurate and complete information for each patient, thus avoiding high missingness.24 All patient records were deidentified, and the study was approved by the University Hospitals of Cleveland Institutional Review Board.
The cohort (supplemental eFigure 1, available with this article at JNCCN.org) included females aged ≥18 years diagnosed with breast cancer (all stages), determined using tumor registry data or ICD-9 and ICD-10 codes obtained from EMRs (ie, C50.xx, C79.81, 174.x, 175.0, 175.9, 198.81, and 217) between January 1, 2010, and December 31, 2019, providing a minimum follow-up of 2 years by 2022, which was the year that data were collected.25,26 Patients were excluded from the analysis if they were of male sex and/or had carcinoma in situ. Hispanic individuals were excluded due to the low number of patients with this ethnicity. We included all patients who had SDOH information available; patients without SDOH information were excluded. Demographic information from the catchment area, primarily based on US Census and American Community Survey data, was included to demonstrate the representativeness of the study cohort.27
Exposure
Individual SDOH features were obtained from LexisNexis, the world’s largest electronic database for legal and public records–related information, consisting of groups of variables in 4 domains (supplemental eTable 1): social and community context (marital status, number of household members, distance to closest relatives), neighborhood and built environment (crime index, burglary index, car theft index, murder index, neighborhood median household income, neighborhood median home value), education access and quality (education institution rating, college attendance), and economic stability (address stability, residence status, property status, annual income, properties owned, wealth index, household income, total number of transportation properties owned).28,29 Access to healthcare is the fifth SDOH domain; however, because all patients were able to seek care at our institution, that metric was not analyzed.
The LexisNexis dataset includes a combination of all adult patients discharged from a UH facility over a 2.5-year time frame and all adult patients who are members of an Accountable Care Organization.30 This dataset is composed of a combination of multiple public and private records that are updated at different frequencies. The data obtained referred to the most current records available.
Outcomes
The co-primary endpoints were diagnosis and time-to-event for 2-year MACE following the cancer diagnosis. The MACE included in this study were heart failure (HF), acute coronary syndrome (ACS), atrial fibrillation (A-fib), and ischemic stroke (IS), determined using ICD-9 and ICD-10 codes obtained from the entire history present in each patient’s EMR (supplemental eTable 1).31 A-fib was included in our MACE definition because it is a commonly unaccounted event in patients with cancer.32
Covariates
Demographics, risk factors, tumor characteristics, and treatment data were obtained for all eligible patients. Demographic characteristics included data from the patient’s EMR, such as age at diagnosis, race (white, Black, other), ethnicity (Hispanic, non-Hispanic), and payer (Medicaid, Medicare, private insurance, self-pay, other). Risk factors were extracted from the comorbidities list based on ICD codes that were presented in the chart before the MACE diagnosis. These risk factors included EMR-based information, such as self-reported smoking status (yes, no, former, unknown), Charlson comorbidity index, and cardiovascular history/risk factors (yes, no).33,34 Cardiovascular history/risk factors were considered positive if one or more of the following diagnoses were present in the patient’s EMR: hyperlipidemia, cardiomyopathy, known coronary artery disease, previous myocardial infarction, carotid artery disease, previous transient ischemic attack/stroke, and/or chronic kidney disease (supplemental eTable 1). These factors were combined into a single variable due to the strong correlation between them, with the objective of generating a variable that characterized patients with high cardiovascular risk.35 The number of cardiovascular history/risk factors was defined as the sum of diagnoses of each component of this variable for each patient.
Tumor characteristics included EMR-based information regarding date of cancer diagnosis, hormone receptor status (estrogen receptor [ER], progesterone receptor [PR], and HER2), histologic type (ductal or lobular, not otherwise specified, other, or unknown], and TNM staging group (stage 0–IV). Treatment characteristics included appointment completion rates and use of single or combination therapy during a lifetime, such as radiation of the breast (right, left), chemotherapy, endocrine therapy, and immunotherapy. Specific medication groups included anthracyclines, PIK3CA/mTOR inhibitors, HER2-targeted agents, ER antagonists, luteinizing hormone-releasing hormone agonists, aromatase inhibitors, and newer therapies (supplemental eTable 1).
Descriptive Analysis
Data were stratified according to race/ethnicity (NHB, non-Hispanic whites [NHW]) and were presented as absolute values and percentages for categorical variables and as medians and quartiles for continuous variables. The category “other” race was considered/included for general demographics but not for the racial comparison analyses. Pearson’s chi-square test was used to compare categorical variables by race/ethnicity. Data distribution assumptions for continuous variables were confirmed using histograms and the Kolmogorov-Smirnov test. We then performed Student t tests for normally distributed factors, and nonparametric Kruskal-Wallis tests for nonnormal factors.
