Abstract
Purpose :
Diabetic retinopathy (DR) is a leading cause of blindness in adults worldwide. Secondary use of electronic health records (EHR) has been used extensively to retrospectively study prevalence, risk factors, and effects of treatment for DR. We aimed to analyze variations in EHR-based definitions of DR cohorts in the literature.
Methods :
Using a keyword based search (diabetic retinopathy, EHR, ophthalmology, ICD-9/10 codes) of academic databases (Google Scholar, PubMed, etc), we extracted data regarding the following variables: study design and purpose, number of patients, dataset source (EHR, claims database, etc), geographic scope (single center or multicenter), DR cohort (new or known DR), inclusion or outcome of DR, and data terminology (ICD-9, ICD-10, etc); as described in Table 1. To explore variations in defining DR, we generated descriptive statistics for all variables and additionally performed Chi-squared testing regarding terminologies used for DR cohort definitions.
Results :
Thirty-nine papers between 2008 and 2022 were identified. The study design of a majority of the papers were cohort (64.1%), with 11 (28.2%) cross sectional and 3 (7.7%) case control studies. EHR (61.5%) and claims databases (33.3%) were the most common dataset sources, with 2 (5.1%) studies using both. 35 (89.7%) and 4 (10.3%) of the studies were multicenter and single center, respectively. For DR cohorts, 13 (33.3%) of 39 studies consisted of new DR, 25 (64.1%) with known DR, 1 (2.6%) included both. 20 (51.3%) of 39 studies used ICD-9 codes only, ICD-10 codes only, or both ICD-9 and ICD-10 codes. It is also worthwhile to note the variations in ICD-9/ICD-10 codes used to define DR cohorts, which include a range of diagnoses related to diabetic ophthalmic complications, retinal disorders, severity of DR, etc. The remaining 19 (48.7%) studies utilized a combination of ICD-9, ICD-10, and/or other coding systems such as Current Procedural Terminology codes (CPT), Healthcare Common Procedure Coding System (HCPCS), etc. There was a statistically significant difference in terminologies used in these studies (p < 0.001).
Conclusions :
Significant variation exists in the use of EHR data to define DR cohorts. This work highlights the need for standards in defining cohorts in diseases such as DR.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.