Trends in Retina Specialist Imaging Utilization from 2012 to 2016 in the United States Medicare Fee-for-Service Population

Human selection bias drives the linear nature of the more ground truth effect in explainable deep learning optical coherence tomography image segmentation

Supervised deep learning (DL) algorithms are highly dependent on training data for which human graders are assigned, for example, for optical coherence tomography (OCT) image annotation. Despite the tremendous success of DL, due to human judgment, these ground truth labels can be inaccurate and/or ambiguous and cause a human selection bias. We therefore investigated the impact of the size of the ground truth and variable numbers of graders on the predictive performance of the same DL architecture and repeated each experiment three times. The largest training dataset delivered a prediction performance close to that of human experts. All DL systems utilized were highly consistent. Nevertheless, the DL under-performers could not achieve any further autonomous improvement even after repeated training. Furthermore, a quantifiable linear relationship between ground truth ambiguity and the beneficial effect of having a larger amount of ground truth data was detected and marked as the more-ground-truth effect.

Trends in the Diagnosed Prevalence and Incidence of Major Eye Diseases in Medicare Part B Fee-for-Service Beneficiaries 68 Years of Age or Older

PURPOSE

To study contemporary trends in the diagnosed prevalence and incidence of age-related eye diseases among Medicare Fee-for-Service (FFS) beneficiaries.

DESIGN

Analysis of Medicare administrative claims data.

PARTICIPANTS

Medicare FFS beneficiaries 68 years of age and older from 2005 through 2020 who were enrolled continuously in both Part A and Part B for 3 years, including the index year and a 2-year lookback period.

METHODS

Annual cross-sectional diagnosed prevalence and incidence rates were calculated. Age standardization was performed using the direct standardization method to account for changes in the age structure of the study population. Rates stratified by demographics (age, sex, race, and ethnicity) also were calculated.

MAIN OUTCOME MEASURES

Annual prevalence and incidence of diagnosed age-related macular degeneration (AMD), diabetic retinopathy (DR) (among those with diabetes), and glaucoma.

RESULTS

At baseline, in 2005, 60% of included beneficiaries were female, 20% were 85 years of age or older, 86% were non-Hispanic White, and one-quarter had a diagnosis of diabetes. From 2005 through 2019, the prevalence of a diagnosis of any of the conditions studied increased from 16.4% (n = 3 628 996) to 17.9% (n = 3 731 281). Diagnosed incidence decreased over this period from 4.9% (n = 954 878) in 2005 to 4.2% in 2019 (n = 757 696). The diagnosed prevalence of AMD increased from 6.8% (n = 1 504 770) to 9.4% (n = 1 965 176); the diagnosed prevalence of any DR among those with diabetes decreased from 9.3% (n = 504 135) to 9.0% (n = 532 859), although the diagnosed prevalence of vision-threatening DR increased from 2.0% to 3.4%; and the diagnosed prevalence of any diagnosed glaucoma decreased from 8.8% (n = 1 951 141) to 8.1% (n = 1 692 837). In 2020, the diagnosed prevalence and incidence of all diagnoses decreased. During the study period, we detected demographic differences in the prevalence and incidence of diagnosis of each condition.

CONCLUSIONS

This study presents updated data on the prevalence and incidence of diagnosed major chronic, age-related eye diseases among Medicare FFS beneficiaries. Compared with older epidemiologic estimates, we found that the diagnosed prevalence of each condition studied was higher in more recent years. These findings may inform public health and policy planning and resource allocation to address the eye health of an increasingly older United States population.

FINANCIAL DISCLOSURE(S)

The author(s) have no proprietary or commercial interest in any materials discussed in this article.

Trends in Imaging Utilization Among United States Medicare Beneficiaries and the Impact of the COVID-19 Pandemic

This study characterizes trends in ophthalmic imaging volume and utilization in the United States among and assesses the potential impact of the COVID-19 pandemic using data from the Centers for Medicare and Medicaid Services. Utilization of macular optical coherence tomography (OCT), optic nerve OCT, and fundus photography steadily increased from 2013 to 2020 and was not impacted by the COVID-19 pandemic. At the same time, there was minimal adoption of anterior segment OCT and a decline in the utilization of dye-based angiography. Utilization patterns may be impacted by the advent of new technologies, role of the clinician, and alignment with treatment paradigms. [Ophthalmic Surg Lasers Imaging Retina 2023;54:661-665.].

Vitreoretinal Specialist Use of Ancillary Testing: An IRIS® Registry Analysis

Purpose

To investigate patterns of ancillary imaging testing among vitreoretinal specialists for patients with vitreoretinal disease in the United States (US).

