Telemedicine Screening Research Articles

Unmasking biases and navigating pitfalls in the ophthalmic artificial intelligence lifecycle: A narrative review.

Over the past 2 decades, exponential growth in data availability, computational power, and newly available modeling techniques has led to an expansion in interest, investment, and research in Artificial Intelligence (AI) applications. Ophthalmology is one of many fields that seek to benefit from AI given the advent of telemedicine screening programs and the use of ancillary imaging. However, before AI can be widely deployed, further work must be done to avoid the pitfalls within the AI lifecycle. This review article breaks down the AI lifecycle into seven steps-data collection; defining the model task; data preprocessing and labeling; model development; model evaluation and validation; deployment; and finally, post-deployment evaluation, monitoring, and system recalibration-and delves into the risks for harm at each step and strategies for mitigating them.

Evaluating the Feasibility of a Telescreening Program for Retinopathy of Prematurity (ROP) in Denmark.

Objectives: This study investigates the feasibility of implementing telescreening for retinopathy of prematurity (ROP) using the ICON GO® widefield camera operated by a non-physician healthcare professional (NPHP). We hypothesized that images captured by an NPHP are adequate to evaluate ROP changes without further examinations. Secondly, the level of agreement between independent ROP graders were evaluated based on the fundus photographs. Methods: National ROP screening criteria were gestational age (GA) < 32 weeks or birthweight (BW) < 1500 g. Exclusion criteria were children hospitalized and born outside the Capital Region and examinations not performed by an NPHP. The screenings were performed using the ICON GO®. The NPHP selected the best images for evaluation by an on-site ophthalmologist, regarding whether re-examination was necessary and if so, whether the re-examination was beneficial. Lastly, the images were re-evaluated by an independent off-site ophthalmologist. Results: A total of 415 screening sessions on 165 patients performed by an NPHP were included. Re-examination was necessary in three screening sessions and beneficial in two. The level of agreement between the on-site and off-site ophthalmologists regarding ROP screening outcome was k = 0.82, ROP stage k = 0.69, plus disease k = 0.69, and lastly ROP zone k = 0.37. Of the screened children, ninety-seven (58.8%) had no ROP at any time points, sixty-two (37.6%) had some stage of ROP not requiring treatment, and six (3.6%) received ROP treatment. Conclusions: Telemedicine screening for ROP with the ICON GO® camera performed by an NPHP was feasible with an almost-perfect agreement and negligible need for re-examinations. The approach effectively identified children needing treatment, supporting the use of telescreening in ROP management.

Comparison of Machine Learning Models to a Novel Score in the Identification of Patients at Low Risk for Diabetic Retinopathy

PurposeTo develop an easily applicable predictor of patients at low risk for diabetic retinopathy (DR). DesignAn experimental study on the development and validation of machine learning models and a novel retinopathy risk score to detect patients at low risk for diabetic retinopathy. SubjectsAll individuals aged 18 years or older who participated in the telemedicine retinal screening initiative through Temple University Health Systems from October 1, 2016 through December 31, 2020. The subjects must have documented evidence of their diabetes mellitus diagnosis as well as a documented hemoglobin A1C recorded in their chart within 6 months of the retinal screening photograph. MethodsThe charts of 1930 subjects (1590 evaluable) undergoing telemedicine screening for DR were reviewed, and 30 demographic and clinical parameters were collected. DR is a dichotomous variable where low risk is defined as no or mild retinopathy using the International Clinical Diabetic Retinopathy severity score. Five machine learning models (MLM) were trained to predict patients at low risk for DR using 1050 subjects and further underwent 10-fold cross validation to maximize its performance indicated by the area under the receiver operator characteristic curve (AUC). Additionally, a novel retinopathy risk score (RRS) is defined as the product of hemoglobin A1c closest to screening and years with diabetes mellitus. RRS was also applied to generate a predictive model. Main Outcome MeasuresThe performance of the trained machine learning models and the RRS model was compared using DeLong’s test. The models were further validated using a separate unseen test set of 540 subjects.The performance of the validation models were compared using DeLong’s test and chi-square tests. ResultsUsing the test set, the AUC for the RRS was not statistically different from four out of five MLM. The error rate for predicting low-risk patients using the RRS was significantly lower than the naive rate (0.097 vs 0.19; p<0.0001), and it was comparable to the error rates of the machine learning models. ConclusionThis novel retinopathy risk score is a potentially useful and easily deployable predictor of patients at low risk for DR.

