Electronic Health Record Well Suited For Pragmatic Clinical Trials

By Deborah Borfitz

June 10, 2021 | A pair of clinical trials implemented through the Epic electronic health record (EHR) system were highlighted during a keynote address at the recent Colorado Pragmatic Research in Health Conference (COPRH Con) by Andrea Troxel, director of biostatistics and professor of population health at New York University (NYU). “There is enormous potential for innovation in this space,” she says, and the enabling technologies include an EHR-integrated Way to Health research platform developed by her colleagues at the University of Pennsylvania.

She began with the story of how Dr. James Lind showed that oranges and lemons were a cure for scurvy while aboard a naval ship in 1747. Though not exactly a trial, Troxel says, it was the most pragmatic experiment ever conducted in that he assigned different therapies to patients having the same level of symptoms and, in doing so, conquered a killer disease. 

Today, she continues, the defining features of pragmatic trials are a diverse and relatively unselected patient population; clinically relevant interventions that require no specialized experts and are widely scalable; simple, easily measured outcomes that come from existing data sources; and non-standard randomization structure (e.g., cluster randomization or unbalanced randomization ratios) to allow for the largest possible number of individuals or care sites to try an intervention. 

COPRH Con was co-sponsored by the Adult and Child Consortium for Health Outcomes Research and Delivery Science (ACCORDS) education program at the University of Colorado School of Medicine and the Colorado Clinical and Translational Sciences Institute. ACCORDS is the recipient of a three-year conference grant from the Agency for Healthcare Research and Quality.

As defined by Robert Califf and Jeremy Sugarman in 2015, pragmatic clinical trials are intended to inform decision-making rather than elucidate a mechanism, enroll populations relevant to that decision and be representative of groups for whom the decision is relevant, and either streamline procedures and data collection or measure a broad range of outcomes, Troxel shares. 

They are effective especially for the study of behavioral interventions and the effect of patient selection in a drug trial (biological mechanism of action) or behavioral trial (impact of motivation on enrollment and response to an intervention), she continues. Standard randomized controlled trials may provide high-confidence answers to the wrong questions or the questions themselves might be less relevant than those pragmatic trials might address.     

Digital platforms are proliferating rapidly, says Troxel, pointing to the upside of larger and faster computers, smartphones and tablets, wearable devices such as FitBits and actigraphy devices, home health monitoring systems to monitor blood pressure and glucose levels, personal devices for self-monitoring, and, most recently, virtual care via telemedicine. 

Health information is also readily available from EHRs ubiquitous in outpatient, inpatient, and specialty service settings, she adds. Claims information can be “fruitful but hard to wrangle sometimes,” and patient registries are a growing source of disease-specific information. Social media data, “for better or worse,” is also increasingly being paired with machine learning in ongoing attempts to track health trends and predict disease outbreaks. 

For the “coolest technology” to be scalable requires that it be user-centered, she says, “which is easy to lose sight of when you are talking to engineers in the back room.” Implementing technology in pragmatic trials might be for purposes of randomization, participant scheduling and tracking, trial activity tracking, safety monitoring and reporting to the data safety monitoring board, device linkage, participants outreach, ecological momentary assessment (e.g., texting participants in an osteoarthritis study several times a day to assess the pain they are experiencing), a platform-enabled survey, or for patient- and provider-facing communication.


Enhancing Recovery 

Epic, the most used EHR system, has an assortment of built-in features for research purposes and integrates with many third-party platforms, says Troxel. It was therefore used for both an EMPOWER (Electronic Monitoring of Patient Offers Ways To Enhance Recovery) and BE-EHR (Behavioral Economics in the Electronic Health Record) pragmatic trials.  

For the EMPOWER study, Troxel worked with colleagues at Penn to lower the risk of decompensation among patients with congestive heart failure, which would require intensive response in a hospital. To incentivize patients to do daily weigh-ins and adhere to diuretics, the intervention group was enrolled in a daily lottery for a chance to win $3 (if they did one or the other) or $50 (if they did both). Compliance was tracked using wireless devices, including a scale and electronic pill bottle. 

Adherence to the intervention and lottery wins was tracked on a chart, Troxel says. The odds of a lottery win were good (19%) and the “regret feature” was important (patients were notified when a win was forfeited due to noncompliance).

Alerts also went out to providers of substantial weight gain (more than three pounds in 24 hours or five pounds in 72 hours) that heightens decompensation risk. When interviewed, providers indicated they had “robust responses” to the notifications sent via Epic, she says.

The Way to Health research platform was used for randomization, tracking, and scheduling, and automated interactions with patients and providers, Troxel says. Patients were sent tips for reducing their salt intake, for example, and the alerts to providers appeared as part of the clinical workflow.

Critical features of EMPOWER were electronic delivery of the intervention, including daily engagement of patients and extra communication when triggered by automated data, and the fact that the study was embedded within the health system and incorporated into the existing workflow to minimize the burden on busy clinicians and enhance management within the context of care. 

Traditional in-person informed consent was used, since participants enrolled while hospitalized with an acute heart failure episode, she says. Outcomes, including hospitalizations and primary diagnoses, were measures using EHR data.


Choosing Wisely

The BE-EHR study was a pragmatic randomized controlled trial in older patients with diabetes and “tailored advisory nudges” embedded within the EHR were used to enhance awareness and compliance among providers with Choosing Wisely treatment guidelines, says Troxel. Recommendations were aimed at reducing low-value care that offers little benefit and has associated risk.

Here, the focus was on one of 10 Choosing Wisely guidelines published by the American Geriatric Association specifying that older patients with type 2 diabetes do not need to have their hemoglobin levels as tightly controlled as their younger counterparts, as it comes with the risk of hyperglycemia and falls. The nudges were embedded in the drug refill and lab results section of the EHR with a caution sign and message about how HbA1c targets are tied to life expectancy for patients over 75, Troxel says. 

Peer comparison emails were also sent to providers showing their rate of overtreatment relative to their practice and NYU colleagues overall, with suggestions for how to improve, she says. Top performers instead got an affirming congratulatory message. 

A separately campaign email with a Jeopardy theme had animated video components and providers answer a quiz about geriatric diabetes in the form of a question, Troxel says. The communication went out monthly, triggered by automated evaluation of patient status.

Troxel says a waiver of consent was allowed for the BE-EHR study, which served to broaden eligibility, since participants were still receiving the standard intervention by their physician. Outcomes were measured using EHR data, reducing the burden on both patients and providers. 

Choosing a comparator group for behavioral trials is “not that straightforward,” Troxel later adds. One question that emerged early on with the EMPOWER study is whether patients might do better because of all the daily, lottery-related interaction. She suggests “attention control” might be achieved by reaching out to a comparator group with tips on weight loss at the same time the intervention group is being messaged about a lottery winning. 

Load more comments