Effects of a screening and treatment protocol with haloperidol on post-cardiotomy delirium: a prospective cohort study

Abstract

OBJECTIVES

Post-cardiotomy delirium is common and associated with increased morbidity and mortality. No gold standard exists for detecting delirium, and evidence to support the choice of treatment is needed. Haloperidol is widely used for treating delirium, but indication, doses and therapeutic targets vary. Moreover, doubt has been raised regarding overall efficacy. The purpose of this study was to assess the effect of a combination of early detection and standardized treatment with haloperidol on post-cardiotomy delirium, with the hypothesis that the proportion of delirium- and coma-free days could be increased. Length of stay (LOS), complications and 180-day mortality are reported.

METHODS

Prospective interventional cohort study. One hundred and seventeen adult patients undergoing cardiac surgery were included before introduction of a screening and treatment protocol with haloperidol for delirium, and 123 patients were included after. Nurses screened patients using validated tools (the Delirium Observation Screening (DOS) scale and confusion assessment method for the intensive care unit (CAM-ICU)). In case of delirium, a checklist to eliminate precipitating/ inducing factors and a protocol for standardized dosing with haloperidol was applied. Group comparison was done using non-parametric tests and analysis of fractions, and associations between delirium and predefined covariates were analysed with logistic regression.

RESULTS

Incidence of delirium after cardiac surgery was 21 (14–29) and 22 (15–30) %, onset was on postoperative day 1 (1–4) and 1 (1–3), duration was 1 (1–4) day and 3 (1–5) days, respectively, with no significant difference (Period 1 vs 2, all values are given as the median and 95% confidence interval). The proportion of delirium- and coma-free days was 67 (61–73) and 65 (60–70) %, respectively (ns). There was no difference in LOS or complication rate. Delirium was associated to increasing age, increased length of stay and complications.

CONCLUSIONS

We observed no increase in the proportion of delirium- and coma-free days after introduction of a combination of early detection and standardized treatment with haloperidol on post-cardiotomy delirium. Most patients were not severely affected, and the few who were, proved difficult to treat, indicating that a simple treatment protocol with haloperidol was ineffective.

INTRODUCTION

Delirium is an acute disturbance of consciousness with a fluctuating course [1]. It occurs frequently, especially in elderly hospitalized patients and is associated with an increased risk of complications, increased length of stay (LOS), long-term cognitive impairment and mortality [2–4]. The reported incidence varies widely, depending on diagnostic criteria, detection method, study design and population. Patients in intensive care units (ICUs) have a high risk of developing delirium [5]. In a Danish study of ICU patients, an incidence rate of 40% was found, but up to 80% have also been reported [2, 6]. In general, it is assumed that delirium occurs in 1 of 5 hospitalized patients. Cardiac surgery patients have several risk factors for developing postoperative delirium, e.g. high age and male gender, and are exposed to precipitating factors such as surgery, ICU admission, anaesthesia and medications [7, 8]. Delirium is divided into three subtypes: hypoactive delirium (25%), the hyperactive form of delirium (30%) and the mixed type, which is reportedly the most common (45%).

According to the Diagnostic and Statistical Manual of Mental Disorders, 4th (DSM-IV), delirium has an organic cause, and is therefore a direct consequence of an altered general medical condition [1]. The pathophysiological background for delirium is, however, not well elucidated. It seems that harmful stimuli leading to acute brain dysfunction are often clinically expressed as delirium independently of the underlying cause, and inflammatory mediators, subclinical cerebral infarctions, generalized hypoperfusion and cellular hypoxia may all play a role [9]. Delirium may be the sentinel sign of global impaired oxygen delivery and prompt attention towards basic stabilization and maintenance of adequate oxygen supply is warranted. Prevention and detection have proved difficult. Various strategies for identification of patients at risk and prevention of postoperative delirium have been shown to reduce the incidence, but without a reduction of LOS [10]. Diagnostic tools are available for different patient populations [11]. The confusion assessment method for the intensive care unit (CAM-ICU) is well established, and had been in use at our ICU for some time. The Delirium Observation Screening (DOS) scale has been validated in a population of elective cardiac surgery patients, with a high specificity (97%) and sensitivity (100%) [12, 13]. It consists of 13 items, is used by nurses in wards and can be completed in a few minutes. It, therefore, seemed well suited to be introduced in our institution.