Excellent accuracy of trained neonatal nurses in the detection of referral-warranted retinopathy of prematurity within an established telemedicine screening programme.

To evaluate level of agreement of specialist trained retinopathy of prematurity (ROP) nurses compared with an experienced paediatric ophthalmologist in detection of referral-warranted ROP (RWROP) using wide-field digital retinal imaging. This is a prospective, observational, blinded study of neonates in a level III neonatal intensive care unit, from July 2020 to November 2022. Image capture using wide-field digital retinal imaging followed by ROP grading and staging was completed by trained ROP nurses. This was then compared with findings by an experienced paediatric ophthalmologist. The primary outcome was presence of RWROP in either eye. One hundred and ninety-five neonates (55% male) with a total of 768 screening visits were included. At the initial screen, nurse and ophthalmologist agreed about presence of RWROP for 191 of 195 neonates (98%, kappa = 0.79, P < 0.0001), with 100% sensitivity for RWROP detection. Including all 768 screening episodes, agreement was 98% for RWROP. There was disagreement in 16 screenings (2%) for 11 (6%) neonates. Of the five screenings (0.7%) that the ophthalmologist thought were RWROP and the nurse did not, three were disagreements about whether the zone was posterior zone 2 or zone 1. We found excellent levels of agreement and add evidence that interpretations by specialist trained nurses could be safely integrated into a 'hybrid ROP screening system'.

Research protocol and recruitment redesign of a study of pregnant women and their infants during the COVID-19 pandemic: Childhood Allergy and the NeOnatal Environment (CANOE)

BackgroundRecruitment for research studies is a challenging endeavor that was further complicated by the COVID-19 pandemic. We launched a new NIH-supported multi-center birth cohort, “Childhood Allergy and the NeOnatal Environment” (CANOE) in January 2020 across four sites. While the pandemic temporarily halted clinical research, we restructured the study and instituted novel recruitment methods that we hypothesized would enable brisk enrollment when research activities resumed. ObjectiveDevelop protocol modifications and recruitment methods that promote successful recruitment of diverse populations in clinical research despite a global pandemic. MethodsWhile study activities were suspended, we modified recruitment strategies to limit in-person contact, shifting toward alternative HIPAA-compliant methods like clinician referrals, institutional social media, and telemedicine screening and consent procedures. Protocol changes included reducing the frequency of in-person visits, leveraging clinical care visits to collect biospecimens, expanded self-collection of samples at home, and making study materials available online. ResultsRemote methods, including targeted social media posts, mailed letters, and email, combined with in-clinic recruitment with modifications for social distancing led to successful recruitment at all sites. Rates of consent have been similar across recruitment sites, with the highest rates of enrollment of mother-infant dyads realized by sites that implemented multiple recruitment strategies. ConclusionsStudy procedures that prioritize health and safety measures such as social distancing, study participant convenience, and diverse recruitment strategies enable successful enrollment of pregnant women and their newborns into clinical research while adhering to public health restrictions during a global pandemic.

Telemedicine for Screening and Follow-Up of Glaucoma: A Descriptive Study.

Introduction: Glaucoma is a leading cause of irreversible blindness. It is a prevalent disease worldwide, affecting ∼70 million people and expected to reach up to 112 million by 2040. Purpose: The aim of this study is to describe the implementation and initial experience of a telemedicine program to monitor glaucoma and glaucoma suspect patients in a large, integrated health care system during the COVID-19 pandemic. Methods: A retrospective chart review of established glaucoma or glaucoma suspect patients who participated in a telemedicine evaluation at the ophthalmic center of a large, Colombian health care system between June 2020 and April 2023 was conducted. Clinical and sociodemographic variables were analyzed. Generated clinical orders for additional testing, surgical procedures, follow-ups, and referrals, as well as changes in medical treatment, were evaluated. Results: A total of 11,034 telemedicine consults were included. The mean ± standard deviation age of this group was 63 ± 17.2 years and 67% were female. Of the patients who attended teleconsults, 49% were glaucoma suspects and 38.5% were followed with a diagnosis of open-angle glaucoma. After the consult, 25% of patients were referred to a glaucoma specialist, 40% had additional testing ordered, and 8% had a surgical procedure ordered, mainly laser iridotomy (409 cases). Almost a third of patients returned for subsequent telemedicine visits after the initial encounter. Despite some technical difficulties, 99.8% of patients attended and completed their scheduled telemedicine appointments. Conclusions: A telemedicine program aimed to monitor established glaucoma patients can be successfully implemented. Established patients within an integrated health care system have high adherence to the virtual model. Further research by health care institutions and government agencies will be key to expand coverage to additional populations. Clinical Trial Registration Number: CEIFUS 1026-24.

