Data Source and Study Population

The data compiled for this study used the Medical Expenditure Panel Survey (MEPS) database. The MEPS is federally funded by the Agency for Healthcare Research and Quality and the National Center for Health Statistics. MEPS is a nationally representative survey of the noninstitutionalized American civilian population, compiling national surveys of families and individuals, their medical providers, and employers across the United States. MEPS collects patient data, such as medical expenses, demographic characteristics, health conditions, and access to care, using a series of 5 questionnaires completed over a span of 2 years. Further information regarding MEPS designs and methods can be found herein.

This study is a retrospective analysis of 3 years of data from the MEPS database (2016-2018), which was used to assess health care system expenditures associated with patients with underlying ophthalmic conditions. To analyze the expenditures specifically related to medical diagnoses and corresponding demographic differences, the MEPS database household component and medical condition files were linked. Expenditures are recorded as part of the billing data shared by the providers to the MEPS survey and accounts for costs to the health care system including the insurance component of the costs. Patients with ophthalmic conditions with ≥1 outpatient visit were identified with complete International Classification of Diseases, 10th revision, Clinical Modification (ICD-10-CM) codes as part of this database. ICD-10-CM codes are recorded in this database primarily from billing data recorded by providers and supplemented by telephone surveys where patients can self-identify diagnoses. The medical condition file is updated with diagnoses and carries over for every subsequent year if a patient is diagnosed with any conditions (eg, depression diagnosed in 2013 will be stored in the 2016 condition file). MEPS clinical classification codes, which correlate with ICD-10-CM codes, considered as ophthalmic conditions included hordeolum and chalazion, other disorders of eyelid, disorders of lacrimal system, conjunctivitis, other disorders of cornea, retinal detachment and breaks, other retinal disorders, glaucoma, disorders of vitreous body, disorders of globe, other disorders of optic nerve and visual pathways, disorders of refraction and accommodation, visual disturbances, blindness and low vision, and other disorders of eye and adnexa. Patients with ophthalmic conditions without outpatient visits were excluded to better include patients receiving care for chronic ophthalmic conditions. Ophthalmic patients were further stratified by the presence of depression (ICD-10-CM code F32.x) as a medical diagnosis. This methodology was used to stratify patients with depression in previous studies based upon further communication with Agency for Healthcare Research and Quality personnel. ^, As this information is publicly available, a waiver of approval was obtained from the institutional review board. Data collection was in conformity with all federal and state laws.

Study Variables

Primary outcome measures used for this study were health care utilization and expenditures. All expenditures were adjusted to 2018 U.S. dollars using the urban Consumer Price Index of the U.S. Bureau of Labor Statistics. The study accounted for the complex survey design (sampling weights, clustering, and stratification) to more accurately estimate aggregate health care expenditures. For each patient, total health care system expenditures and number of visits both in aggregate and according to the health care service sectors, including inpatient, outpatient, emergency room, home health visits, prescription medications, and all payments by third-party payers and out-of-pocket costs, were recorded. Per capita expenditures and health care utilization were assessed for patients with ophthalmic conditions with and without the presence of depression and secondarily stratified based on the presence of comorbidities and certain sociodemographic characteristics.

Sociodemographic information included self-identified race/ethnicity, age, educational attainment, income, insurance status, and geographic location of the patient. Self-identified race/ethnicity were categorized as non-Hispanic white, Hispanic, Black, Native American, or Asian. Educational attainment was categorized as individuals with no degree obtained, high school diploma or equivalent, or a college degree. Income was calculated in relation to the 2018 federal poverty level (FPL). Individuals were categorized as poor (<100% of FPL), near poor (100%-124% of FPL), low income (125%-199% of FPL), middle income (200%-399% of FPL), or high income (>399% of FPL). Insurance status was categorized as private (includes patients with concurrent Medicare coverage), public (Medicare or Medicaid patients with no private insurance), or uninsured. For regional analysis, patients were categorized as living in the Northeast, South, Midwest, or West based on census regions.

Comorbidities were scored based on a modified comorbidity index based heavily on the Charlson Comorbidity Index. The Charlson Comorbidity Index is calculated as a sum of various comorbidities of an individual, such as congestive heart failure, myocardial infarction, and diabetes, because these comorbidities independently contribute to costs and health outcomes and defined by ICD-10-CM diagnosis codes. Additional comorbidities were included in this index based on the most common causes for inpatient admissions for these patients, such as pregnancy, hypertension, and pneumonia that is not included in the Charlson Comorbidity Index (Supplemental Table 1; Supplemental Material available at AJO.com). The total number of inpatient admissions along with pathologies that were directly associated with the most (defined by ICD-10-CM diagnosis codes) admissions (but not necessarily the primary reason for admission) were recorded for ophthalmic patients and for ophthalmic patients with and without depression. In addition, the frequency of depression for each aforementioned ophthalmic condition (as defined by the first 3 characters of the ICD-10 code) were recorded.

Statistical Analysis

Rao–Scott χ ² tests were used to compare sociodemographic characteristics and comorbidity indices of ophthalmic patients that had outpatient visits with and without depression. Incremental burden, otherwise known as the individual economic burden faced caused by patients on the health care system solely because of their diagnosis of depression and not confounded by other variables, such as age, race, income, and our modified Comorbidity Index, was calculated using a 2-part model consisting of logistic regression and gamma distribution with log-link function to address health care expenditure data. ^, Health care expenditure data contain a high concentration of zero expenditures and positive skewing of expenditures. Therefore, we used a 2-part model rather than log ordinary least squares models because of reduced normality and homoscedasticity assumptions. ^, Logistic regression modeling was used to estimate the health care expenditure ratio for ophthalmic patients with and without depression over the pooled period (2016-2018), adjusting for sociodemographic characteristics (age, sex, race/ethnicity, insurance coverage, education, and geographic region), modified comorbidity indices, and year of care received.

The expenditure ratio was then used to estimate the direct incremental economic burden of depression for ophthalmic patients in the United States, using population-level weights assigned to each patient by MEPS. Mean expenditures for these patients were extrapolated to the civilian noninstitutionalized U.S. population. Odds and expenditure ratios were generated to compare health care service sector (inpatient, outpatient, emergency room, home health, and prescription medications) utilization and expenditures respectively between ophthalmic patients with and without depression, adjusted for comorbidities and sociodemographic characteristics. The odds ratio compares utilization between the 2 groups. An adjusted odds ratio of 3 for utilization for inpatients with depression indicates 3 times the inpatient admissions for this cohort. The expenditure ratio is an odds ratio that is used specifically to compare expenditures.

P < .05 was considered statistically significant and all statistical analyses were performed using R software (R Foundation, Vienna, Austria).