When detected, several treatment options of delirium, both pharmacological and non-pharmacological, exist. Combined approaches seem to be more effective, but also labour-demanding. Haloperidol has been used for decades in treating delirium; it has a well-described safety profile; however, use is mostly based on clinical experience. It is an antipsychotic drug selectively blocking the dopamine D2 receptor, thereby counteracting anticholinergic and dopaminergic hyperactivity in the brain [14]. The effect may therefore mostly be disease modifying, by controlling agitation. Numerous studies have addressed the effect of haloperidol and other antipsychotics in preventing and treating delirium in different patient populations. Prophylactic haloperidol was used to reduce the incidence and severity of delirium in ICU patients with a predicted risk of delirium ≥50% [15]; however, an earlier meta-analysis of three randomized controlled trials (RCTs) of ICU patients at risk of developing delirium concluded that the incidence rate did not decrease with the use of prophylactic antipsychotics (haloperidol versus ziprasidone versus placebo) [16]. A Cochrane study addressing the use of haloperidol compared with other antipsychotics in treating delirium, found no difference in effect, but severity was decreased [17]. In an RCT from 2010 testing haloperidol versus ziprasidone versus placebo for ICU patients with delirium, neither haloperidol nor ziprasidone decreased the duration of delirium compared with placebo [18]. A recently published study where haloperidol was tested against saline in ICU patients with delirium found no difference in delirium and coma free days (DCFDs), and the authors conclude that the use of haloperidol may only serve a purpose in patients with agitated delirium [19]. Morphine, however, is also effective in treating agitation in conjunction with delirium after cardiac surgery [20].

Taken together, the evidence supporting routine use of haloperidol for patients suffering from postoperative delirium, in wards as well as in intensive care units (ICUs), is equivocal. The bundle concept, i.e. the combination of early detection, elimination of precipitating/inducing factors and a protocol of medication may be beneficial. Few data can be found on occurrence of delirium from admission to discharge, including periods in the ICU. Since delirium is a condition that encompasses the entire admission course, this seems a relevant matter to address. We hypothesized that a protocol for early detection and standardized treatment with haloperidol for delirium after cardiac surgery would increase the proportion of DCFDs. The protocol was used both in the ICU and on cardiac surgical wards. The purpose of the study was to evaluate the effect of this screening and treatment protocol in a population of Danish cardiac surgery patients.

MATERIALS AND METHODS

This is a prospective interventional cohort study with a before/after design. We investigated the potential effect of a combined protocol consisting of early detection of delirium using a screening tool, a checklist to eliminate precipitating factors and standardized dosing with haloperidol on delirium after cardiac surgery. The study was conducted on the cardiac surgical ward and dedicated ICU in a large tertiary university hospital in Copenhagen, Denmark, where an annual number of 1500 patients undergo cardiac surgery. Adult patients scheduled for open cardiac surgery with use of cardiopulmonary bypass (CPB) were included in two periods.

Inclusion criteria

Patients ≥18 years scheduled for open cardiac surgery with use of CPB.

Exclusion criteria

Death or transfer to another department within 24 h after surgery, coma or heavy sedation throughout the admission.

In May 2012 (Period 1), the first cohort underwent cardiac surgery, and was screened for delirium with the DOS scale (Fig. 1) in wards and with the CAM-ICU if admitted to the ICU. The staff in wards were not aware of the cut-off values for a positive test of delirium, and diagnosing and treatment in this period was done according to ‘standard care’, i.e. decided by the attending physician (cardiothoracic surgeon on wards, cardiothoracic anaesthesiologist in the ICU). These data serve as a baseline for comparison. During the summer of 2012 a combined protocol consisting of early detection of delirium, a checklist to eliminate eliciting factors and standardized dosing of haloperidol was introduced to the staff. In wards, nurses and doctors were now educated in the interpretation of DOS scores, where values >2 indicates delirium. They were also informed of the rest of the protocol and how to use it. The staff at the ICU had been using the CAM-ICU prior to the study, and was introduced to the remainder of the protocol. In September 2012 (Period 2), the second cohort underwent cardiac surgery, and the staff were using the screening tools to identify delirious patients and the protocol to eliminate eliciting factors and support treatment decisions. Data from Period 2 were compared with baseline data.