Machine Learning Modeling
The impact/weight of SDOH in 2-year MACE was determined via machine learning (ML) models (overall data with race as a variable [race-agnostic model] and race-specific data in NHW and NHB patients separately; supplemental eFigure 2). The ML approach was applied because of its superior performance compared with the traditional regression models and its capacity to learn from data and to deal with multiple data structures.36–38 Specifically, we used the tree-based Extreme Gradient Boosting (XGBoost) method from mlr3proba, an R package, for ML in survival analysis.39 This method is widely used with clinical data and was selected because of its explainability, offering crucial perspectives into clinical decision-making.38,40 In addition, XGBoost is notable for being 10 times quicker than other widely used solutions and for its capacity to handle sparse datasets and process hundreds of millions of instances/observations.41 On tabular-style datasets with characteristics that are individually meaningful, tree-based models outperform deep learning, which is another ML option.40
The data were chronologically partitioned as 60% for training, 20% for testing, and 20% for validation.42 Feature selection was performed comparing the variables according to MACE (yes vs no) in the training set, selecting those with P<.30 (supplemental eTable 2).43 The testing set was used for hyperparameter tuning (supplemental eTable 3) applying a 10 times 10-fold cross-validation with 100 iterations, prioritizing the concordance index (C-index). This approach avoids overfitting (when a model is too adapted to the peculiarities of a dataset) and allows the model to learn and improve based on multiple iterations, consequently increasing external validity.44 Subsequently, the tuned model was applied in the validation set with 10 times 10-fold cross-validation. The ML performance was measured via the mean C-index (the most precise and appropriate technique for calculating prediction error) and its 95% confidence interval.45 Variable importance scores for the predictors were obtained using SHapley Additive Explanations (SHAP).40,46 The SHAP score (SS) for each feature shows how the model prediction changes when that feature is taken into consideration, illustrating how that factor contributes to explaining the discrepancy between the average model prediction and the instance’s actual prediction.40,46 An ML prediction, f(x), is represented as a fixed base value plus the sum of SHAP values: f(x) = base value + sum(SHAP values).46 Finally, the SDOH were ranked according to the SS from each model.
MACE Risk
The Fine-Gray method for Cox proportional hazards models was used to examine racial disparities in the risk of 2-year MACE accounting for the competing risk of all-cause mortality. The variables selected for the multivariable models were among those that received the top 15 feature importance scores from the ML model. Sensitivity analysis was performed using only the Fine-Gray modeling approach, selecting variables that achieved P<.15 in bivariate analyses and those deemed to have clinical importance by study investigators. Both approaches were presented side-by-side to ensure the robustness of the approach and clarity for the reader regarding conclusions drawn from the data. To account for the healthcare access domain, a subgroup sensitivity analysis was performed in patients with private insurance. Results were presented as subdistribution hazard ratios (SHRs) and 95% confidence intervals.
Adversity/Unfavorable SDOH Conditions
To determine and quantify the level of adverse SDOH (according to the ML model’s ranking), adversity markers (the point at which the variable becomes associated with a positive prediction of MACE) were defined via partial dependence plots (PDPs).47 PDPs show the change in the average prediction for the outcome as a specified feature varies and also demonstrate what the relationship between the target and the feature is.47,48 We presented the racial stratification of SDOH using the PDP approach.
Statistical Considerations
Independent variable correlations were checked by correlation plots, and the variables found to be statistically significantly correlated were not included simultaneously. P<.05 was considered significant in the final models, and missing values were not included in the analysis. All analyses were performed using RStudio software. We used the STROBE cohort checklist to assess and report outcomes.49
Results
Population
We included 4,309 patients with breast cancer, of whom 765 were NHB females. The cohort’s median age at diagnosis was 63 years (IQR, 53–72 years); 49.2% of the diagnoses were ductal carcinoma, 5.7% were stage III, and 1.9% were stage IV. ER positivity was present in 44.9% of the cases, PR positivity in 40.2%, and HER2 positivity in 6.8%. Most of the patients were never-smokers (50.6%) and had a cardiovascular history/risk factors (74.6%), with a median Charlson comorbidity score of 4 (IQR, 2–7). Surgery was performed in 60% of the cohort, whereas 28.2% received chemotherapy, 46% received endocrine therapy, 4.7% received immunotherapy, and 39.4% received radiotherapy.
