The incidence of microbial colonization of injectate was studied in open and closed delivery systems used for thermodilution measurement of cardiac output. Patients with pulmonary artery catheters were prospectively randomized to receive either open two-bottle systems or closed systems. Paired cultures of injectate solution were withdrawn through the stopcock at the time of the initial setup and every 12 h for 48 h. Injectate from eight (35%) of 23 patients with the open system and one (5%) of 20 patients with the closed system yielded a positive culture (p <.05). Thirteen (10%) of 129 pairs of cultures were positive from the open system, compared to one (0.9%) of 111 pairs from the closed system (p <.01). There was no statistically significant difference in the number of cardiac output injections between the two groups, but patients having more than the mean number of injections for cardiac output measurement had a significantly (p <.05) greater likelihood of positive injectate cultures.