Medical decisionmaking: let's not forget the physician.

In the past 2 decades, the literature has been flooded with studies that attempt to define historical, physical, or laboratory findings that alone or in some combination can be used to riskstratify emergency department (ED) patients. The rationale for this work is that such risk-stratification tools will identify lowrisk patients who do not require additional evaluation during their ED visit. Because the elements of the tool (eg, history, physical, laboratory testing) are typically easier to acquire than the alternative action (eg, computer imaging), the tool will ostensibly decrease use, reduce costs, and avoid adverse effects of the alternate test (eg, consequences of ionizing radiation). In this issue, Zongo et al present their work on S100-B, a marker they posit may identify patients at such low risk for clinically important head injury that cranial computed tomography (CT) can be avoided. Their article is similar to many other risk-stratification tool efforts in that it systematically discounts the physician’s capacity to stratify patients without such heuristics. Research that fails to test the decision rule against physician judgment is implicitly predicated on 3 assumptions: (1) in the absence of the tool, the physician will order the alternate test or therapy; (2) the prediction tool will be more accurate than physician judgment; and (3) the tool neutralizes subjectivity either by posing defined clinical criteria or by introducing biomarkers as the arbiter of further resource use. The first of these assumptions is sometimes true. Some tests are reflexively ordered for the majority of patients with a given chief complaint. For most tests, however, practice is highly variable both among individuals and among institutions. We certainly should not assume, however, that in the absence of a decision rule the alternate test will be performed. For example, the use of head CT in minor trauma is highly variable and constantly changing according to such factors as availability of CT, societal beliefs about benefits and harms, and reimbursement—factors that often have little to do with the medical value of the test. The second assumption, that physicians make less accurate decisions without a rule, is systematically unstudied and lacks face validity. Consider an analogy: there is often vocal disapproval in the medical community when a new drug or a