The catchment area is comprised of a 17.4% Black/African American population and a 94.4% non-Hispanic population, and 91.5% of adults have health insurance. The prevalence of comorbidities in this region are 32.4% for high cholesterol, 35% for hypertension, and 12.5% for diabetes. Median household income is $62,780; there is an average of 2.3 persons per household; the crime rate is 586.1 crimes per 100,000 persons; 31.4% of people aged >65 years live alone; 29.8% of individuals live in a single-parent household; and 9.5% of the households do not have a vehicle.
Two-year MACE was diagnosed in 11.4% of patients, with a median time-to-event of 177 days (IQR, 45–414 days). HF was the most frequently diagnosed event (6.9%), followed by A-fib (3.7%), IS (2.4%), and ACS (2.3%). NHB patients, when compared with NHW patients, had higher rates of MACE (19.2% vs 9.9%), HF (13.1% vs 5.5%), and ACS (4.8% vs 1.7%) (all P<.001). No racial differences were noted in time-to-event. Race-stratified descriptions are summarized in Table 1 and supplemental eTables 4 and 5.
Race/Ethnicity-Stratified Population Characteristics of Patients With Breast Cancer
Predictors of MACE and the Impact of SDOH
With an excellent performance (C-index, 0.79; 95% CI, 0.78–0.80; supplemental eTable 3), the race-agnostic model classified the number of cardiovascular history/risk factors (SS = 0.59), age at diagnosis (SS = 0.26), previous cardiomyopathy (SS = 0.17), time to surgery (SS = 0.11), and neighborhood median household income (SS = 0.07) as the top 5 important variables for predicting 2-year MACE (supplemental eFigure 3, supplemental eTable 6); Black race ranked as the eighth most important variable (SS = 0.05). Among the SDOH variables, however, the 5 most important variables for predicting 2-year MACE were neighborhood median household income (SS = 0.07), neighborhood crime index (SS = 0.06), number of transportation properties in the household (SS = 0.05), neighborhood burglary index (SS = 0.04), and neighborhood median home values (SS = 0.03) (supplemental eTable 7).
The NHB-specific model achieved a fair performance (C-index, 0.66; 95% CI, 0.63–0.69; supplemental eTable 3), and classified the total number of cardiovascular history/risk factors (SS = 1.10), neighborhood median household income (SS = 0.62), age at diagnosis (SS = 0.54), time to surgery (SS = 0.53), and time to chemotherapy (SS = 0.53) as the top 5 important variables (supplemental eTable 6). The model developed for NHW patients achieved an excellent performance (C-index, 0.78; 95% CI, 0.76–0.79; supplemental eTable 3). Among the important variables, the top 5 were total number of cardiovascular history/risk factors (SS = 0.30), age at diagnosis (SS = 0.15), previous cardiomyopathy (SS = 0.07), neighborhood crime index (SS = 0.06), and time to surgery (SS = 0.05) (supplemental eTable 6).
MACE Risk
Including SDOH as adjustments and ethnicity/race as a covariate, Fine-Gray competing risk models showed no racial differences in the risk of 2-year MACE (adjusted SHR [aSHR] for NHB, 1.22; 95% CI, 0.91–1.64), HF (1.45; 95% CI, 0.96–2.17), ACS (1.75; 95% CI, 0.91–3.38), IS (0.98; 95% CI, 0.50–1.90), and A-fib (0.72; 95% CI, 0.39–1.30). The traditional modeling approach and subgroup analysis in patients with private insurance showed similar results (supplemental eTable 8).
Adversity
Using the PDP methodology, neighborhood median household income <$60,000, neighborhood crime index >160, <2 transportation properties in the household, neighborhood crime index >160, and neighborhood median home value <$400,000 were considered adversity cutoffs (higher association with the prediction of 2-year MACE) for the top 5 SDOH variables from the ML model (Table 2, supplemental eFigures 4 and 5). Applying the adversity cutoffs, NHB patients were more likely to be in adversity (P<.05) in 80% of the top 10 SDOH variables from the ML predictive model (Table 2).
Patients Living in Adversea SDOH Conditions, Stratified by Race
Discussion
This is the first study to demonstrate and rank the impact of individual-level SDOH factors from different domains in the development of MACE among patients with breast cancer using race-agnostic and race-specific ML models. We demonstrated that neighborhood and built environment variables were the most important SDOH for predicting 2-year MACE. Thus, racial differences noted in CVD outcomes in women with breast cancer may be explained by adverse SDOH, as shown by this single-institution cohort. Policies and efforts focused on increasing equity may be able to reduce the burden of CVD in women with breast cancer.
