Prediction Models for Glaucoma in a Multicenter Electronic Health Records Consortium: The Sight Outcomes Research Collaborative.

Purpose: Advances in artificial intelligence have enabled the development of predictive models for glaucoma. However, most work is single-center and uncertainty exists regarding the generalizability of such models. The purpose of this study was to build and evaluate machine learning (ML) approaches to predict glaucoma progression requiring surgery using data from a large multicenter consortium of electronic health records (EHR). Design: Cohort study. Participants: Thirty-six thousand five hundred forty-eight patients with glaucoma, as identified by International Classification of Diseases (ICD) codes from 6 academic eye centers participating in the Sight OUtcomes Research Collaborative (SOURCE). Methods: We developed ML models to predict whether patients with glaucoma would progress to glaucoma surgery in the coming year (identified by Current Procedural Terminology codes) using the following modeling approaches: (1) penalized logistic regression (lasso, ridge, and elastic net); (2) tree-based models (random forest, gradient boosted machines, and XGBoost), and (3) deep learning models. Model input features included demographics, diagnosis codes, medications, and clinical information (intraocular pressure, visual acuity, refractive status, and central corneal thickness) available from structured EHR data. One site was reserved as an "external site" test set (N = 1550); of the patients from the remaining sites, 10% each were randomly selected to be in development and test sets, with the remaining 27 999 reserved for model training. Main Outcome Measures: Evaluation metrics included area under the receiver operating characteristic curve (AUROC) on the test set and the external site. Results: Six thousand nineteen (16.5%) of 36 548 patients underwent glaucoma surgery. Overall, the AUROC ranged from 0.735 to 0.771 on the random test set and from 0.706 to 0.754 on the external test site, with the XGBoost and random forest model performing best, respectively. There was greatest performance decrease from the random test set to the external test site for the penalized regression models. Conclusions: Machine learning models developed using structured EHR data can reasonably predict whether glaucoma patients will need surgery, with reasonable generalizability to an external site. Additional research is needed to investigate the impact of protected class characteristics such as race or gender on model performance and fairness. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Predicting Glaucoma Progression to Surgery with Artificial Intelligence Survival Models.

Purpose: Prior artificial intelligence (AI) models for predicting glaucoma progression have used traditional classifiers that do not consider the longitudinal nature of patients' follow-up. In this study, we developed survival-based AI models for predicting glaucoma patients' progression to surgery, comparing performance of regression-, tree-, and deep learning-based approaches. Design: Retrospective observational study. Subjects: Patients with glaucoma seen at a single academic center from 2008 to 2020 identified from electronic health records (EHRs). Methods: From the EHRs, we identified 361 baseline features, including demographics, eye examinations, diagnoses, and medications. We trained AI survival models to predict patients' progression to glaucoma surgery using the following: (1) a penalized Cox proportional hazards (CPH) model with principal component analysis (PCA); (2) random survival forests (RSFs); (3) gradient-boosting survival (GBS); and (4) a deep learning model (DeepSurv). The concordance index (C-index) and mean cumulative/dynamic area under the curve (mean AUC) were used to evaluate model performance on a held-out test set. Explainability was investigated using Shapley values for feature importance and visualization of model-predicted cumulative hazard curves for patients with different treatment trajectories. Main Outcome Measures: Progression to glaucoma surgery. Results: Of the 4512 patients with glaucoma, 748 underwent glaucoma surgery, with a median follow-up of 1038 days. The DeepSurv model performed best overall (C-index, 0.775; mean AUC, 0.802) among the models studied in this article (CPH with PCA: C-index, 0.745; mean AUC, 0.780; RSF: C-index, 0.766; mean AUC, 0.804; GBS: C-index, 0.764; mean AUC, 0.791). Predicted cumulative hazard curves demonstrate how models could distinguish between patient who underwent early surgery and patients who underwent surgery after > 3000 days of follow-up or no surgery. Conclusions: Artificial intelligence survival models can predict progression to glaucoma surgery using structured data from EHRs. Tree-based and deep learning-based models performed better at predicting glaucoma progression to surgery than the CPH regression model, potentially because of their better suitability for high-dimensional data sets. Future work predicting ophthalmic outcomes should consider using tree-based and deep learning-based survival AI models. Additional research is needed to develop and evaluate more sophisticated deep learning survival models that can incorporate clinical notes or imaging. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

Differences in Ideal Cardiovascular Health Between Sexual Minority and Heterosexual Adults.