Figure 1:

Nurses screened patients for delirium using the DOS scale. The DOS scale consists of 13 items. Any answer of ‘sometimes’ or ‘always’ yields 1 point. A total score of 3 or more points indicates delirium. Reproduced from Ann Thorac Surg. 2012; 93(3):705-11, with permission.

The primary outcome variable is the proportion of DCFDs, calculated as the number of DCFDs divided by the number of admission days for patients with delirium, reported as percentages. To be able to adjust for different incidence rates, we also report the proportion of DCFDs for all patients in the cohorts. We hypothesized that a protocol for systematic screening and treatment of delirium would increase the proportion of DCFDs. Sample size was calculated from an assumption of a 20% increase in DCFDs (from 50 to 70%), P = 0.05 and 1 − β = 0.20. Using MedCalc®, a sample size of 91 patients in each group was calculated. We planned to include 100 patients in each group. The study was stopped after inclusion of 117 and 123 patients in each period of 1 month. Secondary outcome variables were LOS (days) and all-cause complication rate (percentage of patients who suffered complications). All-cause 180-day mortality (numbers) is reported. Patient characteristics, complications, LOS, discharge status (home/other ward/dead), status at 180 days (alive/dead), prolonged stay in the ICU (>24 h) and use of medication were extracted from existing databases and clinical information systems.

The screening tool used every 8 h by nurses in wards was the DOS scale (Fig. 1). The ICU nurses used the CAM-ICU. Patients were screened from admission until discharge from the department of cardiac surgery, covering periods in wards as well as in the ICU. Scores were noted on separate charts that were collected daily by the research team. Elective patients were usually admitted the day before surgery, and a preoperative DOS score was obtained. If patients were admitted for acute surgery, a preoperative DOS score was obtained if possible. On the day of surgery (Day 0), patients were not evaluated (premedication and postoperative cognitive affection).

Surgical procedures included coronary artery bypass grafting (CABG), valve replacement and repair, procedures on the ascending aorta and the aortic arch, resection of tumours, operations for atrial fibrillation and combinations (Table 1).

Table 1:

Patient characteristics and clinical features before (Period 1) and after (Period 2) introduction of a protocol for early detection and standardized treatment with haloperidol for post-cardiotomy delirium.

	Period 1 (n = 117)	Period 2 (n = 123)	P-value
Age (years)	66 (24–85)	67 (22–88)	0.18
Male	91 (78%)	82 (67%)	0.30
EuroSCORE	5 (0–15)	6 (0–14)	0.13
Type of surgery
CABG	52 (44%)	50 (41%)	0.47
Aortic valve repair	32 (27%)	37 (30%)	0.48
CABG + valve	14 (12%)	22 (18%)	0.06
Mitral valve repair	9 (8%)	7 (5%)	0.41
Aneurysm	4 (3%)	2 (2%)	0.59
Misc.	7 (6%)	5 (4%)	0.45
Priority
Acute	11 (9%)	5 (4%)	0.06
Subacute	31 (27%)	24 (20%)	0.07
Elective	75 (64%)	94 (76%)	0.07
ICU >24 h	20 (17%)	19 (15%)	0.73
Discharge to
Home	77 (66%)	88 (72%)	0.24
Hospital	40 (34%)	32 (26%)	0.07
Dead	0 (0%)	3 (2%)	NA

Variables given as median (range) or number (%). Comparison was done with Student’s t-test for age, Mann–Whitney U-test for the rest of the numerical variables and χ² test for comparing frequency distributions. CABG: coronary artery bypass grafting; Aneurysm; Aneurism: surgery on ascending aorta and arch; Priority: elective surgical priority: procedure planned >72 h in advance, subacute priority: procedure planned 12–72 h in advance, acute procedure: planned <12 h ahead of surgery; ICU: intensive care unit; NA: not applicable.

