Purpose
To determine the incidence of central serous chorioretinopathy (CSC) stratified by age, sex, and diagnosis with obstructive sleep apnea (OSA), and to determine whether some patients with newly diagnosed CSC may be candidates for OSA evaluation.
Design
Retrospective cohort study.
Methods
We used the IBM MarketScan database to select 59,016,145 commercially insured patients in the United States between 2007 and 2016. We identified patients’ first diagnosis with CSC, and defined patients as having OSA if they had a diagnosis following a sleep study. We specified Cox proportional hazard models with interactions between age, sex, and OSA status to determine patients’ risk of developing CSC. We estimated the positive predictive value (PPV) that a new diagnosis of CSC would have in predicting a subsequent diagnosis of OSA.
Results
Risk of CSC increased with age in years (hazard ratio [HR] = 1.030, P < .001) and OSA diagnosis (HR = 1.081, P < .033), and was lower in women (HR = 0.284, P < .001). We estimated the annual incidence of CSC was 9.6 and 23.4 per 100,000 women and men, respectively. Incidence was higher in women and men with OSA (17.2 and 40.8 per 100,000). The PPV of CSC diagnosis as a predictor of OSA was highest in the fifth decade of life.
Conclusion
The incidence of CSC in our patient sample is higher than previously reported. Risk of CSC is higher in men than in women, and OSA increases risk of CSC in both men and women. Some patients, particularly older male patients, may be good candidates for OSA evaluation following a CSC diagnosis.
Central serous chorioretinopathy (CSC) is a disorder characterized by the formation of a localized neurosensory retinal detachment caused by leakage of fluid at the level of the retinal pigment epithelium (RPE). It is the fourth most common nonsurgical retinopathy after age-related macular degeneration, diabetic retinopathy, and branch retinal vein occlusion. It is more common in men than in women and primarily affects patients between the ages of 20 and 50. , The true incidence of CSC is unknown. The only population-based study in the United States reported an annual incidence of 9.9 cases per 100,000 men and 1.7 per 100,000 women in Olmstead County, Minnesota, USA; a population-based study in Taiwan estimated an annual incidence of 27 cases per 100,000 men and 15 cases per 100,000 women. ,
The pathophysiology of CSC remains poorly understood despite advances in imaging techniques and numerous studies of the disease. Many risk factors have been described in the literature, including psychological stress, , type A personality, corticosteroid medication use, elevation of endogenous steroids owing to endocrine disorders such as Cushing syndrome or steroid-producing tumors, pregnancy, hypertension, and psychopharmacologic medication use.
Obstructive sleep apnea (OSA) is characterized by recurrent episodes of partial or complete upper airway obstruction during sleep, leading to repetitive oxygen desaturation and/or cortical arousals with consequent autonomic nervous system activation. OSA has been identified as a possible risk factor for the development of CSC, , but the extent of this association is not fully understood. The intermittent episodes of asphyxia are believed to trigger increased sympathetic activity and the episodes of sudden arousal are physical stressors that may lead to elevated catecholamine levels. , This increased sympathetic drive is also present in CSC and can cause vascular endothelial dysfunction. , It has also been hypothesized that the elevated cortisol levels may increase the likelihood of CSC development. Additionally, in both diseases, there is increased oxidative stress, platelet aggregation, and hypercoagulability. Previous studies have noted co-occurrence of OSA in patients with CSC, but the magnitude of the association has been variable, , , , and the extent to which the association is driven by other comorbidities, such as obesity, is unclear. ,
We used administrative claims data to estimate the incidence of CSC in a national population of patients with employer-provided health insurance, the absolute risk of CSC conferred by an OSA diagnosis, and the feasibility of using a CSC diagnosis in an ophthalmologist’s office to identify undiagnosed cases of OSA. This is the first estimate of the incidence of CSC in the United States drawn from a national sample, stratified by sex, age, and OSA diagnosis.
Methods
Data Source and Sample Selection
We used administrative insurance claims data from the IBM MarketScan database, one of the largest healthcare databases available for patients with employer-provided health insurance, including data from over 245 million unique patients since 1995, and data on patients in all 50 states. Our version of the database extends from 2007 to 2016. This database has been used in prior ophthalmic studies and in prior estimates of national disease incidence. MarketScan data are longitudinal, extending from the time patients first receive insurance coverage until they discontinue insurance, and include demographic information, diagnosis, procedural, and billing codes for patient encounters with health care providers. The institutional review board at Stanford University determined this secondary analysis of deidentified administrative data to be exempt from review.
