Predicting Incident Heart Failure in Women With Machine Learning: The Women's Health Initiative Cohort

AI's ongoing impact: Implications of AI's effects on health equity for women's healthcare providers

Objective

To assess the effects of the current use of artificial intelligence (AI) in women's health on health equity, specifically in primary and secondary prevention efforts among women.

Methods

Two databases, Scopus and PubMed, were used to conduct this narrative review. The keywords included "artificial intelligence," "machine learning," "women's health," "screen," "risk factor," and "prevent," and papers were filtered only to include those about AI models that general practitioners may use.

Results

Of the 18 articles reviewed, 8 articles focused on risk factor modeling under primary prevention, and 10 articles focused on screening tools under secondary prevention. Gaps were found in the ability of AI models to train using large, diverse datasets that were reflective of the population it is intended for. Lack of these datasets was frequently identified as a limitation in the papers reviewed (n = 7).

Conclusions

Minority, low-income women have poor access to health care and are, therefore, not well represented in the datasets AI uses to train, which risks introducing bias in its output. To mitigate this, more datasets should be developed to validate AI models, and AI in women's health should expand to include conditions that affect men and women to provide a gendered lens on these conditions. Public health, medical, and technology entities need to collaborate to regulate the development and use of AI in health care at a standard that reduces bias.

Sex-Based Performance Disparities in Machine Learning Algorithms for Cardiac Disease Prediction: Exploratory Study

BACKGROUND

The presence of bias in artificial intelligence has garnered increased attention, with inequities in algorithmic performance being exposed across the fields of criminal justice, education, and welfare services. In health care, the inequitable performance of algorithms across demographic groups may widen health inequalities.

OBJECTIVE

Here, we identify and characterize bias in cardiology algorithms, looking specifically at algorithms used in the management of heart failure.

METHODS

Stage 1 involved a literature search of PubMed and Web of Science for key terms relating to cardiac machine learning (ML) algorithms. Papers that built ML models to predict cardiac disease were evaluated for their focus on demographic bias in model performance, and open-source data sets were retained for our investigation. Two open-source data sets were identified: (1) the University of California Irvine Heart Failure data set and (2) the University of California Irvine Coronary Artery Disease data set. We reproduced existing algorithms that have been reported for these data sets, tested them for sex biases in algorithm performance, and assessed a range of remediation techniques for their efficacy in reducing inequities. Particular attention was paid to the false negative rate (FNR), due to the clinical significance of underdiagnosis and missed opportunities for treatment.

RESULTS

In stage 1, our literature search returned 127 papers, with 60 meeting the criteria for a full review and only 3 papers highlighting sex differences in algorithm performance. In the papers that reported sex, there was a consistent underrepresentation of female patients in the data sets. No papers investigated racial or ethnic differences. In stage 2, we reproduced algorithms reported in the literature, achieving mean accuracies of 84.24% (SD 3.51%) for data set 1 and 85.72% (SD 1.75%) for data set 2 (random forest models). For data set 1, the FNR was significantly higher for female patients in 13 out of 16 experiments, meeting the threshold of statistical significance (-17.81% to -3.37%; P<.05). A smaller disparity in the false positive rate was significant for male patients in 13 out of 16 experiments (-0.48% to +9.77%; P<.05). We observed an overprediction of disease for male patients (higher false positive rate) and an underprediction of disease for female patients (higher FNR). Sex differences in feature importance suggest that feature selection needs to be demographically tailored.

CONCLUSIONS

Our research exposes a significant gap in cardiac ML research, highlighting that the underperformance of algorithms for female patients has been overlooked in the published literature. Our study quantifies sex disparities in algorithmic performance and explores several sources of bias. We found an underrepresentation of female patients in the data sets used to train algorithms, identified sex biases in model error rates, and demonstrated that a series of remediation techniques were unable to address the inequities present.

Pregnancy-Induced Hypertensive Disorder and Risks of Future Ischemic and Nonischemic Heart Failure

BACKGROUND

Although adverse pregnancy outcomes are associated with an increased risk of cardiovascular disease, studies on timing and subtypes of heart failure after a hypertensive pregnancy are lacking.

OBJECTIVES

The goal of this study was to assess the association between pregnancy-induced hypertensive disorder and risk of heart failure, according to ischemic and nonischemic subtypes, and the impact of disease characteristics and the timing of heart failure risks.

