Abstract
Purpose :
We have previously shown that glaucoma (GL) patients are more likely to be frail than glaucoma suspects (GS) using an electronic frailty index (eFI) which can be automatically estimated at the point of care, however, eFI generation requires ≥2 years of electronic health record (EHR) outpatient data. New technology can combine responses from a fall questionnaire with inertial sensor readings from a timed up and go test to generate a statistical estimate of frailty (FE). The purpose of this clinical pilot study was 1) to compare the proportion of GL/GS patients in whom a questionnaire/sensor-based FE vs. eFI could be estimated in clinic and 2) to determine the agreement of these two frailty scores.
Methods :
Cross-sectional pilot of 68 GL/GS patients age ≥65 years. Participants were invited to complete a questionnaire and timed up and go test while wearing inertial sensors and these data were used to generate the statistical FE based on the Fried Frailty Index. An automated eFI was calculated if adequate data was available in the electronic health record. Kappa statistic was used to estimate the agreement of the frailty categories.
Results :
Mean age was 73.9±5.5 years, with 36% black or mixed race, 46% female, and 90% GL vs 10% GS. All patients were able to complete the sensor-based FE but only 54/68 patients had a calculable eFI. The 21% without a calculable eFI were more likely to be black or mixed race (p=0.038). The kappa score was -0.53 (p=0.99) indicating poor agreement of the two frailty scores (Table 1).
Conclusions :
The sensor-based FE provided additional frailty scores in the 21% of patients without an eFI score. However, since a higher proportion were categorized as pre-frail/transitional by the sensor-based FE compared to eFI which categorized a higher proportion as frail, it is possible the FE might miss cases of frailty. This is concerning since non-white patients were also more likely to have inadequate historic EHR data to generate an eFI. Future studies should address whether other data captured at the point of care can be used to provide more accurate frailty estimates in those without adequate historic EHR data.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.