An innovative approach that incorporates a massive, de-identified data set of EHR records is transforming the research landscape.
By John Halamka, M.D., Diercks President, Mayo Clinic Platform and Paul Cerrato, MA, senior research analyst and communications specialist, Mayo Clinic Platform
Without a steady influx of high-quality clinical research studies, medical practice is little more than a haphazard collection of expert opinions and provider experience. Unfortunately, generating this kind of evidence is usually expensive and time consuming—at least until now. With the introduction of multi-modal AI models, de-identified data from electronic patient records, and state-of-the-art analytical tools, this scenario is slowly changing. A recent report in npj Health Systems describes four research projects launched by Mayo Clinic Platform that illustrate this new paradigm. This innovative approach provides physicians and nurses with the intelligence needed to make more informed day-to-day decisions.
Over the years, several digital initiatives have been put in place to make this paradigm a reality. As Yu et al point out, this included the development of the Observational Medical Outcomes Partnership Common Data Model (OMOP)(CDM). It’s a standardized data format that enables hospitals, insurers, diverse EHR vendors, and other organizations to easily share patient data. For instance, while Epic, Cerner, and Blue Cross may use different terms to describe patient demographics and disease descriptions, OMOP serves as a universal translator that lets third parties compare all these data elements. Other invaluable resources that are enabling users to analyze patient data are the All to Use Research Program in the US and the UK Biobank, which offer investigators the ability to review EHR content in the OMOP format. Mayo Clinic Platform has taken advantage of these tools to help researchers accelerate project development and generate much needed insights for frontline clinicians.
Using EHR data from Mayo Clinic, Yu et al developed projects that 1) use AI simulation to emulate a randomized controlled trial (RCT) that evaluated how effective heart failure medication is, 2) develop a second RCT emulation to find the relationship between drugs that lower blood pressure and the risk of dementia, 3) create a deep learning algorithm designed to predict the progression of mild cognitive impairment to Alzheimer’s disease, and 4) launch a project that attempts to predict the likelihood of serious cardiovascular complications after patients have undergone liver transplantation. The researchers explain: “In this study, we demonstrated that Mayo Clinic Platform has played a critical role in enabling clinical studies using real-world EHR data. While multiple platforms have enabled advances in AI research, this paper focuses on the Mayo Clinic Platform environment to illustrate practical workflows, collaborations, and outcomes.” This is accomplished by “integrating federated, multi-institutional data with standardized OMOP CDM formatting and embedding comprehensive research tools within a single cloud-based environment.” External end users can take advantage of such tools by applying to join Mayo Clinic Platform and going through a detailed approval process.
The research projects described here are similar to one we described in an earlier column. Gelareh Zadeh, M.D., Ph.D., David C. and Flora C. Pratt Distinguished Chief Medical Officer for Mayo Clinic Platform, has leveraged Mayo Clinic Platform's capability to help answer questions like: What factors influence outcomes for brain tumors such as blood sugar, hemoglobin, white blood count, and many multiple other clinical parameters? What fraction of meningioma patients received radiation? How long is the wait time from diagnosis to surgery the safe limit of care? To answer the question about radiation therapy, she initially collected her own cohort of 2,154 patients, which took five years to build. On the other hand, it took her only a few minutes when she utilized Mayo Clinic Platform's discovery capabilities to access its range of de-identified clinical data and created a cohort of 6,545 patients.
Traditional clinical trials will always play an important role in evidence-based medicine, but with the help of large de-identified data sets and advanced analytical tools, we are making bold moves into a more innovative future. Patients will ultimately benefit from this paradigm shift.
Share this:
