Shifts in ophthalmic care utilization during the COVID-19 pandemic in the US

Abstract

Background: Healthcare restrictions during the COVID-19 pandemic, particularly in ophthalmology, led to a differential underutilization of care. An analytic approach is needed to characterize pandemic health services usage across many conditions.

Methods: A common analytical framework identified pandemic care utilization patterns across 261 ophthalmic diagnoses. Using a United States eye care registry, predictions of utilization expected without the pandemic were established for each diagnosis via models trained on pre-pandemic data. Pandemic effects on utilization were estimated by calculating deviations between observed and expected patient volumes from January 2020 to December 2021, with two sub-periods of focus: the hiatus (March-May 2020) and post-hiatus (June 2020-December 2021). Deviation patterns were analyzed using cluster analyses, data visualizations, and hypothesis testing.

Results: Records from 44.62 million patients and 2455 practices show lasting reductions in ophthalmic care utilization, including visits for leading causes of visual impairment (age-related macular degeneration, diabetic retinopathy, cataract, glaucoma). Mean deviations among all diagnoses are 67% below expectation during the hiatus peak, and 13% post-hiatus. Less severe conditions experience greater utilization reductions, with heterogeneities across diagnosis categories and pandemic phases. Intense post-hiatus reductions occur among non-vision-threatening conditions or asymptomatic precursors of vision-threatening diseases. Many conditions with above-average post-hiatus utilization pose a risk for irreversible morbidity, such as emergent pediatric, retinal, or uveitic diseases.

Conclusions: We derive high-resolution insights on pandemic care utilization in the US from high-dimensional data using an analytical framework that can be applied to study healthcare disruptions in other settings and inform efforts to pinpoint unmet clinical needs.

Fig. 1. A common analytical framework for the high-dimensional characterization of care utilization patterns across many conditions.

Full specifications of candidate counterfactual models (Step 2a) are available in Supplementary Note 1. Abbreviations: IRIS Registry American Academy of Ophthalmology Intelligent Research in Sight Registry®, RMSPE root-mean-squared percentage error, MSE mean squared error.

Fig. 2. Deviations from expectation for common non-severe eye conditions.

Time series of observed (purple line) and predicted (teal line for mean predictions; teal shading for 95% prediction intervals) monthly numbers of patients documented with four eye conditions that represent the least severe forms of the leading causes of low vision and blindness: a early-stage dry age-related macular degeneration (dry AMD, early-stage), (b) non-proliferative diabetic retinopathy without diabetic macular edema (NDPR w/o DME), (c) age-related cataract, and (d) suspect glaucoma, from January 2017 to December 2021. The black vertical line at January 2020 denotes the start of the pandemic study period, the period for which monthly deviations from expectation are computed (e). A diverging color scale in the heatmap (e) is used to illustrate the direction, magnitude, and statistical significance of each monthly deviation, with the darkness of each cell a function of the product between the magnitude of the deviation and the negative log of its adjusted p-value. Abbreviations: AMD age-related macular degeneration, NPDR non-proliferative diabetic retinopathy, DME diabetic macular edema, w/o without.

Fig. 3. Hiatus vs. post-hiatus deviations across all diagnosis entities.

Deviations during the nadir of the hiatus period (April 2020) are plotted against deviations in the post-hiatus period (June 2020– December 2021), for all 261 diagnosis entities individually (red points in (a)) and averaged across the diagnosis entities that belong to each diagnosis category (pink circles in (b)). In (a), the 95% normal data ellipse (red oval) represents an estimated probability contour that is expected to contain 95% of all plotted diagnosis entities, and a line of equality (dashed gray line) represents no change in deviations over time (i.e., deviations during April 2020 are equal to post-hiatus deviations). In (b), the size of each point corresponds to the cumulative number of average monthly patients for the diagnosis entities within a category. The blue diamond represents the average of all deviations for April 2020 (−0.67, standard deviation (SD): 0.14) and the post-hiatus period (−0.13, SD: 0.09) across all 261 diagnosis entities. Abbreviations: OGI/IOFB ocular globe injury/intraocular foreign bodies, Postop complications postprocedural or postoperative eye complication, Oculoplastics and orbital conditions oculofacial plastics and orbital conditions, SD standard deviation.

