The challenge of managing severe delirium: an observational study of the effects of pharmacological and supportive care on delirium duration and resolution in 602 patients

Background: Delirium is the most common neuropsychiatric disorder seen in hospitalised patients. Current guidelines recommend only using antipsychotics with distressed patients. Nonetheless, in routine clinical practice, multiple psychotropics are commonly administered. More evidence on the short-term benefits of various management approaches in patients with delirium is required. Methods: In this observational cohort study, 602 delirious patients were followed for twenty days. Supportive care was provided to all patients in addition to either no psychotropic therapy, monotherapy, dual therapy, or polytherapy defined as three or more psychotropic drugs. The effectiveness of interventions regarding delirium resolution and symptom severity was determined by Cox proportional hazards regression and generalized estimating equation models. Results: Psychotropics were commonly used to manage delirium. In total, 12.1% of patients received polytherapy, 37.2% dual therapy, 37.7% monotherapy, and 12.1% supportive care alone (i.e., almost half of the patients received ≥ two psychotropics). Patients who received polytherapy had higher initial baseline delirium severity and exposed the mixed subtype more often; with the latter delirium lasted longer and recovery was less frequent than in mild delirium. Providing supportive care alone in mild delirium was superior to all psychotropic approaches. Conclusions: In routine clinical practice, the use of multiple psychotropics is common. In our study, however, despite combined supportive management and polypharmacy, patients with severe delirium suffer longer from delirium and have lower resolution rates. For the management of patients with mild delirium

alterations; but, ultimately, they merge into a common pathway of system integration failure [2].
The management of delirium consists of both non-pharmacological and pharmacological strategies to prevent or, once evident, ameliorate the symptoms of delirium. The first priority in delirium management is to eliminate underlying aetiologies. Non-pharmacological strategies include providing sufficient pain management and adequate fluid intake, striving to preserve or restore normal circadian rhythms, ensuring a calm environment and familiar caregivers, and enacting early patient activation and mobilisation. Pharmacological strategies include psychotropic drugs Administration of haloperidol has long been considered the primary option to treat symptoms of delirium [12]. Current clinical guidelines from the National Institute for Health and Care Excellence (NICE) recommend short-term haloperidol or olanzapine.
Alternatively, the cautious use of antipsychotic drugs is recommended in patients with severe distress and those patients at risk to harm themselves or their surroundings after non-pharmacological interventions were not effective or not applicable [13]. This approach is supported by the results of recent randomized controlled trials [14,15]. Despite this evidence, the administration of multiple psychotropics remains common practice [16].
To date, the effects of concomitant or sequential use of multiple psychotropics to manage delirium are unknown. Therefore, the aim of the current study was to evaluate the benefits of various delirium management approaches in routine clinical practice. We hypothesised that patients show no clinically relevant benefit from multiple-drug regimens. We also compared the use of psychotropics versus supportive care alone.

Patients
All patients in this study were prospectively enrolled at the University Hospital of Zurich, Switzerland, a tertiary care centre, from regular units across all departments between July 2012 and June 2015. The results are presented in accordance with the STROBE statement [17].
Inclusion criteria were being a consenting adult, having an episode of delirium, and having Delirium Observation Screening scale scores, DOS, over 72 hours or longer [1,18]. Exclusion criteria were age 18 or younger, short-term hospitalisation (< 24 hours), and missing DOS scores.
The study was approved by the Canton of Zurich Ethics Committee (PB2016-01264).

Procedures
All enrolled patients were assessed three times per day with the Delirium Observation Screening scale over a period of twenty days or until discharge from the hospital. Each of the 13 DOS items, reflecting DSM-IV criteria (American Psychiatric Association, 1994), is rated as either 0 (never) or 1 (occasional to always). For our purposes, delirium was defined as either a single score of three and more, or a 24-hour summation score of seven and more. Upon suspicion of incident delirium, the short version of the Confusion Assessment Method (CAM), with its four items reflecting DSM-III-R criteria (American Psychiatric Association, 1987), was used to confirm the diagnosis. Subtyping of delirium was based on the patient's psychomotor presentation, into hypoactive, hyperactive or mixed subtype, as reflected by DOS items 10, 11 and 13 [19]. Delirium resolution was defined as a DOS-score < 3 points in three assessments over 24-hours.
All patients were managed according to an algorithm already established at the University Hospital Zurich (Appendix 1). Non-pharmacological care basically followed the long-standing concept of reality orientation for patients with cognitive disorders [20]. Additionally, supportive means (e.g. adequate pain management, light-management for the restoration of sleep-wake cycle) were provided by constant caregivers.