Anaesthesia was induced with propofol, fentanyl and cisatracurium and maintained with sevoflurane and remifentanil. Cardiopulmonary bypass was performed using commercial equipment (regular heart–lung machine with standard hosing, reservoir, oxygenator and arterial filter size 40 µm). Haemodilution to a haematocrit of 22–24 was allowed before transfusion of red blood cells, flow was regulated to 2.4 l m⁻² min⁻¹+ 0–20% and perfusion pressure was kept above 45 mmHg with the use of phenylephrine or norepinephrine. Haemodynamic function after CPB weaning was monitored with a pulmonary artery catheter and/or left atrial catheter and ScVO₂ measurements. Inotropes, fluids and blood products were given according to department standards. Pain management consisted of 0.15 mg/kg morphine intravenous (IV) at the end of surgery and as needed postoperatively, and all patients received paracetamol 1 g times 4 per day orally. All patients were admitted to the cardiac surgery ICU, where extubation usually took place within a few hours. Observation continued until the next morning, where discharge to wards took place if predefined criteria were fulfilled.

In the cardiac surgical wards, screening for delirium with the DOS scale started when patients returned from the ICU, usually 24 h after surgery, and was repeated every 8 h until discharge from the department. If patients stayed at the ICU, or were readmitted, screening was performed with the Richmond Agitation and Sedation Scale and CAM-ICU [21]. In case of missing values or doubt as to whether a patient had delirium or not, a trained psychiatrist was consulted to evaluate the patient by examination if possible, or retrospectively by charts. ‘No delirium’ was assumed if DOS scores were consistently below 3 points, the CAM-ICU was negative, no delirious behaviour was noted in the charts and no antipsychotics had been administered. After collecting baseline data, the protocol consisting of early detection of delirium using a screening tool, a checklist to eliminate eliciting factors and standardized dosing of haloperidol was introduced to the nursing staff by a 30-min presentation on each section of the surgical wards and the ICU. Written information was sent to all doctors by e-mail. All material was made visible and available in each ward and furthermore as a part of the electronic guideline library of the hospital. Each day all wards were visited by the research team to assist in the use of screening tools and treatment protocol and to collect filled out charts.

The protocol consisted of a combination of (i) early detection with a screening tool (the DOS in wards and the CAM-ICU at the ICU); (ii) a decision support tool (the guideline) and (iii) suggestions on pharmacological treatment and a computer-based standard dosing tool for haloperidol. The guideline had a checklist for elimination of eliciting factors: surgical complications, affected vital parameters, medication, insomnia and suggestions of non-pharmacological supportive measures (re-establishment of normal sleep–wake cycle, reorientation, involvement of relatives, mobilization and use of extra staff). The protocolled treatment regimen with haloperidol was incorporated in an electronic medication module where a recommended dose, administration times, duration and 2-day tapering was predefined, according to age and weight. After approval by the physician, medication was administered by nurses, and doses (2.5–5 mg) of oral haloperidol was given three times a day for 1.5 days, then tapered over 2 days, with the morning doses taken away first. For patients with agitation and/or hallucinations IV haloperidol in small incremental doses (1.255 mg) was recommended every 15–30 min until the patient was calm. If patients reported severe anxiety, a benzodiazepine (diazepam IV or lorazepam orally) was offered. For insomnia, promethazine or a short-acting benzodiazepine could be added by demand.

Ethics

The study was evaluated and filed at the Committees on Biomedical Research Ethics for the Capital Region in Denmark, who decided that informed consent was not needed (protocol no. H-2-2011-166). Data were collected and handled according to national law and the project was approved by the Danish Data Protection Agency (journal no. 2012-41-0358). The study was filed at clinicaltrials.gov (protocol identifier NCT01774240).

Statistical analysis

Non-parametric desciptive statistics were used to present patient characteristics. When comparing groups, Student's t-test was use for the variable ‘age’, Mann-Whitney U-test was used for other numerical values and χ² test for comparing proportions. Logistic regression was used for identifying associations between delirium and predefined variables (age, gender, EuroSCORE, surgical priority [elective/subacute/acute], all-cause complications (yes/no), prolonged stay in the ICU (yes/no), LOS, discharge status (home/other). Elective surgical priority was defined as procedure planned >72 h in advance, subacute priority as planned 12–72 h in advance, acute as planned <12 h ahead of surgery. Prolonged stay in the ICU was defined as >24 h. Discharge status was registered as home/other wards/death. The following complications were registered: neurological (central and peripheral neurological damage, coma), cardiac (cardiac failure, tamponade, acute myocardial infarction and arrhythmia), pulmonary (pneumonia, pleural effusions, pneumothorax), reoperation, infections (sternal, graft site and sepsis) and sternal instability. Complication rate is reported as the number and percentage of patients with one or more complications. Age was analysed in three strata of approximately the same sample size: 18–60, 61–70 and 71–90 years. Simple logistic regression analysis was performed for each covariate individually (‘crude’) and a multiple logistic regression analysis was done (‘adjusted’). All tests were two-sided and a significance level of 5% was used. Data were analysed using SPSS (v. 19.0, Statistical Package for the Social Sciences Software, Inc. Chicago, IL, USA) and Graph Pad Prism (v. 6, GraphPad Software, Inc., La Jolla, CA, USA). The Department of Biostatistics, University of Copenhagen gave statistical assistance.