Methods

Optical coherence tomography (OCT), color fundus photography (CFP), and fluorescein angiography (FA), ordered by vitreoretinal specialists and documented within the American Academy of Ophthalmology IRIS® Registry (Intelligent Research in Sight) between 01 January 2018 and 31 December 2020, were retrospectively assessed. Trends in imaging modality choice were analyzed by payer type, geographic region, and practice type. Sub-analyses were conducted according to categorization of vitreoretinal specialists into those treating a high versus low volume of patients with neovascular age-related macular degeneration (nAMD).

Results

OCT was the most common imaging modality used, followed by CFP and FA. Following normalization, the highest volume of OCT procedures performed were identified among Medicare Advantage and Medicare Fee-for-Service beneficiaries, within the South of the US, and at medium and large practices. Minimal differences were observed for CFP and FA volume across payer types and regions. Across practice types, the largest volume of CFP and FA procedures were identified in small and private equity owned practices, respectively. Vitreoretinal specialists with a high nAMD volume more frequently performed OCT than those with a low nAMD volume.

Conclusion

Vitreoretinal specialists demonstrated a strong preference for OCT, with real-world associations according to payer type, geographic location, and practice type. Volume of nAMD patients seen impacted the likelihood of specialists ordering OCTs.

A Validated Method to Identify Neuro-Ophthalmologists in a Large Administrative Claims Database

BACKGROUND

Validated methods to identify neuro-ophthalmologists in administrative data do not exist. The development of such method will facilitate research on the quality of neuro-ophthalmic care and health care utilization for patients with neuro-ophthalmic conditions in the United States.

METHODS

Using nationally representative, 20% sample from Medicare carrier files from 2018, we identified all neurologists and ophthalmologists billing at least 1 office-based evaluation and management (E/M) outpatient visit claim in 2018. To isolate neuro-ophthalmologists, the National Provider Identifier numbers of neuro-ophthalmologists in the North American Neuro-Ophthalmology Society (NANOS) directory were collected and linked to Medicare files. The proportion of E/M visits with International Classification of Diseases-10 diagnosis codes that best distinguished neuro-ophthalmic care ("neuro-ophthalmology-specific codes" or NSC) was calculated for each physician. Multiple logistic regression models assessed predictors of neuro-ophthalmology specialty designation after accounting for proportion of ophthalmology, neurology, and NSC claims and primary specialty designation. Sensitivity, specificity, and positive predictive value (PPV) for varying proportions of E/M visits with NSC were calculated.

RESULTS

We identified 32,293 neurologists and ophthalmologists who billed at least 1 outpatient E/M visit claim in 2018 in Medicare. Of the 472 NANOS members with a valid individual National Provider Identifier, 399 (84.5%) had a Medicare outpatient E/M visit in 2018. The model containing only the proportion of E/M visits with NSC best predicted neuro-ophthalmology specialty designation (odds ratio 1.05 [95% confidence interval 1.04, 1.05]; P < 0.001; area under the receiver operating characteristic [AUROC] = 0.91). Model predictiveness for neuro-ophthalmology designation was maximized when 6% of all billed claims were for NSC (AUROC = 0.89; sensitivity: 84.0%; specificity: 93.9%), but PPV was low (14.9%). The threshold was unchanged when limited only to neurologists billing ≥1% ophthalmology claims or ophthalmologists billing ≥1% neurology claims, but PPV increased (33.3%).

CONCLUSIONS

Our study provides a validated method to identify neuro-ophthalmologists who can be further adapted for use in other administrative databases to facilitate future research of neuro-ophthalmic care delivery in the United States.

Agreement of optical coherence tomography thickness measurements between Heidelberg Eye Explorer and ImageJ software

OBJECTIVE

To investigate the agreement and correlation of manual central foveal thickness (CFT) measurements using the Heidelberg Eye Explorer and ImageJ software.

METHODS

Optical coherence tomography (OCT) images were identified. CFT was measured using the Heidelberg Eye Explorer caliper tool. Two independent graders measured the CFT of the same images using ImageJ software. A Pearson correlation coefficient, intraclass correlation coefficient, Bland-Altman plots, and coefficient of repeatability were used to assess the agreement and correlation between the Heidelberg Eye Explorer and ImageJ measurements.