Statistical methods
Analyses were performed using the Statistical Package for the Social Sciences, version 25 (SPSS, v. 25) and R version 3.5.3 and the "survival" package (R Foundation for Statistical Computing, Vienna, Austria).
Baseline characteristics were described depending on their parametric properties, either as medians and inter-quartile ranges, or as means and standard deviations.
Subjects were categorised into the four pre-defined management approaches: (1) non-pharmacological therapy/supportive care alone; (2) one psychotropic drug (monotherapy); (3) two psychotropic drugs (dual therapy); or (4) three or more psychotropic drugs (polytherapy). Only psychotropics deemed relevant to delirium management were included. Those deemed irrelevant -like antidepressants, mood stabilizers and stimulants -were excluded.
Psychotropic medications were categorised into either: (a) antipsychotics, including haloperidol, pipamperone, risperidone, olanzapine, and quetiapine; or (b) benzodiazepines, including lorazepam and midazolam. Mean daily dosages were determined by dividing the cumulative dose administered by the total number of days administered. Doses considered clinically relevant were: ≥1 mg haloperidol; ≥20 mg pipamperone; ≥1 mg risperidone; ≥2.5 mg olanzapine; ≥25 mg quetiapine; ≥1 mg lorazepam; and ≥ 7.5 mg midazolam. Medications dosed below these ranges were deemed irrelevant and omitted from further analysis. 'delirium resolution' as the status variable (y axis). Proportional hazards (PH-) assumption was tested with Schoenfeld residuals. Schoenfeld residuals approve the PH-assumption for management groups (ρ = -0.0795, P = 0.088) in the presence of baseline delirium severity (DOS-score day 1) as a covariate.
With respect to the data distribution, i.e. higher baseline severity in polytherapy compared to supportive care only and monotherapy, to further investigate the relation between delirium severity and management, a generalized-estimatingequations (GEE) model was implemented with gamma-distribution and log-link. A within-subject effect delirium score per day was set, and baseline DOS-score introduced as split-half group factor. This means, that DOS-scores of all patients were dichotomized, for this analysis, according to their baseline delirium severity, and categorized into the higher-or lower-severity group (split-half). Model effects were tested with χ2 analysis. The day of delirium resolution was tested with regards to baseline severity irrespective of the management group (DOS-score day 1 * Days of observation) and for the management groups (DOS-score day 1 * Days of observation * Management groups). For missing values, the 'last observation carried forward' method was applied. For all inferential testing, statistical significance was defined as P < 0.05.

Sociodemographic and clinical characteristics
A total of 602 delirious patients were identified who met inclusion criteria and were sub-categorized into the four management approaches: supportive care alone (n = 78; 13.0%); monotherapy (n = 227; 37.7%); dual therapy (n = 224; 37.2%); polytherapy (n = 73; 12.1%). Mean ages for the four treatment groups were similar, ranging from 69.4 to 72.1 years (P = 0.24). In all groups, subjects were predominantly male, the percentage ranging between 57 and 70%. The only statistically significant between-group difference in gender distribution was comparing those on dual therapy (70%) versus monotherapy or supportive care alone (each 57%, P = 0.033 and P = 0.006, respectively). Sociodemographic and delirium characteristics are summarised in Table 1. Table 1 Demographics, baseline clinical characteristics and outcome measures, *mean (standard deviation (SD), range), **Days to resolution of delirium was estimated in split-half groups and not in whole groups, due to distribution of delirium severity. In terms of reasons for admission, neurological and medical disorders were most common, accounting for almost three-quarters of admissions, followed by oncological diseases (15%) and trauma (7%). Psychiatric disorders were present in one fifth of patients: dementia in 7%, an affective disorder in 5%, a substance use disorder in 5%, and psychosis in 2.5%. All psychiatric comorbidities occurred equally across the four management groups (P = 0.43).
Prior to admission, 8.6% of patients had received at least one antidepressant (serotonin-reuptake inhibitor, tricyclic antidepressant or atypical antipsychotic).
Transfers to inpatient psychiatry and mortality in non-pharmacological management versus monotherapy, dual therapy and polytherapy were without statistically significant differences (Table 1). Patients with high baseline delirium severity were prone to receive more medications. Furthermore, fewer patients with high baseline delirium severity were diagnosed with hypoactive subtype (Table 1). Within the first day of management, patients on polytherapy were more delirious (mean DOS score 17.7) than those managed with support only (12.1; P = 0.002), and those on single psychotropics (14.4; P = 0.027). Comparing those on poly-and dual therapy, baseline delirium severity was the same (P = 0.24). In patients on dual therapy, delirium was more severe than among those receiving supportive care alone (P < 0.001) or monotherapy (P = 0.01). Furthermore, patients on monotherapy were more delirious than those receiving supportive care alone (P = 0.003).
There were some differences in the prevalence of delirium subtypes between management approaches: the mixed subtype was more common in patients on two or more medications versus either monotherapy (P = 0.02) or supportive care alone (P = 0.007). In those patients on dual-versus poly-therapy no differences were noted in the prevalence of the mixed subtype (P = 0.35). There were no group differences in the prevalence of hyper-and hypoactive subtypes.   Table 2 Psychotropic drugs given. IQR: interquartile range.

