Features Associated with Visible Lamina Cribrosa Pores in Individuals of African Ancestry with Glaucoma: Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) Study.

There are scarce data regarding the rate of the occurrence of primary open-angle glaucoma (POAG) and visible lamina cribrosa pores (LCPs) in the eyes of individuals with African ancestry; the potential impact of these features on disease burden remains unknown. We recruited subjects with POAG to the Primary Open-Angle African American Glaucoma Genetics (POAAGG) study. Through regression models, we evaluated the association between the presence of LCPs and various phenotypic features. In a multivariable analysis of 1187 glaucomatous eyes, LCPs were found to be more likely to be present in eyes with cup-to-disc ratios (CDR) of ≥0.9 (adjusted risk ratio (aRR) 1.11, 95%CI: 1.04-1.19, p = 0.005), eyes with cylindrical-shaped (aRR 1.22, 95%CI: 1.11-1.33) and bean pot (aRR 1.24, 95%CI: 1.13-1.36) cups versus conical cups (p < 0.0001), moderate cup depth (aRR 1.24, 95%CI: 1.06-1.46) and deep cups (aRR 1.27, 95%CI: 1.07-1.50) compared to shallow cups (p = 0.01), and the nasalization of central retinal vessels (aRR 1.33, 95%CI: 1.23-1.44), p < 0.0001). Eyes with LCPs were more likely to have a higher degree of African ancestry (q0), determined by means of SNP analysis (aRR 0.96, 95%CI: 0.93-0.99, p = 0.005 for per 0.1 increase in q0). Our large cohort of POAG cases of people with African ancestry showed that LCPs may be an important risk factor in identifying severe disease, potentially warranting closer monitoring by physicians.

Prevalence and factors associated with optic disc grey crescent in the Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) Study.

AIM: To investigate the prevalence and factors associated with optic disc grey crescent (GC) in African Americans with glaucoma. METHODS: Stereo optic disc image features from subjects with glaucoma in the Primary Open-Angle African Ancestry Glaucoma Genetics Study were evaluated independently by non-physician graders and discrepancies adjudicated by an ophthalmologist. Risk factors for GC were evaluated by logistic regression models with intereye correlation accounted for by generalised estimating equations. Adjusted ORs (aORs) were generated. RESULTS: GC was present in 227 (15%) of 1491 glaucoma cases, with 57 (3.82%) bilateral and 170 (11.4%) unilateral. In multivariable analysis, factors associated with GC were younger age (aOR 1.27, 95% CI 1.11 to 1.43 for every decade younger in age, p=0.001), diabetes (aOR 1.46, 95% CI 1.09 to 1.96, p=0.01), optic disc tilt (aOR 1.84, 95% CI 1.36 to 2.48, p<0.0001), a sloping retinal region adjacent to the outer disc margin (aOR 2.37, 95% CI 1.74 to 3.32, p<0.0001) and beta peripapillary atrophy (aOR 2.32, 95% CI 1.60 to 3.37, p<0.0001). Subjects with GC had a lower mean (SD) value of the ancestral component q0 than those without GC (0.22 (0.15) vs 0.27 (0.20), p=0.001), consistent with higher degrees of African ancestry. CONCLUSIONS: More than 1 in 10 glaucoma cases with African ancestry have GC, occurring more frequently in younger subjects, higher degrees of African ancestry and diabetes. GC was associated with several ocular features, including optic disc tilt and beta peripapillary atrophy. These associations should be considered when evaluating black patients with primary open-angle glaucoma.

Utility of a machine-guided tool for assessing risk behaviour associated with contracting HIV in three sites in South Africa.

