OBJECTIVES: To compare the usefulness of procalcitonin (PCT) in detecting infection in elderly patients with that of other clinical and biological markers. DESIGN: Prospective observational study to compare PCT levels in infected and uninfected patients. SETTING: Geriatric teaching hospital in Switzerland. PARTICIPANTS: Two hundred eighteen elderly patients aged 75 and older admitted to an acute geriatric care unit. MEASUREMENTS: Demographic characteristics, comorbidities, Charlson index, general signs (respiratory rate, temperature, pulse rate, confusion, falls, shivering), presence of systemic inflammatory response syndrome (SIRS), sepsis, severe sepsis, septic shock, functional score (Functional Independence Measurement (FIM)) biological parameters (PCT, C-reactive protein (CRP), leukocytes, albumin), and definite diagnosis at admission were collected prospectively for each patient. RESULTS: Long-term corticotherapy, chronic immune diseases, fever of 38 degrees C or higher, white blood cell count, pulse rate, FIM, SIRS, sepsis, CRP of 3 mg/mL or higher, and PCT of 0.5 ng/mL or higher were associated with an infection at admission. In multivariate analysis, only sepsis and CRP of 3 mg/mL or higher were still associated with an infection; PCT levels do not show any significant association in the multivariate analysis. In addition, when PCT had good specificity (94%), it had low sensitivity (24%). False-negative PCT was related to lower severity of infection (lower inflammatory reaction and lower acute renal failure) than true-positive PCT. This finding may also be related to aging per se. CONCLUSION: PCT may be useful to identify severely ill elderly patients admitted to an acute geriatric ward but not to discriminate patients with infection from those without.