The Association among Blood Pressure, Blood Pressure Medications, and Glaucoma in a Nationwide Electronic Health Records Database.

PURPOSE: To measure the association among blood pressure (BP), BP medications, and glaucoma using the All of Us Research Program database. DESIGN: A retrospective, longitudinal cohort study leveraging a national electronic health record (EHR) database administered by the National Institutes of Health. PARTICIPANTS: Eye patients in the All of Us Research Program database with at least 15 months of follow-up and 1 BP measurement. METHODS: Univariable and multivariable Cox regression models predicted the risk of developing incident open-angle glaucoma (OAG). Mean arterial pressure (MAP) and the number of BP medication classes were entered as time-varying predictors to account for changes over time. MAIN OUTCOME MEASURES: The risk of developing incident OAG, as defined by billing diagnosis codes. RESULTS: Of 20 815 eligible eye patients who qualified for this study, 462 developed OAG. Low BP (MAP < 83.0 mmHg) was associated with increased risk of developing OAG (hazard ratio [HR], 1.32; 95% confidence interval [CI], 1.04-1.67). High BP (MAP > 101.3 mmHg) and the number of BP medication classes were not associated with OAG after adjustment for covariates. Other risk factors associated with OAG included being Black (HR, 3.31, 95% CI, 2.63-4.17), Hispanic or Latino (HR, 2.53, 95% CI, 1.94-3.28), Asian (HR, 2.22, 95% CI, 1.24-3.97), older in age (80+ years, HR, 20.1, 95% CI, 9.10-44.5), and diabetic (HR, 1.32, 95% CI, 1.04-1.67). Female gender was associated with decreased hazard of developing OAG (HR, 0.66, 95% CI, 0.55-0.80). No significant interaction was observed between MAP and the number of BP medications on the risk of developing OAG. CONCLUSIONS: We found that low BP is associated with increased risk of developing OAG in a national longitudinal EHR database. We did not find evidence supporting a differential effect of medically treated and untreated low BP. This study adds to the body of literature implicating vascular dysregulation as a potential etiology for the development of OAG, particularly emphasizing the lack of influence of BP medications on this relationship.

Determination of Patient Sentiment and Emotion in Ophthalmology: Infoveillance Tutorial on Web-Based Health Forum Discussions.

BACKGROUND: Clinical data in social media are an underused source of information with great potential to allow for a deeper understanding of patient values, attitudes, and preferences. OBJECTIVE: This tutorial aims to describe a novel, robust, and modular method for the sentiment analysis and emotion detection of free text from web-based forums and the factors to consider during its application. METHODS: We mined the discussion and user information of all posts containing search terms related to a medical subspecialty (oculoplastics) from MedHelp, the largest web-based platform for patient health forums. We used data cleaning and processing tools to define the relevant subset of results and prepare them for sentiment analysis. We executed sentiment and emotion analyses by using IBM Watson Natural Language Understanding to generate sentiment and emotion scores for the posts and their associated keywords. The keywords were aggregated using natural language processing tools. RESULTS: Overall, 39 oculoplastic-related search terms resulted in 46,381 eligible posts within 14,329 threads. Posts were written by 18,319 users (117 doctors; 18,202 patients) and included 201,611 associated keywords. Keywords that occurred ≥500 times in the corpus were used to identify the most prominent topics, including specific symptoms, medication, and complications. The sentiment and emotion scores of these keywords and eligible posts were analyzed to provide concrete examples of the potential of this methodology to allow for a better understanding of patients' attitudes. The overall sentiment score reflects a positive, neutral, or negative sentiment, whereas the emotion scores (anger, disgust, fear, joy, and sadness) represent the likelihood of the presence of the emotion. In keyword grouping analyses, medical signs, symptoms, and diseases had the lowest overall sentiment scores (-0.598). Complications were highly associated with sadness (0.485). Forum posts mentioning body parts were related to sadness (0.416) and fear (0.321). Administration was the category with the highest anger score (0.146). The top 6 forum subgroups had an overall negative sentiment score; the most negative one was the Neurology forum, with a score of -0.438. The Undiagnosed Symptoms forum had the highest sadness score (0.448). The least likely fearful posts were those from the Eye Care forum, with a score of 0.260. The overall sentiment score was much more negative before the doctor replied. The anger, disgust, fear, and sadness emotion scores decreased in likelihood, whereas joy was slightly more likely to be expressed after doctors replied. CONCLUSIONS: This report allows physicians and researchers to efficiently mine and perform sentiment analysis on social media to better understand patients' perspectives and promote patient-centric care. Important factors to be considered during its application include evaluating the scope of the search; selecting search terms and understanding their linguistic usages; and establishing selection, filtering, and processing criteria for posts and keywords tailored to the desired results.