Introduction: Digital data collection and the associated mobile health technologies have allowed for the recent exploration of artificial intelligence as a tool for combatting the HIV epidemic. Machine learning has been found to be useful both in HIV risk prediction and as a decision support tool for guiding pre-exposure prophylaxis (PrEP) treatment. This paper reports data from two sequential studies evaluating the viability of using machine learning to predict the susceptibility of adults to HIV infection using responses from a digital survey deployed in a high burden, low-resource setting. Methods: 1036 and 593 participants were recruited across two trials. The first trial was a cross-sectional study in one location and the second trial was a cohort study across three trial sites. The data from the studies were merged, partitioned using standard techniques, and then used to train and evaluate multiple different machine learning models and select and evaluate a final model. Variable importance estimates were calculated using the PIMP and SHAP methodologies. Results: Characteristics associated with HIV were consistent across both studies. Overall, HIV positive patients had a higher median age (34 [IQR: 29-39] vs 26 [IQR 22-33], p < 0.001), and were more likely to be female (155/703 [22%] vs 107/927 [12%], p < 0.001). HIV positive participants also had more commonly gone a year or more since their last HIV test (183/262 [70%] vs 540/1368 [39%], p < 0.001) and were less likely to report consistent condom usage (113/262 [43%] vs 758/1368 [55%], p < 0.001). Patients who reported TB symptoms were more likely to be HIV positive. The trained models had accuracy values (AUROCs) ranging from 78.5% to 82.8%. A boosted tree model performed best with a sensitivity of 84% (95% CI 72-92), specificity of 71% (95% CI 67-76), and a negative predictive value of 95% (95% CI 93-96) in a hold-out dataset. Age, duration since last HIV test, and number of male sexual partners were consistently three of the four most important variables across both variable importance estimates. Conclusions: This study has highlighted the synergies present between mobile health and machine learning in HIV. It has been demonstrated that a viable ML model can be built using digital survey data from an low-middle income setting with potential utility in directing health resources.

Prevalence and Factors Associated with Optic Disc Tilt in the Primary Open-Angle African American Glaucoma Genetics Study.

PURPOSE: To investigate the prevalence and factors associated with optic disc tilt in the eyes of Black Americans with glaucoma. DESIGN: Cross-sectional. PARTICIPANTS: Subjects with glaucoma participating in the Primary Open-Angle African American Glaucoma Genetics (POAAGG) study. METHODS: Stereo pairs of optic disc images were assessed independently by POAAGG-certified nonphysician graders for quantitative features including maximum and minimum linear disc diameters, and qualitative features including gradeability of images, shape of the cup, rim plane position, ß-peripapillary atrophy, sloping region adjacent to the outer disc margin, and rim pallor. Discrepancies were adjudicated by an ophthalmologist. Descriptive statistics and P values were generated for associations of tilt with demographic and ocular characteristics. Stepwise multivariable analysis was performed with logistic regression using Generalized Estimating Equations (GEEs) to account for inter-eye correlation within subjects. MAIN OUTCOME MEASURES: Tilt Ovality Index (TOI) of >1.30 and Stereoscopically Identified optic disc Tilt (SIT). RESULTS: Among 1251 subjects with data on both eyes, 104 (8.3%) had TOI. Subjects with TOI were less likely to be male (adjusted odds ratio [aOR], 0.46, 95% confidence interval [CI], 0.29-0.74, P < 0.001). Eyes with TOI were less likely to have large cup disc ratios (aOR, 0.18, 95% CI, 0.06-0.53, P < 0.001) and less likely to have cylinder-shaped cups compared with conical-shaped cups (aOR, 0.31, 95% CI, 0.19-0.49, P < 0.001). Among 1007 subjects with data on both eyes, 254 (25.2%) had SIT. Subjects with SIT were younger (aOR, 0.95, 95% CI, 0.93-0.96, P < 0.001), and eyes with SIT were more likely to have oval-shaped discs compared with round discs (aOR, 1.82, 95% CI, 1.32-2.52, P < 0.001), more likely to have a sloping region adjacent to the outer disc margin instead of being flat (aOR, 3.26, 95% CI, 2.32-4.59, P < 0.001), and less likely to have cylinder-shaped cups compared with conical-shaped cups (aOR, 0.59, 95% CI, 0.41-0.85, P < 0.001). Both TOI and SIT were not associated with myopia. CONCLUSIONS: There are substantial numbers of tilted optic discs in glaucoma patients with African ancestry. They occur more frequently in female subjects and younger subjects and are associated with several ocular features but not with myopia.

The role of machine learning in HIV risk prediction.

