Abstract
The mortality attributable to ICU-acquired bloodstream infection (BSI) differs between studies due to statistical methods used for cohort matching. Propensity-score matching has never been used to avoid eventual bias when studying BSI attributable mortality in the ICU. We conducted an observational prospective study over a 4-year period, on patients admitted for at least 48 h in 2 intensive care units. Based on risk factors for death in the ICU and for BSI, each patient with BSI was matched with 3 patients without BSI using propensity-score matching. We performed a competitive risk analysis to study BSI mortality attributable fraction. Of 2464 included patients, 71 (2.9%) had a BSI. Propensity-score matching was highly effective and group characteristics were fully balanced. Crude mortality was 36.6% in patients with BSI and 21.6% in propensity-score matched patients (p=0.018). Attributable mortality of BSI was 2.3% [1.2-4.0] and number needed to harm was 6.7. With Fine and Gray model, a higher risk for death was observed in patients with BSI than in propensity-score matched patients (sub distribution Hazard Ratio (sdHR) = 2.11; 95% CI [1.32-3.37] p = 0.002). Patients with BSI had a higher risk for death and BSI attributable mortality fraction was 2.3%.