Management of the glaucoma patient progressing at low normal intraocular pressure.

PURPOSE OF REVIEW: Patients with glaucoma with disease progression despite low or normal intraocular pressure (IOP) present special challenges to the treating clinician. Treatment goals may depend on whether patients have apparent low IOP with concurrent treatment or have low IOP at baseline without treatment. We review the diagnostic and therapeutic approaches to these patients. RECENT FINDINGS: Apparent progression at low IOP should start with confirmation of IOP, made easier by devices enabling patient home self-tonometry. Suspected visual field progression should be confirmed by repeat testing prior to advancement of therapy. Trabeculectomy remains the most effective surgical method of achieving long-term success, particularly when there is a low starting IOP. Drainage tube implantation or the use of novel micro-incisional non-bleb-forming procedures are less likely to be successful in achieving low IOP goals. SUMMARY: Diagnostic testing is important in confirming progressive glaucomatous disease at low IOP levels. The most effective way of slowing the progression of glaucoma in a patient with low IOP is to lower the IOP further, sometimes to single digit levels, which is most often achievable with trabeculectomy.

Big data requirements for artificial intelligence.

PURPOSE OF REVIEW: To summarize how big data and artificial intelligence technologies have evolved, their current state, and next steps to enable future generations of artificial intelligence for ophthalmology. RECENT FINDINGS: Big data in health care is ever increasing in volume and variety, enabled by the widespread adoption of electronic health records (EHRs) and standards for health data information exchange, such as Digital Imaging and Communications in Medicine and Fast Healthcare Interoperability Resources. Simultaneously, the development of powerful cloud-based storage and computing architectures supports a fertile environment for big data and artificial intelligence in health care. The high volume and velocity of imaging and structured data in ophthalmology and is one of the reasons why ophthalmology is at the forefront of artificial intelligence research. Still needed are consensus labeling conventions for performing supervised learning on big data, promotion of data sharing and reuse, standards for sharing artificial intelligence model architectures, and access to artificial intelligence models through open application program interfaces (APIs). SUMMARY: Future requirements for big data and artificial intelligence include fostering reproducible science, continuing open innovation, and supporting the clinical use of artificial intelligence by promoting standards for data labels, data sharing, artificial intelligence model architecture sharing, and accessible code and APIs.

Incidence and Risk Factors for Developing Diabetic Retinopathy among Youths with Type 1 or Type 2 Diabetes throughout the United States.