Despite advances in reducing HIV-related mortality, persistently high HIV incidence rates are undermining global efforts to end the epidemic by 2030. The UNAIDS Fast-track targets as well as other preventative strategies, such as pre-exposure prophylaxis, have been identified as priority areas to reduce the ongoing transmission threatening to undermine recent progress. Accurate and granular risk prediction is critical for these campaigns but is often lacking in regions where the burden is highest. Owing to their ability to capture complex interactions between data, machine learning and artificial intelligence algorithms have proven effective at predicting the risk of HIV infection in both high resource and low resource settings. However, interpretability of these algorithms presents a challenge to the understanding and adoption of these algorithms. In this perspectives article, we provide an introduction to machine learning and discuss some of the important considerations when choosing the variables used in model development and when evaluating the performance of different machine learning algorithms, as well as the role emerging tools such as Shapely Additive Explanations may play in helping understand and decompose these models in the context of HIV. Finally, we discuss some of the potential public health and clinical use cases for such decomposed risk assessment models in directing testing and preventative interventions including pre-exposure prophylaxis, as well as highlight the potential integration synergies with algorithms that predict the risk of sexually transmitted infections and tuberculosis.

A Precise Method to Evaluate 360 Degree Measures of Optic Cup and Disc Morphology in an African American Cohort and Its Genetic Applications.

(1) Background: Vertical cup-to-disc ratio (CDR) is an important measure for evaluating damage to the optic nerve head (ONH) in glaucoma patients. However, this measure often does not fully capture the irregular cupping observed in glaucomatous nerves. We developed and evaluated a method to measure cup-to-disc ratio (CDR) at all 360 degrees of the ONH. (2) Methods: Non-physician graders from the Scheie Reading Center outlined the cup and disc on digital stereo color disc images from African American patients enrolled in the Primary Open-Angle African American Glaucoma Genetics (POAAGG) study. After converting the resultant coordinates into polar representation, the CDR at each 360-degree location of the ONH was obtained. We compared grader VCDR values with clinical VCDR values, using Spearman correlation analysis, and validated significant genetic associations with clinical VCDR, using grader VCDR values. (3) Results: Graders delineated outlines of the cup contour and disc boundaries twice in each of 1815 stereo disc images. For both cases and controls, the mean CDR was highest at the horizontal bisector, particularly in the temporal region, as compared to other degree locations. There was a good correlation between grader CDR at the vertical bisector and clinical VCDR (Spearman Correlation OD: r = 0.78 [95% CI: 0.76-0.79]). An SNP in the MPDZ gene, associated with clinical VCDR in a prior genome-wide association study, showed a significant association with grader VCDR (p = 0.01) and grader CDR area ratio (p = 0.02). (4) Conclusions: The CDR of both glaucomatous and non-glaucomatous eyes varies by degree location, with the highest measurements in the temporal region of the eye. This method can be useful for capturing innate eccentric ONH morphology, tracking disease progression, and identifying genetic associations.

BETA-PERIPAPILLARY ATROPHY AND GEOGRAPHIC ATROPHY IN THE COMPARISON OF AGE-RELATED MACULAR DEGENERATION TREATMENTS TRIALS.

PURPOSE: To determine associations between beta-peripapillary atrophy (B-PPA) and incidence and growth of geographic atrophy (GA) in eyes treated with anti-vascular endothelial growth factor agents in the Comparison of Age-Related Macular Degeneration Treatments Trials (CATT). METHODS: We included 245 cases with incident GA and 245 controls matched by baseline demographics and characteristics associated with development of GA in the CATT. Baseline color images were graded for the type of B-PPA, defined as presence of hypopigmentation with visible choroidal vessels and sclera that is adjacent to the optic disk. Beta-peripapillary atrophy was further classified as scleral ring, sclera, sclera/choroidal blood vessels, or combination. Areas of each type of B-PPA and the circumferential extent of B-PPA were measured. RESULTS: Beta-peripapillary atrophy was present in 58% of eyes developing GA and in 52% without GA (P = 0.17). The greater circumferential extent of sclera/choroidal blood vessels B-PPA in relation to the optic disk was associated with incident GA (P = 0.02) and the GA size at first observation (P = 0.047). Beta-peripapillary atrophy was not associated with GA growth rates (P>0.05). Patients without B-PPA had a higher number of GA-associated risk alleles of ARMS2 (P = 0.0003) and HTRA1 (P = 0.001). CONCLUSION: The extent of sclera/choroidal blood vessel B-PPA was associated with the GA incidence and size but not with the growth rate in eyes treated for neovascular age-related macular degeneration. Beta-peripapillary atrophy and GA may share some common pathophysiologic pathways unrelated to the GA-associated risk alleles evaluated.