RESULTS

In May 2012 (Period 1, baseline), 123 patients underwent cardiac surgery, of which 117 were included, and in September 2012 (Period 2, intervention), 126 patients underwent cardiac surgery, of which 123 were included. A flowchart is shown in Fig. 2. Patient characteristics and clinical variables are given in Table 1.

Figure 2:

Flowchart with patients from the two observation periods. Period 1: May 2012 (baseline). Period 2: September 2012. Between periods a protocol for early detection and standardized treatment with haloperidol on post-cardiotomy delirium was introduced.

Regarding the primary endpoint, no significant change was observed in the proportion of DCFDs when comparing patients with delirium in the two periods. The observed LOS and all-cause complication rate were also unchanged between periods (Table 2). Likewise, we observed no significant difference in incidence, onset or duration of delirium between the two periods (Table 2). Fifty-one patients were identified as delirious (24 and 27, respectively), the majority being affected in wards (14 and 20 patients, respectively), 4 patients in each period were delirious only when in the ICU and 3 in each period were delirious both in wards and in the ICU. Three patients in Period 1 were classified as delirious retrospectively (Fig. 3). The median duration of delirium was 1 and 3 days, but with wide ranges (1–27 and 1–19 days, Periods 1 and 2, respectively).

Table 2:

Incidence, onset and duration of post-cardiotomy delirium before (Period 1) and after (Period 2) introduction of a protocol for early detection and standardized treatment with haloperidol on post-cardiotomy delirium

	Period 1 (n = 117)	Period 2 (n = 123)	P-value
Incidence (%)	21 (14–29)	22 (15–30)	0.66
Onset (POD)	1 (1–4)	1 (1–3)	1.0
Duration (days)	1 (1–4)	3 (1–5)	0.23
DCFDD (%)	67 (61–73)	65 (60–70)	0.41
DCFDA (%)	91 (89–93)	90 (88–92)	0.16
LOSD (days)	9 (7–12)	9 (8–12)	0.23
LOSA (days)	8 (7–8)	8 (7–8)	0.49
All-cause complications (%)	37 (28–46)	33 (24–42)	0.35
180-day mortality (%)	4 (3)	5 (4)	0.23

The proportion of DCFDs, LOS, all-cause complications and 180-day mortality were also calculated. Variables are given as a median (95% CI) or number (%). Comparison was done with the Mann–Whitney U-test for numerical variables and χ² test for comparing frequency distributions. Patients were screened by nurses with the DOS scale or the CAM-ICU. Treatment decisions were supported by a guideline and a computer-based standard dosing tool for haloperidol. POD: postoperative day; LOS_A: length of stay, all patients; DCFD_A: proportion of delirium- and coma-free days, all patients; LOS_D: length of stay, patients with delirium; DCFD_D: Proportion of delirium- and coma-free days for patients with delirium.

Figure 3:

An overview of patients tested positive for delirium, false positive/negative patients and patients who were evaluated by a psychiatrist. Patients were classified as having delirium if they had DOS Scale value >2 in wards [12], or had a positive CAM-ICU test [5] while in the ICU, or were assessed as delirious by a trained psychiatrist. The majority were affected in the wards (DOS positive), 4 patients in each period were only delirious in the ICU, and 3 in each period were delirious both in the ward and in the ICU. Three patients in Period 1 were classified as delirious retrospectively.

A total of 9 patients died within the observation period of 180 days. Of these, 4 cases were ‘operative’ (occurred within 30 days of the operation or had a clear correlation to surgery). Mortality data from 7 (of 240) patients are missing due to foreign citizenship.

