Evidence Based Medicine

The Old Way EBM Overview EBM and Computers

The Old Way

The term physician comes from the Greek word PHYSICOS, meaning a scientist. In Greek mythology physicians were the messengers of the gods Iris and Mercury, who carried the Caduceus with intertwined serpents. Modern Medicine, as a profession, cultivated the elevated role of the physician. Until about 1880 it was possible to know virtually everything there was to know in Medicine. Master clinicians were expected to know it.

Since ancient times Medicine has been viewed as both an art and a science. Combined with the notion of an all knowing physician, it is not suprising that expert opinions were treasured.

Evidence Based Medicine (EBM)

In the last decade, the value of expert opinion has been challenged by careful scientific evaluation of the evidence. Usually expert opinions have serious deviations from such evidence. The complexity and sheer volume of data now available overwhelms the most competent physicians. Specialization was a temporary compensatory strategy, but now, even specialists cannot assemilate everything in their domain.

EBM has developed its own style. Evidence is classified into categories based on their objectivity and the strength of evidence. Research study designs has evolved to accomodate this logic. The results are translated into guideline outline the evidence and offer specific recommendations based on the evidence. The rigor of the methods has elevated EBM to a lofty pinnacle that rivals the old role of the master clinician.

Several problems are emerging with this approach.

• A guideline is only as good as the data. EBM uses carefully controlled studies with constraints on selection of participants. These studies answer a narrowly focused question. In real life situations there are general extenuating circumstances and complexities that are simply not addressed by the available guidelines.
• Guidelines have dead ends. Complex problems have numerous branches in decision logic. When you analyze a guideline you discover that there are many gray areas where there are simply no recommendations possible. The science has not generated the data to bridge the gap.
• Most guidelines are not algorithmic and thus difficult to incorporate into computerized systems for purposes of decision support and alerts.
• Guidelines have a limited lifespan. New data requires updating and modifications.

Because of these limitations, there is still a demand for expert opinions and master clinicians. But knowing their limitations, and those of current EBM, one seeks a better way. At the core of this problem is the desire to make decisions that are intuitively correct and scientifically sound.

EBM and Computers

Because most guidelines are text documents, they are not readily translated into algorithms that can be incorporated into computer supported decision support. The deadends also provide no rational decision point for computerized systems. There have been several efforts for overcome these shortcomings:

1. Guideline Elements Model (GEM). GEM Cutter is a Yale Medical School Informatics Program XML software product that parses existing text guidelines.
2. GuideLine Interchange Format (GLIF). Developed by the Harvard Decision Support Group and collaborators, GLIF is described at one of their websites: "GLIF defines a logical structure and text-based exchange format that can be used as a basis for shared representation. The model includes a range of guideline step types and wrappers for patient data and action models that may be associated with a guideline."
3. Bayesian systems. Discussed in the "Data: Maximizing Value with Advanced Methods" document (click link).

David A Stumpf, MD, PhD