In addition to our main findings, number of cardiovascular history/risk factors, age at diagnosis, time to treatment, previous cardiomyopathy, and appointment completion rates were important predictors of MACE. Examining racial disparities, NHB females with breast cancer, when compared with their NHW counterparts, were diagnosed at later stages, received higher rates of treatment, had lower rates of appointment completion, lived in poorer SDOH conditions, and had higher rates of 2-year MACE, without racial differences in time-to-event. Moreover, NHB females with breast cancer were shown to be in adversity conditions in 8 of the 10 most important SDOH variables to predict MACE.
NHB females have a higher incidence of triple-negative breast cancer, known as the most aggressive subtype, and are diagnosed at later stages.50,51 It is also known that risk factors and modifiable social behaviors (eg, smoking, alcoholism, sedentarism) are higher in NHB females when compared with NHW females, and this may be explained by lifestyle, income differences, and environmental factors.52–54 Moreover, NHB females have a lower probability of receiving the most suitable treatment and care approach compared with their NHW counterparts.22,50,55,56 Our results reinforce these reports, with the higher treatment rates in NHB females likely a result of diagnosis occurring at more aggressive stages in this population.
A combination of factors contributes to the racial disparity in the incidence of MACE. Black patients are estimated to have one of the highest rates of hypertension in the world, with hyperaldosteronism having a significant correlation with cardiovascular risk factors.53,57,58 This population is more likely to have symptoms of and functional impairment from ACS, which can lead to a bias in diagnosis.59,60 Later breast cancer stages at diagnosis also play a role, with recent studies showing a significant correlation with A-fib.61 Different rates in breast cancer treatment are another important factor, because chemotherapy, immunotherapy, endocrine therapy agents, and radiotherapy are known to cause a variety of MACE.4,62 Age at diagnosis is also a key factor corroborated by our study, as younger age at cancer diagnosis is linked with higher CVD risk.63 This association may be due to an early exposure to risk factors for cancer and CVD, both of which are strongly correlated with social conditions.64–66
Racism is a central topic when analyzing racial disparities.67 It is defined as “an organized system premised on the categorization and ranking of racial/ethnic groups into social hierarchies, assigned differential values and access to power, opportunities, and resources, resulting in disadvantage.”68–72 The existence of racism is due to a historical factor, determined by slavery, which began in the United States in the 17th century, and the attempt to classify Black individuals as an inferior race subject to inferior rights and opportunities.73 There are different forms of racism, and many directly and indirectly affect health, with studies already showing direct links between the self-reported experience of personally mediated racism and negative physical and mental health outcomes.68,74 Examples of this are the reports of inequities in factors such as income, education, employment, and living standards, in concordance with our findings, which have an impact on living environments and exposure to risk and protective factors.68–70,75,76
SDOH play a central role in elucidating some of the mechanisms underlying racial disparities.77 Factors such as poverty, cultural and social injustice, overall lower income, and education level, mediated by structural racism, influence conditions such as lifestyle and healthcare access.50,78,79 These poor conditions that lead to limited access to healthcare are demonstrated by our findings of lower rates of appointment completion in NHB patients, despite the higher treatment rates in this population. Regarding cardiovascular health, a recent study showed that the addition of SDOH parameters improved the prognostic utility of prediction models in Black patients with HF.36 The SDOH and structural factors were reported to be significant drivers of the racial/ethnic disparities seen in coronary artery disease and stroke.80 In dilated cardiomyopathy, the interplay of social and economic factors was identified as a driver of the poor outcomes in Blacks.81 Moreover, a recent review called for the need to identify the biological markers associated with SDOH that predict CVD and to develop personalized interventions for patients at highest risk.65 In agreement, the results from our study show clearly the social construct of race, because the addition of SDOH as covariates equalized the previously related racial difference in MACE risk.11,12 Thus, the biological concept of race has a small role in the higher risk of MACE in NHB patients.82 Taking this all into consideration, it is clear that interventions and programs focused on improving healthcare quality and safety should address the drivers of disparities in health outcomes both within and outside healthcare systems to make these programs more effective.83
Our results demonstrate an important role of neighborhood and built environment in the prediction of MACE, especially in NHB females. Historically, places with a large Black population were segregated as a result of social divestment in local infrastructure, perpetuating a disadvantage for this population.65,84,85 This segregation has been associated with higher levels of neighborhood violence, crime, and poverty as well as reduced work possibilities, economic stability, and access to a high-quality education.65,86 Reports demonstrate that Black individuals living in areas with increased segregation have a 12% higher risk of CVD.87 Annual income, related to economic stability, also played an important role in our model, supporting findings that show an association between low socioeconomic status and atherogenesis and a proinflammatory state.72,88,89
This study has several limitations. Our institutional database is EMR-based, and some of the information in the EMRs may be incomplete. Additionally, due to the retrospective nature of this study, some variables were not available (eg, use of cardiovascular medications prior to the breast cancer diagnosis). Because this was a single-institution study, some patients may have been lost to follow-up or sought emergency care at other institutions, but because the institution is an oncology center, it maintains close follow-up with patients. The criteria for data availability in LexisNexis may have generated a selection bias. The results presented may be reflective of the catchment area and its characteristics, and the sample may be representative of individuals with greater healthcare-seeking behavior. The inclusion of both patients with curable and incurable breast cancer may have impacted the reported rates of MACE. The definition of MACE did not include a wider range of conditions (eg, arrythmias other than A-fib, valvular heart disease, cerebrovascular conditions). On the other hand, the integration of disparate sources, including individual-level SDOH, allowed access to detailed information on the patients’ longitudinal trajectory not commonly available in other datasets. The inclusion of census data for the catchment area allowed interpretation of the representativeness of the study results, considering that no dataset is fully representative. These facts, added to an ML approach improved through multiple iterations and validations, with the ability to deal with missingness, provided a robust reliability for our results.
