In the present study, we demonstrated that the agreement rate between CAM-ICU and ICDSC is in general moderate, but varies with the type of ICU admission and severity of disease.
The agreement between scales for delirium diagnosis was the object of a few studies 
but, to our knowledge, we provide the first investigation attempting to analyze in separate the agreement between CAM-ICU and ICDSC in medical and surgical patients (elective and emergency surgery) admitted to the ICU and stratified by severity of illness. In a study comparing the agreement between these scales in general ICU patients, kappa’s coefficients ranged from 0.65 to 0.92 
. We had previously observed a kappa agreement rate of 0.59 both in a single center study and a multicenter study 
. Interestingly, even between subsets of surgical patients the agreement rate varies. There are some different characteristics observed in patients that can be related to this variability. Medical and emergency / urgency patients are usually more severely ill at ICU admission as compared to elective surgery patients. In addition, these patients are more prone to use sedation and we can suppose that these differences can interfere in the agreement rate between CAM-ICU and ICDSC. In fact, according to our data it seems that the observed differences in the agreement rates between medical and urgency surgery compared to elective surgery are mainly related to disease severity.
We demonstrated that in the present study population the incidence of delirium did not differ significantly between medical (26%), elective (35%) and emergency surgery (28%) when delirium was evaluated by the ICDSC. In contrast, when evaluated by the CAM-ICU there was a higher delirium incidence in medical patients (20%) when compared to elective surgical patients (10%) and emergency surgical (13%) patients. Several studies had shown that the occurrence of delirium in postoperative patients is common 
, as it is in the general ICU patients 
. Patients who were exposed to major surgeries or emergency surgery and developed delirium had more postoperative complications than the patients who never develop delirium 
. In addition, medical patients also presented worse outcomes when develop delirium 
. Nevertheless, delirium is probably under diagnosed 
. Thus it seems that the low positivity of CAM-ICU in surgical patients indicates that, for this subset of patients, the ICDSC can be a better screening tool. These differences in the performance of the scales also seem to be related to disease severity. In patients presenting with less severe disease delirium positivity was similar in both medical and surgical patients independent on the diagnosis tool that was used. In contrast, in patients presenting with APACHE II score higher than 14 the positivity of CAM-ICU, but not ICDSC, was significantly more frequent in the medical group. The application of CAM-ICU, differently from ICDSC, is more dependent on the interaction between the interviewer and patient, thus is an active diagnosis tool. It is plausible to suggest that as more severely ill, more difficult the interaction between the interviewer and patient (mainly in patients presenting with RASS -3) leading to more difficult tool application. In contrast, ICDSC seems to be more subjective when compared to CAM-ICU, suggesting that its higher positivity is associated a low specificity of delirium diagnosis.
Some limitations of our study must be pointed out. Despite the large sample size this is a single center study. Second, we do not include evaluation of delirium using gold-standard diagnosis by the DSM-IV criteria, thus we can not evaluate sensitivity and specificity of these tool nor ascertain that the differences on CAM-ICU and ICDSC positivity really reflects differences on diagnosis of delirium. This is minimized by the results from a multicenter study demonstrating similar kappa values when comparing CAM-ICU and ICDSC 
. Third, no statistical analyses were done to compare agreement rates, nor if disease severity is an independent risk factor for agreement of the two delirium assessment tools. We had tried to assess this, but the regression for the concordance that we had performed have, in general, poor discriminative capacity. In addition, kappa analyses are subject to “kappa paradox” which in turn limits the interpretation of agreement through its estimation and a formal (statistic) comparison between two kappa values. We tried to minimize this performing two different agreement analyses, the kappa and the AC1. In addition, from the clinical point of view there is no meaning to determine the variables associated with the agreement between the scales, but we just need to know which tool works better for a determined patient.
In conclusion, agreement rates between CAM-ICU and ICDSC may vary between different groups of ICU patients and seems to be affected by disease severity.
- The agreement rate between CAM-ICU and ICDSC is in general moderate, but varies depending on the type of ICU admission and severity of disease.
- Medical and emergency / urgency patients have more severe disease at ICU admission, and they more prone to use sedation and this can interfere in the agreement rate between CAM-ICU and ICDSC.