Can Financial Incentives Improve Medication Adherence and Patient Outcomes After Involuntary Hospitalization?
Can Financial Incentives Improve Medication Adherence and Patient Outcomes After Involuntary Hospitalization?
This study is a randomized controlled trial evaluating the impact of financial incentives on medication adherence among individuals with schizophrenia, schizoaffective disorder, or bipolar disorder and/or co-occurring substance use disorder who are recently discharged from involuntary hospitalization or are at high risk of future involuntary hospitalization. Participants will be randomized to receive financial incentives for adherence to long-acting injectable medications or to a control group.
The investigators seek to understand the effectiveness of financial incentives at reducing medication non-adherence and costly involuntary hospitalizations for individuals with acute mental health needs. The focus will be on patients diagnosed with schizophrenia, schizoaffective disorder, and/or bipolar disorder who have recent experiences with involuntary hospitalization and possibly a co-occurring substance use disorder. This population has been documented to have particularly high healthcare expenditures and mortality, as well as incidence of criminal charges. Partnering with UPMC Western Psychiatric Hospital (WPH), Pittsburgh Mercy, and the Allegheny County Department of Human Services (ACDHS), the investigators will run a randomized controlled trial (RCT) where the treatment group is offered financial incentives for long-acting injectable (LAI) versions of the antipsychotic and substance use disorder medications they need. If effective, financial incentives for LAIs can be a scalable tool for addressing a range of costly problems facing the whole country.
Investigators will enroll eligible individuals at discharge from involuntary hospitalization at WPH, in WPH outpatient clinics, and in the field with Pittsburgh Mercy's community treatment teams. Eligible individuals will be those who have a diagnosis of schizophrenia, schizoaffective disorder, and/or bipolar disorder, who were involuntary hospitalized within the last year or who are deemed by ACDHS models to be at high risk of involuntary hospitalization, and whose prescribed medication has an LAI formulation. Investigators will hire and train a research team to enroll patients in the study, randomize patients into treatment and control groups, and work with their physicians to keep track of who is taking up LAIs and who needs to receive the incentives. Each participant will experience a study duration of 12 months.
The primary hypotheses are that financial incentives for LAIs can increase medication adherence in this particularly hard-to-treat population and (2) increased medication adherence will decrease the incidence of involuntary hospitalization. Additional hypotheses, based on ongoing research in Allegheny County (Emanuel, Welle, and Bolotnyy, 2025), are that criminal charges and mortality will also decline if medication adherence improves and there will be improvements in housing stability, employment, subjective well being, relationships with others, and attitudes towards long-acting injections and the healthcare system.
Outcomes will be measured using ACDHS administrative data, including Medicaid claims and prescription refill records, death and Medical Examiner records, county court records, utilization data for homeless shelters in Allegheny County, and employment and income data from Pennsylvania's Department of Labor and Industry. Additionally, study participants will be surveyed every month during the study period and 6 months and 12 months after the end of the study. Surveys will collect information on participants' subjective wellbeing, the quality of their relationships, as well as their attitudes towards long-acting injections and the healthcare system. Outcomes will be measured and effects assessed at the end of the study period for all participants and after the follow-up survey data are collected. Priebe et al. (2013) and Noordraven et al. (2017) work with patients with a baseline adherence rate of 70-80%; the comparable adherence rate in our setting is 52%.
Randomization Plan
A trained research team will work with physician partners to approach an eligible patient, explain the study to them, confirm their eligibility, and enroll them in the study. Everyone enrolling will receive a debit card preloaded with $25. Right after enrollment, the research team will utilize a random number generator to determine whether the individual is in the treatment or the control group. The study participant will then be informed of their group status. Like those in the treatment group, those in the control group will be eligible to receive $10 for each monthly survey they complete, with the money being deposited to their debit card.