We restricted this sample to include all patients in MarketScan with continuous insurance coverage, with at least 365 days of enrollment to avoid including patients with unobserved care. In sensitivity analyses, we limited the sample to patients with longer enrollment periods. In order to ensure that the diagnoses observed in our analysis were new diagnoses, rather than old diagnoses that were recoded after patients changed insurers, we used a “lookback” period of 90 days, excluding patients whose first diagnosis of CSC or OSA occurred within their first 90 days of enrollment in the database.
Variable Definitions
We categorized patients as having OSA if they received a diagnosis of OSA following a sleep study. OSA diagnosis was defined using International Classification of Diseases, 9th and 10th editions (ICD-9 and ICD-10) diagnosis codes ( Supplemental Table 1 ; Supplemental Material available at AJO.com ). As a sensitivity analysis, we also created a more restrictive (but less inclusive) definition of sleep apnea, in which we defined patients as having OSA only if they received a diagnosis following a sleep study and then also had a record of receiving treatment with a continuous positive airway pressure (CPAP) device. Receipt of sleep studies and CPAP devices was identified using Current Procedural Terminology (CPT) codes ( Supplemental Table 1 ). Diagnosis with CSC was identified as the first date in a patient’s record where an ICD-9 or ICD-10 code for CSC was listed ( Supplemental Table 1 ).
Statistical Analysis: Estimating Overall Incidence
To estimate the overall incidence of CSC in this sample, without adjusting for patient age, sex, or OSA, we constructed basic life tables by calculating the number of new diagnoses of CSC in each year, and dividing that by the total “time at risk” for all patients in that year of the dataset. We calculated “time at risk” in each year of the dataset according to the number of “person-years” in that year. “Time at risk” excludes patients who have already been diagnosed with CSC as they are no longer at risk for the disease. For example, if a patient began their insurance enrollment halfway through 2008, was diagnosed with CSC at the end of 2009, and then discontinued enrollment at the end of 2010, they would contribute 0.5 person-years to total time at risk in 2008, 1.0 person-years in 2009, and 0 person-years in 2010. These life tables allowed us to calculate overall incidence of CSC in the entire sample and compare those results with other studies that have estimated absolute risk of CSC in select populations.
Statistical Analysis: Estimating Risk of Central Serous Chorioretinopathy
To estimate the absolute risk of CSC for patients depending on their age, sex, and OSA diagnosis, we conducted a time-to-event analysis beginning from a patient’s initial enrollment in the database, and extending until the date of their first diagnosis with CSC or, if they were never diagnosed with CSC, extending until they discontinued enrollment. We specified a Cox proportional hazards model, with patients’ age at enrollment, sex, and OSA diagnosis as covariates. As we were interested in estimating the effect of OSA on subsequent risk of developing CSC, patients were only coded as having OSA if their first diagnosis with OSA preceded their first diagnosis with CSC. To allow for a nonlinear association between age and risk of CSC, we included a quadratic term for age. Additionally, to account for the possibility that age, sex, and OSA diagnosis may each affect risk of CSC differently depending on the values of the other covariates, we included interaction terms for all variables; for example, it is possible that age has a different effect on CSC risk for men and for women, and that OSA affects risk of CSC in young men differently than it affects risk for older women. We used the estimates from this model to calculate the annual risk of CSC for each patient.