Psychopharmacological characteristics: doses and treatment duration
For patients on dual therapy, the doses of lorazepam, haloperidol and pipamperone were higher than among those on monotherapy. Conversely, no dose-differences were noted between those on dual-and polytherapy.
With respect to haloperidol dosing, patients on dual-versus monotherapy (median: 6 vs 2 mg; P = 0.003) and on poly-versus mono-therapy (median: 9 vs 2 mg; P < 0.001) received more haloperidol. For those on dual-and polytherapy, the doses of haloperidol were the same (P = 0.86).
Higher doses of lorazepam were given in dual-(median: 4.5 mg; P = 0.003) and poly-therapy patients (median: 6 mg; P < 0.001) than in those on monotherapy (median: 2 mg). There was no further increase between those on dual-and polytherapy (P = 0.86).
In terms of treatment duration, there were no statistically significant differences for haloperidol (P = 0.206), pipamperone (P = 0.616) and lorazepam (P = 0.12) between the three drug-management approaches.
Irrespective of the multiple regimens used, doses for risperidone (median: 7.5- Delirium resolution and severity over the course of management In a Cox regression model, all three psychopharmacological approaches appeared inferior to supportive care only, which was associated with a delirium resolution rate of 92.5%. Notably, polytherapy achieved the lowest delirium resolution rate (60%), followed by dual (80%), and monotherapy (88.5%).
In the GEE model, irrespective of management approach, patients with milder baseline delirium did not show a relevant decrease in DOS scores on day 2, compared those with severe baseline delirium (X 2 = 54.620, df = 5, P < 0.001).
Further, patients with severe delirium recovered more rapidly on supportive care alone than with any of the other management approaches (X 2 = 43.995, df = 15, P < 0.001).
For patients on psychotropic medications, recovery differed between those less versus more severely delirious (P = 0.047). In patients with mild delirium and on monotherapy, the initial decrease in delirium severity occurred on day #6 (P < 0.001). In those less delirious on polytherapy, delirium peaked on day 2 (P = 0.039) and subsequently improved. In patients with more severe delirium, more rapid recovery was noted with the initiation of management, irrespective of the approach (P < 0.001) (Fig. 2).  Table 2 Psychotropic drugs given. IQR: interquartile range.

Administration of psychotropics: antipsychotics and benzodiazepines
The psychotropics most commonly used in monotherapy were pipamperone (48.

Psychopharmacological characteristics: doses and treatment duration
For patients on dual therapy, the doses of lorazepam, haloperidol and pipamperone were higher than among those on monotherapy. Conversely, no dose-differences were noted between those on dual-and polytherapy.
With respect to haloperidol dosing, patients on dual-versus monotherapy (median: 6 vs 2 mg; P = 0.003) and on poly-versus mono-therapy (median: 9 vs 2 mg; P < 0.001) received more haloperidol. For those on dual-and polytherapy, the doses of haloperidol were the same (P = 0.86).
Higher doses of lorazepam were given in dual-(median: 4.5 mg; P = 0.003) and poly-therapy patients (median: 6 mg; P < 0.001) than in those on monotherapy (median: 2 mg). There was no further increase between those on dual-and polytherapy (P = 0.86).
In terms of treatment duration, there were no statistically significant differences for haloperidol (P = 0.206), pipamperone (P = 0.616) and lorazepam (P = 0.12) between the three drug-management approaches.