Telemedicine Screening for Eye Disorders in Federally Qualified Health Centers: Relationship to Vision-Targeted Health-Related Quality of Life.

Introduction: Federally Qualified Health Centers (FQHCs) play a crucial role as safety-net primary health care clinics in the United States, serving medically underserved areas and populations. However, eye services are rarely offered at FQHCs. We examined how telemedicine-generated ocular diagnoses impacted vision-targeted health-related quality of life at FQHCs in rural Alabama. Methods: We focused on patients who are at risk for glaucoma. Both visual function and retinal imaging were assessed. The telemedicine vision screening protocol performed by a remote ophthalmologist evaluated eyes for glaucoma, diabetic retinopathy, cataract, age-related macular degeneration, and a measurement of habitual visual acuity. The National Eye Institute Visual Function Questionnaire-9 (VFQ-9) was administered. Results: Using stepwise regression, the best-fitting model for predicting VFQ-9 scores incorporated visual acuity 20/40 or worse, a diabetic retinopathy diagnosis, and sociodemographic variables (gender, transportation, insurance type/status, and employment status). Conclusion: Vision-targeted, health-related quality of life in our FQHC settings was related to the visual acuity impairment and the diagnosis of diabetic retinopathy but was also influenced by a variety of sociodemographic factors.

Telemedicine Screening for Diabetic Retinopathy Using Digital Technology: Foreign Experience

Background. Timely detection and treatment of diabetic retinopathy (DR) could prevent the development of blindness. However, large number of patients are diagnosed at late stages when effective treatment becomes impossible. In this context, active screening for DR plays an important role. Various screening programs have been developed, including programs using digital technologies.Objective. To analyze and summarize the literature data on digitalization of DR screening.Materials and methods. The study was carried out based on the results of a search in the PubMed / Medline bibliographic database and the Google system. The used keywords were “diabetes”, “retinopathy”, “telemedi-cine”, “digital technologies”, “artificial intelligence” and others. Relevant publications were selected by exam-ining the titles and abstracts of articles.Results. The study addressed the approaches to organizing the telemedicine programs of DR screening. The examples of innovative digital technologies for performing diagnostic studies as well as the characteristics of various electronic systems and devices are presented. It was noted that the DR screening could be performed in a local center with data transfer to a regional specialized diagnostic (analytical) center where telemedicine analysis is done. All diagnostic information is available to the diabetologist who prescribes and adjusts ther-apy. This operating principle facilitates the interaction between various specialists and thus increases the efficiency of screening.Discussion. Although the telemedicine screening for diabetic retinopathy has already been introduced in a number of countries, it is still at its early stages of development. Many modern technologies such as portable mobile devices could be adapted and used in the implementation of telemedicine programs. Deep machine learning algorithms and other artificial intelligence-based approaches demonstrate excellent results. Some systems have already received regulatory approval. Conclusions. The early diagnostics of diabetic retinopathy could be improved by integrating innovative screening technologies in a unified digital platform implemented in routine clinical practice.

A Pilot Study to Evaluate a New Hep-GRP Care Pathway to Improve Outcomes Among Canadian Older Adults with Liver Cirrhosis.