Multiple regression analysis was performed (n = 240), where age, complications and discharge to other than home were significantly associated with delirium (Table 3). Compared with patients aged 18-60 years, the 71-90 year old patients had 4.4 times increased risk of developing delirium, and the 61-70 year old patients had a 3.8 times increased risk.. Patients with complications had a 3.4 times increased risk of delirium. Preoperative DOS scores were obtained in 203 of 240 patients (85%); 9 patients had a baseline DOS score >2, equally distributed between periods. Of these patients, 4 developed postoperative delirium. Numbers for false positive and negative patients are shown in Fig. 3.

Table 3:

Age, EuroSCORE, LOS, complication rates, discharge status (± home) and status at 180 days (± alive) for patients with and without delirium

	Never delirium (n = 189)	Delirium (n = 51)	P-value
Age (years)	64 (62–65)	72 (70–75)	0.03
EuroSCORE	5 (4.7–5.7)	7 (6.2–8.2)	0.48
LOS (days)	7 (7–8)	9 (8–10)	0.08
All-cause complications (%)	27 (21–34)	63 (48–76)	0.003
Discharge to home (%)	77 (71–83)	37 (24–52)	0.003
180-day mortality (%)	3.3 (1.2–7.0)	5.9 (1.2–16.2)	0.23

Values are given as a median (95% CI) or number (%). Regarding 180-day mortality: 3 of 51 patients with delirium died before 180 days, and 6 of 182 patients without delirium. P-values from the multiple logistic regression analysis are given.

In the first period, 13 of 24 delirious patients (54%) were agitated, defined as having points on item 10, 11 or 13 on the DOS scale or rated as agitated when evaluated by the CAM-ICU. Six of the 13 agitated patients in this first period were treated with haloperidol; only 3 of these regimens were categorized as adequate (regarding dosing, intervals and tapering). In the second period, 21 of 27 delirious patients (78%) were agitated, and 14 (67%) of these were treated with haloperidol. All treatments were recorded as adequate and in accordance with the protocol. The 20 patients treated with haloperidol were all agitated and/or suffering from unpleasant hallucinations. There were significantly more agitated patients in the second period (P < 0.001, χ² test). For both periods combined, the daily oral dose of haloperidol was 13 (8–19) mg, highest daily dose was 27 (14–40) mg and the total cumulative individual dose was 78 (24–132) mg (values are given as the mean value and 95% confidence interval (CI); IV doses were converted to oral doses by multiplying by 2). The mean daily dose of haloperidol was significantly higher in the first period [21 (5–50) vs 8 (5–12) mg, P < 0.05, Period 1 vs 2, Mann–Whitney U-test]. There were no reports of adverse reactions or events.

Patients with delirium were older, and had a higher EuroSCORE, an increased LOS and rate of complications compared with patients who never had delirium (Table 3).

DISCUSSION

This study shows that delirium after cardiac surgery in most cases has a benign course, where treatment protocol with small doses of haloperidol did not reduce the duration of delirium. A small subset of patients (the oldest and cerebrally vulnerable) suffered from extended periods of delirium, refractory to a simple multimodal treatment regimen with environmental adjustments and haloperidol. Haloperidol was exclusively used to control agitation, and the protocolled prescription regimen reduced the administered dose by two-thirds.

A cohort study in a before/after design carries a risk of bias, since confounding variables may be unevenly distributed in the two groups. The two cohorts are separated by 3 months, and there may be undetected time-related differences between them. In the first period, nurses in wards collected DOS scores without knowing the cut-off value for delirium, to enable calculation of baseline values for incidence, duration and DCFDs. However, awareness of delirium may have been affected which could unintentionally influence treatment and care before the protocol was put to use. Furthermore, delirious patients in Period 2 may not have received the intended standardized treatment. The protocol was introduced and use was supported on a daily basis, but compliance was not evaluated. Moreover, the pharmacological intervention was pragmatic, since the decision to treat was left to the attending physician. The inclusion and exclusion criteria were pragmatic, and the cohorts represent almost all eligible patients in the two periods. All patients included are accounted for and assessed for the primary outcome. The observation period covered periods in cardiac surgical wards as well as in the ICU, thus offering valuable information about debut and duration of delirium, even in case of transfer between units. The primary outcome variable, DCFDs might be a relevant patient-related outcome, since increased duration of delirium has been found to be associated with increased 1-year mortality [22]. The observed 95% CIs of the primary outcome variable found in our study indicate an acceptable precision of measurement (Table 2). The study was not blinded, and one member of the study group supported the staff in performing DOS and CAM-ICU evaluations. Administrative staff without relation to the study group performed blinded data extraction regarding LOS, complications and mortality. The DOS scale was adopted well by nursing staff in the cardiac surgical wards, generating relevant clinical information without excessive extra workload. False-positive DOS scores were recorded from patients suffering from insomnia and/or postoperative pain, where medication resulted in unintended sedation or fatigue. Elderly patients with pre-existing cerebral and cognitive disturbances could be fully oriented when evaluated by a psychiatrist, even though the DOS score was >2. If communication was difficult or impossible, nonsense DOS scores >2 were observed.

There was no significant increase in DCFDs after introduction of early detection and protocolled treatment, and the periods were remarkably similar, with 67 (61–73)% DCFD in Period 1 vs 65 (60–70)% in Period 2. Given the observed baseline event rate and sample size, an 18% increase in DCFDs could be detected with a power of 90%. We acknowledge the fact that there is a potential risk of uncontrolled confounding bias in this study due to the fact that groups are not randomly generated, potentially refuting the 0-hypothesis regarding the primary endpoint (proportion of DCFDs). We have done a multivariate logistic regression to identify factors associated with delirium, where age, LOS and discharge to other than home came out significant (Table 3). These factors are statistically equally distributed in the two groups (Periods 1 and 2), as given in Table 1, where comparison is done for each variable. There are of course other covariates of importance (smoking status, alcohol consumption, pre-existing cognitive dysfunction, drug use, blood product transfusions, CPB run times, blood pressure levels, acid/base status), but we did not record these parameters. Even if these covariates were unevenly distributed, it is very unlikely that the point estimates of the primary endpoint (proportion of DCFDs) would separate significantly, since the 95% CIs are found to be almost completely overlapping. The potential contribution of confounding factors would have to be very substantial to generate a different result (refute the 0-hypothesis), when comparing the primary endpoint.

The observed incidence of delirium was 21 (14–29) and 22 (15–30) % (Periods 1 and 2, respectively, reported as the mean and 95% CI). Similar incidence rates were found in a Dutch study using the DOS scale, albeit the population being more selected (elective cardiac surgery patients) [12]. A recent study of cardiac surgery ICU patients found an incidence rate of delirium of 24% [23]. Onset of delirium was early after surgery, when the patients had returned from the ICU to the wards, and there were no difference between Periods 1 and 2. If patients in wards had not developed delirium within the fifth postoperative day (POD), it was very unlikely to occur. Patients with a prolonged ICU stay seemed to be continuously at risk of delirium since it occurred as late as 17 days after surgery. Others have reported that delirium started on the second POD (±1.6 days) after cardiac surgery, and on the fifth POD at the latest [12]. In a study of medical ICU patients, the observed onset of delirium was on the second (±1.7) day after admission [5]. Delirium therefore seems to occur early after the ‘physiological insult’, but with unpredictable onset if patients stay or are readmitted to the ICU. Only few studies report duration of delirium, maybe due to limited observation periods. In our study, most patients were only affected for a few days but some were affected during most of their admission course. In a Dutch study, the mean duration of delirium was 2.5 days (SD 2.1), with a maximum of 8 days [12], but periods in the ICU was not included. In a study of elderly patients in a general ICU, a median duration of 4 days and a range 1–46 days were found [22], reflecting a heterogenic population much more at risk. In our study, the most severely affected patients were delirious almost throughout their stay, and they had a prolonged admission course due to an increased rate of complications, reoperations, readmissions to the ICU, need for extra staff and increased use of antipsychotics and sedation. We found no significant association between LOS and delirium, after adjusting for other covariates, in contrast to other studies [2, 3]. One-third of the patients in our study were not discharged home, and patients with delirium had a higher risk than others. The recorded LOS for these patients is underestimated, possibly masking a correlation. We observed an association between the occurrence of complications and delirium, as previous studies have, but we did not record time-to-event relations, and therefore the causality remains unclear.