PURPOSE: Despite the increasing prevalence of type 2 diabetes mellitus (T2DM) among children and adolescents, little is known about their risk of developing diabetic retinopathy (DR). We sought to identify risk factors for DR in youths with diabetes mellitus, to compare DR rates for youths with type 1 diabetes mellitus (T1DM) and those with T2DM, and to assess whether adherence to DR screening guidelines promoted by the American Academy of Ophthalmology, American Academy of Pediatrics, and American Diabetes Association adequately capture youths with DR. DESIGN: Retrospective observational longitudinal cohort study. PARTICIPANTS: Youths aged ≤21 years with newly diagnosed T1DM or T2DM who were enrolled in a large US managed-care network. METHODS: In this study of youths aged ≤21 years with newly diagnosed T1DM or T2DM who were under ophthalmic surveillance, we identified the incidence and timing of DR onset. Kaplan-Meier survival curves assessed the timing of initial diagnosis of DR for participants. Multivariable Cox proportional hazard regression modeling identified factors associated with the hazard of developing DR. Model predictors were age and calendar year at initial diabetes mellitus diagnosis, sex, race/ethnicity, net worth, and glycated hemoglobin A1c fraction (HbA1c). MAIN OUTCOME MEASURES: Hazard ratios (HRs) with 95% confidence intervals (CIs) for developing DR. RESULTS: Among the 2240 youths with T1DM and 1768 youths with T2DM, 20.1% and 7.2% developed DR over a median follow-up time of 3.2 and 3.1 years, respectively. Survival curves demonstrated that youths with T1DM developed DR faster than youths with T2DM (P < 0.0001). For every 1-point increase in HbA1c, the hazard for DR increased by 20% (HR = 1.20; 95% CI 1.06-1.35) and 30% (HR = 1.30; 95% CI 1.08-1.56) among youths with T1DM and T2DM, respectively. Current guidelines suggest that ophthalmic screening begin 3 to 5 years after initial diabetes mellitus diagnosis, at which point in our study, >18% of youths with T1DM had already received ≥1 DR diagnosis. CONCLUSIONS: Youths with T1DM or T2DM exhibit a considerable risk for DR and should undergo regular screenings by eye-care professionals to ensure timely DR diagnosis and limit progression to vision-threatening disease.

Patient Attitudes Toward Telemedicine for Diabetic Retinopathy.

INTRODUCTION: Diabetic retinopathy (DR) is the leading cause of new-onset blindness in adults. Telemedicine is a validated, cost-effective method to improve monitoring. However, little is known of patients' attitudes toward telemedicine for DR. Our study explores factors that influence patients' attitudes toward participating in telemedicine. MATERIALS AND METHODS: Ninety seven participants in a university and the Veterans Administration setting completed a survey. Only people with diabetes mellitus (DM) were included. The main outcome was willingness to participate in telemedicine. The other outcomes were perceived convenience and impact on the patient-physician relationship. Participants reported demographic information, comorbidities, and access to healthcare. Analysis was performed with t-tests and multivariable logistic regression. RESULTS: Demographic factors were not associated with the outcomes (all p > 0.05). Patients had decreased odds of willingness if they valued the patient-physician relationship (adjusted odds ratio [OR] = 0.08, confidence interval [CI] = 0.02-0.35, p = 0.001) or had a longer duration of diabetes (adjusted OR = 0.93, CI = 0.88-0.99, p = 0.02). Patients had increased odds of willingness if they perceived increased convenience (adjusted OR = 8.10, CI = 1.77-36.97, p = 0.01) or had more systemic comorbidities (adjusted OR = 1.85, CI = 1.10-3.11, p = 0.02). DISCUSSION: It is critical to understand the attitudes of people with DM where telemedicine shows promise for disease management and end-organ damage prevention. Patients' attitudes are influenced by their health and perceptions, but not by their demographics. Receptive patients focus on convenience, whereas unreceptive patients strongly value their patient-physician relationships or have long-standing DM. Telemedicine monitoring should be designed for people who are in need and receptive to telemedicine.

Association between visual field defects and quality of life in the United States.

PURPOSE: To investigate the association between visual field defects and quality of life in the United States population. DESIGN: Cross-sectional study. PARTICIPANTS: A total of 5186 participants in the 2005 through 2008 National Health and Nutrition Examination Survey 40 years of age and older without a self-reported history of age-related macular degeneration or prior refractive surgery who had undergone frequency doubling technology perimetric testing. METHODS: Frequency doubling technology perimetry was performed in both eyes. Results from the better eye were used to categorize subjects as normal or having mild, moderate, or severe visual field loss. Subjects completed surveys about their visual and physical functioning ability. MAIN OUTCOME MEASURES: Disability pertaining to 6 vision-related activities, 2 visual function questions, and 5 physical functioning domains. RESULTS: Eighty-one percent of subjects had normal visual fields and 10%, 7%, and 2% demonstrated mild, moderate, and severe visual field defects, respectively. Subjects with greater severity of visual field defects had greater difficulty with vision-related activities. Subjects with severe visual field defects demonstrated the greatest odds of difficulty with all 6 activities. The 2 activities impacted most adversely were daytime driving in familiar places (odds ratio [OR], 12.4; 95% confidence interval [CI], 6.1-25.1) and noticing objects off to the side when walking (OR, 7.7; 95% CI, 4.7-12.7). Subjects with severe visual field defects had greater odds of worrying about eyesight (OR, 3.4; 95% CI, 2.0-5.8) and being limited by vision in the time spent on daily activities (OR, 5.1; 95% CI, 3.0-8.5). Subjects with severe visual field defects demonstrated the greatest odds of difficulty with 3 physical function domains, including activities of daily living (OR, 2.45; 95% CI, 1.37-4.38), instrumental activities of daily living (OR, 2.45; 95% CI, 1.37-4.38), as well as leisure and social activities (OR, 3.29; 95% CI, 1.87-5.77). CONCLUSIONS: Greater severity of visual field abnormality was associated with significantly greater odds of disability with vision-related function and physical function. These findings support the necessity of routine screening to find those who may benefit from therapy to prevent progressive glaucomatous vision loss.