METHODS

This was a population-based matched cohort study, comprising all primiparous women without a history of cardiovascular disease included in the Swedish Medical Birth Register between 1988 and 2019. Women with pregnancy-induced hypertensive disorder were matched with women with normotensive pregnancies. Through linkage with health care registers, all women were followed up for incident heart failure, classified as ischemic or nonischemic.

RESULTS

In total, 79,334 women with pregnancy-induced hypertensive disorder were matched with 396,531 women with normotensive pregnancies. During a median follow-up of 13 years, rates of all heart failure subtypes were more common among women with pregnancy-induced hypertensive disorder. Compared with women with normotensive pregnancies, adjusted HRs (aHRs) with 95% CIs were as follows: heart failure overall, aHR: 1.70 (95% CI: 1.51-1.91); ischemic heart failure, aHR: 2.28 (95% CI: 1.74-2.98); and nonischemic heart failure, aHR: 1.60 (95% CI: 1.40-1.83). Disease characteristics indicating severe hypertensive disorder were associated with higher heart failure rates, and rates were highest within the first years after the hypertensive pregnancy but remained significantly increased thereafter.

CONCLUSIONS

Pregnancy-induced hypertensive disorder is associated with an increased short-term and long-term risk of incident ischemic and nonischemic heart failure. Disease characteristics indicating more severe forms of pregnancy-induced hypertensive disorder amplify the heart failure risks.

A systematic comparison of machine learning algorithms to develop and validate prediction model to predict heart failure risk in middle-aged and elderly patients with periodontitis (NHANES 2009 to 2014)

Periodontitis is increasingly associated with heart failure, and the goal of this study was to develop and validate a prediction model based on machine learning algorithms for the risk of heart failure in middle-aged and elderly participants with periodontitis. We analyzed data from a total of 2876 participants with a history of periodontitis from the National Health and Nutrition Examination Survey (NHANES) 2009 to 2014, with a training set of 1980 subjects with periodontitis from the NHANES 2009 to 2012 and an external validation set of 896 subjects from the NHANES 2013 to 2014. The independent risk factors for heart failure were identified using univariate and multivariate logistic regression analysis. Machine learning algorithms such as logistic regression, k-nearest neighbor, support vector machine, random forest, gradient boosting machine, and multilayer perceptron were used on the training set to construct the models. The performance of the machine learning models was evaluated using 10-fold cross-validation on the training set and receiver operating characteristic curve (ROC) analysis in the validation set. Based on the results of univariate logistic regression and multivariate logistic regression, it was found that age, race, myocardial infarction, and diabetes mellitus status were independent predictors of the risk of heart failure in participants with periodontitis. Six machine learning models, including logistic regression, K-nearest neighbor, support vector machine, random forest, gradient boosting machine, and multilayer perceptron, were built on the training set, respectively. The area under the ROC for the 6 models was obtained using 10-fold cross-validation with values of 0 848, 0.936, 0.859, 0.889, 0.927, and 0.666, respectively. The areas under the ROC on the external validation set were 0.854, 0.949, 0.647, 0.933, 0.855, and 0.74, respectively. K-nearest neighbor model got the best prediction performance across all models. Out of 6 machine learning models, the K-nearest neighbor algorithm model performed the best. The prediction model offers early, individualized diagnosis and treatment plans and assists in identifying the risk of heart failure occurrence in middle-aged and elderly patients with periodontitis.

Contributions of the Women's Health Initiative to Cardiovascular Research: JACC State-of-the-Art Review

The WHI (Women's Health Initiative) enrolled 161,808 racially and ethnically diverse postmenopausal women, ages 50-79 years, from 1993 to 1998 at 40 clinical centers across the United States. In its clinical trial component, WHI evaluated 3 randomized interventions (menopausal hormone therapy; diet modification; and calcium/vitamin D supplementation) for the primary prevention of major chronic diseases, including cardiovascular disease, in older women. In the WHI observational study, numerous clinical, behavioral, and social factors have been evaluated as predictors of incident chronic disease and mortality. Although the original interventions have been completed, the WHI data and biomarker resources continue to be leveraged and expanded through ancillary studies to yield novel insights regarding cardiovascular disease prevention and healthy aging in women.