BACKGROUND: Transmission of microorganisms from the hands of health care workers is the main source of cross-infection in hospitals and can be prevented by handwashing. OBJECTIVE: To identify predictors of noncompliance with handwashing during routine patient care. DESIGN: Observational study. SETTING: Teaching hospital in Geneva, Switzerland. PARTICIPANTS: Nurses (66%), physicians (10%), nursing assistants (13%), and other health care workers (11%). MEASUREMENTS: Compliance with handwashing. RESULTS: In 2834 observed opportunities for handwashing, average compliance was 48%. In multivariate analysis, noncompliance was higher among physicians (odds ratio [OR], 2.8 [95% CI, 1.9 to 4.1]), nursing assistants (OR, 1.3 [CI, 1.0 to 1.6]), and other health care workers (OR, 2.1 [CI, 1.4 to 3.2]) than among nurses and was lowest on weekends (OR, 0.6 [CI, 0.4 to 0.8]). Noncompliance was higher in intensive care than in internal medicine units (OR, 2.0 [CI, 1.3 to 3.1]), during procedures that carry a high risk for contamination (OR, 1.8 [CI, 1.4 to 2.4]), and when intensity of patient care was high (compared with < or = 20 opportunities for handwashing per hour of care, 21 to 40 opportunities: OR, 1.3 [CI, 1.0 to 1.7]; 41 to 60 opportunities: OR, 2.1 [CI, 1.5 to 2.9]; and > 60 opportunities: OR, 2.1 [CI, 1.3 to 3.5]). CONCLUSIONS: Compliance with handwashing was moderate. Variation across hospital ward and type of health care worker suggests that targeted educational programs may be useful. Even though observational data cannot prove causality, the association between noncompliance and intensity of care suggests that understaffing may decrease quality of patient care.