Delirium resolution and severity over the course of management
In a Cox regression model, all three psychopharmacological approaches appeared inferior to supportive care only, which was associated with a delirium resolution rate of 92.5%. Notably, polytherapy achieved the lowest delirium resolution rate (60%), followed by dual (80%), and monotherapy (88.5%).
In the GEE model, irrespective of management approach, patients with milder baseline delirium did not show a relevant decrease in DOS scores on day 2, compared those with severe baseline delirium (X 2 = 54.620, df = 5, P < 0.001).
Further, patients with severe delirium recovered more rapidly on supportive care alone than with any of the other management approaches (X 2 = 43.995, df = 15, P < 0.001).
For patients on psychotropic medications, recovery differed between those less versus more severely delirious (P = 0.047). In patients with mild delirium and on monotherapy, the initial decrease in delirium severity occurred on day #6 (P < 0.001). In those less delirious on polytherapy, delirium peaked on day 2 (P = 0.039) and subsequently improved. In patients with more severe delirium, more rapid recovery was noted with the initiation of management, irrespective of the approach (P < 0.001) (Fig. 2).

Summary of main findings
In this observational cohort study involving 602 patients, we evaluated the role of psychotropic medications combined with supportive care versus supportive care alone in the management of delirium, and specifically compared one-, two-and poly-drug regimens. As part of routine clinical practice at our hospital, multiple regimens involving psychotropic medications were used. In total, 73 patients were managed with polytherapy, 224 with dual therapy, 227 with monotherapy, and 78 patients with supportive care alone, meaning that roughly seven out of eight patients received pharmacotherapy, whereas almost one in two received two or more psychotropic drugs.
Patients who were managed with two or more psychotropics in addition to supportive care had higher baseline delirium severity (mean DOS scores: 16.2-17.7) and they exposed the mixed subtype more often than those with lower baseline delirium severity (mean DOS score: 12.1). Patients who were managed with supportive care alone showed the best course of delirium with faster resolution than patients on pharmacological management. However, those patients had lower initial delirium severity. This indicates that mild delirium was manageable with supportive care alone. Despite combined supportive care and pharmacological management, initially severe delirious patients recovered slower and resolution rates were lower.
The pharmacological management groups differed significantly in terms of drug administration, in particular those on polytherapy. Relative to those given a single psychotropic medication, patients on polytherapy had 17.2 times the odds of receiving haloperidol, 17.2 times the odds of receiving at least one benzodiazepine, and 12.1 times the odds to receive pipamperone. Conversely, doses of lorazepam, haloperidol and pipamperone were not different between dual-therapy and polytherapy indicating that rather than adjusting the dose, further psychotropics were added.
Comparing our findings against existing guidelines and the scientific literature The foremost step in the management of delirium is preemptive management: preventing delirium from starting in the first place [6]. Once delirium becomes manifest, identifying and eliminating as many underlying aetiologies as possible is required, and both non-pharmacological and pharmacological strategies may be employed.
Past American Psychiatric Association [21] guidelines for the treatment of delirium from 1999 recommended haloperidol as the mainstay of delirium management [12].
Conversely, the NICE guideline -published in 2010 -underlines the application of non-pharmacological management and recommends the judicious use of psychotropics for distressed patients only [13]. Although the benefits and disadvantages of these two approaches continue to be debated [22], the NICE guideline concurs with the importance of non-pharmacological, environmental, and supportive interventions -like providing a calm environment, regularly reorienting the patient, managing metabolic imbalances, reducing mind-altering medications like opioids, and providing visual and hearing aids, when necessary.
The management algorithm we used for the current study followed the APA guidelines, favouring the use of psychotropics for patients with delirium (Appendix 1), and emphasizing the use of haloperidol, pipamperone, and benzodiazepines.
Our findings support those of a pivotal study which evaluated haloperidol and risperidone versus placebo in palliative care patients [15]. In that study, neither psychopharmacological management approach matched the effectiveness of supportive care alone. The investigators concluded that placebo was superior to either haloperidol or risperidone and that antipsychotics potentially increased mortality.
The liberal administration of psychotropics in patients with delirium was recently challenged in a systematic review that incorporated 19 randomized controlled trials [23]. However, authors of another earlier systematic review of 15 randomized controlled trials came to the opposing conclusion that antipsychotics were superior to non-pharmacological management, with antipsychotics achieving higher resolution rates and decreasing overall delirium severity [24]. Patients in these two reviews were quite comparable with respect to age, amount of haloperidol used and mortality rates.
The management approach used in our study partly reflected those from a recent multicentre postal survey evaluating delirium management strategies [25]. olanzapine, risperidone and quetiapine). The use of haloperidol was more common in these two afore-mentioned studies than in our own (> 80% versus 41.8%).
Although the current study sheds some light on delirium management, several questions remain unanswered. In our cohort, patients on polytherapy were more delirious at baseline and delirium severity increased towards the second day of management. One plausible explanation is that polytherapy was initiated due to these patients' more severe delirium, which alone could explain their worsening over the first 24-48 hours. On the other hand, it also is possible that the increase in delirium severity patients experienced was actually caused by their polytherapy. It also is conceivable that the worsening was caused by some other factor altogether.
Unfortunately, which of these explanations is correct cannot be determined by our data.