Older adults with cirrhosis have complex medical needs that are not satisfied by organ specific management. Interdisciplinary approach may mitigate comorbidity and improve patient satisfaction. A pilot study consisted of dual specialist interdisciplinary referral pathway and mixed virtual care delivery model are prospectively evaluated in older adults (65 years and older) with cirrhosis during the COVID-19 pandemic between September and December 2022. Participant attitudes towards telemedicine were surveyed. 68 participants with cirrhosis were consecutively assessed by hepatology. The mean age was 73 years. 39 (57%) screened positive for one or more geriatric syndrome(s). Comprehensive geriatric assessments were conducted via telemedicine in 18 participants, with additional referrals to physiotherapy and nutritional education. Compared to a historic cohort matched for age, sex, and Child-Pugh class, acute health service utilization measured by ER visits among those received dual specialist interdisciplinary consultation were lowered by 1.11 per patient at three-month follow up period (p = .0006, 95% CI 0.47-1.74). Majority participants (87.6%) preferred telemedicine or mixed method visits. An interdisciplinary approach to older adults with cirrhosis will likely be beneficial, and routine screening for geriatric syndrome may lead to reduced acute health-care utilization in the short term. Telemedicine and virtual screening tools in seniors should be fully explored to improve access to care.

Telemedicine and Depression Screening After the Start of the COVID-19 Pandemic

This cohort study investigates the probability of depression screening by visit type and by patient demographic characteristics in a large health system during the early COVID-19 pandemic.

Unilateral Retinopathy in a Preterm Infant With 4q Duplication: Description and Management.

A 33-5/7, 1570 g dichorionic diamniotic twin presented with cryptorchidism, failed hearing examination (both ears), poor feeding, profound hypoglycemia, coagulopathy, conjugated hyper-bilirubinemia, hydronephrosis, and hypotension. Microarray sent with results of whole genome SNP microgray analysis detected an interstitial duplication of the chromosomal segment 4q35 1q35.2. On this basis, telemedicine screening was performed to evaluate for ocular abnormalities in association with abnormal gene testing. Unilateral advanced retinopathy was noted affecting the right eye, with mature vascularization in the left eye. This infant was managed in concordance with retinopathy of prematurity guidelines, despite not making screening criteria. Off-label intravitreal bevacizumab injection (0.625 mg in 0.025 mL) resulted in full vascular maturation assessed by fluorescein angiography 6 months later. This represents the first description and management of retinopathy in 4q duplication syndrome. [Ophthalmic Surg Lasers Imaging Retina 2024;55:228-230.].

Integrated Telemedicine and E-Health with ERP System

In Medicals and medicine is very essential and most required thing in the world .so here is the resource to reach to everyone i.e. Medical solution for all the roles required in the main purpose of this integrated system is to keep the expire medicines record as well as to notify with itsneed and expire date. I’ll deal with the proof of concept of the most important part in telemedicine by doing many simulations on Software including the platform development, and those demonstrations are based on the actual data. In telemedicine you can buy medicine and also concern the doctor. It helps medical store owner to about medicine expiries and store details to maintain the medicine.The specific proposals are considered for the implementation of telemedicine screening complex on the application

Alabama Screening and Intervention for Glaucoma and Eye Health through Telemedicine (AL-SIGHT): Baseline Results

To describe baseline results of the Alabama Screening and Intervention for Glaucoma and eye Health through Telemedicine (AL-SIGHT) for patients at federally qualified health centers (FQHC's). Candidates were persons at-risk for glaucoma associated diseases (GAD) based on age, race/ethnicity, current diagnosis of GAD, family history, and diabetes. Baseline screening visit followed by remote diagnosis and referral for follow-up examinations. Patients presenting to FQHCs ≥18 years old were enrolled and underwent screening for acuity, autorefraction, intraocular pressure, visual field testing, and fundus imaging. Results were transmitted to an ophthalmologist at UAB for diagnosis who made referrals for follow-up; follow-up attendance was noted. Questionnaires assessed participants' perspectives on screening. Primary outcomes were rates of disease detection, referral for follow-up, follow-up attendance, and participant satisfaction. Of 500 participants enrolled (mean age 58), 45.6% were African American and 51.6% White. Remote diagnostic evaluation of ocular screening by ophthalmologist revealed 30% GAD, 6.8% diabetic retinopathy, 37.6% cataract, 68.4% refractive error, 9.2% other eye conditions. 47.2% were referred for follow-up examination and for acuity 20/40 or worse or IOP ≥ 23mmHG in one or both eyes. Follow-up examination attendance was 76.7% for those referred. Participants reported being very satisfied with screening (85.8%) and the convenience of screening in their primary care clinic (92.2%). The high percentage of patients diagnosed with treatable eye conditions at telemedicine screening suggest these programs in FQHCs can be effective and scalable nationwide. Attendance when referred for follow-up examination was high. Participants welcomed screenings in their communities.