Fig. 4. Hiatus vs. post-hiatus deviations for selected ophthalmic conditions.

Deviations during the nadir of the hiatus period (April 2020) are plotted against deviations in the post-hiatus period (June 2020–December 2021), for diagnosis entities that correspond to common eye diseases (a), and those that belong to the neuro-ophthalmic disease category (b). The size of each point corresponds to the average number of monthly patients for the diagnosis entity. Points were selectively labeled based on relevance (e.g., we excluded labels of diagnosis entities with names that contain key words such as other or unspecified). A Pearson’s product moment correlation coefficient (r) is reported for the distribution of points in each plot (blue text; upper left corner), and a line of equality (dashed gray line) represents no change in deviations over time. Abbreviations: DR diabetic retinopathy, AMD age-related macular degeneration, CNV choroidal neovascularization, PDR proliferative diabetic retinopathy, NPDR non-proliferative diabetic retinopathy, DME diabetic macular edema, PACG primary angle-closure glaucoma, OAG open-angle glaucoma, w with, w/o without.

Fig. 5. Deviations by severity level for ophthalmic emergencies and common conditions.

Boxplots depicting distributions of deviations from expectation, during the hiatus (orange boxplots) and post-hiatus (blue boxplots) periods, stratified by (a) increasing levels of severity (derived from the aggregate BaSe SCOrE’s compiled by Bourges et al.) for a set of n = 36 diagnosis entities considered as common ocular emergencies, and (b) vision-threatening (VT) vs. non-vision-threatening (NVT) status for age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma diagnoses (Table S4) (n = 28 diagnosis entities in total). Outliers, indicated as black dots, are data points that are located at a distance greater than 1.5 times the interquartile range from either the lower quartile or the upper quartile of the boxplot. The ‘whiskers’ of the boxplots, which extend from the boxes as black vertical lines, represent the range of values that lie within 1.5 times the interquartile range from the lower and upper quartiles. To test for statistically significant differences in the central tendencies between distributions of deviations, we used Kruskal-Wallis (a) and Mann-Whitney U (b) tests to compute p-values (gray text). Abbreviations: BaSe SCOrE BAsic SEverity Score for Common OculaR Emergencies, DR diabetic retinopathy, AMD age-related macular degeneration, VT vision-threatening, NVT non-vision-threatening.

Fig. 6. High-dimensional characterization of pandemic care utilization patterns.

Patterns of pandemic care utilization over time are illustrated for all 261 diagnosis entities, partitioned by 13 diagnosis categories. Beginning with the innermost (first) ring, barplots represent the number of months, starting from January 2020, it took for each diagnosis entity to achieve partial or sustained recovery of patient volumes. The absence of a barplot denotes no recovery. The second ring depicts the raw values of the deviations for each condition during April 2020 (sequential color scale). In the third ring, a cluster heatmap shows diagnoses’ aggregate deviations for each quarter in the post-hiatus period, from the third quarter (Q3) of 2020 (innermost track of the heatmap) to the fourth quarter (Q4) of 2021 (outermost track), with the darkness of each heatmap cell a function of the product between the magnitude of the estimated quarterly deviation and the negative log of its adjusted p-value (diverging color scale). Diagnosis entities are clustered to group together conditions that exhibited similar patterns of post-hiatus deviations. In the outermost (fourth) ring, the magnitudes of each diagnosis entity’s counterfactual model performance error (RMSPE) are visualized as barplots. Abbreviations: Blind blindness and vision defects, Cataract cataract and other lens disorders, Cornea/External cornea and external disease, Neuro neuro-ophthalmology, Oculoplastics oculofacial plastics and orbital conditions, OGI ocular globe injury/intraocular foreign bodies, Other other specified eye disorders, PO postprocedural or postoperative eye complication, Refra refractive error, Retina/Vitreous retina and vitreous conditions, Strab strabismus, Uveitis uveitis and ocular inflammation, RMSPE root mean squared percentage error, Q1 first quarter, Q2 second quarter, Q3 third quarter, Q4 fourth quarter.