Importance: Research on the cardiovascular health (CVH) of sexual minority adults has primarily examined differences in the prevalence of individual CVH metrics rather than comprehensive measures, which has limited development of behavioral interventions. Objective: To investigate sexual identity differences in CVH, measured using the American Heart Association's revised measure of ideal CVH, among adults in the US. Design, Setting, and Participants: This cross-sectional study analyzed population-based data from the National Health and Nutrition Examination Survey (NHANES; 2007-2016) in June 2022. Participants included noninstitutional adults aged 18 to 59 years. We excluded individuals who were pregnant at the time of their interview and those with a history of atherosclerotic cardiovascular disease or heart failure. Exposures: Self-identified sexual identity categorized as heterosexual, gay/lesbian, bisexual, or something else. Main Outcomes and Measures: The main outcome was ideal CVH (assessed using questionnaire, dietary, and physical examination data). Participants received a score from 0 to 100 for each CVH metric, with higher scores indicating a more favorable CVH profile. An unweighted average was calculated to determine cumulative CVH (range, 0-100), which was recoded as low, moderate, or high. Sex-stratified regression models were performed to examine sexual identity differences in CVH metrics, disease awareness, and medication use. Results: The sample included 12 180 participants (mean [SD] age, 39.6 [11.7] years; 6147 male individuals [50.5%]). Lesbian (B = -17.21; 95% CI, -31.98 to -2.44) and bisexual (B = -13.76; 95% CI, -20.54 to -6.99) female individuals had less favorable nicotine scores than heterosexual female individuals. Bisexual female individuals had less favorable body mass index scores (B = -7.47; 95% CI, -12.89 to -1.97) and lower cumulative ideal CVH scores (B = -2.59; 95% CI, -4.84 to -0.33) than heterosexual female individuals. Compared with heterosexual male individuals, gay male individuals had less favorable nicotine scores (B = -11.43; 95% CI, -21.87 to -0.99) but more favorable diet (B = 9.65; 95% CI, 2.38-16.92), body mass index (B = 9.75; 95% CI, 1.25-18.25), and glycemic status scores (B = 5.28; 95% CI, 0.59-9.97). Bisexual male individuals were twice as likely as heterosexual male individuals to report a diagnosis of hypertension (adjusted odds ratio [aOR], 1.98; 95% CI, 1.10-3.56) and use of antihypertensive medication (aOR, 2.20; 95% CI, 1.12-4.32). No differences in CVH were found between participants who reported their sexual identity as something else and heterosexual participants. Conclusion and Relevance: Results of this cross-sectional study suggest that bisexual female individuals had worse cumulative CVH scores than heterosexual female individuals, whereas gay male individuals generally had better CVH than heterosexual male individuals. There is a need for tailored interventions to improve the CVH of sexual minority adults, particularly bisexual female individuals. Future longitudinal research is needed to examine factors that might contribute to CVH disparities among bisexual female individuals.

Social Determinants of Health and Perceived Barriers to Care in Diabetic Retinopathy Screening.

Importance: Regular screening for diabetic retinopathy often is crucial for the health of patients with diabetes. However, many factors may be barriers to regular screening and associated with disparities in screening rates. Objective: To evaluate the associations between visiting an eye care practitioner for diabetic retinopathy screening and factors related to overall health and social determinants of health, including socioeconomic status and health care access and utilization. Design, Setting, and Participants: This retrospective cross-sectional study included adults aged 18 years or older with type 2 diabetes who answered survey questions in the All of Us Research Program, a national multicenter cohort of patients contributing electronic health records and survey data, who were enrolled from May 1, 2018, to July 1, 2022. Exposures: The associations between visiting an eye care practitioner and (1) demographic and socioeconomic factors and (2) responses to the Health Care Access and Utilization, Social Determinants of Health, and Overall Health surveys were investigated using univariable and multivariable logistic regressions. Main Outcome and Measures: The primary outcome was whether patients self-reported visiting an eye care practitioner in the past 12 months. The associations between visiting an eye care practitioner and demographic and socioeconomic factors and responses to the Health Care Access and Utilization, Social Determinants of Health, and Overall Health surveys in All of Us were investigated using univariable and multivariable logistic regression. Results: Of the 11 551 included participants (54.55% cisgender women; mean [SD] age, 64.71 [11.82] years), 7983 (69.11%) self-reported visiting an eye care practitioner in the past year. Individuals who thought practitioner concordance was somewhat or very important were less likely to have seen an eye care practitioner (somewhat important: adjusted odds ratio [AOR], 0.83 [95% CI, 0.74-0.93]; very important: AOR, 0.85 [95% CI, 0.76-0.95]). Compared with financially stable participants, individuals with food or housing insecurity were less likely to visit an eye care practitioner (food insecurity: AOR, 0.75 [95% CI, 0.61-0.91]; housing insecurity: AOR, 0.86 [95% CI, 0.75-0.98]). Individuals who reported fair mental health were less likely to visit an eye care practitioner than were those who reported good mental health (AOR, 0.84; 95% CI, 0.74-0.96). Conclusions and Relevance: This study found that food insecurity, housing insecurity, mental health concerns, and the perceived importance of practitioner concordance were associated with a lower likelihood of receiving eye care. Such findings highlight the self-reported barriers to seeking care and the importance of taking steps to promote health equity.