Abstract
Purpose :
Group-Based Trajectory Modeling (GBTM) is a statistical approach that clusters persons with similar behavioral and developmental trajectories. When applied to administrative claims data which document the frequency of ocular hypotensive prescription refills, GBTM has revealed distinct patterns of medication adherence. In this study, we applied GBTM to electronically monitored data which document daily eye drop use in order to identify distinct patterns of medication adherence in glaucoma patients.
Methods :
We performed GBTM with ancillary adherence data from 100 participants enrolled in an NIH-funded glaucoma progression study at the University of Alabama at Birmingham. Participants were included if they were above age 18, used hypotensive eye drops, had 2 or more reliable visual field tests, and had visual acuity better than 20/40 at baseline. Adherence was monitored for 6 months using Medication Event Monitoring Systems (Aardex, Switzerland) which record the date and time at which the devices are opened to use the eye drops stored inside. Using the Proc Traj macro (SAS Institute, Cary NC), quadratic functions [RL(1] [sp2] were fitted to weekly mean adherence data to estimate trajectory models with 2, 3, 4, 5, and 6 groups. The trajectory model that included 5% or more of participants in each group and had the highest Bayesian Information Criterion (BIC) was identified as optimal.
Results :
Mean age was 68.6 ± 8.3 years, and mean baseline intraocular pressure was 25 ± 4.8 mmHg (right eye) and 23.9 ± 6.5 mmHg (left eye). Mean adherence was 80 ± .21% and the BIC for the final model was -437.5. Five trajectory groups were estimated: Near-perfect adherence (30.7% of participants), Good adherence (32.7%), Initially Good but Declining adherence (6% of participants), Consistently Poor adherence (18.7% of participants), and Poor and Declining adherence (11.9% of participants).
Conclusions :
We identified “Near-perfect” adherence and “Initially Good but Declining” adherence as two patterns that had not been previously captured in administrative data. Utilizing different metrics to quantify adherence may provide additional insight into the dynamic nature of medication adherence, which may help to predict adherence to prescribed hypotensive therapy.
This is a 2021 ARVO Annual Meeting abstract.