Predicting Glaucoma Surgical Outcomes Using Neural Networks and Machine Learning on Electronic Health Records.

Purpose: To develop machine learning (ML) and deep learning (DL) models to predict glaucoma surgical outcomes, including postoperative intraocular pressure, use of ocular antihypertensive medications, and need for repeat surgery. Methods: We identified glaucoma surgeries performed at Stanford from 2013-2024, with two or more postoperative visits with intraocular pressure (IOP) measurement. Patient features were identified from the electronic health record (EHR), including demographics, prior diagnosis and procedure codes, medications and eye exam findings. Classical ML and DL models were developed to predict which glaucoma surgeries would result in surgical failure, defined as (1) IOP not reduced by more than 20% of preoperative baseline on two consecutive postoperative visits, (2) increased classes of glaucoma medications, and (3) need for additional glaucoma surgery or revision of original surgery. Results: A total of 2398 glaucoma surgeries of 1571 patients were included, of which 1677 surgeries met failure criteria. Random forest performed best for prediction of overall surgical failure, with accuracy of 75.5% and area under the receiver operator curve (AUROC) of 76.7%, similar to the deep learning model (accuracy 75.5%, AUROC 76.6%). Across all models, prediction performance was better for IOP outcomes (AUROC 86%) than need for an additional surgery (AUROC 76%) or need for additional glaucoma medication (AUC 70%). Conclusions: ML and DL algorithms can predict glaucoma surgery outcomes using structured data inputs from EHRs. Translational Relevance: Models that predict outcomes of glaucoma surgery may one day provide the basis for clinical decision support tools supporting surgeons in personalizing glaucoma treatment plans.

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.

Impact of Type 2 diabetes mellitus and insulin use on progression to glaucoma surgery in primary open angle glaucoma.

PURPOSE: To investigate outcomes of primary open-angle glaucoma (POAG) patients with and without type 2 diabetes mellitus (T2DM). METHODS: Retrospective observational study using U.S. nationwide healthcare insurance claims database. Patients ≥40 years old with at least one HbA1c within one year of POAG diagnosis were included. Diabetic factors associated with POAG progression requiring glaucoma surgery were evaluated using multivariable Cox proportional hazards regression models adjusted for demographic, diabetic and glaucoma factors. T2DM diagnosis and use of either oral hypoglycaemic agents or insulin therapy were assessed in association with POAG progression requiring glaucoma surgery. RESULTS: 104,515 POAG patients were included, of which 70,315 (67%) had T2DM. The mean age was 68.9 years (Standard deviation 9.2) and 55% were female. Of those with T2DM, 93% were taking medication (65,468); 95% (62,412) taking oral hypoglycaemic agents, and 34% (22,028) were on insulin. In multivariable analyses, patients with T2DM had a higher hazard of requiring glaucoma surgery (Hazard ratio, HR 1.15, 95% CI 1.09-1.21, p < 0.001). Higher mean HbA1c was also a significant predictor of progression requiring glaucoma surgery (HR 1.02, 95% CI 1.01-1.03, p < 0.001). When evaluating only patients who were taking antidiabetic medication, after adjusting for confounders, insulin use was associated with a 1.20 higher hazard of requiring glaucoma surgery compared to oral hypoglycaemic agents (95% CI 1.14-1.27, p < 0.001), but when stratified by HbA1c, this effect was only significant for those with HbA1c > 7.5%. CONCLUSIONS: Higher baseline HbA1c, particularly in patients taking insulin may be associated with higher rates of glaucoma surgery in POAG.