At the patient level, the key clinical and practical implications of this study are the need for proactive screening and management of cardiovascular risk factors and CVD in patients at high risk of MACE who are being considered for anticancer treatments. For the screening phase, allostatic load—a measure of chronic stress that accounts for SDOH factors—has been shown to be an effective marker of CVD risk in patients with cancer.90 Moreover, at a population level, our results reinforce the increased need for cardio-oncology services, especially in specific (underserved) populations.
Conclusions
Our findings showed that neighborhood and built environment played an important role in the development of 2-year MACE in patients with breast cancer and showed that NHB females in our study live in unfavorable SDOH conditions. Race increasingly needs to be understood as a social construct, and public health policies must focus on equity to mitigate the effects of racial disparities in health outcomes, including cardiovascular outcomes. Future studies should focus on increasing and diversifying the covariates analyzed, using multicenter designs, using national cohorts of data, developing specific models for each CVD/MACE, and examining the geographical variation of SDOH within different countries and regions in order to provide personalized care according to specific local needs.
References
- 1.↑
Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–249.
- 2.↑
Surveillance, Epidemiology, and End Results Program. Cancer stat facts: female breast cancer. Accessed June 1, 2022. Available at: https://seer.cancer.gov/statfacts/html/breast.html
- 4.↑
Mehta LS, Watson KE, Barac A, et al. Cardiovascular disease and breast cancer: where these entities intersect: a scientific statement from the American Heart Association. Circulation 2018;137:e30–66.
- 5.↑
American Cancer Society. Cancer treatment & survivorship facts & figures. Accessed September 9, 2021. Available at: https://www.cancer.org/research/cancer-facts-statistics/survivor-facts-figures.html
- 6.↑
Coughlin SS, Ayyala D, Majeed B, et al. Cardiovascular disease among breast cancer survivors. Cardiovasc Disord Med 2020;2:10.31489.
- 7.↑
Mery B, Fouilloux A, Rowinski E, et al. Cardiovascular disease events within 5 years after a diagnosis of breast cancer. BMC Cancer 2020;20:337.
- 8.↑
Chou YH, Huang JY, Kornelius E, et al. Major adverse cardiovascular events after treatment in early-stage breast cancer patients receiving hormone therapy. Sci Rep 2020;10:1408.
- 9.↑
Ellis L, Canchola AJ, Spiegel D, et al. Racial and ethnic disparities in cancer survival: the contribution of tumor, sociodemographic, institutional, and neighborhood characteristics. J Clin Oncol 2018;36:25–33.
- 10.↑
Sparano JA, Brawley OW. Deconstructing racial and ethnic disparities in breast cancer. JAMA Oncol 2021;7:355–356.
- 11.↑
Brothers RM, Fadel PJ, Keller DM. Racial disparities in cardiovascular disease risk: mechanisms of vascular dysfunction. Am J Physiol Heart Circ Physiol 2019;317:H777–789.
- 12.↑
Benjamin EJ, Muntner P, Alonso A, et al. Heart disease and stroke statistics-2019 update: a report from the American Heart Association. Circulation 2019;139:e56–528.
