Johnson et al reviewed charts in a retrospective case-control cohort study on 754 consecutive patients at Barnes-Jewish with severe sepsis or shock due to Gram-negative bacteremia. The exposure was receipt of antibiotics in the previous 90 days.

310 of the bacteremic patients had received antibiotics previously. Compared to unexposed controls, the previously antibiotic-exposed had a greater likelihood of inadequate coverage for their bacteremia (45% vs 21%) and mortality (51% vs 34%), p<0.001. Recent antibiotics were independently associated with mortality (odds ratio 1.70). Rates of resistance were higher for each antibiotic tested in the exposed group -- usually by a factor of 3 or 4 times the resistance rates in the controls (e.g., 29% vs 7% for cefepime).  Crit Care Med 2011;39:1859-1865.