The Association Between Frailty and Visual Field Loss in US Adults.

PURPOSE: To describe the association between visual field loss and frailty in a nationally representative cohort of US adults. DESIGN: Retrospective cross-sectional study. METHODS: The cohort included adults 40 years or older with complete eye examination data from the 2005-2006 and 2007-2008 National Health and Nutrition Examination Surveys (NHANES). Visual field loss (VFL) was determined by frequency doubling technology and a 2-2-1 algorithm. A 36-item deficit accumulation-based frailty index was used to divide subjects into 4 categories of increasing frailty severity. RESULTS: Of the 4897 participants, 4402 (93.2%) had no VFL, 301 (4.1%) had unilateral VFL, and 194 (2.73%) had bilateral VFL. Within the sample, 2 subjects197 (53.1%) were categorized as non-frail, 1659 (31.3%) as vulnerable, 732 (11.3%) as mildly frail, and 312 (4.3%) as most frail. In multivariable models adjusted for demographics, visual acuity, and history of cataract surgery, subjects with unilateral VFL had higher adjusted odds of being in a more frail category (adjusted odds ratio [aOR], 2.07; 95% CI, 1.42-3.02) than subjects without VFL. Subjects with bilateral VFL also had higher odds of a more frail category compared to subjects without VFL (aOR, 1.74; 95% CI, 1.20-2.52). CONCLUSIONS: In the 2005-2008 NHANES adult population, VFL is associated with higher odds of frailty, independent of central visual acuity loss. Frail individuals may be more susceptible to diseases that can cause VFL, and/or VFL may predispose to frailty. Additional studies are needed to determine the directionality of this relationship and to assess potential interventions.

Automated Recognition of Visual Acuity Measurements in Ophthalmology Clinical Notes Using Deep Learning.

Purpose: Visual acuity (VA) is a critical component of the eye examination but is often only documented in electronic health records (EHRs) as unstructured free-text notes, making it challenging to use in research. This study aimed to improve on existing rule-based algorithms by developing and evaluating deep learning models to perform named entity recognition of different types of VA measurements and their lateralities from free-text ophthalmology notes: VA for each of the right and left eyes, with and without glasses correction, and with and without pinhole. Design: Cross-sectional study. Subjects: A total of 319 756 clinical notes with documented VA measurements from approximately 90 000 patients were included. Methods: The notes were split into train, validation, and test sets. Bidirectional Encoder Representations from Transformers (BERT) models were fine-tuned to identify VA measurements from the progress notes and included BERT models pretrained on biomedical literature (BioBERT), critical care EHR notes (ClinicalBERT), both (BlueBERT), and a lighter version of BERT with 40% fewer parameters (DistilBERT). A baseline rule-based algorithm was created to recognize the same VA entities to compare against BERT models. Main Outcome Measures: Model performance was evaluated on a held-out test set using microaveraged precision, recall, and F1 score for all entities. Results: On the human-annotated subset, BlueBERT achieved the best microaveraged F1 score (F1 = 0.92), followed by ClinicalBERT (F1 = 0.91), DistilBERT (F1 = 0.90), BioBERT (F1 = 0.84), and the baseline model (F1 = 0.83). Common errors included labeling VA in sections outside of the examination portion of the note, difficulties labeling current VA alongside a series of past VAs, and missing nonnumeric VAs. Conclusions: This study demonstrates that deep learning models are capable of identifying VA measurements from free-text ophthalmology notes with high precision and recall, achieving significant performance improvements over a rule-based algorithm. The ability to recognize VA from free-text notes would enable a more detailed characterization of ophthalmology patient cohorts and enhance the development of models to predict ophthalmology outcomes. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Validation of Artificial Intelligence (AI)-Assisted Flow Cytometry Analysis for Immunological Disorders.

