Abstract
Purpose :
Meeting quality measures (QM) is an important for ophthalmic care and reimbursement. QMs use data recorded in the electronic health record (EHR); ideally they are calculated using automated data queries. The quality of EHR data can limit this automatic calculation and require manual chart review for full measure determination. This study examines the ability of EHRs to adequately capture necessary clinical data to calculate QMs.
Methods :
Thirty ophthalmic QMs developed by the American Academy of Ophthalmology were assessed for required data elements by reviewing published specifications. The location of each data element was determined in our institution’s EHR (Epic, Verona, WI). Challenges for each of the data elements were noted by manually reviewing samples of each data element.
Results :
We found 15 categories of data elements in the ophthalmic exam, imaging, visit/procedure, and patient data. Most/all of the QMs used patient age, and procedure & diagnosis codes that were easily extracted. Nevertheless, lack of specificity in diagnosis code data and incomplete problem lists made patient identification for some QMs difficult. Exam data was frequently used; most were stored in free-text fields and needed text processing (e.g. extracting and converting Snellen visual acuity fields to LogMar scale) but non-standard entries were impossible to automatically process (i.e. text instead of numeric data). Two elements were only in notes and required natural language processing (NLP): lid position and marginal reflex distance. Similarly, data from OCT and visual field imaging were in free-text fields and notes; these, too, required additional processing. Finally, patient medications were easily extracted from lists, which were often not complete and required NLP or manual review of notes. In summary, for the study EHR, most measures required text processing or NLP and over half of the measures would require manual review if medication or problem lists were not complete.
Conclusions :
The study EHR was unable to capture and report most data elements important for ophthalmic clinical care, quality measure calculation, and research without customization. Generalizability of these results to other EHRs merits studying. Ultimately, tools for automatically extracting this data, as well as improvements and standardization in EHR design and use are required to address this problem.
This is a 2021 ARVO Annual Meeting abstract.