Open Access Highly Accessed Open Badges Research

Assessment of candidemia-attributable mortality in critically ill patients using propensity score matching analysis

Francisco J González de Molina1*, Cristóbal León2, Sergio Ruiz-Santana3, Pedro Saavedra4 and the CAVA I Study Group

Author Affiliations

1 Intensive Care Department, Hospital Universitari Mútua Terrassa, Universitat de Barcelona, Plaça Dr. Robert, (08221) Terrassa, Barcelona. Spain

2 Intensive Care Unit, Hospital Universitario de Valme, Universidad de Sevilla, Carretera de Cádiz s/n, (41014) Seville, Spain

3 Intensive Care Unit, Hospital Universitario Dr. Negrín, Universidad de las Palmas de Gran Canaria, Barranco de la Ballena s/n (35010) Las Palmas de Gran Canarias. Spain

4 Mathematics Department, Universidad de las Palmas de Gran Canaria Campus Universitario de Tafira (35017), Las Palmas de Gran Canaria. Spain

For all author emails, please log on.

Critical Care 2012, 16:R105  doi:10.1186/cc11388

Published: 14 June 2012



Candidemia in critically ill patients is usually a severe and life-threatening condition with a high crude mortality. Very few studies have focused on the impact of candidemia on ICU patient outcome and attributable mortality still remains controversial. This study was carried out to determine the attributable mortality of ICU-acquired candidemia in critically ill patients using propensity score matching analysis.


A prospective observational study was conducted of all consecutive non-neutropenic adult patients admitted for at least seven days to 36 ICUs in Spain, France, and Argentina between April 2006 and June 2007. The probability of developing candidemia was estimated using a multivariate logistic regression model. Each patient with ICU-acquired candidemia was matched with two control patients with the nearest available Mahalanobis metric matching within the calipers defined by the propensity score. Standardized differences tests (SDT) for each variable before and after matching were calculated. Attributable mortality was determined by a modified Poisson regression model adjusted by those variables that still presented certain misalignments defined as a SDT > 10%.


Thirty-eight candidemias were diagnosed in 1,107 patients (34.3 episodes/1,000 ICU patients). Patients with and without candidemia had an ICU crude mortality of 52.6% versus 20.6% (P < 0.001) and a crude hospital mortality of 55.3% versus 29.6% (P = 0.01), respectively. In the propensity matched analysis, the corresponding figures were 51.4% versus 37.1% (P = 0.222) and 54.3% versus 50% (P = 0.680). After controlling residual confusion by the Poisson regression model, the relative risk (RR) of ICU- and hospital-attributable mortality from candidemia was RR 1.298 (95% confidence interval (CI) 0.88 to 1.98) and RR 1.096 (95% CI 0.68 to 1.69), respectively.


ICU-acquired candidemia in critically ill patients is not associated with an increase in either ICU or hospital mortality.