Study strengths and limitations
This observational study had several strengths. In total, 602 inpatients were included, which makes it, to our knowledge, the largest delirium cohort study to investigate the effects of psychotropic polypharmacy on delirium resolution in a published study to date. Patients also had a broad spectrum of medical diagnoses and were recruited from diverse clinical settings. Their age and gender distribution, as well as the nature of their psychotropic management, were comparable to prior studies. Obtaining delirium scores thrice daily captured fluctuations in delirium and allowed for the identification of delirium subtypes. Recent reports also have highlighted the need for longer observation times in delirium trials [27], and our patients were followed for up to twenty days.
Despite these study strengths, several limitations must be noted. First, due to the study's design, patients were not enrolled until the delirium manifested. That is why there was no basal cognitive assessment and no screening for dementia. According to the evaluation of patient records, there was no difference in the prevalence of dementia between groups. The actual extent of mild to moderate cognitive impairment in this cohort might be underestimated, however. The presence of cognitive impairment is of importance for both the course of delirium and the effect of psychotropic medication.
Then, pipamperone is rarely used outside Germany and Switzerland and may not be relevant for some audiences. It is low-potency antipsychotic, an antagonist of 5HT 2A − C -, D 2 − 4 -and α 1− 2 -receptors with much higher activity against the D 4 and 5HT 2A than D 2 receptor and insignificant activity against histamine 1 -and muscarinicanticholinergic receptors [28].
Even though benzodiazepines were administered in a considerable number of patients in our cohort, it is usually not considered a first line treatment in the pharmacological management of delirium. On the contrary, benzodiazepine administration might be a significant risk factor for delirium. Therefore, benzodiazepine administration might have confounded antipsychotic treatment and even worsen delirium.
Most importantly, since our patients were not randomised to their treatment groups, it may be that group allocation bias was an issue, with those in the polytherapy group destined to do worse primarily on the basis of their delirium severity, rather than their treatment being less effective. Similarly, it may be that those who were treated non-pharmacologically were so for various reasons, including the treatment team's expectations of patient recovery. One finding against this latter argument is that our patients with severe baseline delirium improved significantly more with non-drug than drug therapy. Why this is so clearly warrants further study. In the end, further randomised controlled clinical trials assessing delirium management are required to clarify the role of psychotropic and non-psychotropic therapies in delirium management.

Conclusions
The management of patients with delirium is challenging, particularly of those patients with severe delirium. This study provides further information on delirium management approaches and appears to favour supportive care alone in mild delirium, as well as the judicious, rather than routine, use of psychotropics.
Certainly, patients in severe distress require psychotropics. In our cohort, physicians tend to add new substance classes instead of increasing the dosages of those drugs already administered; this strategy did not improve delirium management in terms of resolution rates and duration. In some instances, optimising the dose of the psychotropic drugs already being used, rather than adding new ones, might be prudent in seemingly treatment-resistant patients. Our findings indicate that patients with severe delirium are prone to receive more psychotropic medication and that those patients, despite supportive and pharmacological management, suffer longer from delirium. Therefore, strategies for patients with severe delirium are highly requested. Randomised clinical trials for the investigation of no-drug, single-drug, dual-drug and poly-drug therapy remain necessary.  Generalized estimating equation (GEE) models for the four delirium-management regimens. T

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download. Delir_Polypharmacy_APPENDIX1_FINAL.docx