Efficacy of Smartphone-Based Telescreening for Retinopathy of Prematurity With and Without Artificial Intelligence in India

Retinopathy of prematurity (ROP) telemedicine screening programs have been found to be effective, but they rely on widefield digital fundus imaging (WDFI) cameras, which are expensive, making them less accessible in low- to middle-income countries. Cheaper, smartphone-based fundus imaging (SBFI) systems have been described, but these have a narrower field of view (FOV) and have not been tested in a real-world, operational telemedicine setting. To assess the efficacy of SBFI systems compared with WDFI when used by technicians for ROP screening with both artificial intelligence (AI) and human graders. This prospective cross-sectional comparison study took place as a single-center ROP teleophthalmology program in India from January 2021 to April 2022. Premature infants who met normal ROP screening criteria and enrolled in the teleophthalmology screening program were included. Those who had already been treated for ROP were excluded. All participants had WDFI images and from 1 of 2 SBFI devices, the Make-In-India (MII) Retcam or Keeler Monocular Indirect Ophthalmoscope (MIO) devices. Two masked readers evaluated zone, stage, plus, and vascular severity scores (VSS, from 1-9) in all images. Smartphone images were then stratified by patient into training (70%), validation (10%), and test (20%) data sets and used to train a ResNet18 deep learning architecture for binary classification of normal vs preplus or plus disease, which was then used for patient-level predictions of referral warranted (RW)- and treatment requiring (TR)-ROP. Sensitivity and specificity of detection of RW-ROP, and TR-ROP by both human graders and an AI system and area under the receiver operating characteristic curve (AUC) of grader-assigned VSS. Sensitivity and specificity were compared between the 2 SBFI systems using Pearson χ2testing. A total of 156 infants (312 eyes; mean [SD] gestational age, 33.0 [3.0] weeks; 75 [48%] female) were included with paired examinations. Sensitivity and specificity were not found to be statistically different between the 2 SBFI systems. Human graders were effective with SBFI at detecting TR-ROP with a sensitivity of 100% and specificity of 83.49%. The AUCs with grader-assigned VSS only were 0.95 (95% CI, 0.91-0.99) and 0.96 (95% CI, 0.93-0.99) for RW-ROP and TR-ROP, respectively. For the AI system, the sensitivity of detecting TR-ROP sensitivity was 100% with specificity of 58.6%, and RW-ROP sensitivity was 80.0% with specificity of 59.3%. In this cross-sectional study, 2 different SBFI systems used by technicians in an ROP screening program were highly sensitive for TR-ROP. SBFI systems with AI may be a cost-effective method to improve the global capacity for ROP screening.

Photographic Gel Artifact Simulating International Classification of Retinopathy of Prematurity Notch: Stanford University Network for Diagnosis of Retinopathy of Prematurity (SUNDROP) Report no. 16

To report a series of fundus photographs taken for retinopathy of prematurity (ROP) screening that contain artifacts with imaging characteristics mimicking a notch, a recently refined classification metric in the International Classification of Retinopathy of Prematurity, third edition. Retrospective case series. Infants requiring ROP screening in neonatal intensive care units from the Stanford University Network for Diagnosis of Retinopathy of Prematurity (SUNDROP) and TeleROP telemedicine screening programs. Preterm infants meeting ROP examination criteria were screened with 130° wide-angle imaging systems. The images were taken by a trained nurse in the neonatal intensive care unit and transferred to an ROP specialist using a Health Insurance Portability and Accountability Act-compliant picture archiving and communication system for interpretation. Presence of an artifact that appeared consistent with a notch. We identified a total of 17 cases in ROP screening with artifact findings that had imaging characteristics similar to a notch. The artifactual appearance of the pseudo-notch was created by the camera illumination system within the gel-lens interface when the lens was not well apposed to the cornea. In telemedicine screening for ROP, we present fundus images of eyes with a pseudo-notch appearance; review of overlapping images can help differentiate between notch and artifact. Pediatric retinal specialists need to be aware that artifacts play a confounding role in screening for ROP, that can be mitigated through the use of overlapping and redundant images. Proprietary or commercial disclosure may be found after the references.