Progression from preplus to plus disease in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study: incidence, timing, and predictors.

PURPOSE: To determine the incidence of and timing and predictors for progression from pre-plus to plus disease, based on evaluation of images. METHODS: Two trained readers independently evaluated posterior pole images of infants from 13 North American centers for pre-plus/plus disease, stage, and zone of retinopathy of prematurity (ROP). Discrepancies between readers were adjudicated. To be eligible for analysis, eyes had to have at least two imaging sessions, the earlier one with pre-plus disease. RESULTS: Of 681 eyes of 444 infants with pre-plus first detected at mean postmenstrual age (PMA) of 35.5 ± 2.1 weeks, 54 (7.9%) progressed to plus disease at a mean PMA of 37.6 ± 2.4 weeks with the mean interval for progression of 2.7 weeks (range, 0.4-8.9 weeks). Progression rate was higher for eyes with larger number of quadrants of pre-plus (44% for eyes with four quadrants vs 4% with one quadrant [P < 0.0001]), earlier PMA with pre-plus (18% for 32 weeks' PMA vs 3% for PMA of >37 weeks [P = 0.02]), higher ROP stage (12% for stage 3, 2.5% for no ROP [P < 0.0001]), lower ROP zone (24% for zone I, 6% for zone II or no ROP [P < 0.0001]) at the time of first pre-plus detection. CONCLUSIONS: Based on image evaluation, 8% of eyes progressed from pre-plus to plus disease at a mean interval of 3 weeks. Pre-plus in multiple quadrants, higher stages of ROP, and lower zones of ROP were associated with higher risk of progression. Image evaluation for pre-plus may help in the identification of high-risk eyes for developing plus disease.

Grading and baseline characteristics of meibomian glands in meibography images and their clinical associations in the Dry Eye Assessment and Management (DREAM) study.

PURPOSE: To describe and evaluate a comprehensive grading system for meibomian gland (MG) digital infrared images developed for the Dry Eye Assessment and Management (DREAM) Study. METHODS: Cross-sectional study. Reading Center (RC) certified readers independently evaluated MG features of both lids from meibography images of dry eye disease subjects. Dropout areas were measured using planimetry software. Inter-reader and grade-regrade agreement and comparison of meiboscale scores (Meiboscale©; Pult) from clinical centers to RC percent dropout and of MG features with clinical parameters were evaluated. RESULTS: Among 551 eyes of 277 patients at baseline, 62 (11%) upper lid and 5 (1%) lower lid images were missing. Lid eversion was poor in 63 (13%) of upper lids compared to 15 (3%) of lower lids. Intraclass correlation for inter-reader and grade-regrade agreement was moderate to substantial for most MG features. MG features were more frequent in the upper lid (p < 0.001), except for dropout glands, gaps, fluffy gland areas and dropout areas. Clinic meiboscale score was associated with RC percent dropout (p < 0.001), a clinic score of 0% having a mean RC score of 19%, and a clinic score of >75% having a mean RC score of 66%. MG plugging was associated with ghost glands (p = 0.009), dropout glands (p < 0.001) and a composite severity score (p = 0.02); turbid and absent secretions were associated with ghost glands (p = 0.046). CONCLUSION: RC readers identified MG features with good reproducibility. Upper lids had more MG features. RC dropout areas correlated well with clinic meiboscale scores. Ghost glands were associated with paste like and absent meibomian secretions.

Asymmetry of Retinopathy of Prematurity Border in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity Study.

