Objectives: The objective of the study was to estimate the length of stay of patients with hospital-acquired infections hospitalized in ICUs using a multistate model. Design: Active prospective surveillance of hospital-acquired infection from January 1, 1995, to December 31, 2012. Setting: Twelve ICUs at the University of Lyon hospital (France). Patients: Adult patients age greater than or equal to 18 years old and hospitalized greater than or equal to 2 days were included in the surveillance. All hospital-acquired infections (pneumonia, bacteremia, and urinary tract infection) occurring during ICU stay were collected. Interventions: None. Measurements and Main Results: The competitive risks of in-hospital death, transfer, or discharge were considered in estimating the change in length of stay due to infection(s), using a multistate model, time of infection onset. Thirty-three thousand four-hundred forty-nine patients were involved, with an overall hospital-acquired infection attack rate of 15.5% (n = 5,176). Mean length of stay was 27.4 (+/- 18.3) days in patients with hospital-acquired infection and 7.3 (+/- 7.6) days in patients without hospital-acquired infection. A multistate model-estimated mean found an increase in length of stay by 5.0 days (95% CI, 4.6-5.4 d). The extra length of stay increased with the number of infected site and was higher for patients discharged alive from ICU. No increased length of stay was found for patients presenting late-onset hospital-acquired infection, more than the 25th day after admission. Conclusions: An increase length of stay of 5 days attributable to hospital-acquired infection in the ICU was estimated using a multistate model in a prospective surveillance study in France. The dose-response relationship between the number of hospital-acquired infection and length of stay and the impact of early-stage hospital-acquired infection may strengthen attention for clinicians to focus interventions on early preventions of hospital-acquired infection in ICU.