HCLF: The powers and pitfalls of evidence-based medicine

CHICAGO—Evidence-based medicine offers power and pitfalls to clinicians, said Mark S. Roberts, MD, MPP, professor and chair of the department of health policy & management at the University of Pittsburgh Graduate School of Public Health, speaking during the Healthcare Leadership Forum on Nov. 14.

Roberts cited an elderly patient he cared for from the age of 90 until his death at 104. The EHR alert system always offered up a reminder that he was due for a colonoscopy and Roberts always rejected that reminder because there is no evidence that colonoscopy is safe and effective for elderly patients.

Most evidence and guidelines are based on the latest randomized controlled trials and they are not adequate for the complexity of the task at hand. Most large trials aim for FDA approval of a new drug that does not need to be better than what’s already available, just better than placebo.

The studies are rarely designed to answer common clinical questions and consequently, the guidelines built from these trials are less useful than what's possible.

“Homogeneity is not what you see in real-life practice,” Roberts said. “By its very nature, EBM is correct only on average in select populations.” Randomized controlled trials are difficult to apply.

EBM “breaks down a lot,” Roberts said. It became important because of the substantial variability in patient care. The problem, he said, is “our recommendations are blunt, and we need to develop methods for making them more personalized. We need to be able to estimate individual benefits and harms and we need to be able to predict the effect of therapies as a function of comorbidities.”

IT can help drive personalized guidelines at the point of care, Roberts said, citing a test of tailored guidelines conducted by Kaiser Hawaii. Calculation of modifiable risk (calculated directly from the EHR) was presented to both patient and doctor. The results showed that they decreased the five-year cardiovascular risk by 13 percentage points (22 percent to 9 percent), experienced a six-time increase in closing care gaps and achieved high patient and provider satisfaction.

EBM was initiated to solve a real problem with variation in care and poor care quality, Roberts said. “However, EBM continues to provide blunt, population-based answers for individual care decisions which not even be correct.”

IT and predictive modeling may help tailor recommendations to individuals. “We need a paradigm shift in what we consider useful evidence. Most research is done in isolation but we need heterogeneity to be part of the reality.”

Health IT can connect data with risk prediction tools and provide “really, really personalized information to doctors and patients so they can make informed decisions about that information as it applies to them, not how it applies on average.”

The true value of health IT with predictive modeling and the statistics and data we have from different sources, he said, is in its ability to understand modifiable risk rather than overall risk.

The Healthcare Leadership Forum was sponsored by ClinicalKey and presented by Clinical Innovation + Technology.