Prescribing of potentially unsafe medications for older adults is extremely common; benzodiazepines and sedative hypnotics are, for example, key drug classes frequently implicated in adverse health consequences for vulnerable older adults, such as confusion or sedation, leading to hospitalizations, falls, and fractures. Fortunately, most of these consequences are preventable. Physicians' lack of awareness of alternatives, ambiguous practice guidelines, and perceived pressure from patients or caregivers are among the reasons why these drugs are used more than might be optimal. Reducing inappropriate use of these drugs may be achieved through decision support tools for providers that are embedded in electronic health record (EHR) systems. While EHR strategies are widely used to support the informational needs of providers, these tools have demonstrated only modest effectiveness at improving prescribing. The effectiveness of these tools could be enhanced by leveraging principles of behavioral economics and related sciences.
This is an adaptive cluster randomized control trial (RCT) to evaluate whether newly designed EHR-based tools designed using behavioral principles reduce inappropriate prescribing and adverse outcomes among older adults. This study will be conducted in outpatient and acute care practices of Atrius Health, a large integrated delivery network in eastern and central Massachusetts, which uses the Epic EHR system. In Stage 1, approximately 200 primary care providers at Atrius Health will be randomized to receive usual care or an active intervention. Providers randomized to the active intervention will be randomly assigned to one of 15 active intervention arms. They will then be followed for 6 months. Providers randomized to one of the 15 active intervention arms will receive a newly-designed EHR tool to guide their care of eligible patients. Providers randomized to usual care will receive no newly-designed EHR tool. Providers will receive these EHR tools for their patients who meet the following criteria: 1) older adults (aged 65 years or more) and 2) who have been prescribed at least 90 pills of benzodiazepine or sedative hypnotic in the last 180 days. At the end of Stage 1 follow-up, the 15 active intervention arms will be ranked based on their observed effectiveness at reducing prescribing high-risk medications and select up to the 5 more promising arms for Stage 2. In Stage 2, the providers assigned in Stage 1 to usual care will be randomized with equal probability to be assigned to one of the 5 most promising treatment arms identified or usual care. Providers randomized to one of up to the 5 selected treatment arms will receive an EHR tool to guide their care of eligible patients. After this analysis, the Stage 1 providers in the "winning" arms (i.e., the promising arms) will be randomly assigned to continue to receive their original treatment assignments or to usual care. Similarly, the Stage 1 providers assigned to treatment arms determined to be statistically inferior will be randomly assigned in equal proportions to one of the winning arms or to usual care. After Stage 2, we will evaluate the effectiveness of the tools by comparing the effectiveness of the behavioral principles contained within the tools on outcomes, combining data across both Stages. This is our primary analytic approach. No participant (patient or provider) provided consent for participation, as this trial received a waiver of informed consent and authorization for use of study data. The Mass General Brigham trial described in the protocol is described in another clinicaltrials.gov record (NCT05538065).
Study Type
INTERVENTIONAL
Allocation
RANDOMIZED
Purpose
HEALTH_SERVICES_RESEARCH
Masking
NONE
Enrollment
216
An alert will display in the electronic health record when the provider orders one of the high-risk medications for an eligible patient.
An alert will display in the electronic health record when the provider opens the chart of an eligible patient.
Once the order entry or open encounter alert displays, the provider will have the option to schedule a follow-up message that will be sent 4 weeks after the alert is triggered
An in-basket message will be sent to the eligible provider 2 days before the eligible patient is scheduled for an in-person visit.
The alert language itself will be simplified.
An alert will display in the electronic health record when the medication is sent to sign-off for providers.
A two-staged pre-commitment electronic health record alert will be used. In the 1st alert, the providers will be prompted to discuss risks of these high-risk medications and share a handout about the risks with their patients, at their own discretion. The second alert will be either an order entry or open encounter alert, depending on the arm the provider is assigned to.
Risks of the high-risk medications will be framed differently.