- 13.↑
Howlader N, Noone A, Krapcho M, et al. SEER Cancer Statistics Review, 1975-2014, National Cancer Institute. Bethesda, MD. Based on November 2016 SEER data submission, posted to the SEER web site, April 2017. Accessed July 6, 2022. Available at: https://seer.cancer.gov/csr/1975_2014/
- 14.↑
Al-Sadawi M, Hussain Y, Copeland-Halperin RS, et al. Racial and socioeconomic disparities in cardiotoxicity among women with HER2-positive breast cancer. Am J Cardiol 2021;147:116–121.
- 15.↑
Stabellini N, Dmukauskas M, Cullen J, et al. Racial disparities in zip-code level social determinants of health and cardiovascular outcomes in females with breast cancer. Circulation 2022;146(Suppl 1):Abstract 11546.
- 16.↑
National Cancer Institute. Cancer Moonshot. Accessed October 6, 2022. Available at: https://www.cancer.gov/research/key-initiatives/moonshot-cancer-initiative
- 17.↑
Battisti NML, Welch CA, Sweeting M, et al. Prevalence of cardiovascular disease in patients with potentially curable malignancies: a national registry dataset analysis. JACC CardioOncol 2022;4:238–253.
- 18.↑
University Hospitals. Annual report 2021. Accessed March 5, 2023. Available at: https://www.uhhospitals.org/about-uh/publications/UH-Annual-Report/2021-annual-report
- 20.↑
Stabellini N, Bruno DS, Dmukauskas M, et al. Sex differences in lung cancer treatment and outcomes at a large hybrid academic-community practice. JTO Clin Res Rep 2022;3:100307.
- 21.↑
Stabellini N, Chandar AK, Chak A, et al. Sex differences in esophageal cancer overall and by histological subtype. Sci Rep 2022;12:5248.
- 22.↑
Stabellini N, Cullen J, Cao L, et al. Racial disparities in breast cancer treatment patterns and treatment related adverse events. Sci Rep 2023;13:1233.
- 23.↑
Stabellini N, Tomlinson B, Cullen J, et al. Sex differences in adults with acute myeloid leukemia and the impact of sex on overall survival. Cancer Med 2023;12:6711–6721.
- 24.↑
EMERSE. Electronic Medical Record Search Engine. Accessed May 16, 2022. Available at: https://project-emerse.org/index.html
- 25.↑
ICD-10 Version:2019. Accessed July 23, 2021. Available at: https://icd.who.int/browse10/2019/en
- 26.↑
Centers for Disease Control and Prevention. International Classification of Diseases, Ninth Revision (ICD-9). Accessed July 23, 2021. Available at: https://www.cdc.gov/nchs/icd/icd9.htm
- 27.↑
Healthy Northeast Ohio. Community homepage. Accessed March 6, 2023. Available at: https://www.healthyneo.org/
- 28.↑
LexisNexis. Social determinants of health. Accessed May 3, 2022. Available at: https://risk.lexisnexis.com/healthcare/social-determinants-of-health
- 29.↑
Healthy People 2030. Social determinants of health. Accessed May 3, 2022. Available at: https://health.gov/healthypeople/priority-areas/social-determinants-health
- 30.↑
Centers for Medicare & Medicaid Services. Accountable care organizations (ACOs): general information. Accessed October 19, 2022. Available at: https://innovation.cms.gov/innovation-models/aco
- 31.↑
Bonsu JM, Guha A, Charles L, et al. Reporting of cardiovascular events in clinical trials supporting FDA approval of contemporary cancer therapies. J Am Coll Cardiol 2020;75:620–628.
- 32.↑
Rahman F, Ko D, Benjamin EJ. Association of atrial fibrillation and cancer. JAMA Cardiol 2016;1:384–386.
- 33.↑
Guha A, Dey AK, Arora S, et al. Contemporary trends and outcomes of percutaneous and surgical aortic valve replacement in patients with cancer. J Am Heart Assoc 2020;9:e014248.
- 34.↑
Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–383.
- 35.↑
Kerr AJ, Broad J, Wells S, et al. Should the first priority in cardiovascular risk management be those with prior cardiovascular disease? Heart 2009;95:125–129.
- 36.↑
Segar MW, Hall JL, Jhund PS, et al. Machine learning-based models incorporating social determinants of health vs traditional models for predicting in-hospital mortality in patients with heart failure. JAMA Cardiol 2022;7:844–854.
- 37.↑
Lewis EF. Machine learning and social determinants of health-an opportunity to move beyond race for inpatient risk prediction in patients with heart failure. JAMA Cardiol 2022;7:854–855.
- 38.↑
Azuaje F. Artificial intelligence for precision oncology: beyond patient stratification. NPJ Precis Oncol 2019;3:6.