Flow cytometry is a vital diagnostic tool for hematologic and immunologic disorders, but manual analysis is prone to variation and time-consuming. Over the last decade, artificial intelligence (AI) has advanced significantly. In this study, we developed and validated an AI-assisted flow cytometry workflow using 379 clinical cases from 2021, employing a 3-tube, 10-color flow panel with 21 antibodies for primary immunodeficiency diseases and related immunological disorders. The AI software (DeepFlow™, version 2.1.1) is fully automated, reducing analysis time to under 5 min per case. It interacts with hematopatholoists for manual gating adjustments when necessary. Using proprietary multidimensional density-phenotype coupling algorithm, the AI model accurately classifies and enumerates T, B, and NK cells, along with important immune cell subsets, including CD4+ helper T cells, CD8+ cytotoxic T cells, CD3+/CD4-/CD8- double-negative T cells, and class-switched or non-switched B cells. Compared to manual analysis with hematopathologist-determined lymphocyte subset percentages as the gold standard, the AI model exhibited a strong correlation (r > 0.9) across lymphocyte subsets. This study highlights the accuracy and efficiency of AI-assisted flow cytometry in diagnosing immunological disorders in a clinical setting, providing a transformative approach within a concise timeframe.

Using Natural Language Processing to Identify Different Lens Pathology in Electronic Health Records.

PURPOSE: Nearly all published ophthalmology-related Big Data studies rely exclusively on International Classification of Diseases (ICD) billing codes to identify patients with particular ocular conditions. However, inaccurate or nonspecific codes may be used. We assessed whether natural language processing (NLP), as an alternative approach, could more accurately identify lens pathology. DESIGN: Database study comparing the accuracy of NLP versus ICD billing codes to properly identify lens pathology. METHODS: We developed an NLP algorithm capable of searching free-text lens exam data in the electronic health record (EHR) to identify the type(s) of cataract present, cataract density, presence of intraocular lenses, and other lens pathology. We applied our algorithm to 17.5 million lens exam records in the Sight Outcomes Research Collaborative (SOURCE) repository. We selected 4314 unique lens-exam entries and asked 11 clinicians to assess whether all pathology present in the entries had been correctly identified in the NLP algorithm output. The algorithm's sensitivity at accurately identifying lens pathology was compared with that of the ICD codes. RESULTS: The NLP algorithm correctly identified all lens pathology present in 4104 of the 4314 lens-exam entries (95.1%). For less common lens pathology, algorithm findings were corroborated by reviewing clinicians for 100% of mentions of pseudoexfoliation material and 99.7% for phimosis, subluxation, and synechia. Sensitivity at identifying lens pathology was better for NLP (0.98 [0.96-0.99] than for billing codes (0.49 [0.46-0.53]). CONCLUSIONS: Our NLP algorithm identifies and classifies lens abnormalities routinely documented by eye-care professionals with high accuracy. Such algorithms will help researchers to properly identify and classify ocular pathology, broadening the scope of feasible research using real-world data.

Ethnic differences in intraocular pressure reduction and changes in anterior segment biometric parameters following cataract surgery by phacoemulsification.

BACKGROUND: To determine the association between ethnicity and changes in intraocular pressure and anterior segment biometric parameters following cataract surgery by phacoemulsification in nonglaucomatous subjects. DESIGN: Prospective clinical cohort study. PARTICIPANTS: Caucasian and Asian subjects. METHODS: Customized software was used to calculate parameters from anterior segment optical coherence tomography images obtained preoperatively and at 3 months following cataract surgery by phacoemulsification. The percentage changes in intraocular pressure and anterior segment biometric parameters following cataract surgery by phacoemulsification were modelled as a function of ethnicity using linear mixed-effects regression, a likelihood ratio test function that adjusted for age, sex and the use of both eyes in the same subject, to determine the association between ethnicity and postoperative outcomes. MAIN OUTCOME MEASURES: Intraocular pressure, angle opening distance, anterior chamber depth, anterior chamber volume, and angle recess area. RESULTS: Fifty Asian and 23 Caucasian nonglaucomatous eyes were analysed. Postoperative decrease in intraocular pressure and increases in angle opening distance, anterior chamber depth, anterior chamber volume and angle recess area were observed within each ethnic group (P < 0.005). The percent changes in intraocular pressure, angle opening distance, anterior chamber depth, anterior chamber volume and angle recess area did not differ between ethnic groups (P > 0.05). CONCLUSIONS: In this study, regardless of ethnic classification, subjects who received cataract surgery by phacoemulsification experienced a significant postoperative decrease in intraocular pressure and increases in angle opening distance, anterior chamber depth, anterior chamber volume and angle recess area. The percent changes in postoperative outcomes did not differ significantly by ethnicity.

