Precisely capturing eye drop use behavior

Abstract

Purpose : Because poor medication adherence impacts at least half of glaucoma patients, there is a need to monitor glaucoma eye drop medication use precisely and communicate that information in real time to the eye care team.

Methods : We developed a prototype of an electronic eye drop use monitor that assesses: 1) when the eye drop cap is removed (reed switches); 2) when the medication is inverted (inertial measurement unit); 3) when the bottle has been squeezed (capacitance sensor). The capacitance sensor also measures fluid level. The monitor uses Bluetooth Low Energy to transmit the use event and fluid level to a server. To compare the accuracy of the use detection to video-recorded use behavior, ten participants ≥age 65 were enrolled in a pilot study. Each participant: 1) instilled eye drops standing up; 2) instilled eye drops laying down; 3) pretended to instill eye drops by removing the cap, inverting it and not squeezing it; 4) took the cap off without using the drops; 5) walked with the bottle and 6) shook the bottle. Each task was repeated 5 times. A rule-based algorithm was developed to identify a use event by checking for: 1) cap off 2) orientation > 100 degrees of tilt on the z-axis 3) positive slope to capacitance. Sensitivity of capacitance assessment of fluid level was assessed for a 15 mL bottle of lubricant eye drops. The bottle mass was recorded between each step for verification of the remaining fluid level. First, capacitance readings were taken as fluid was removed in 1mL increments until empty. Second, the bottle was filled to 8 mL and 2 ml of fluid was removed with the micropipette in 0.2 mL increments. Each test was repeated five times.

Results : The electronic monitor had a sensitivity of 92% and specificity of 88.5%. The lower specificity came from the algorithm’s inability to correctly classify the “non-use” event in 23/50 trials when the participants pretended to instill an eye drop. When these 23 trials where participants pretended to instill the drops were removed, the algorithm had a specificity of 100%. Fluid level was reliably assessed at the 0.4mL resolution.

Conclusions : This electronic eye drop adherence monitor can assess use behavior in almost all usual use cases and transmit the information to the eye care team. Manufacturing such a monitor at scale could be a low-cost way of accurately assessing glaucoma medication use in order to intervene and improve adherence.

This is a 2020 ARVO Annual Meeting abstract.