In order to make a reliable diagnosis of recovery room delirium in the very busy PACU setting, we chosen CAM-ICU based on following consideration: (1), CAM-ICU Flowsheet was proved to be a most reliable instrument for delirium assessment in many settings under various cultures including surgical ICU settings in Germany (Guenther et al., 2010, Luetz et al., 2010), a Swedish ICU setting (Larsson et al., 2007), a mixed medical-surgical ICU setting in Netherlands (Spronk et al., 2009), ICU setting in Chinese populations (Chuang WL et al., 2007) and PACU setting (Card et al., 2015); (2), CAM-ICU Flowsheet allow a quick assessment that need only need 50 seconds (interquartile range, 40–120 seconds) in patients with delirium vs 45 seconds (interquartile range, 40–75 seconds) in those without delirium to complete assessments (Guenther et al., 2010), which would be a great advantage for use of CAM-ICU in the busy settings such as PACU; (3), PACU settings are similar with surgical ICU settings in our hospital, and Chinese version of CAM-ICU was tested in a prior study in Chinese population shown good validity and reliability (Chuang WL et al., 2007).
According to CAM-ICU Flowsheet, one quarter of the patients in our study experienced RRD after general anaesthesia for elective non-cardiovascular surgery. Our results are similar to previous studies. Using the confusion assessment method (CAM) score, Sharma et al. reported 45% of elderly patients have RRD after hip-fracture repair surgery (Sharma et al., 2005). Using the Riker sedation–agitation scale, Lepouse et al. reported a delirium rate of 4.7% in adults in the PACU (Lepouse et al., 2006). Radtke et al reported that delirium in the recovery room was seen in 21 patients (14%) with the Diagnostic and Statistical Manual of Mental Disorders -IV (DSM-IV) criteria, in 11 patients (7%) with the CAM score, in four patients (3%) with the Delirium Detection Score (DDS), and in 37 patients (24%) with the Nursing Delirium Screening Scale (Nu-DESC) in the same patient population (Radtke et al., 2008). The prevalence rate for delirium is greatly affected by the diagnostic formulation used (Voyer et al., 2009). The definition of the outcome measure, the length of the post-operative observation period and the patient population also cause differences in observed delirium rates. Exclusion criteria may also affect delirium rates, as cardiac surgery and neurosurgery are major contributors to postoperative delirium (Oh et al., 2008, Rudolph et al., 2009). Development of a widely accepted scale for detecting RRD in the postoperative setting would improve the timely diagnosis and management of RRD.
In the present study, a greater proportion of patients who received isoflurane or sevoflurane for maintenance anaesthesia experienced RRD than patients who received TIVA. Multivariate logistic regression analysis confirmed that isoflurane or sevoflurane for maintenance anaesthesia was the strongest risk factor for RRD. Previous studies have shown that inhalation anaesthetics such as isoflurane and sevoflurane are associated with postoperative delirium during recovery, particularly in young children or elderly patients (Aono et al., 1997). Very few studies have compared the incidence of delirium in adults anaesthetised with inhalation anaesthetics and those anaesthetised with propofol (Lepouse et al., 2006, Nishikawa et al., 2004). Lepouse et al. found more agitated patients had been anaesthetised with inhalation anaesthetics (62%) than with propofol (37%), but multivariate analysis did not confirm this result (Lepouse et al., 2006). Several studies have demonstrated a protective effect of propofol on postoperative delirium in children (Aouad et al., 2007), although this is controversial (Konig et al., 2009). Old rats are more profoundly influenced than young adult rats by isoflurane anaesthesia with regard to reductions in acetylcholine release and stress responses (Jansson et al., 2004). In addition, isoflurane-induced beta-amyloid protein oligomerization and apoptosis may contribute to the risk of postoperative cognitive dysfunction (Xie et al., 2006). Inhalation anaesthetic agents may thus increase the risk of postoperative delirium in specific populations. Testing this hypothesis in a well-designed prospective study may give further evidence in this direction.
Our data showed that patients undergoing surgery for malignant disease had higher incidence rate of RRD than patients with benign disease, and our multivariate logistic regression analysis confirmed malignant primary disease as an independent risk factor for RRD. Delirium occurs in 26–44% of cancer patients (Centeno et al., 2004), and 74% of patients with advanced cancer experience an episode of delirium (Bruera et al., 2009). Structural brain lesions and toxic or metabolic encephalopathy are thought to be causes of delirium in cancer patients (Doriath et al., 2007). Our data suggest that cancer patients undergoing surgery are at increased risk of RRD. Whether interventions for the prevention of delirium in cancer patients result in better short- or long-term outcomes after surgery are unknown. Prevention of delirium, however, is desirable for cancer patients and their anaesthetists (Siddiqi et al., 2007).