This alert will be representative of the alerts currently firing in the Atrius system and not incorporate any functionality.
An enhanced alert (known as a Best Practice Advisory \[BPA\]) will appear on the provider's electronic health record screen and will contain several standard components such as information about why the medication is dangerous for their patient and an order set that will allow providers to order a gradual dose taper for their patient, order alternative medications, place a referral to a behavioral health specialist, provide instructions on how to make lifestyle modifications to improve patient symptoms, and add patient instructions for how to gradually taper off benzodiazepines and sedative hypnotics, as applicable.
Atrius Health
Braintree, Massachusetts, United States
Change in Inappropriate Prescribing by Study Arm: Descriptive Comparison Across Study Arms
This outcome is measured as a composite of 1) discontinuation of study high-risk medications (i.e., active discontinuation or lack of an order during follow-up) or 2) ordering a dose taper (for benzodiazepine or sedative hypnotics) using EHR data by the primary care provider included in the study arm. If any of these actions occurred by the primary care provider, then the patient was considered to have had a change in prescribing (i.e., a reduction in inappropriate prescribing). This outcome was measured as a binary outcome. As noted in the Outcome Measure Time Frame below, these outcomes were measured and presented across both stages of the adaptive trial at Atrius Health rather than separately, as this was part of the pre-specified analytic plan.
Time frame: 22 months (2 stages)
Change in Inappropriate Prescribing by the Seven Behavioral Intervention Factors (Open Encounter Timing, Boostering, Cold-state Priming, Simplification, Sign-off Approval, Pre-commitment, and Risk Framing): Primary Analysis Approach
This outcome is measured as a binary composite of 1) discontinuation of study high-risk medications (i.e., active discontinuation or lack of an order during follow-up) or 2) ordering a dose taper using EHR data by the primary care provider in the arm. If the provider performed any of these actions, then the patient was considered as having a change in prescribing (i.e., a reduction in inappropriate prescribing). As noted in the Time Frame below, these outcomes were measured and analyzed across both Stages of the trial in the primary analysis. The following outcomes are shown stratified by patients who were in arms containing one of the seven behavioral intervention factors and also among those in arms that did not contain an intervention factor. The primary analysis conducts analyses by intervention factor; the model treats patients in arms that do not contain any of the behavioral factors (i.e., Base Order Entry Alert, Standard Epic Basic Alert, and Usual Care) as the referent.
Time frame: 22 months (2 stages)
Quantity of Prescribing by Study Arm: Descriptive Comparison Across Study Arms
This outcome is measured on the patient level as the number of cumulative lorazepam milligram equivalents of high-risk medications prescribed to patients by Atrius providers over the follow-up for the relevant adaptive trial stage, measured within the EHR system. This outcome was measured as a continuous outcome. As noted in the Outcome Measure Time Frame below, these outcomes were measured and presented across both stages of the adaptive trial at Atrius Health rather than separately, as this was part of the pre-specified analytic plan.
Time frame: 22 months (2 stages)
Quantity of Prescribing by the Seven Behavioral Intervention Factors (Open Encounter Timing, Boostering, Cold-state Priming, Simplification, Sign-off Approval, Pre-commitment, and Risk Framing): Primary Analysis Approach
This outcome is measured on the patient level as the cumulative number of lorazepam milligram equivalents of high-risk medications prescribed to patients by Atrius providers over the follow-up for the relevant adaptive trial stage, measured in the EHR system. This outcome was measured as a continuous outcome. As noted in the Time Frame below, these outcomes were measured and analyzed across both Stages of the trial in the primary analysis. The following outcomes are shown stratified by patients who were in arms containing one of the seven behavioral intervention factors and also among those in arms that did not contain an intervention factor. The primary analysis conducts analyses by intervention factor; the model treats patients in arms that do not contain any of the behavioral factors (i.e., Base Order Entry Alert, Standard Epic Basic Alert, and Usual Care) as the referent.
Time frame: 22 months (2 stages)
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.