Empiric antibiotic selection by physicians: evaluation of reasoning strategies.

The objectives of the study were to evaluate the appropriateness of empiric antibiotic selection by housestaff treating medical patients with bacteremia. The design was a prospective, observational study at a university-affiliated hospital. Seventy-eight patients with bacteremia were evaluated. A clinical grade of acceptable or not acceptable was assigned to each antibiotic prescription by a consensus panel. The consensus panel found that 34.6% of antibiotic prescriptions were unacceptable (clinical grade). At least one flaw in the chain of reasoning was found in 56.4% of the 78 cases evaluated. Assessment of the clinical setting was correct in 94.9% of the cases; the portal of entry was identified in 91%; adequate knowledge of the bacterial flora at the suspected site of infection was found in 69%; the diagnostic workup was appropriate in 81%, and the correct antibiotic susceptibility patterns were given in 72%. A correct chain of reasoning was more likely to result in an acceptable clinical grade than flawed reasoning (p less than 0.005). However, an appropriate antibiotic selection was made by some physicians despite flawed reasoning, and inappropriate antibiotic selection occurred in a few cases despite fautless reasoning. In 3.8% of cases, unexpected organisms appeared in blood culture. Prescription of broad spectrum antibiotics may then be learned response. If so, educational efforts that emphasize narrow, rather than broad spectrum prescribing may be inadequate to change physician prescribing habits.