Univariate analyses showed a higher proportion of patients with RRD were ASA-PS III-V. Multivariate logistic regression analyses confirmed higher ASA-PS to be an independent risk factor for RRD. Clinical studies of such differences have produced conflicting results. Higher ASA-PS was identified as a risk factor after abdominal surgery in univariate but not in multivariate analysis in a previous study with a small patient population (Koebrugge et al., 2009). Illness severity was also associated with risk of delirium in a prospective study in hospitalised elderly (Francis et al., 1990). Moreover, delirium was the most common neuropsychiatric complication experienced by patients with advanced illness, occurring in up to 85% of patients in the last weeks of life (Breitbart and Alici, 2008). Consistent with our study, Zakriya reported ASA physical status > II to be one of three significant predictors of postoperative delirium in geriatric patients (OR = 11.3, 95% CI 2.6–49.2, p < 0.001) (Zakriya et al., 2002).
Elevated serum total or direct bilirubin was more frequent in the RRD group, and multivariate analysis confirmed elevated total or direct serum bilirubin as an independent risk factor for RRD. Literature examining the relationship between bilirubin and delirium is limited. Dubois et al. demonstrated that abnormal bilirubin levels were associated with delirium in the intensive care unit (Dubois et al., 2001). Direct bilirubin is also assumed to play a role in the pathogenesis of hepatic encephalopathy (Muller et al., 1994). Due to the small number of patients with elevated bilirubin in our population (n = 30), we recommend caution in interpreting this result.
We showed that invasive surgery was an independent risk factor for RRD, in accordance with many previous studies. Low operative stress procedures such as cataract surgery resulted in delirium in 4.4% cases (Milstein et al., 2002), whereas higher stress procedures such as acute hip fracture surgery resulted in delirium in 40% of cases (Marcantonio et al., 2002). Shiiba reported that postoperative delirium was associated with extensive surgery for oral carcinoma (Shiiba et al., 2009). The degree of operative stress may be one of factors affecting RRD. Mini-invasive endoscopic surgery may prevent RRD in high-risk patients, but a proper randomised trial would be required to test this hypothesis.
Several studies demonstrated that older age (Koebrugge et al., 2009), abnormal preoperative sodium, potassium, or glucose levels (Galanakis et al., 2001, Marcantonio et al., 1994), diabetes mellitus (Gao et al., 2008), haemoglobin < 100 g/L (Gao et al., 2008), hypoalbuminemia (Robinson et al., 2009), longer operation time (Yildizeli et al., 2005), massive blood transfusion (Katznelson et al., 2009b), abnormal postoperative sodium, potassium, or glucose levels (Yildizeli et al., 2005), and postoperative haematocrit < 30% (Marcantonio et al., 1998), were important in influencing postoperative delirium. In our study, older age, decreased preoperative serum calcium, elevated preoperative serum glucose, decreased preoperative serum total protein or albumin, location of surgery, total intraoperative body fluid loss and intraoperative fluid application, and duration of surgery were significant in univariate but not multivariate analyses. Difference in study design, study population and the definition of outcome parameters may account for the variance. Some of these parameters seem to play a role, however, and should be included in future prospective studies.
We did not include some variables reported to influence postoperative delirium, such as a history of central nervous system disorder (Gao et al., 2008), pre-existing dementia (Robinson et al., 2009), preoperative depression (Katznelson et al., 2009b), preoperative alcohol use (Williams-Russo et al., 1992), postoperative pain (Oh et al., 2008), and preoperative medication such as beta-blockers (Katznelson et al., 2009a) in our analyses. Many of these variables are not included in our routine clinical data with enough reliability and we excluded patients with central nervous system disease. We are thus unable to report on the relative contribution of these factors in our patients.
Delirium in the surgical/trauma ICU cohort is associated with more days of mechanical ventilation and more days in ICU and hospital (Lat et al., 2009). Elderly subjects with postoperative delirium have a greater hospital LOS, are more likely to be institutionalised after discharge and have a higher six-month mortality than those without delirium (Robinson et al., 2009). As with former studies, our univariate analyses demonstrated that patients with RRD stayed longer in PACU, had longer hospital and postoperative stays, and had higher daily and total healthcare costs.
Our study has several limitations. Due to the observational design, a causal link between the proposed risk factors and RRD cannot be inferred. Choosing exclusion criteria to reduce the possibility of confounding factors may have influenced the results, as excluding patients undergoing cardiac surgery, as well as neurosurgery, may have reduced the incidence of RRD. The patient group we investigated was not a consecutive sample so selection bias is possible. However, the patients in this study were representative of the type of patients treated in our hospital.