Postoperative delirium: are there modifiable risk factors?

Delirium is an acute confusional state with alterations in attention and consciousness.1 Delirium occurs in 14–50% of hospitalized medical patients and its associated mortality rate is 10–65%.2,3 After major noncardiac surgery, 10–60% of patients have delirium.4 Delirium is a serious problem for hospitalized geriatric patients. It may be caused by an underlying medical illness, but often, the exact cause is not identifiable.5 The course of delirium can vary considerably and depends on resolution of the causative factors. From an economic standpoint, postoperative delirium constitutes a significant healthcare expenditure adding nearly $4000 in costs per patient.6 A recent study similarly reported that healthcare costs in patients who developed delirium in the intensive care unit were 31% higher than those without delirium ($41 836 versus $27 105).7 These descriptions of the costs and burden of delirium do not even include the costs associated with nursing home placement and the economic and psychological burden to the family. Understanding the pathophysiology of postoperative delirium is critical to allow the development of management strategies to reduce its occurrence. The development of delirium is thought to be a multifactorial process in which there is a complex interrelationship between baseline patient vulnerability and precipitating factors or insults.8 In the perioperative period, the potential precipitating risk factors for the development of postoperative delirium in older patients include pain, stress of surgery such as blood loss and/or haemodynamic changes, immobility, sleep disturbance, and exposure to medications such as opioids and benzodiazepines that have the potential for profound effects on the central nervous system. This issue of the European Journal of Anaesthesiology contains a study suggesting that two additional precipitating risk factors for postoperative delirium may actually be iatrogenic and potentially preventable.9 The study by Radtke et al.9 used a prospective cohort study design to identify ‘modifiable’ risk factors for early postoperative delirium after elective general anaesthesia. The authors enrolled over 1000 patients undergoing surgery at a university medical centre in a 6-month period. Postoperative delirium occurring in the postoperative anaesthesia recovery unit prior to discharge to the ward and on the first postoperative day was measured using the Nursing Delirium Screening Scale. In addition to patient demographics such as age, sex, the American Society of Anesthesiology (ASA) classification, the study also measured the duration of preoperative fluid and solids fasting, type of anaesthetic agent, choice of intraoperative opioid, site of surgery and duration of surgery, to determine their independent effect on early postoperative delirium. In this study, patients who had preoperative fluid fasting of 2–6 h had a significantly reduced incidence of delirium in the recovery room and in the ward compared with those who fasted for more than 6 h. Patients who were given intraoperative fentanyl also had higher rates of postoperative delirium in the ward compared with those who received remifentanil. When taking into consideration other potential covariates for early postoperative delirium, the authors concluded that the duration of preoperative fluid fasting and choice of intraoperative opioid remained two independent modifiable risk factors for early postoperative delirium. How should these results be interpreted and what are the clinical implications? One of the more difficult tasks in clinical research is to assess whether associations between predictors and outcome derived from observational cohort studies are of a causal nature or not. Observational cohort studies such as the one described here are subject to the influence of factors over which the investigators most often do not have full control, such as the duration of fluid fasting or the choice of intraoperative opioid. In this study, it is unclear why some patients fasted longer than others. It is curious that despite the longer period of fluid fasting, patients received similar amount of fluid during anaesthesia as compared with those who fasted for a shorter period. Customary practice would suggest that more fluid should have been given to replete the longer duration of fluid fasting. Regarding intraoperative opioid, although the authors made a comparison between fentanyl and remifentanil, it appears that additional opioids, including piritramid or morphine were also given to some patients. It is unclear whether these additional opioids affected the occurrence of early postoperative delirium. Second, one of the major determinants of the degree to which chance affects the findings is sample size. Although most clinical readers focus on the P value, a statistically significant result does not mean that chance cannot have accounted for the finding; only that such an explanation is unlikely. A more useful way to examine the role of chance is to use confidence interval, which reveals the true magnitude of the effect with a certain degree of assurance. The wider the confidence interval, the