Physicians have always felt that their decisions were based on evidence; thus, the current term “evidence-based medicine” is somewhat of a misnomer. However, for many clinicians, the “evidence” is often a vague combination of recollected strategies effective in previous patients, advice given by mentors and colleagues, and a general impression of “what is being done” based on random journal articles, abstracts, symposia, and advertisements. This kind of practice results in wide variations in strategies for diagnosing and managing similar conditions, even when strong evidence exists for favoring one particular strategy over another. Variations exist among different countries, different regions, different hospitals, and even within individual group practices. These variations have led to a call for a more systematic approach to identifying the most appropriate strategy for an individual patient; this approach is called evidence-based medicine (EBM). EBM is built on reviews of relevant medical literature and follows a discrete series of steps.
EBM is not the blind application of advice gleaned from recently published literature to individual patient problems. It does not imply a "one size fits all" model of care. Rather, EBM requires the use of a series of steps to gather sufficiently useful information to answer a carefully crafted question for an individual patient. Fully integrating the principles of EBM also incorporates the patient’s value system, which includes such things as costs incurred, the patient’s religious or moral beliefs, and patient autonomy. Applying the principles of EBM typically involves the following steps:
Questions must be specific. Specific questions are most likely to be addressed in the medical literature. A well-designed question specifies the population, intervention (diagnostic test, treatment), comparison (treatment A vs treatment B), and outcome. “What is the best way to evaluate someone with abdominal pain?” is not an overly useful question to pursue in the literature. A better, more specific question may be “Is CT or ultrasonography preferable for diagnosing acute appendicitis in a 30-year-old male with acute lower abdominal pain?”
Not all scientific studies are of equal value. Different types of studies have different scientific strengths and legitimacy, and for any given type of study, individual examples often vary in quality of the methodology, internal validity, generalizability of results, and applicability to a specific patient (external validity).
Levels of evidence are graded 1 through 5 in decreasing order of quality. Types of studies at each level vary somewhat with the clinical question (eg, of diagnosis, treatment, or economic analysis), but typically consist of the following:
Level 1 (the highest quality): Systematic reviews or meta-analyses of randomized controlled trials and high-quality, single, randomized controlled trials
Level 2: Well-designed cohort studies
Level 3: Systematically reviewed case-control studies
Level 4: Case series and lesser-quality cohort and case-control studies
Level 5: Expert opinion not based on critical appraisal, but based on reasoning from physiology, bench research, or underlying principles
For EBM analysis, the highest level of evidence available is selected. Ideally, a significant number of large, well-conducted level 1 studies are available. However, because the number of high-quality, randomized, controlled trials is vanishingly small compared with the number of possible clinical questions, less reliable level 4 or 5 evidence is very often all that is available. Lower-quality evidence does not mean that the EBM process cannot be used, just that the strength of the conclusion is weaker.
Because the best available evidence may have come from patient populations with different characteristics from those of the patient in question, significant judgment is required when applying results from a randomized trial to a specific patient. Additionally, patients’ wishes regarding aggressive or invasive tests and treatment must be taken into account as well as their tolerance for discomfort, risk, and uncertainty. For example, even though an EBM review may definitively show a 3-month survival advantage from an aggressive chemotherapy regimen in a certain form of cancer, patients may differ on whether they prefer to gain the extra time or avoid the extra discomfort. The cost of tests and treatments may also influence physician and patient decision making, especially when some of the alternatives are significantly costlier for the patient. Two general concerns are that patients who voluntarily participate in clinical trials are not the same as patients in general practice, and care delivered in a clinical trial environment is not identical to general care in the medical community.
Dozens of clinical questions are faced during the course of even one day in a busy practice. Although some of them may be the subject of an existing EBM review available for reference, most are not, and preparing a formal EBM analysis is too time-consuming to be useful in answering an immediate clinical question. Even when time is not a consideration, many clinical questions do not have any relevant studies in the literature.
Clinical guidelines have become widely available across the practice of medicine; many specialty societies have published such guidelines. Most well-conceived clinical guidelines are developed using a specified method that incorporates principles of EBM and consensus or Delphi process recommendations made by a panel of experts. Although clinical guidelines may describe idealized practice, clinical guidelines alone cannot establish the standard of care for an individual patient.
Some clinical guidelines follow “if, then” rules (eg, if a patient is febrile and neutropenic, then institute broad-spectrum antibiotics). More complex, multistep rules may be formalized as algorithms. Guidelines and algorithms are generally straightforward and easy to use but should be applied only to patients whose clinical characteristics (eg, demographics, comorbidities, clinical features) are similar to those of the patient group used to create the guideline. Furthermore, guidelines do not take into account the degree of uncertainty inherent in test results, the likelihood of treatment success, and the relative risks and benefits of each course of action. To incorporate uncertainty and the value of outcomes into clinical decision making, clinicians must often apply the principles of quantitative or analytical medical decision making (see also Clinical Decision-Making Strategies). Additionally, many entities that publish guidelines require that only randomized trial data be used, which is often a significant limitation.