We performed several sensitivity analyses to determine whether the risks estimated in this model changed substantially when OSA was coded differently, or when minimum follow-up time for patients in the model was adjusted. First, we re-estimated the model using the more restrictive definition of OSA, in which patients were only coded as having OSA if they completed a sleep study and also received a CPAP device. Second, we re-estimated the model restricting the sample to patients with several different minimum follow-up times, beginning with patients who had at least 1 year of enrollment in the database, and then increasing that requirement in 1-year increments, finishing with a model that only includes patients who had 10 years of enrollment in the database. Increasing the minimum enrollment for the analysis reduced the sample size in terms of the number of patients the model used to estimate risk, but ensured that our calculations of risk were drawn from a sample of patients that were observed for a longer period, giving us more confidence that diagnoses of CSC really did represent patients’ first diagnosis, rather than a secondary diagnosis that was preceded by an unobserved index diagnosis occurring under a different insurance plan. Varying the follow-up time also allowed us to determine whether our estimated incidence of CSC was associated with how long patients were observed in the dataset. For example, patients with shorter periods of enrollment in the dataset may represent patients with poor insurance continuity, who may not receive appropriate preventive care, and may therefore have a higher burden of disease. Alternatively, it is possible that these patients may be less likely to be diagnosed with a disease even if it is present, owing to less stable access to health services.
Statistical Analysis: Estimating the Value of Central Serous Chorioretinopathy as a Screen for Obstructive Sleep Apnea
After estimating the overall incidence of CSC in our cohort and estimating the risk of CSC for individual patients depending on their age, sex, and OSA diagnosis, we reconsidered the potential relationship between CSC and OSA from the perspective of an ophthalmologist seeing a patient in clinic with newly diagnosed CSC. If OSA increases patients’ risk of developing CSC, then it is possible that a patient who presents to an ophthalmology clinic with new-onset CSC may have undiagnosed OSA. We analyzed whether CSC diagnosis might be used to screen for OSA in ophthalmology clinics, and whether ophthalmologists should consider referring patients with CSC for evaluation of suspected OSA. To assess whether ophthalmologists should consider referral in this population, we calculated the positive predictive value (PPV) of a CSC diagnosis in predicting a subsequent diagnosis of OSA. Our models of CSC risk relied on OSA diagnoses that precede CSC diagnosis, but in this analysis we considered the possibility that patients with CSC might have latent or undiagnosed OSA, and we determined how many patients with newly diagnosed CSC would go on to develop their first diagnosis of OSA following their first CSC diagnosis. We stratified our estimates of PPV by patient age and sex, to determine whether CSC might be more useful for screening some populations than others.
Results
We studied a population of 59,016,145 commercially insured patients in the United States between 2007 and 2016. Women comprised 54% of the sample (n = 32,018,719), and the mean age at the time of enrollment in the database was 33.1 (standard deviation 18.6) years. A total of 39,254 patients were diagnosed with CSC. A total of 1,539,006 patients were diagnosed with OSA, using the more inclusive definition of OSA (an ICD-9/10 diagnosis following a CPT code for a sleep study), and 1,000,130 patients were diagnosed with OSA under the most restrictive definition (ICD-9/10 diagnosis following a sleep study and a CPT code for a CPAP device).
Incidence of Central Serous Chorioretinopathy
The overall incidence of CSC in this analysis was 18.3 per 100,000 person-years. The estimated incidence was quite stable between 2007 and 2016, ranging from a minimum of 16.9 per 100,000 person-years in 2007 to a maximum of 20.69 per 100,000 person-years in 2016 ( Table , Figure 1 ).
| Year | Patients | Mean Days at Risk | Person-Years at Risk | New Diagnoses of CSC | Rate of CSC per 100,000 Person-Years |
|---|---|---|---|---|---|
| 2007 | 16,763,128 | 340.1 | 15,619,882.7 | 2,636 | 16.88 |
| 2008 | 24,293,723 | 333.6 | 22,201,367.9 | 3,952 | 17.80 |
| 2009 | 26,537,499 | 333.2 | 24,221,873.2 | 4,352 | 17.97 |
| 2010 | 27,054,825 | 332.0 | 24,611,729.8 | 4,224 | 17.16 |
| 2011 | 28,829,932 | 329.9 | 26,053,593.8 | 4,656 | 17.87 |
| 2012 | 29,139,349 | 332.5 | 26,545,515.8 | 5,001 | 18.84 |
| 2013 | 24,736,057 | 332.4 | 22,524,883.9 | 4,054 | 18.00 |
| 2014 | 23,153,082 | 332.5 | 21,090,542.0 | 4,017 | 19.05 |
| 2015 | 18,325,421 | 332.4 | 16,687,665.6 | 3,150 | 18.88 |
| 2016 | 16,404,166 | 345.4 | 15,524,336.4 | 3,212 | 20.69 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