Perception of Telemedicine in Health Care Workers of a Tertiary Care Hospital

Objective: This investigation aimed to determine the feasibility of using telemedicine for screening and to observe healthcare provider’s opinion about telemedicine. In addition to that, explore the barriers in use of telemedicine Study Design and setting: A cross-sectional study was carried out at Central Park Medical College and Teaching Hospital. Methodology: The study was conducted between November, 2020 to April,2021 from 176 doctors and nurses working at Central Park Teaching Hospital using a convenience sampling technique. Convenience sampling was used because there is always a high probability of non-response or unavailability of clinical doctors and nurses. The minimum sample size was calculated as 88 for the group of doctors. The minimum sample size was doubled to collect the data from two cadres. The questionnaire was based on demographic information, general opinion about telemedicine such as reliability, comparison with standard examination procedure, nature, popularity, knowledge requirement, cost-effectiveness, designed frame and barriers for telemedicine. Results: Nearly, 80.1% of the participants were doctors with clinical experience while the remaining were nurses. Approximately 72.5% of the doctors and 58.5% of the sampled participants said that telemedicine does not fulfill the need for standard examination procedures. Conclusion: The findings of our study showed that the general opinion of healthcare professionals about telemedicine was unfavorable. Doctors had negative views but the nursing staff was optimistic. There was a mixed response about convenience. The main barrier to telemedicine was the poor communication skills of patients and availability and knowledge for using infrastructure.

In a Large Healthcare System in the Bronx, Teleretinal Triaging Was Found to Increase Screening and Healthcare Access for an Underserved Population with a High Incidence of T2DM and Retinopathy

The early treatment of diabetic retinopathy (DR) prevents vision-threatening proliferative retinopathy (PDR) and macular edema (DME). Our study evaluates telemedicine (teleretinal) screening for DR in an inner-city healthcare network with a high ethnic diversity and disease burden. Fundus photographs were obtained and graded in a centralized reading center between 2014 and 2016. Patients with positive screenings were referred to a retina specialist. An analysis of sensitivity and specificity and a subgroup analysis of prevalence, disease severity, and follow-up adherence were conducted. In 2251 patients, the '1-year' and 'Overall' follow-ups were 35.1% and 54.8%, respectively. Severe grading, male gender, and age were associated with better follow-up compliance. The DR, PDR, and DME prevalence was 24.9%, 4.1%, and 5.9%, respectively, and was significantly associated with HbA1c. The sensitivity and specificity for DR, PDR, and DME were 70% and 87%, 87% and 75%, and 37% and 95%, respectively. No prevalence differences were noted between ethnicities. Annual diabetic eye exam adherence increased from 55% to 85% during the study period. Teleretinal triaging is sensitive and specific for DR and improved diabetic eye exam compliance for underserved populations when integrated into large healthcare networks. The adherence to follow-up recommendations was better among older patients and among those with more severe retinopathy.

Telemedicine screening program for diabetic retinopathy in patients with type 1 diabetes mellitus

PurposeTo analyze the results of the telemedicine screening program for diabetic retinopathy (DR) in patients with type 1 diabetes conducted by the Endocrinology and Nutrition Management Unit of Virgen del Rocío University Hospital. MethodsThis cross-sectional study comprised patients with type 1 diabetes mellitus (DM) in our DR screening program from January 2018 to November 2020. Fundus photographs are performed by trained nurses and reviewed by a trained endocrinologist. Those suggestive of pathology are sent to ophthalmology through a telematic program for review. ResultsOf the 995 fundus photographs evaluated, 646 (65.3%) showed no evidence of DR, 327 (33.1%) presented possible DR, and 16 (1.6%) were not gradable. The diagnosis was confirmed in 254 patients after reviewing by ophthalmology, and the screening program achieved a positive predictive value for DR of 77.7%. Seventy-three were excluded by ophthalmology due to the absence of DR (false positive rate – 22.3%). In 92.5% of the cases classified by the ophthalmologist, the degree of DR was mild or very mild. ConclusionOur telemedicine screening program for DR in patients with type 1 DM is consistent with the literature. Effective screening for DR is performed, with patients diagnosed in the early stages. Telemedicine programs facilitate efficient communication among healthcare personnel.