PURPOSE: To measure the location of the retinopathy of prematurity (ROP) border in a sample of premature infants developed ROP and to determine the location for predicting subsequent development of referral-warranted ROP (defined as stage 3 ROP, zone I ROP, or plus disease) or treated ROP. DESIGN: Secondary analysis of data from an observational cohort study, the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity study. PARTICIPANTS: Infants with birth weight (BW) of less than 1251 g with at least 1 examination and imaging session at 34 weeks postmenstrual age (PMA) or younger with ROP not meeting referral-warranted ROP and at least 1 subsequent examination at 36 weeks PMA or older. METHODS: The 5-image set from the first imaging session for each eligible eye was mosaicked, and measurements of the ROP border were made using Image J. Measurements were compared among 3 groups of eyes with ROP: (1) those that never developed referral-warranted ROP or received treatment, (2) those that developed referral-warranted ROP but did not receive treatment, and (3) those that received treatment. MAIN OUTCOME MEASURES: Referral-warranted ROP and treated ROP. RESULTS: Three hundred seventeen eyes from 217 infants with mean BW of 755 g, mean gestational age (GA) of 25 weeks, and mean PMA of 33 weeks at first examination met study criteria. Of 211 eyes (66.6%) with sufficient-quality images for grading, 147 (69.7%) did not develop referral-warranted ROP or receive treatment, 36 (17.1%) developed referral-warranted ROP not requiring treatment, and 28 (13.3%) received treatment for ROP. Among all eyes, the disc-to-ROP border distance followed a consistent pattern, with nasal less than inferior less than superior less than temporal. In multivariate analysis adjusted by BW and GA, nasal ROP border distance was a significant predictor of the subsequent development of ROP that was treated (odds ratio, 0.86 for every 10-pixel increase in distance; P = 0.002). CONCLUSIONS: Retinopathy of prematurity is located asymmetrically around the optic disc and is closest to the optic disc nasally. Location of ROP nasally at first imaging is a significant predictor for subsequent development of ROP that was treated.

A new scale for the assessment of conjunctival bulbar redness.

PURPOSE: Current scales for assessment of bulbar conjunctival redness have limitations for evaluating digital images. We developed a scale suited for evaluating digital images and compared it to the Validated Bulbar Redness (VBR) scale. METHODS: From a digital image database of 4889 color corrected bulbar conjunctival images, we identified 20 images with varied degrees of redness. These images, ten each of nasal and temporal views, constitute the Digital Bulbar Redness (DBR) scale. The chromaticity of these images was assessed with an established image processing algorithm. Using 100 unique, randomly selected images from the database, three trained, non-physician graders applied the DBR scale and printed VBR scale. Agreement was assessed with weighted Kappa statistics (Kw). RESULTS: The DBR scale scores provide linear increments of 10 from 10-100 when redness is measured objectively with an established image processing algorithm. Exact agreement of all graders was 38% and agreement with no more than a difference of ten units between graders was 91%. Kw for agreement between any two graders ranged from 0.57 to 0.73 for the DBR scale and from 0.38 to 0.66 for the VBR scale. The DBR scale allowed direct comparison of digital to digital images, could be used in dim lighting, had both temporal and nasal conjunctival reference images, and permitted viewing reference and test images at the same magnification. CONCLUSION: The novel DBR scale, with its objective linear chromatic steps, demonstrated improved reproducibility, fewer visualization artifacts and improved ease of use over the VBR scale for assessing conjunctival redness.

Single grading vs double grading with adjudication in the telemedicine approaches to evaluating acute-phase retinopathy of prematurity (e-ROP) study.

PURPOSE: To evaluate the sensitivity and specificity of single, independent, nonphysician trained reader (TR) gradings in the Telemedicine Approaches to Evaluating Acute-phase Retinopathy of Prematurity (e-ROP) study. METHODS: Secondary analyses of image grading results from 1,235 infants of birth weights <1251 g. Two of three TRs independently graded image sets; discrepancies were adjudicated by the reading center director (an ophthalmologist) to reach final grading. Sensitivity and specificity of each TR grading and final grading was calculated by comparing gradings to clinical examination results. RESULTS: Of 7,808 double graded image sets, TR1 graded 5,165; TR2, 3,787; and TR3, 6,664. Compared to final grading for referral warranted retinopathy of prematurity (RW-ROP), two TRs had relatively lower sensitivity (TR1, 75% vs 79% [P = 0.03]; TR2, 73% vs 77% [P = 0.02]) and specificity (TR1, 80% vs 83% [P < 0.001]; TR2, 82% vs 83% [P = 0.09]). TR3 had similar sensitivity (83% vs 83% [P = 0.78]) and specificity (83% vs 84% [P = 0.02]). Compared to final grading, TR1 had lower sensitivity for zone I ROP (47% vs 56% [P = 0.04]) and stage ≥3 ROP (71% vs 77% [P = 0.002]); TR2 had lower sensitivity for stage ≥3 ROP (69% vs 77% [P < 0.001]) and lower specificity for all three components (P < 0.001); TR3 had lower sensitivity for detecting plus disease (23% vs 35% [P < 0.001]) and similar sensitivity for zone I ROP and stage ≥3 ROP. CONCLUSIONS: There is a small but significant decrease in the sensitivity and specificity for RW-ROP when single-reader grading is compared to double adjudicated grading.

