Cally important bacteremia, while in 13 patients, the blood cultures were judged to be contaminants. The clinical criteria of MacGregor and Beaty4 were used as indi¬ cators of true bacteremia. Agreement between both authors was required for any judgments. An episode of polymicro¬ bial bacteremia is defined as bacteremia where more than one organism was grown from a single blood culture. All other posi¬ tive blood cultures taken at the same time are included in the single episode. During the study period, routing blood cultures were collected by sterile vena puncture after skin preparation with an iodophor antibacterial agent. A volume of blood equivalent to 10% of the total volume of media was inoculated into two bottles containing Columbia broth. The aerobic bottle contained sodium polyane- tholesulfonate and carbon dioxide, while the anaerobic bottle contained sodium polyanetholesulfonate, carbon dioxide, hemin, menadione, and resazurin as an indicator. The aerobic bottle was vented, and both were incubated at 37 °C under 10% carbon dioxide in a humidified atmo¬ sphere. Blind subculture and Gram's stain were done at 24 hours, 72 hours, and ten days. All subcultures were placed on chocolate agar and incubated in both 10% carbon dioxide and under anaerobic condi¬ tions. Organism identification was accom¬ plished by standard microbiological tech¬ niques.5 The relationship between mortality and appropriateness of antibiotic therapy was assessed, fatal outcome being defined as death at any time during hospitalization. Correct therapy was determined to be one or more doses of any antibiotic to which the organisms were susceptible by stan- dard disk diffusion methodss given within 24 hours of culture evidence of infection. Anaerobic Gram-negative rods were as¬ sumed to be penicillin resistant, although susceptibility testing was not done. RESULTS During the 18-month period, a total of 26,961 blood cultures were taken, with 1,715 (6% ) being positive. Nine¬ ty-four patients had 113 cultures containing more than one organism. These polymicrobial cultures ac¬ counted for 6.5% of the total positive cultures and 0.4% of the total cultures taken. Medical records of 61 patients with clinically important polymicrobial bacteremia were avail¬ able for evaluation in this study. MORTALITY Of the 61 patients with polymicro¬ bial bacteremia, 28 were alive at the end of the hospitalization during which the bacteremia occurred. Of the 33 deaths, 19 (58%) could be directly attributed to an episode of