greater the variability in assessing effect size, and the smaller the sample size. In the study presented here, in which delirium on postoperative day 1 was the primary outcome variable, the authors found preoperative fluid fasting to be one of the independent variables predicting delirium. However, the 95% confidence interval for preoperative fluid fasting was large (1.42–78.62) reflecting the small number of patients with delirium in the ward (4.2%). This large confidence interval tells us that the estimate of the true effect of fluid fasting is imprecise and such a study is ‘low powered’. One of the potential explanations that the rate of delirium was much lower in this study than that previously reported is likely that the present study included younger patients and only evaluated delirium up to the first postoperative day. Furthermore, as the authors evaluated only patients discharged from the recovery room to the regular ward, selection bias likely occurred in this study, because sicker patients, particularly those who needed to be admitted to the intensive care unit, were excluded in the study. As ageing and preoperative burden of illness increase the occurrence of postoperative delirium, it would be of interest to determine whether dehydration has a more pronounced effect in the older and sicker surgical patients. A third factor that is important in interpreting the cause–effect relationship in cohort study is the role of confounding. A confounding factor is one that is associated with the predictor, and independent of the predictor, but also a predictor of the outcome variable. In this study, the authors also reported ‘intraoperative fentanyl was an independent risk factor for delirium on the first postoperative day’. The comparison group was the use of remifentanil. It is not clear how the selection between fentanyl and remifentanil was made in this study as the choice of opioid was not controlled. If patients who were expected to have more postoperative pain were more likely to receive fentanyl than remifentanil, then pain is a potential confounding factor in the study that should have been considered. In fact, work from our group suggests that pain is an independent and potentially reversible precipitating factor for early postoperative delirium.10–12 Therefore, studies on postoperative delirium should consider and include the assessment of pain. Despite these potential limitations, the current study does add to the existing body of knowledge on postoperative delirium and raises the question whether dehydration in particular leads to an increased risk for developing early postoperative delirium. Although dehydration has been linked to impaired cognitive performance in both young and older patients,13,14 a study on older patients who had bowel preparation-induced dehydration showed no change in cognitive function compared with those who had no bowel preparation.15 As the duration and magnitude of hydration are critical factors affecting cognitive function, future studies examining the role of dehydration and cognitive performance should consider whether there is a critical level of water deficit affecting cognition. Results from the current study provide a potentially interesting research hypothesis for a future prospective randomized trial to investigate whether dehydration leads to an increased incidence of early postoperative delirium. In surgical patients, successful interventions of postoperative delirium are limited; the most successful study was that by Marcantonio et al.16 In this study, elderly patients admitted for emergency surgical repair of hip fracture were randomly assigned to an intervention (a ‘proactive geriatrics consultation’) or the usual care. Delirium occurred in 32% of the intervention patients and in 50% of the usual-care patients. Despite this reduction in delirium, the length of hospital stay did not differ significantly between the two groups. Other intervention such as prophylactic treatment with pharmacologic agents did not prevent the occurrence of delirium. For example, haloperidol did not prevent delirium occurrence but did reduce its severity and duration, and possible hospital stay.17 This result suggests that haloperidol is effective in the treatment of delirium when it occurs, but not preventing its occurrence. Therefore, although previous clinical trials have been effective to a certain degree in the interventions of postoperative delirium, there is substantial room for more target-oriented interventions. Therefore, if dehydration is proven to be one of the etiologic factors for early postoperative delirium, treatment or amelioration of the dehydrated state should lead to a reduction in the occurrence of postoperative delirium. Radtke et al.9 should be congratulated for conducting a cohort study that is hypothesis generating, which hopefully will stimulate further work in this important area. Acknowledgement Dr Leung has no conflict of interests relating to the review of this article and has not been supported by, nor maintained any financial interest in, any commercial activity that may be associated with the topic of this article. The review is supported in part by NIH 1RO1AG031795-01A2.