Delirium is an important and frequently occurring complication in intensive care unit (ICU) patients [15] associated with a longer ICU stay, longer duration of mechanical ventilation, and higher mortality rates [6]. Consequently, the development of delirium leads to higher costs of care [7]. In view of the strong impact of delirium on clinical outcomes, the early recognition of delirium may result in better outcome if therapy could be started earlier. Nevertheless, the incidence of delirium may be severely underestimated [8], which is an important reason for why specific screening instruments have been developed over time [9]. Both the Confusion Assessment Method for the ICU (CAM-ICU) and the Intensive Care Delirium Screening Checklist (ICDSC) [9] may be used as reliable daily diagnostic instruments for the detection of delirium [10]. Recently, the Dutch version of the CAM-ICU was validated as well [11].

In our unit, the nurses and intensivists thought that they were very able to detect delirium in their patients due to the continuous bed-side care in the ICU setting. In addition, no protocol was present in the ICU with respect to the way patients were screened for delirium. Agitated patients were frequently suspected of delirium; in such cases, a geriatrician or psychiatrist was consulted to establish or rule out the presence of delirium. However, the clinical approach to considering delirium seemed to be strongly dependent on the attending physicians and nurses.

We investigated whether ICU nurses and intensivists can clinically identify the presence of delirium in ICU patients during daily care in comparison to the CAM-ICU.


Project location

We report here on the clinical recognition of delirium in a university-affiliated teaching hospital in the Netherlands (Gelre Hospitals [GH], Location Lukas, Apeldoorn). In the GH, the ICU is a ten-bed mixed medical–surgical ‘closed format’ department in which all patients are under the direct care of the ICU team. This team comprises two full-time ICU physicians, five ICU physicians who participate in evening and night shifts, one resident, and 46 full-time equivalent ICU nurses.


Only patients with an ICU stay longer than 48 h were included. Patients with pre-existing neurocognitive dysfunction, documented signs of dementia after psychiatric or geriatric evaluation, language barriers or deafness, an active psychiatric disorder like psychosis, and severe neurological disorders like stroke or meningitis were excluded.

Diagnosing a delirious state

Over a period of 3 months, both attending ICU physicians and bed-side nurses were asked to score daily every morning between 9:00 a.m. and 10:00 a.m. whether they had the clinical impression that, during that particular day, the patient had signs of delirium, were not delirious, or that the delirious state could not be evaluated, e.g., due to a sedative state. Although they asked the patients involved questions and tried to make contact with them, no structural validated screening instrument for delirium (like the CAM-ICU or the ICDSC) was used, which left them with their ‘clinical impression’ of whether a delirium was present or not. The level of sedation was routinely evaluated as part of the local sedation protocol using the Richmond Agitation and Sedation Scale (RASS) score [12]. In this protocol, a specific RASS score is recorded in the patient record as the daily goal. A comatose state was defined when the patient had an RASS score <−3. Using an objective and validated sedation scale is particularly important when patients develop a hypoactive delirium (in this study, arbitrarily defined as a RASS score from −3 to −1), which might be easily missed, whereas a hyperactive delirium (in this study, defined as an RASS score from 2 to 5) may be more obvious. Separately, in the same time window, patients were also evaluated daily by delirium working group research nurses for the presence of delirium using the CAM-ICU. These nurses had been specifically trained in applying the CAM-ICU (see the appendix in the ESM). This evaluation was not disclosed to ICU team members involved in bed-side patient care. In view of previous validation studies of the CAM-ICU in relation to psychiatric evaluation according the DSM-IV criteria [4, 11, 1315], the CAM-ICU was considered as the gold standard for the diagnosis of delirium in this study.

Statistical analysis

The completed data were analyzed using the SPSS version 12.0 statistical software package. Continuous data are expressed as medians with interquartile range (IQR) and categorical data in percentages. Comparisons were made using the Mann–Whitney and Kruskal–Wallis tests for continuous data and the Chi-square test for categorical data. All tests were two-tailed and differences with a P-value <0.05 were considered as significant.


During the 3-month study period, 53 patients with an ICU length of stay (LOS-ICU) >48 h were eligible for the study. Seven patients were excluded, i.e., due to language barriers (n = 3), deafness (n = 2), established dementia (n = 1), and stroke (n = 1), which left 46 patients for the final analysis. The characteristics of those patients are shown in Table 1.

