Eighty-two episodes of polymicrobial bacteremia in two time periods, 1986-87 and 1996-97, were compared to assess differences in risk factors and outcome to mortality. A prospective, concurrent, anterograde study with univariate analysis of all episodes of polymicrobial bacteremia was performed in Hospital de la Princesa. Logistic regression analysis was applied to all significant variables (p < 0.05) in the univariate analysis in either of the two time periods. Variables showing statistically significant differences in incidence between the two time periods included the following: hospital acquired bacteremia; previous use of antibiotics; genitourinary, respiratory and cardiovascular manipulations; septic metastases; and absence of leukocytosis. These factors were more frequently present during 1986-87 than during 1996-97. The overall RR of outcome to mortality was five-fold greater during the first period than the second: RR 5.6 (CI 1.76-17.56) p < 0.001. The clinical characteristics at the onset of bacteremia associated with mortality in the first period were: underlying disease - < RR 2.20 (CI 1.18-4.08), steroid treatment - < RR 4.24 (CI 0.68-26.59), hypotension - < RR 2.05 (CI 1.0-4.17), and disseminated intravascular coagulation - < RR 2.31 (CI 1.69-3.35). Clinical characteristics at the onset of bacteremia associated with mortality in the second period were: hypotension - < RR 1.44 (CI 1.01-2.08), underlying disease - < RR 1.16 (CI 1.02-1.34), and disseminated intravascular coagulation - < RR 6.40 (CI 1.15-35.69). The variables independently associated with mortality in polymicrobial bacteremia were: period - < RR 2.05 (CI 1.50-2.10), underlying disease - < RR 7.05 (CI 2.68-7.50), hypotension - < RR 7.06 (CI 3.80-7.29), and (probably) vascular manipulations - < RR 3.41 (CI 0.85-4.53). Polymicrobial bacteremia-associated mortality was five-fold greater in 1986-87 than in 1996-97. The variables independently associated with mortality risk were underlying disease, hypotension, the period studied (which would include a number of variables not analyzed in this work) and, probably, vascular manipulations.