- 39.↑
Sonabend R, Király FJ, Bender A, et al. mlr3proba: an R package for machine learning in survival analysis. Bioinformatics 2021;37:2789–2791.
- 40.↑
Lundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell 2020;2:56–67.
- 41.↑
Chen T, Guestrin C. XGBoost: a scalable tree boosting system. Accessed May 3, 2022. Available at: https://dl.acm.org/doi/pdf/10.1145/2939672.2939785
- 42.↑
Xu Y, Goodacre R. On splitting training and validation set: a comparative study of cross-validation, bootstrap and systematic sampling for estimating the generalization performance of supervised learning. J Anal Test 2018;2:249–262.
- 43.↑
Pudjihartono N, Fadason T, Kempa-Liehr AW, et al. A review of feature selection methods for machine learning-based disease risk prediction. Accessed September 5, 2022. Available at: https://www.frontiersin.org/articles/10.3389/fbinf.2022.927312/full
- 44.↑
Refaeilzadeh P, Tang L, Liu H. Cross-validation. In: Liu L, Özsu MT, eds. Encyclopedia of Database Systems. Springer; 2009:532–538.
- 45.↑
Harrell FE Jr, Califf RM, Pryor DB, et al. Evaluating the yield of medical tests. JAMA 1982;247:2543–2546.
- 46.↑
Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Accessed July 25, 2022. Available at: https://dl.acm.org/doi/10.5555/3295222.3295230
- 47.↑
Friedman JH. Greedy function approximation: a gradient boosting machine. Ann Statist 2001;29:1189–1232.
- 48.↑
Goldstein A, Kapelner A, Bleich J, et al. Peeking inside the black box: visualizing statistical learning with plots of individual conditional expectation. J Comp Graph Stat 2015;24:44–65.
- 49.↑
von Elm E, Altman DG, Egger M, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008;61:344–349.
- 50.↑
Yedjou CG, Sims JN, Miele L, et al. Health and racial disparity in breast cancer. Adv Exp Med Biol 2019;1152:31–49.
- 51.↑
Akinyemiju T, Moore JX, Ojesina AI, et al. Racial disparities in individual breast cancer outcomes by hormone-receptor subtype, area-level socio-economic status and healthcare resources. Breast Cancer Res Treat 2016;157:575–586.
- 52.↑
Muncan B. Cardiovascular disease in racial/ethnic minority populations: illness burden and overview of community-based interventions. Public Health Rev 2018;39:32.
- 53.↑
Kurian AK, Cardarelli KM. Racial and ethnic differences in cardiovascular disease risk factors: a systematic review. Ethn Dis 2007;17:143–152.
- 54.↑
Fei K, Rodriguez-Lopez JS, Ramos M, et al. Racial and ethnic subgroup disparities in hypertension prevalence, New York City Health and Nutrition Examination Survey, 2013–2014. Prev Chronic Dis 2017;14:E33.
- 55.↑
Tammemagi CM. Racial/ethnic disparities in breast and gynecologic cancer treatment and outcomes. Curr Opin Obstet Gynecol 2007;19:31–36.
- 56.↑
Hirschman J, Whitman S, Ansell D. The black:white disparity in breast cancer mortality: the example of Chicago. Cancer Causes Control 2007;18:323–333.
- 57.↑
Mensah GA, Mokdad AH, Ford ES, et al. State of disparities in cardiovascular health in the United States. Circulation 2005;111:1233–1241.
- 58.↑
Kidambi S, Kotchen JM, Grim CE, et al. Association of adrenal steroids with hypertension and the metabolic syndrome in blacks. Hypertension 2007;49:704–711.
- 59.↑
Kaul P, Lytle BL, Spertus JA, et al. Influence of racial disparities in procedure use on functional status outcomes among patients with coronary artery disease. Circulation 2005;111:1284–1290.
- 61.↑
Guha A, Fradley MG, Dent SF, et al. Incidence, risk factors, and mortality of atrial fibrillation in breast cancer: a SEER-Medicare analysis. Eur Heart J 2022;43:300–312.
- 62.↑
Guha A, Stabellini N, Montero AJ. Commentary: longitudinal changes in circulating metabolites and lipoproteins after breast cancer treatment. Front Cardiovasc Med 2022;9:962698.
- 63.↑
Paterson DI, Wiebe N, Cheung WY, et al. Incident cardiovascular disease among adults with cancer: a population-based cohort study. JACC CardioOncol 2022;4:85–94.
- 64.↑
Koene RJ, Prizment AE, Blaes A, et al. Shared risk factors in cardiovascular disease and cancer. Circulation 2016;133:1104–1114.
