Adherence Trends from Objective Dispensing Data: Evaluation Results from a Telemedicine Intervention

Introduction: Individuals with Serious Mental Illnesses (SMI) have a substantially lower life expectancy due to cardiovascular disease, hypertension, diabetes, and suicide. Pharmacological treatments exist but have limited effectiveness because of low adherence to medications. Good adherence is 80% or more of medication taken as prescribed, but the average patient with SMI takes 50-70% of prescribed somatic and psychiatric medications. Currently intrusive interventions like residential treatment and direct observation of medication consumption are used to improve adherence. Telehealth offers an opportunity to support adherence while increasing consumer independence. Medherent is a telehealth adherence platform being used to dispense medication and monitor consumer adherence in community settings.

Methods: Medherent devices record if an individual dispenses their medication within a two-hour window set by the consumer and their prescriber. Medherent sends data to the pharmacy server that includes dose and dispense time. We downloaded server data for all consumers using the device between 1/1/2019 and 6/20/2021. We constructed indicators for weekends and number of months using the Medherent device. Individuals were coded as missing a dose if they did not dispense a dose within the two-hour window. Using a random intercept model nested in consumers and robust standard errors to adjust for heteroskedasticity, the dispensing data was analyzed to understand medication dispensing averages and individual variability. A random coefficient model was estimated including the weekend indicator and time on the device.

Results: The data did not include any demographic information about consumers but all individuals using the device were clients of community mental health agencies and most consumers lived in residential treatment facilities. 121 individuals had at least two weeks of device use. There were 82,860 scheduled doses and 73,125 (88%) dispensed doses. Using a random intercept model, we found that 14% of the variance was between individuals. Individual adherence ranged from 36% to 99% with 79% of participants having adherence above the recommended 80%. We included indicators for weekend to understand the effect of lower agency support staff on weekends and length of time on the device to measure the effect of device novelty on adherence. There was no evidence that time of the device influenced adherence. On average individuals had a 1.7% lower adherence on the weekend but there was statistically significant ( p < 0.001) variability between participants with the top 95% having a 1.7% higher weekend adherence.

Conclusion: The evidence from this study suggest that Medherent has potential to help individuals with serious mental illness adhere to complicated medication regimens in independent living settings. In this sample 79% of individuals had high rates of adherence and the 21% with lower adherence can be identified for targeted interventions to address those adherence issues. On weekends, when support staffing levels are lower, adherence decreased on average less than 2% suggesting that individuals will be able to maintain adequate adherence in more independent settings. This evidence suggests that telehealth approaches can help mental health systems support recovery and consumer independence while maintain high levels of medication adherence.