[ICU performance: results of a French study involving 80,000 ICU stays].

Objective: The Standard Mortality Ratio (SMR), comparing the observed in-hospital mortality to the predicted, may measure the intensive care units (ICU) performance.

Study Design: Multicentric retrospective national study.

Methods: A probability model using a severity score such SAPS II calculated the predicted mortality rate. A national French study has been undertaken to compare the SMR of ICUs and looked for explanation.

Results: One hundred six units, 34 were medical (32%), 18 surgical (17%) and 57 medical/surgical (51%) participated to the study. Forty-six ICUs (43%) were located in teaching hospitals. The SMR of the 87,099 stays was 0.84 (0.82-0.85). The SMR of ICUs varied from 0.41 to 1.55. Ten units had a SMR>0.85, which suggested a low performance. They had more stays for cardiovascular failures, as compared with others. The best units (SMR<0.82) had more stays for drug overdose. The SMR increased with the number of organ failures, from 0.47 with zero failure to 1.11 with 4 or more organ failures. The stays with cardiovascular failure, either unique or associated, had a higher SMR. The 7935 stays with a drug overdose had a SMR of 0.12 (0.10-0.14), which suggested a bad calibration of the model in theses cases.

Conclusion: The case mix must be taken in account when comparing the ICUs performance by the mean of SMR, particularly when the units admitted a lot of drug overdoses.