If randomly selected into the treatment group, individuals will receive a base reward of $2 for each day of LAI duration to treat schizophrenia, schizoaffective disorder, bipolar disorder, or substance use disorder. Structuring the incentives this way avoids nudging physicians into prescribing LAIs of a particular duration; the investigators want them to pick the duration that they think is right for the patient, incentives aside. Individuals will also be rewarded with $0.25 for each consecutive day on medication. For example, if they receive two consecutive 4-week antipsychotic LAIs in a row, they will receive $2*28 days + $0.25*28 days = $63 the first time and $2*28 + $0.25*56 days = $70 the second time. Doing so allows for streaks to be incentivized, with the understanding that having people be on medication for more consecutive days has significant benefits and can lead to habit formation that might be sustainable even after the study is over.
Missing a scheduled LAI will result in no reward and the next LAI being rewarded again at the base payment of $2 per day, bringing the bonus reward streak down to 0 days. At the end of the study, treated individuals will be able to receive a $25 reward for signing a pledge to continue with their treatment plan even after the rewards in this study are no longer offered. This is intended to prevent some of the reversion to non-adherence that previous studies of incentives for LAIs have documented.
The maximum amount that an individual receiving an LAI every 4 weeks (the most common duration) can receive over the course of the 12-month study is $1,218 for injections + $25 for enrollment + $25 for pledge at completion + $10*12 completed surveys = $1,388. If an individual has co-occurring substance use disorder and takes up an LAI every 4 weeks for that as well, they can receive $1,218 x 2 = $2,436 for injections over the course of the year. The largest possible incentive received in a day for someone taking up an LAI every 4 weeks would be $140 at their 12th consecutive injection ($280 for someone who also receives their 12th consecutive LAI for substance use disorder).
Sample Size
The investigators are targeting a total of 1,000 enrolled participants, with 500 people randomized into the treatment group and 500 into the control group. Assessment of the administrative data suggests that there are currently 700 eligible individuals receiving care in the participating outpatient clinics and another 300 individuals per year who would be eligible through the inpatient discharge recruitment process at WPH. Given uncertainty about the take-up rates of enrollment in the study, the expectation is to recruit 1,000 individuals within 2 years.
The investigators obtained this enrollment target from statistical power calculations using an intent-to-treat framework and three 12-month outcomes of interest: medication adherence (base rate of 50%), inpatient use (base rate of 30%), and involuntary hospitalization (base rate of 20%). In two prior studies, the effect sizes of financial incentives for LAIs were 11.5 percentage points and 14.9 percentage points. Since this will be the first study using financial incentives for LAIs with an especially adversely affected and non-adherent population, it is hard to know whether the effect sizes will be larger or smaller. They could be larger in part because there is so much room for improvement, but they could also be lower due to, among other things, immovable distrust of the healthcare system. The investigators performed power calculations for effect sizes of 10% and 12.5% to ensure that they are powered to measure even lower-end effect sizes. Note that these power calculations are additionally conservative because they do not assume the inclusion of covariates from the administrative data. It is estimated that 500 individuals in the treatment and 500 in the control group is both feasible in this setting and would give the investigators confidence for measuring effects on the primary outcomes of interest. If the effect size is large enough, such enrollment numbers would also provide an opportunity to measure effects on secondary outcomes (criminal charges, mortality, etc.).
Plans to Account for Bias
Given the randomization plan, the investigators expect balance in treatment and control groups. However, takeup of the financial incentives or of the surveys might look different across the treatment and control groups. The administrative data will allow the investigators to control for a number of observable characteristics across the two groups and to adjust for or at the least understand the nature of the bias in our measurements.
While a fall in response rates for the 6-month and 12-month follow up surveys is expected, the hope is that continued incentives for survey completion ($10 per survey) and an established relationship with participants will keep response rates high. It is possible, for example, that improved wellbeing among some in the treatment group will result in survey respondents in the treatment group being positively selected, perhaps especially so in the follow-up surveys, and unrepresentative of the whole treatment group sample. Here too the investigators should be able to use administrative data on participant outcomes to understand and account for such bias.
Inclusion Criteria:
Exclusion Criteria: