How can delirium best be prevented and managed in older patients in hospital?

Multicomponent interventions

All 3 trials of multicomponent interventions for the prevention of delirium involved specialists in geriatrics and multicomponent strategies to target risk factors for delirium (Table 2). Two of the 3 studies conducted intention-to-treat analyses. The outcome assessors in all 3 trials were blinded. Two trials used the Confusion Assessment Method 26 to diagnosis delirium; the third trial used a modified version of the Organic Brain Syndrome Scale. 27 The Confusion Assessment Method is a tool designed for diagnosing delirium. It can be completed in less than 5 minutes, includes 4 criteria (acute onset and fluctuating course, inattention, disorganized thinking, and altered level of consciousness), is very accurate (sensitivity 94%, specificity 90%) and has high interobserver reliability (kappa > 0.8). 26 The Organic Brain Syndrome Scale was designed to detect and monitor confusion. It consists of 2 parts (disorientation subscale and confusion subscale) and has excellent agreement with the Confusion Assessment Method. 27

The multicomponent interventions appeared to be effective in preventing delirium among patients admitted to hospital with hip fracture (summary RR 0.75, 95% CI 0.64–0.88; p for heterogeneity = 0.58). The number needed to treat to prevent 1 case of delirium was 7 (95% CI 4–20). In one of the trials, there were nonsignificant (p > 0.1) differences in baseline characteristics, including prefracture dementia and impairment in activities of daily living. 23 In multivariable analyses to adjust for these differences, they found similar effect sizes, but the results were no longer significant.

In only 1 of the 3 studies did the intervention have a significant effect on postoperative length of hospital stay (28 days in the intervention group v. 38 days in the control group; p = 0.03). 22 However, the intervention in this study also involved extensive staff education and teamwork around other issues such as the prevention of osteoporosis, rehabilitation goals and discharge planning. There was no difference in discharge location between the intervention and control groups in the 2 studies reporting this outcome. 22^,23

Two studies reported on mortality. 22^,25 Only one trial found a significant decrease in hospital mortality (0.6% [1/155] in the intervention group v. 5.5% [9/164] in the control group; p = 0.03). 25 The intervention in this trial primarily involved a geriatric team comprising a geriatrician, a rehabilitation specialist and a social worker that helped the usual-care group of surgeons and orthopedic nurses coordinate and provide care. However, the details around the care provided by this team were not well documented. As for other secondary outcomes, the multicomponent interventions appeared to reduce the incidence of medical complications, including pressure ulcers, 22^,25 urinary tract infections, 22 sleeping problems, 22 nutritional complications 22 and falls. 22 Despite the limitations of these trials, there appears to be a role for multicomponent interventions to prevent delirium. There is further evidence in support of this from an earlier study involving 852 medical in-patients (RR 0.66, 95% CI 0.46–0.95). 28 In this trial, the intervention was clearly outlined and involved many of the same preventive strategies used in the 3 trials we included in our review. Although it is one of the largest and best known trials of the prevention of delirium, we did not include it because patient assignment was done using prospective matching instead of randomization.

Recent practice guidelines developed by the British Geriatrics Society in conjunction with the Royal College of Physicians of London recommend that patients at high risk of delirium be identified at the time of hospital admission and that prevention strategies be incorporated into their care plan (grade A evidence). 29 Clinical practice guidelines for the management of delirium in older people in Australia also agree that only hospital-based multicomponent preventive strategies currently have good evidence to support their use. 30

Pharmacologic interventions

There were 5 trials of pharmacologic interventions, 4 of which enrolled fewer than 100 patients (Table 1). Since no 2 trials used the same drug, we were unable to combine data for analysis, and thus each trial is described individually.

In a trial involving 57 patients with hip fracture, rates of delirium were examined between patients who received epidural anesthesia with prilocaine and epinephrine with or without bupivacaine (n = 28) and patients who received general anesthesia with thiopental, succinylcholine, atropine and halothane (n = 29). 17 Delirium was assessed using the Organic Brain Syndrome Scale. The method of randomization was not described. Outcome assessors were blinded and intention-to-treat analysis was completed. No significant difference in the rates of delirium were observed between the 2 study groups (50% [14/28] in the epidural group v. 38% [11/29] in the general anesthetic group).

One trial examined postoperative rates of delirium among 59 older patients undergoing surgical resection of colon or rectal cancers. 16 The intervention group received preoperative intrathecal morphine at L4–L5 (n = 29); the control group received a preoperative subcutaneous injection of saline at L4–L5 (n = 30). Postoperatively both groups received propacetamol intravenously and patient-controlled morphine intravenously for breakthrough pain. Delirium was assessed using the Confusion Assessment Method. Randomization was computer generated; treatment allocation was concealed. The physicians in charge of the patients during the intraoperative and postoperative periods were blinded to group assignment. Three patients were excluded in the intervention group: 1 each because of a major deviation in surgery, early postoperative abdominal sepsis, and aspiration pneumonia requiring intubation. Four patients were excluded in the control group: 2 because of a major deviation in surgery, 1 because of early postoperative abdominal sepsis and 1 because of death from pulmonary embolism. Of those remaining in the study, 9 patients in the intervention group and 10 in the control group had postoperative delirium.

The largest of the pharmacologic trials compared prophylactic haloperidol (0.5 mg 3 times daily starting on admission and continued for 3 days postoperatively) and placebo in 430 patients with hip fracture at moderate or high risk of delirium. 20 Computer-generated randomization was conducted; treatment allocation was concealed. The research team and study patients were blinded to treatment allocation. Delirium was diagnosed on the basis of the DSM-IV criteria and the Confusion Assessment Method. Once delirium was diagnosed, patients were given haloperidol or lorazepam, or both. Twenty of the 212 patients in the intervention group and 28 of the 218 in the control group dropped out. Among those who dropped out, data were missing for 35 patients (11 in the intervention group and 24 in the control group). The rates of delirium did not differ significantly between the 2 groups (15.1% [32/212] in the haloperidol group v. 16.5% [36/218] in the placebo group; RR 0.91, 95% CI 0.59–1.44). However, among the patients in whom delirium developed, those originally assigned to the haloperidol group had a shorter duration of delirium (5.4 days v. 11.8 days in the placebo group; p < 0.001) and a shorter hospital stay (17.1 days v. 22.6 days in the placebo group; p < 0.001). There were no reports of drug-related extrapyramidal symptoms or sedation.

Donepezil (5 mg/d for 14 days preoperatively and 14 days postoperatively), a cholinesterase inhibitor most commonly used in the treatment of dementia, was compared with placebo in a trial involving 90 patients undergoing elective total knee or hip replacement. 21 Randomization was done by the research pharmacist. Study patients, health care providers and the outcome assessor were blinded to treatment allocation. Delirium was diagnosed on the basis of the DSM-IV criteria, the Delirium Symptom Interview 31 and the Confusion Assessment Method. The rate of delirium did not differ significantly between the study groups (20.5% [8/39] in the donepezil group v. 17.1% [7/41] in the placebo group; p = 0.69). The mean length of hospital stay did not differ either (4.4 days in the donepezil group v. 4.2 days in the placebo group; p = 0.09).

A small trial involving 42 older patients undergoing surgical resection of gastric or colon cancer examined the use of intramuscular diazepam combined with a continuous intravenous infusion of flunitrazepam and pethidine administered from 8 pm until 4 am each night for 3 nights postoperatively. 15 The investigators hypothesized that sleep disorders are a critical factor in the development of postoperative delirium and designed this protocol in an attempt to control disturbances in the sleep–wake cycle. The protocol was compared with usual care. The method of randomization was not documented. Delirium was diagnosed on the basis of the DSM-IV criteria by a psychiatrist who was blinded to group assignment. Two patients in the intervention group were excluded from analysis because of incomplete administration of the protocol. The rate of postoperative delirium was lower in the intervention group (5% [1/20] v. 35% [7/20] in the control group; p = 0.02). However, there was no statistically significant difference in length of hospital stay (25.6 days in the intervention group v. 29.9 days in the control group; p = 0.74). The intervention led to morning lethargy in 8 patients.

Overall, there is currently insufficient evidence to support the use of any pharmacologic intervention for the prevention or management of delirium. However, further study into the role of antipsychotic agents in reducing the duration of delirium appears indicated.

How should delirium be managed?

We identified 3 trials (total 489 patients) that studied methods to manage delirium in medical in-patients (Table 1). All 3 trials involved a comprehensive geriatric assessment and multicomponent interventions targeted at precipitants of delirium. There were no pharmacologic trials identified.

The multicomponent interventions varied somewhat between studies but included strategies such as optimizing sensory input, orientation protocols, provision of familiar items and family presence, avoidance of restraints, encouragement of self-care, use of atypical antipsychotic agents for hyperactive and psychotic symptoms, use of nutritional supplements where indicated, screening for potentially treatable causes of cognitive impairment and comprehensive discharge planning. In 2 of the studies a geriatrician or geriatric psychiatrist assessed the patient; 18^,19 it is unclear who conducted the assessment in the third trial. 24 All 3 trials conducted intention-to-treat analysis; the outcome assessors were blinded in 2 trials. In 2 trials, computer-generated randomization and concealed treatment allocation were used. The third trial did not describe the method of randomization. All 3 studies used the Confusion Assessment Method to diagnose delirium.

The multicomponent interventions for the management of delirium did not decrease mortality (summary RR 1.08, 95% CI 0.81–1.44; p for heterogeneity = 0.77). There was no effect on length of hospital stay (summary weighted mean difference 3.25 days, 95% CI −2.85 to 9.34 days; p for heterogeneity = 0.12). 18^,24 There was no impact on postdischarge dependency, 18^,19 function 24 or the need for institutional care. 24