To the Editor.
—Carson et al1 claim to have demonstrated an improvement in outcome in critically ill medical patients managed in a closed ICU compared with those managed in an open unit. However, what is essentially a political position (and one that I also hold) is not supported by their results.The samples of patients studied (124 in the open period, 121 in the closed period) were too small to compare standardized mortality ratios (SMRs) in any meaningful way (SMR, 0.90 and 0.78, respectively; no confidence intervals [CIs] given). Moreover, the investigators provide no evidence that the APACHE II system "calibrated" in their unit (ie, there was a linear relationship between observed and expected mortality across the entire range of risk) and also "discriminated" between survivors and patients who died (ie, gave an area of >80% under a receiver operating characteristic curve). Neither of these seems likely given the small numbers of