Changes in glaucoma management following visual field testing and optical coherence tomography.

BACKGROUND: Optimal utilisation of investigations in glaucoma management remains unclear. We aimed to assess whether a temporal association exists between such testing and management changes. METHODS: Retrospective observational study using nationwide healthcare insurance claims database. Glaucoma outpatient encounters from patients aged ≥40 years with/without Humphrey visual field (HVF) and/or optical coherence tomography (OCT) were identified. An encounter was considered associated with an intervention if surgery occurred within 90 days, or if medication change or laser trabeculoplasty (LT) occurred within 30 days. RESULTS: 12 669 324 outpatient encounters of 1 863 748 individuals from 2003 to 2020 were included. HVF and OCT was performed during 32.8% and 22.2% of encounters respectively. Of the 36 763 (0.3%) encounters preceding surgery, 28.1% included HVF, 11.9% had OCT and 8.5% both. 79 181 (0.6%) visits preceded LT, of which 28.2% had HVF, 13.2% OCT and 9.3% both. Of the 515 899 (4.5%) encounters preceding medication changes, 29.1% had HVF, 16.7% OCT and 12.2% both. Compared with encounters with no investigations, those with HVF and/or OCT were associated with a 49% increased odds of a management change (p<0.001). In multivariate analyses, compared with encounters without investigations, visits with HVF alone had higher odds of subsequent surgery and LT, while HVF and/or OCT were associated with higher odds of medication change (p<0.001 for all). CONCLUSION: Glaucoma therapeutic changes occurred following approximately 5% of outpatient encounters. Surgery and LT were more likely to occur following a visit with a HVF rather than an OCT, while either investigation was associated with a higher odds of medication change.

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.

Predicting near-term glaucoma progression: An artificial intelligence approach using clinical free-text notes and data from electronic health records.

Purpose: The purpose of this study was to develop a model to predict whether or not glaucoma will progress to the point of requiring surgery within the following year, using data from electronic health records (EHRs), including both structured data and free-text progress notes. Methods: A cohort of adult glaucoma patients was identified from the EHR at Stanford University between 2008 and 2020, with data including free-text clinical notes, demographics, diagnosis codes, prior surgeries, and clinical information, including intraocular pressure, visual acuity, and central corneal thickness. Words from patients' notes were mapped to ophthalmology domain-specific neural word embeddings. Word embeddings and structured clinical data were combined as inputs to deep learning models to predict whether a patient would undergo glaucoma surgery in the following 12 months using the previous 4-12 months of clinical data. We also evaluated models using only structured data inputs (regression-, tree-, and deep-learning-based models) and models using only text inputs. Results: Of the 3,469 glaucoma patients included in our cohort, 26% underwent surgery. The baseline penalized logistic regression model achieved an area under the receiver operating curve (AUC) of 0.873 and F1 score of 0.750, compared with the best tree-based model (random forest, AUC 0.876; F1 0.746), the deep learning structured features model (AUC 0.885; F1 0.757), the deep learning clinical free-text features model (AUC 0.767; F1 0.536), and the deep learning model with both the structured clinical features and free-text features (AUC 0.899; F1 0.745). Discussion: Fusion models combining text and EHR structured data successfully and accurately predicted glaucoma progression to surgery. Future research incorporating imaging data could further optimize this predictive approach and be translated into clinical decision support tools.