There was no difference in 180-day mortality between periods, nor did we observe a statistically significant mortality difference when comparing patients with and without delirium. Differences in 6-month mortality have previously been reported, with a 3.2 times increased risk of death for patients with delirium, after adjusting for relevant covariates [2]. These patients were mechanically ventilated medical ICU patients, with a higher background mortality than our population, but even in another study of cardiac surgery patients, the 6-month mortality was significantly higher in patients who had delirium (Odds Ratio 5.1; P = 0.011) [24]. First, our study was not powered for this outcome variable; secondly, survival curves do not seem to separate significantly before 180 days.

Use of haloperidol for delirious patients seemed more frequent and consistent in Period 2 (15 of 27 patients treated (56%) and all treatments regarded ‘adequate’) compared with the baseline period (6 of 24 patients (25%) treated and only half regarded ‘adequate’). There were more agitated patients (78 vs 54%) in Period 2, and a larger proportion of these were treated with haloperidol (67 vs 46%), indicating a certain effect of the protocol and dosing tool. The median individual dose, however, was significantly lower in the second period (8 (5–12) vs 21 (5–50) mg, P < 0.05, Period 2 vs 1, Mann–Whitney U-test). This probably reflects a shift in the way haloperidol was used; instead of sporadic nocturnal IV dosing, it was given in small oral doses over 2–3 days and then tapered. Haloperidol was exclusively used to treat agitation and/or hallucinations; all patients treated with haloperidol were evaluated as agitated and/or suffering from severe hallucinations. Not all agitated patients in the first period were treated with haloperidol, or other drugs targeting agitation, and only 3 of the 6 patients treated, received an adequate regimen. After the intervention, 14 of 21 agitated patients were treated with haloperidol, and all received an adequate regimen. Since we did not ask patients how they felt before and after treatment with haloperidol, there may have been unrecorded beneficial or unpleasant subjective effects.

We did not observe any effect on DCFDs of a simple multimodal treatment regimen with environmental adjustments and haloperidol, which may simply be ineffective in the treatment of post-cardiotomy delirium, although it would require a blinded RCT to determine this properly. The possible pharmacological effect of haloperidol on delirium is probably not curative, but rather disease modifying. This may be observed as a shortening of the agitated periods or a subjective amelioration of unpleasant cognitive disturbances. Since the DOS scale is not designed for quantification of severity of delirium, only the duration of delirium was recorded. We did not record subjective effects and may therefore have overlooked other effects of haloperidol. Our treatment regimen seemed inadequate in cases of prolonged delirium, and less affected patients apparently did well without or with small doses of haloperidol. The mean individual dose of haloperidol was reduced by 13 mg (62%) without an increase in the duration of delirium, and so a dose-dependent effect seems unlikely. Agitation was more frequent in the second period and these agitated patients were more often treated with haloperidol, where dosing was in consistence with the guideline. The mean individual daily dose was lower in the second period, and this may have contributed to more agitation; alternatively, agitated patients may have been treated with morphine or non-pharmacological measures. Either way, this did not seem to influence any of the variables recorded. The present design does not allow further analysis of possible effects in the subgroup of patients with agitation; however, these patients are of particular interest, since it may be the only subgroup of delirious patients that can benefit from treatment with haloperidol. When selectively evaluating haloperidol for agitated delirium, assessment of the subjective cognitive effect is important, especially since morphine may be an alternative treatment option. All charts of haloperidol-treated patients were evaluated for adverse effects or events, and none were observed or reported, apart from (intended) sedation. Others have reported adverse effects or events in 8% [18] to 30% [15] of cases, depending on the population studied.

CONCLUSION

We found no increase in the proportion of DCFDs (67 vs 65%, respectively) after introducing a protocol of early detection and standardized treatment with haloperidol on post-cardiotomy delirium, and it is possible that haloperidol is not effective for reducing the duration of delirium. The incidence rate of post-cardiotomy delirium was 21–22% and it occurred early after surgery with a short duration in most patients. Some patients were severely affected, and difficult to treat satisfactorily. Patients with delirium were older; they had longer LOS, developed complications more frequently and were less often discharged to their home.

Funding

The Heart Center Research Fund, Rigshospitalet, Copenhagen.

Conflict of interest: none declared.