Broadening the Approach to Evidence-Based Medicine

Coloring outside the lines has always been a challenge for healthcare providers. There are several ways to address the problem.

By John Halamka, M.D., President, Mayo Clinic Platform, and Paul Cerrato, senior research analyst and communications specialist, Mayo Clinic Platform.

In our new book, we talk about the need to redefine the boundaries of medicine, including the need to rethink the way medical knowledge is acquired.  As most health professionals know, reason coupled with careful observation and experimentation are the cornerstone of medical research and the foundation upon which high quality patient care rests. But determining how to use these three cognitive tools continues to challenge physicians, nurses, technologists, and investigators alike. And they have sparked debate among thought leaders, clinicians, and philosophers of science for decades.

In the past, too much emphasis was placed on expert opinions in medicine. Diagnostic and treatment protocols were largely based on expert opinion and clinical experience. This combined expertise and experience was often collected by consensus panels, summarized in official-sounding statements, and “canonized” in medical textbooks. This approach too often ignores evidence from comparative clinical trials.

A meta-analysis that compared expert recommendations to randomized controlled trials (RCTs) on the treatment of myocardial infarction, for instance, revealed important differences between the experts’ opinions and the results of RCTs, pointing out that “Review articles often failed to mention important advances or exhibited delays in recommending effective preventive measures. In some cases, treatments that have no effect on mortality or are potentially harmful continued to be recommended by several clinical experts.”

This outdated approach has been largely replaced by evidence-based medicine, with its heavy reliance on RCTs, probability, and statistical analysis. This evidence-based medicine (EBM) approach to clinical medicine has become the primary standard currently used to judge the efficacy of diagnostic and therapeutic regimens. The rationale for placing the most trust in RCTs, systematic reviews, and meta-analyses stems from two critical issues that reduce the reliability of observational studies: expectations and confounding.

However, while RCTs are considered the gold standard for judging new treatments, they too have their weaknesses, and clinicians who place too much emphasis on RCTs will miss many potentially valuable treatment options from which their patients will likely benefit. Observational studies and open clinical trials have a role in choosing the best diagnostic and treatment options, and so does real world data (RWD) that can be directly gleaned from electronic health records. Dutch researchers point out that: “RWD can complement traditional study designs because it captures almost the complete variety of patients, leading to more generalisable results. For rheumatology, these data are particularly interesting as our diseases are uncommon and often take years to develop.” Similarly, investigators from the University of California, Davis and the University of Colorado explain: “Electronic health records (EHRs) and linked biobanks have tremendous potential to advance biomedical research and ultimately improve the health of future generations.”

Mayo Clinic Platform is partnering with Atropos Health, who has developed a clinical informatics consult service powered by a network of real world data. Physicians submit a query through an online portal. Then, using a proprietary search engine, statistical analyses, and by extracting the benefits of clinical data from Mayo Clinic Platform_Discover, Atropos’ clinical informatics team generates an automated Prognostogram report to answer the question. This analysis is often based on aggregated data from millions of patient charts, including procedures, lab results, demographics, vitals, and more.

The querying involves proprietary tools for cohorting and analytics. Behind the scenes, Atropos staff uses a fast temporal query language (TQL) and an advanced cohorting engine (ACE) to generate new patient cohorts from de-identified medical data. For example, a cohort might consist of male patients over 65 years old with no history of stroke who are type II diabetic (defined by at least two occurrences of type II diabetes ICD9 codes or two high A1C lab results) and who went on to have a stroke within three months after administration of glipizide. Atropos can quickly analyze the association of such patient characteristics and interventions on outcomes. This information can help healthcare organizations, individual clinicians, and researchers to solve institution-wide problems or improve the treatment of individual patients.

In one example, a user wanted to know the mortality for hypertensive and hyperlipidemic patients at another health system with elevated troponin who receive clopidigrel in addition to other therapies. More specifically, the question was: For patients aged 55-70 who have a history of elevated lipids and present with an elevated troponin, what is the mortality benefit for patients who receive clopidogrel in the seven days post-presentation vs those who do not?  Atropos “evaluated patients aged 55-70 seen at a single health system that had a history of hypertension and hyperlipidemia and were on statins who presented with their first elevated troponin. Patients were split into those that received clopidogrel as part of their care (n=798) and those that did not (n=4,433). The cohorts were comparable in baseline demographics and comorbidities. It then evaluated the patients unmatched, matched for basic demographics, and matched using high-dimensional propensity scores and found no difference in overall survival for the cohorts after matching for observable confounders.”

Clinicians take a conservative approach to patient care for a reason. Jumping on the bandwagon and embracing every innovation that comes along can do serious harm when additional research proves the new treatment causes long-term complications. But at the same time, being overly cautious slows down the implementation of genuine breakthroughs. We need to find a delicate balance between these two poles. Tapping EHR data, observation studies, and open clinical trials is one way to get us there.