Artificial Intelligence in Community-Based Diabetic Retinopathy Telemedicine Screening in Urban China: Cost-effectiveness and Cost-Utility Analyses With Real-world Data.

Community-based telemedicine screening for diabetic retinopathy (DR) has been highly recommended worldwide. However, evidence from low- and middle-income countries (LMICs) on the choice between artificial intelligence (AI)-based and manual grading-based telemedicine screening is inadequate for policy making. The aim of this study was to test whether the AI model is more worthwhile than manual grading in community-based telemedicine screening for DR in the context of labor costs in urban China. We conducted cost-effectiveness and cost-utility analyses by using decision-analytic Markov models with 30 one-year cycles from a societal perspective to compare the cost, effectiveness, and utility of 2 scenarios in telemedicine screening for DR: manual grading and an AI model. Sensitivity analyses were performed. Real-world data were obtained mainly from the Shanghai Digital Eye Disease Screening Program. The main outcomes were the incremental cost-effectiveness ratio (ICER) and the incremental cost-utility ratio (ICUR). The ICUR thresholds were set as 1 and 3 times the local gross domestic product per capita. The total expected costs for a 65-year-old resident were US $3182.50 and US $3265.40, while the total expected years without blindness were 9.80 years and 9.83 years, and the utilities were 6.748 quality-adjusted life years (QALYs) and 6.753 QALYs in the AI model and manual grading, respectively. The ICER for the AI-assisted model was US $2553.39 per year without blindness, and the ICUR was US $15,216.96 per QALY, which indicated that AI-assisted model was not cost-effective. The sensitivity analysis suggested that if there is an increase in compliance with referrals after the adoption of AI by 7.5%, an increase in on-site screening costs in manual grading by 50%, or a decrease in on-site screening costs in the AI model by 50%, then the AI model could be the dominant strategy. Our study may provide a reference for policy making in planning community-based telemedicine screening for DR in LMICs. Our findings indicate that unless the referral compliance of patients with suspected DR increases, the adoption of the AI model may not improve the value of telemedicine screening compared to that of manual grading in LMICs. The main reason is that in the context of the low labor costs in LMICs, the direct health care costs saved by replacing manual grading with AI are less, and the screening effectiveness (QALYs and years without blindness) decreases. Our study suggests that the magnitude of the value generated by this technology replacement depends primarily on 2 aspects. The first is the extent of direct health care costs reduced by AI, and the second is the change in health care service utilization caused by AI. Therefore, our research can also provide analytical ideas for other health care sectors in their decision to use AI.

Convolutional Neural Networks Accurately Identify Ungradable Images in a Diabetic Retinopathy Telemedicine Screening Program.

Purpose: Diabetic retinopathy (DR) is a microvascular complication of diabetes mellitus (DM). Standard of care for patients with DM is an annual eye examination or retinal imaging to assess for DR, the latter of which may be completed through telemedicine approaches. One significant issue is poor-quality images that prevent adequate screening and are thus ungradable. We used artificial intelligence to enable point-of-care (at time of imaging) identification of ungradable images in a DR screening program. Methods: Nonmydriatic retinal images were gathered from patients with DM imaged during a primary care or endocrinology visit from September 1, 2017, to June 1, 2021. The Topcon TRC-NW400 retinal camera (Topcon Corp., Tokyo, Japan) was used. Images were interpreted by 5 ophthalmologists for gradeability, presence and stage of DR, and presence of non-DR pathologies. A convolutional neural network with Inception V3 network architecture was trained to assess image gradeability. Images were divided into training and test sets, and 10-fold cross-validation was performed. Results: A total of 1,377 images from 537 patients (56.1% female, median age 58) were analyzed. Ophthalmologists classified 25.9% of images as ungradable. Of gradable images, 18.6% had DR of varying degrees and 26.5% had non-DR pathology. 10 fold cross-validation produced an average area under receiver operating characteristic curve (AUC) of 0.922 (standard deviation: 0.027, range: 0.882 to 0.961). The final model exhibited similar test set performance with an AUC of 0.924. Conclusions: This model accurately assesses gradeability of nonmydriatic retinal images. It could be used for increasing the efficiency of DR screening programs by enabling point-of-care identification of poor-quality images.