Table 1 Characteristics of patients with (n = 23) and without (n = 23) delirium during intensive care unit (ICU) stay

Recognition of delirium

CAM-ICU scores were obtained during 425 patient days in 46 patients. Scores were positive (n = 89), negative (n = 172), or non-evaluable because the RASS was ≤−3 (n = 164). Only 28.0% of the delirious days were picked up by attending intensivists (Table 2). Although nurses performed a little bit better (34.8%), their recognition was also poor. Specificity was high for both groups of caregivers (100% and 98.3%, respectively). Hence, the occurrence of delirium was severely underestimated, but if delirium was judged to be present on clinical grounds, it was virtually always confirmed by a positive CAM-ICU testing. Sensitivity of the clinical detection of delirium was lower (doctors and nurses: 23.6%) during screening moments when patients were in an active state (n = 196) when compared to screening moments in patients who were in a hypoactive state (doctors: 28.6%; nurses 42.9%; n = 121; Table 2). The number of screening moments in patients who were in an agitated state was too low to allow any conclusions to be drawn.

Table 2 Comparison between doctors and nurses with respect to the detection of delirium during 425 observation days in 46 ICU patients when the Confusion Assessment Method for the ICU (CAM-ICU) serves as the diagnostic standard

Factors associated with delirium

Considering the CAM-ICU as the gold standard, delirium occurred in 50% of the patients with a median duration of 3 days (IQR 1–9). Patients with delirium stayed longer on the ventilator (6 days, [4–25]) than those without delirium (4 days, [1–6]; P < 0.001; Table 1). Consequently, patients with delirium during their ICU stay remained longer in the ICU (9 days, [6–26]) than those without delirium (5 days, [3–8]; P < 0.001). Interestingly, the hospital stay after ICU discharge was also longer in patients who had been delirious (29 days, [21–41]) compared to those who had not been delirious (19 days, [7–30]; P < 0.001).


This study demonstrates that both ICU nurses and intensivists have difficulty in diagnosing the presence of delirium in their patients during daily care without a validated, reliable, and easy-to-use diagnostic instrument. Despite a potential Hawthorne effect due to the presence of a ‘delirium working group,’ the ICU team was insufficiently capable of detecting delirium in critically ill patients, which led to severe underestimation of its prevalence. Only 35% of the ICU days with delirium were recognized by bed-side nurses, while attending physicians performed comparably. Comparable findings were recently reported by Devlin et al. using the ICDSC, showing that a period of structural training improves the capacity for ICU nurses to diagnose delirium nearly as well as experts [16].

Others have reported that involving ICU physicians in the delirium-screening process by applying tools like the ICDSC [10] and the Neelon and Champagne (NEECHAM) confusion scale [17] improves the diagnostic process. This is important because intensivists play a coordinating role in diagnosing and treating delirium. Indeed, wide variation in the approach towards delirium was reported in a recent Canadian study by Cheung et al. [18]. In particular, the presence or absence of an obvious medical etiology was strongly related to willingness by physicians to call the presence of delirium in that study. The incidence of delirium in this study was comparable to previous reports, although incidences vary between 30 and 80% [5, 6, 15, 19, 20]. It is intriguing to observe that, even in this small study, a significant impact of delirium was found on ventilator days, length of stay in the ICU, and total hospital stay. This underlines the importance of this entity.

Several limitations to our study should be mentioned. First, we report observations from a single center in the Netherlands. However, the fact that the ICU team composition was constant during the observation period strengthens the data in that the recognition of delirium by critical care providers was rather poor. Also, the morbidity associated with delirium was comparable to previous reports, which makes it unlikely that our results would have been different in a much larger group of patients. Second, this is a small observational study, not designed to evaluate the possible association between delirium and outcome parameters. However, the data are in line with previous findings and illustrate that the patients included represent a common ICU population. Third, one might argue that concurrent sedation may have influenced the results. However, this seems unlikely in view of the finding that the sensitivity of detecting delirium was somewhat higher, but at least not lower, during sedative states than during active states.

In conclusion, in view of the strong impact on patient morbidity and resource consumption [6, 7], structural screening for delirium by a validated diagnostic instrument should be a part of routine daily critical care. Further studies should demonstrate whether the implementation of these structural tools results in better patient care and outcome.