Development and Evaluation of Semiautomated Quantification of Lissamine Green Staining of the Bulbar Conjunctiva From Digital Images.

Importance: Lissamine green (LG) staining of the conjunctiva is a key biomarker in evaluating ocular surface disease. The disease currently is assessed using relatively coarse subjective scales. Objective assessment would standardize comparisons over time and between clinicians. Objective: To develop a semiautomated, quantitative system to assess lissamine green staining of the bulbar conjunctiva on digital images. Design, Setting, and Participants: Using a standard photography protocol, 35 digital images of the conjunctiva of 11 patients with a diagnosis of dry eye disease based on characteristic signs and symptoms were obtained after topical administration of preservative-free LG, 1%, solution. Images were scored independently by 2 masked ophthalmologists in an academic medical center using the van Bijsterveld and National Eye Institute (NEI) scales. The region of interest was identified by manually marking 7 anatomic landmarks on the images. An objective measure was developed by segmenting the images, forming a vector of key attributes, and then performing a random forest regression. Subjective scores were correlated with the output from a computer algorithm using a cross-validation technique. The ranking of images from least to most staining was compared between the algorithm and the ophthalmologists. The study was conducted from April 26, 2012, through June 2, 2016. Main Outcomes and Measures: Correlation and level of agreement among computerized algorithm scores, van Bijsterveld scale clinical scores, and NEI scale clinical scores. Results: The scores from the automated algorithm correlated well with the mean scores obtained from the gradings of 2 ophthalmologists for the 35 images using the van Bijsterveld scale (Spearman correlation coefficient, rs = 0.79), and moderately with the NEI scale (rs = 0.61) scores. For qualitative ranking of staining, the correlation between the automated algorithm and the 2 ophthalmologists was rs = 0.78 and rs = 0.83. Conclusions and Relevance: The algorithm performed well when evaluating LG staining of the conjunctiva, as evidenced by good correlation with subjective gradings using 2 different grading scales. Future longitudinal studies are needed to assess the responsiveness of the algorithm to change of conjunctival staining over time.

Use of Crossed Polarizers to Enhance Images of the Eyelids.

PURPOSE: To describe imaging of the external eye with Crossed Polarizers to enhance clinically important features in digital photographs of the eyelids. METHODS: External photographs with and without crossed polarizing filters were taken of patients with blepharitis and controls with no clinical eye pathology. RESULTS: Photographing eyelid skin through Crossed Polarizers decreased reflections on the skin surface and improved visualization of eyelid telangiectasias and blood vessels in patients with a broad range of skin pigmentation and ethnicities. CONCLUSIONS: The use of Crossed Polarizers in imaging the external eye reduces reflections and glare from the eyelid skin and margins, thereby allowing for a more detailed evaluation of underlying structures and analysis of images. These findings suggest that including Crossed Polarizers in clinical photography has informative applications for assessing eyelid disease.

Intraocular Hemorrhages and Retinopathy of Prematurity in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study.

