BACKGROUND: Cross-transmission of microorganisms by the hands of health care workers is considered the main route of spread of nosocomial infections. OBJECTIVE: To study the process of bacterial contamination of health care workers' hands during routine patient care in a large teaching hospital. METHODS: Structured observations of 417 episodes of care were conducted by trained external observers (S.T. and V.S.). Each observation period started after a hand-cleansing procedure and ended when the health care worker proceeded to clean his or her hands or at the end of a coherent episode of care. At the end of each period of observation, an imprint of the 5 fingertips of the dominant hand was taken and bacterial colony counts were quantified. Regression methods were used to model the intensity of bacterial contamination as a function of method of hand cleansing, use of gloves during patient care, duration and type of care, and hospital ward. RESULTS: Bacterial contamination increased linearly with time on ungloved hands during patient care (average, 16 colony-forming units [CFUs] per minute; 95% confidence interval, 11-21 CFUs per minute). Patient care activities independently (P<.05 for all) associated with higher contamination levels were direct patient contact, respiratory care, handling of body fluid secretions, and rupture in the sequence of patient care. Contamination levels varied with hospital location; the medical rehabilitation ward had higher levels (49 CFUs; P=.03) than did other wards. Finally, simple hand washing before patient care, without hand antisepsis, was also associated with higher colony counts (52 CFUs; P=.03). CONCLUSIONS: The duration and type of patient care affect hand contamination. Furthermore, because hand antisepsis was superior to hand washing, intervention trials should explore the role of systematic hand antisepsis as a cornerstone of infection control to reduce cross-transmission in hospitals.