- 65.↑
Powell-Wiley TM, Baumer Y, Baah FO, et al. Social determinants of cardiovascular disease. Circ Res 2022;130:782–799.
- 66.↑
Tucker-Seeley RD. Social determinants of health and disparities in cancer care for Black people in the United States. JCO Oncol Pract 2021;17:261–263.
- 67.↑
Vince RA Jr, Eyrich NW, Mahal BA, et al. Reporting of racial health disparities research: are we making progress? J Clin Oncol 2022;40:8–11.
- 68.↑
Stanley J, Harris R, Cormack D, et al. The impact of racism on the future health of adults: protocol for a prospective cohort study. BMC Public Health 2019;19:346.
- 69.↑
Williams DR, Mohammed SA. Racism and health I: pathways and scientific evidence. Am Behav Sci 2013;57:1152–1173.
- 71.↑
Ganatra S, Dani SS, Kumar A, et al. Impact of social vulnerability on comorbid cancer and cardiovascular disease mortality in the United States. JACC CardioOncol 2022;4:326–337.
- 72.↑
Stabellini N, Cullen J, Moore JX, et al. Racial differences in chronic stress/allostatic load variation due to androgen deprivation therapy in prostate cancer. JACC CardioOncol 2022;4:555–557.
- 73.↑
Equal Justice Initiative. Racial justice. Accessed January 21, 2022. Available at: https://eji.org/racial-justice/
- 74.↑
Paradies Y, Ben J, Denson N, et al. Racism as a determinant of health: a systematic review and meta-analysis. PLoS One 2015;10:e0138511.
- 75.↑
Paradies Y. Defining, conceptualizing and characterizing racism in health research. Crit Public Health 2006;16:143–157.
- 76.↑
Krieger N. Methods for the scientific study of discrimination and health: an ecosocial approach. Am J Public Health 2012;102:936–944.
- 77.↑
Williams DR, Sternthal M. Understanding racial-ethnic disparities in health: sociological contributions. J Health Soc Behav 2010;51(Suppl l):S15–27.
- 78.↑
Freeman HP, Chu KC. Determinants of cancer disparities: barriers to cancer screening, diagnosis, and treatment. Surg Oncol Clin N Am 2005;14:655–669.
- 79.↑
Carnethon MR, Pu J, Howard G, et al. Cardiovascular health in African Americans: a scientific statement from the American Heart Association. Circulation 2017;136:e393–423.
- 80.↑
Mital R, Bayne J, Rodriguez F, et al. Race and ethnicity considerations in patients with coronary artery disease and stroke: JACC Focus Seminar 3/9. J Am Coll Cardiol 2021;78:2483–2492.
- 81.↑
Ntusi NAB, Sliwa K. Impact of racial and ethnic disparities on patients with dilated cardiomyopathy: JACC Focus Seminar 7/9. J Am Coll Cardiol 2021;78:2580–2588.
- 82.↑
Kuzawa CW, Sweet E. Epigenetics and the embodiment of race: developmental origins of US racial disparities in cardiovascular health. Am J Hum Biol 2009;21:2–15.
- 83.↑
Schneider EC, Chin MH, Graham GN, et al. Increasing equity while improving the quality of care: JACC Focus Seminar 9/9. J Am Coll Cardiol 2021;78:2599–2611.
- 84.↑
Rothstein R. The Color of Law: A Forgotten History of How Our Government Segregated America. Liveright Publishing; 2017.
- 85.↑
Bailey ZD, Feldman JM, Bassett MT. How structural racism –orks - racist policies as a root cause of U.S. racial health inequities. N Engl J Med 2021;384:768–773.
- 86.↑
Williams DR, Lawrence JA, Davis BA. Racism and health: evidence and needed research. Annu Rev Public Health 2019;40:105–125.
- 87.↑
Williams DR, Collins C. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Rep 2001;116:404–416.
- 88.↑
Schultz WM, Kelli HM, Lisko JC, et al. Socioeconomic status and cardiovascular outcomes: challenges and interventions. Circulation 2018;137:2166–2178.
- 89.↑
Rosengren A, Hawken S, Ounpuu S, et al. Association of psychosocial risk factors with risk of acute myocardial infarction in 11119 cases and 13648 controls from 52 countries (the INTERHEART study): case-control study. Lancet 2004;364:953–962.
- 90.↑
Stabellini N, Cullen J, Bittencourt MS, et al. Allostatic load and cardiovascular outcomes in males with prostate cancer. JNCI Cancer Spectrum 2023;7:pkad005. https://academic.oup.com/jncics/article/7/2/pkad005/7031248