PURPOSE: To describe the clinical characteristics of intraocular hemorrhages (IOHs) in infants in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study and to evaluate their potential use for prediction of disease severity. DESIGN: Secondary data analysis from a prospective study. PARTICIPANTS: Preterm infants with birth weight (BW) ≤1250 g. METHODS: Infants underwent serial digital retinal imaging in both eyes starting at 32 weeks' postmenstrual age. Nonphysician trained readers (TRs) evaluated all image sets from eyes that ever had IOHs documented on image evaluation or eye examination for the presence, location, type, area, and relation of the IOH to the junction between vascularized and avascular retina. Associations of IOH with demographic and neonatal factors, and with the presence and severity of retinopathy of prematurity (ROP) were investigated by univariate and multivariate analyses. Sensitivity and specificity of the telemedicine system for detecting referral-warranted ROP (RW-ROP) were calculated with and without incorporating hemorrhage into the standardized grading protocol. MAIN OUTCOME MEASURES: Retinal and vitreous hemorrhage. RESULTS: Among 1239 infants (mean [standard deviation] BW = 864 [212] g; gestational age [GA] = 27 [2.2] weeks) who underwent an average of 3.2 imaging sessions, 22% had an IOH in an eye on at least 1 of the e-ROP visits. Classification of IOH was preretinal (57%), blot (57%), dot (38%), flame-shaped (16%), and vitreous (8%); most IOHs were unilateral (70%). The IOH resolved in 35% of eyes by the next imaging session and in the majority (76%) of cases by 8 weeks after initial detection. Presence of IOH was inversely associated with BW and GA and significantly associated (P < 0.0001) with the presence and severity of ROP (BW and GA adjusted odds ratios [ORs] of 2.46 for any ROP, 2.88 for stage 3, and 3.19 for RW-ROP). Incorporating IOH into the grading protocol minimally altered the sensitivity of the system (94% vs. 95%). CONCLUSIONS: Approximately 1 in 5 preterm infants examined had IOHs, generally unilateral. The presence of hemorrhage was directly correlated with both presence and severity of ROP and inversely correlated with BW and GA, although including hemorrhage in the grading algorithm only minimally improved the sensitivity of the telemedicine system to detect RW-ROP.

Validated System for Centralized Grading of Retinopathy of Prematurity: Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study.

IMPORTANCE: Measurable competence derived from comprehensive and advanced training in grading digital images is critical in studies using a reading center to evaluate retinal fundus images from infants at risk for retinopathy of prematurity (ROP). Details of certification for nonphysician trained readers (TRs) have not yet been described. OBJECTIVE: To describe a centralized system for grading ROP digital images by TRs in the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study. DESIGN, SETTING, AND PARTICIPANTS: Multicenter observational cohort study conducted from July 1, 2010, to June 30, 2014. The TRs were trained by experienced ROP specialists and certified to detect ROP morphology in digital retinal images under supervision of an ophthalmologist reading center director. An ROP reading center was developed with standard hardware, secure Internet access, and customized image viewing software with an electronic grading form. A detailed protocol for grading was developed. Based on results of TR gradings, a computerized algorithm determined whether referral-warranted ROP (RW-ROP; defined as presence of plus disease, zone I ROP, and stage 3 or worse ROP) was present in digital images from infants with birth weight less than 1251 g enrolled from May 25, 2011, through October 31, 2013. Independent double grading was done by the TRs with adjudication of discrepant fields performed by the reading center director. EXPOSURE: Digital retinal images. MAIN OUTCOMES AND MEASURES: Intragrader and intergrader variability and monitoring for temporal drift. RESULTS: Four TRs underwent rigorous training and certification. A total of 5520 image sets were double graded, with 24.5% requiring adjudication for at least 1 component of RW-ROP. For individual RW-ROP components, the adjudication rate was 3.9% for plus disease, 12.4% for zone I ROP, and 16.9% for stage 3 or worse ROP. The weighted κ for intergrader agreement (n = 80 image sets) was 0.72 (95% CI, 0.52-0.93) for RW-ROP, 0.57 (95% CI, 0.37-0.77) for plus disease, 0.43 (95% CI, 0.24-0.63) for zone I ROP, and 0.67 (95% CI, 0.47-0.88) for stage 3 or worse ROP. The weighted κ for grade-regrade agreement was 0.77 (95% CI, 0.57-0.97) for RW-ROP, 0.87 (95% CI, 0.67-1.00) for plus disease, 0.70 (95% CI, 0.51-0.90) for zone I ROP, and 0.77 (95% CI, 0.57-0.97) for stage 3 or worse ROP. CONCLUSIONS AND RELEVANCE: These data suggest that the e-ROP system for training and certifying nonphysicians to grade ROP images under the supervision of a reading center director reliably detects potentially serious ROP with good intragrader and intergrader consistency and minimal temporal drift.