RESULTS

One-hundred and twenty-two OCT images from 91 patients were analyzed. Mean CFT as measured using the Heidelberg caliper tool was 264.8 ± 113 µm. Using ImageJ software, graders 1 and 2 obtained mean CFT values of 267 ± 114.3 µm and 270 ± 114.8 µm, respectively. Pearson correlation revealed a strong correlation between the Heidelberg and Image J measurements (r = 0.999, p < 0.0001). Bland-Altman plots noted a less than 4 µm mean difference between Heidelberg and ImageJ CFT values (2.67 ± 3.46 and 3.72 ± 2.78 µm for graders 1 and 2, respectively).

CONCLUSION

There was strong agreement between the Heidelberg Eye Explorer and ImageJ software for manual CFT measurements. ImageJ may be a reliable tool for thickness measurements when proprietary OCT software is unavailable.

Importance of Anatomical Efficacy for Disease Control in Neovascular AMD: An Expert Opinion

Background

Neovascular age-related macular degeneration (nAMD) presents a significant treatment burden for patients, carers and medical retina services. However, significant debate remains regarding how best to manage nAMD when assessing disease activity by optical coherence tomography (OCT), and particularly the significance of different types of fluid and how the understanding of anatomical efficacy can influence treatment strategies. This article provides opinion on the practical implications of anatomical efficacy and significance of fluid in the management of nAMD and proposes recommendations for healthcare professionals (HCPs) to improve understanding and promote best practice to achieve disease control.

Methods

An evidence-based review was performed and an expert panel debate from the Retina Outcomes Group (ROG), a forum of retinal specialists, provided insights and recommendations on the definition, role and practical implications of anatomical efficacy and the significance of fluid at the macula in the management of nAMD.

Results

The ROG has developed recommendations for achieving disease control through a zero-tolerance approach to the presence of fluid in nAMD as patients who avoid fluctuations in fluid at the macula have better visual outcomes. Recommendations cover five key areas: service protocol, training, regimen, multidisciplinary teams and engagement. This approach facilitates more standardised protocol-based treatment strategies.

Conclusions

Targeting a fluid-free macula and aiming for disease control are essential to improve outcomes. As new therapies and technologies become available, drying the macula and maintaining disease control will become even more achievable. The outlined recommendations aim to promote best practice among HCPs and medical retina services to improve patient outcomes.

Agreement of Spectral-Domain OCT with Fluorescein Leakage in Neovascular Age-Related Macular Degeneration: Post Hoc Analysis of the HARBOR Study

PURPOSE

To evaluate the agreement between detection of activity of choroidal neovascularization (CNV) in neovascular age-related macular degeneration (AMD) by fundus fluorescein angiography (FFA) and spectral-domain (SD) OCT in the HARBOR study. Most retina specialists rely on OCT to guide treatment decisions in neovascular AMD. However, OCT may not always detect exudative activity. Traditionally, FFA was frequently performed in clinical practice, but its use has diminished due to reliance on OCT.

DESIGN

Retrospective post hoc analysis of prospective clinical trial (HARBOR; ClinicalTrials.gov identifier, NCT00891735).

PARTICIPANTS

Patients with neovascular AMD in the HARBOR Trial.

METHODS

Baseline to month 24 data from all randomized study eyes in HARBOR with both FFA and SD OCT data were analyzed for (1) evidence of CNV activity on SD OCT (presence of subretinal fluid, intraretinal fluid, and/or cystoid spaces); (2) evidence of CNV activity on FFA identified by the presence of leakage, and (3) cross-tabulation of CNV activity identified by FFA and SD OCT by office visit.

MAIN OUTCOME MEASURES

The percent agreement between FFA and SD OCT in detecting CNV activity and sensitivity and specificity of SD OCT to detect fluorescein leakage in neovascular AMD using FFA as the reference standard.

RESULTS

At baseline, 1094 patients (99.9%) had agreement between SD OCT and FFA in detecting CNV activity. By month 24, of the 779 total active cases, the agreement was only 36% (277 cases). By month 24, most cases (n = 452 [58%]) had evidence of CNV activity on SD OCT only, whereas 6% of cases (n = 50) had CNV activity identified by FFA only. At screening and months 3, 6, 12, and 24, 92% to 100% of cases identified by FFA only were occult CNV lesions. Using FFA as the reference standard, the sensitivity and specificity of SD OCT in detecting CNV activity was 91% (95% confidence interval [CI], 84%-99%) and 13% (95% CI, 4%-22%).

CONCLUSIONS

Spectral-domain OCT alone can be relied upon for detecting CNV activity while monitoring eyes with neovascular AMD. However, FFA may still be of value in those with occult lesions that appear quiescent on SD OCT, as this type of lesion may show leakage on FFA.