Epidemiology of mechanical ventilation: analysis of the SAPS 3 database

Intensive Care Med. 2009 May;35(5):816-25. doi: 10.1007/s00134-009-1449-9. Epub 2009 Mar 14.


Objective: To evaluate current practice of mechanical ventilation in the ICU and the characteristics and outcomes of patients receiving it.

Design: Pre-planned sub-study of a multicenter, multinational cohort study (SAPS 3).

Patients: 13,322 patients admitted to 299 intensive care units (ICUs) from 35 countries.

Interventions: None.

Main measurements and results: Patients were divided into three groups: no mechanical ventilation (MV), noninvasive MV (NIV), and invasive MV. More than half of the patients (53% [CI: 52.2-53.9%]) were mechanically ventilated at ICU admission. FIO2, VT and PEEP used during invasive MV were on average 50% (40-80%), 8 mL/kg actual body weight (6.9-9.4 mL/kg) and 5 cmH2O (3-6 cmH2O), respectively. Several invMV patients (17.3% (CI:16.4-18.3%)) were ventilated with zero PEEP (ZEEP). These patients exhibited a significantly increased risk-adjusted hospital mortality, compared with patients ventilated with higher PEEP (O/E ratio 1.12 [1.05-1.18]). NIV was used in 4.2% (CI: 3.8-4.5%) of all patients and was associated with an improved risk-adjusted outcome (OR 0.79, [0.69-0.90]).

Conclusion: Ventilation mode and parameter settings for MV varied significantly across ICUs. Our results provide evidence that some ventilatory modes and settings could still be used against current evidence and recommendations. This includes ventilation with tidal volumes >8mL/kg body weight in patients with a low PaO2/FiO2 ratio and ZEEP in invMV patients. Invasive mechanical ventilation with ZEEP was associated with a worse outcome, even after controlling for severity of disease. Since our study did not document indications for MV, the association between MV settings and outcome must be viewed with caution.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Cohort Studies
  • Critical Illness / epidemiology
  • Databases, Factual*
  • Female
  • Humans
  • Intensive Care Units / statistics & numerical data
  • Male
  • Middle Aged
  • Respiration, Artificial / statistics & numerical data*