t information on terminal cancer, we used as a proxy a diagnosis of cancer together with a poor Karnofsky score. We then performed a sensitivity analysis with exclusion of these patients, which showed results very similar to those of the main analysis. Although the use of this proxy may have caused the wrong inclusion of some cancer patients who were not terminal but had low Karnofsky scores due selleck catalog to other severe comorbidities, this sensitivity analysis would have detected substantial bias if there were any.The major limitation of this study is inherent to its observational nature, where the decision to accept a patient into ICU or not is influenced by factors associated with the outcome of the patient. Although we adjusted the analyses for important confounding factors, including measures of severity of illness, residual confounding is likely to be present.

Moreover, we used a top-down approach to the costing of ICU and ward, while ideally the costing method should be bottom up (microcosting), where the individual items of cost are accounted for, and totalled. This method, however, is too expensive to be currently viable throughout Europe. The fact that the use of an alternative and more sensitive approach to costing the ICU stay, based on the level of care received by the patient, did not modify our results suggests robustness of the findings. It is also reassuring to see that our estimate of an average daily ICU cost in the 11 centres using a top-down approach of �1,028 is indeed close to the figure of �923 found in 51 ICUs in German teaching hospitals obtained by Moerer et al.

[35] through the use of a bottom-up approach.ConclusionsIt has been widely assumed that ICU care is expensive and this has almost certainly encouraged an unreasonably low level of provision of ICU beds and resources. The average cost per life saved of $103,771 (�82,358) and an average cost per life-year gained of $7,065 (�5,607) observed in our study would suggest that this assumption is incorrect. This information is important for health care providers who need to balance the costs and benefits of ICU against those of other types of health care. Our results may also influence those who sanction the construction of critical care facilities in the future and may allow those who manage critical care to argue more effectively that critical care potentially does represent value for money and should have a similar priority to other therapies where a higher level of financial provision is regarded as the norm.

Key messages? The daily cost of intensive care has been described, but randomised controlled studies to assess cost effectiveness of admission to intensive care are Anacetrapib unlikely to be feasible.? Observational studies looking at the cost effectiveness of admission to intensive care after ICU triage have generally assumed that the result of non-admission would be death, but some patients do survive.? This observational study attempts to determine cost effectiveness for intensive