Abstract
Purpose :
With an ever-increasing volume of electronic health record data, algorithm-driven data extraction must replace manual extraction for large scale analyses. Visual acuity and intraocular pressure (IOP) are important outcome measures commonly extracted from clinical records in vision research. The TOtal Visual acuity extraction Algorithm (TOVA) is presented and validated for automated extraction of best corrected visual acuity and IOP from clinical notes.
Methods :
Consecutive outpatient ophthalmology notes over a 10 day period from the University of Washington healthcare system in Seattle, WA were used for testing and validation of TOVA. The algorithm applied natural language processing (NLP) to recognize Snellen visual acuity targets in each line of free text and assigned laterality for each target. Visual acuity targets were divided into four discrete elements for analysis; numerator (e.g. the first number in 20/40), denominator (e.g. the second number in 20/40), adjustment sign (plus or minus for letters read or missed within a Snellen chart line), and adjustment letters (the number of such letters). The best corrected measurement was determined for each eye. Highest recorded IOP measurements for each eye were also extracted using NLP. The algorithm was validated against the full set of extracted notes.
Results :
A total of 200 clinical records were obtained, giving 1560 data points. Interrater reliabilities by kappa for Snellen numerator, denominator, adjustment sign, and adjustment letters were 0.96, 0.93, 0.87, and 0.84, respectively. Pearson correlation coefficient for IOP was 0.93 (95% CI: 0.92 to 0.95, p < 2.2x10^-16).
Conclusions :
TOVA is a validated tool for extraction of visual acuity and IOP data from free text clinical notes and provides an open source method of accurate, efficient data extraction.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.