Incidence in ICU populations: how to measure and report it?

Incidence of ICU events is mostly measured in one of two ways which differ by the denominator only. Either the number of incident events divided by the number of ICU patients is reported or the number of incident events per 1,000 ICU days is calculated. The difference is relevant, but a connection is rarely made. We give an example where pneumonia diagnosis on admission has no effect on one measure of mortality incidence, but increases the other. We demonstrate how to connect the two measures of incidence. The conclusion is that so-called ‘competing incidences’ should also be reported.

[1]  D. Lanska.  Epidemiology and Biostatistics: An Introduction to Clinical Research , 2010 .

[2]  G. Bernard,et al.  Statistical evaluation of ventilator-free days as an efficacy measure in clinical trials of treatments for acute respiratory distress syndrome , 2002, Critical care medicine.

[3]  Martin Schumacher,et al.  Competing Risks and Multistate Models with R , 2011 .

[4]  P. Michel,et al.  hospitalsadverse events in acute care rates of adverse events and rates of Comparison of three methods for estimating , 2006 .

[5]  Martin Schumacher,et al.  Application of multistate models in hospital epidemiology: Advances and challenges , 2011, Biometrical journal. Biometrische Zeitschrift.

[6]  J. Hanley,et al.  Cultural imagery and statistical models of the force of mortality: Addison, Gompertz and Pearson , 2010 .

[7]  M. Schumacher,et al.  Modeling the effect of time-dependent exposure on intensive care unit mortality , 2009, Intensive Care Medicine.

[8]  David Schoenfeld,et al.  Survival methods, including those using competing risk analysis, are not appropriate for intensive care unit outcome studies , 2005, Critical care.

[9]  Martin Schumacher,et al.  Incidence densities in a competing events analysis. , 2010, American journal of epidemiology.

[10]  M. Schumacher,et al.  Hospital-acquired infections--appropriate statistical treatment is urgently needed! , 2013, International journal of epidemiology.

[11]  S. Chevret.  Logistic or Cox model to identify risk factors of nosocomial infection: still a controversial issue , 2001, Intensive Care Medicine.

[12]  Elie Azoulay,et al.  Attributable mortality of ventilator-associated pneumonia: respective impact of main characteristics at ICU admission and VAP onset using conditional logistic regression and multi-state models , 2010, Intensive Care Medicine.

[13]  J. de Irala-Estévez,et al.  Comparison of different methodological approaches to identify risk factors of nosocomial infection in intensive care units , 2001, Intensive Care Medicine.

[14]  S. Chevret,et al.  Evaluating mortality in intensive care units: contribution of competing risks analyses , 2005, Critical care.

[15]  Martin Schumacher,et al.  Time-dependent covariates in the proportional subdistribution hazards model for competing risks. , 2008, Biostatistics.

[16]  J. Vandenbroucke,et al.  Incidence rates in dynamic populations , 2012, International journal of epidemiology.

[17]  Hugo Sax,et al.  Ventilator-associated pneumonia: caveats for benchmarking , 2003 .

[18]  Martin Schumacher,et al.  Risk factors for the development of nosocomial pneumonia and mortality on intensive care units: application of competing risks models , 2008, Critical care.

[19]  M. Schumacher,et al.  Use of Multistate Models to Assess Prolongation of Intensive Care Unit Stay Due to Nosocomial Infection , 2006, Infection Control & Hospital Epidemiology.