The effect of non-pharmacologic strategies on prevention or management of intensive care unit delirium: a systematic review

Background: Post-operative delirium is a common complication among adult patients in the intensive care unit. Current literature does not support the use of pharmacologic measures to manage this condition, and several studies explore the potential for the use of non-pharmacologic methods such as early mobility plans or environmental modifications. The aim of this systematic review is to examine and report on recently available literature evaluating the relationship between non-pharmacologic management strategies and the reduction of delirium in the intensive care unit. Methods: Six major research databases were systematically searched for articles analyzing the efficacy of non-pharmacologic delirium interventions in the past five years. Search results were restricted to adult human patients aged 18 years or older in the intensive care unit setting, excluding terminally ill subjects and withdrawal-related delirium. Following title, abstract, and full text review, 27 articles fulfilled the inclusion criteria and are included in this report. Results: The 27 reviewed articles consist of 12 interventions with a single-component investigational approach, and 15 with multi-component bundled protocols. Delirium incidence was the most commonly assessed outcome followed by duration. Family visitation was the most effective individual intervention while mobility interventions were the least effective. Two of the three family studies significantly reduced delirium incidence, while one in five mobility studies did the same. Multi-component bundle approaches were the most effective of all; of the reviewed studies, eight of 11 bundles significantly improved delirium incidence and seven of eight bundles decreased the duration of delirium. Conclusions: Multi-component, bundled interventions were more effective at managing intensive care unit delirium than those utilizing an approach with a single interventional element. Although better management of this condition suggests a decrease in resource burden and improvement in patient outcomes, comparative research should be performed to identify the importance of specific bundle elements.


Introduction
Delirium is a multifactorial, acute, confusional state characterized by the disturbance of consciousness and cognition; it is particularly common in the intensive care unit (ICU) with incidence ranging from 19 to 87% with especially high rates in mechanically ventilated patients [1][2][3] . ICU delirium is associated with adverse outcomes including increased mortality, prolonged mechanical ventilation and hospitalization, increased risk of cognitive dysfunction after discharge, and increased cost of care [4][5][6][7] .
While the pathophysiology of delirium is not well understood, there are multiple factors associated with an increased risk for developing delirium including age, neurologic or psychological disorders, polypharmacy, medications, and sensory impairment [8][9][10][11] . Modifiable environmental risk factors including immobilization, use of restraints, isolation, and levels of environmental light and sound are also considered risk factors for the development of delirium in the ICU 8,12 .
The morbidity associated with delirium as well as the multitude of delirium risk factors present in the ICU make delirium prevention and management strategies essential. These strategies have included pharmacological, non-pharmacological, and multicomponent interventions with the aim of decreasing the incidence and duration of delirium. Research into pharmacological interventions has focused on haloperidol and dexmedetomidine, though there has also been limited research into the effects of ramelteon, melatonin, and ziprasidone [13][14][15][16] . Despite continued research, current literature does not support the use of anti-psychotic agents, benzodiazepines, or melatonin in the management of delirium 13, 17 .
Given the lack of evidence supporting pharmacological measures, further research into the efficacy of non-pharmacologic interventions such as early mobilization, environmental modifications, or management bundles is crucial. The implementation of effective delirium management shows promise in decreasing morbidity, mortality, length of stay, and resource burden in the ICU setting. In terms of the PICOS framework (Population, Interventions, Comparisons, Outcomes, Study Design) 18 , our systematic review aims to address the effects of any non-pharmacologic prevention or management strategy on the incidence, prevalence, duration, or severity of delirium in critically ill adult patients compared to control patients, with no restrictions on study design.

Search strategy and data extraction
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed in this review and included as Reporting guidelines 19,20 . The electronic databases of PubMed, Embase, Cochrane Central, Web of Science, The Cumulative Index to Nursing and Allied Health Literature (CINAHL), and ClinicalTrials.gov were systematically searched on May 15, 2019 for articles concerning non-pharmacologic treatments for delirium in the ICU. Search terms were tailored to each database in order to best utilize the individual subject headings, keywords, and medical subject headings (MeSH) terms included in the individual databases. A full list of search terms is shown in the Extended data 20 .
In addition to our search terms, search results were restricted to articles published in English within five years of the date of the search (Jan 1, 2014 to May 15, 2019). After retrieving articles within a ten year window as outlined in Extended Data 20 , the authors decided to further reduce the range to within five years. This date range was chosen in order to provide a review of the most recent developments in this field of research. After search results were compiled and duplicates were removed, a total of 5234 articles were selected for title and abstract review. Four authors (JSC, HH, MMR, RB) screened the titles and abstracts and retrieved articles for eligibility, resulting in 113 articles selected for full text review. Three authors (JSC, HH, JL) then independently reviewed the full text of eligible articles, completed data extraction worksheets adapted from the Cochrane Review Group's Data Extraction Form (https://dplp.cochrane.org/data-extraction-forms), and assessed the articles for risk of bias using the Cochrane risk of bias tool 21 . Elements of the data extraction worksheet included study design and setting, participant characteristics, details of the intervention and control groups, diagnostic tools, and patient outcomes (Extended data, Supplementary Table 1) 20 . Any disagreements were resolved by thoroughly discussing any points of concern. During the full text review 86 articles were removed because they failed to meet our inclusion criteria resulting in a total of 27 included articles ( Figure 1).

Inclusion and exclusion criteria
Our review addresses non-pharmacological management strategies for delirium in the ICU. Included articles were those investigating non-pharmacologic interventions and their impact on delirium incidence, prevalence, duration, or severity in an adult (≥ 18 years) intensive care unit setting. Articles were excluded if they focused on non-human subjects, pediatrics, terminally ill subjects, withdrawal related delirium, case reports, or where no full-text article was available (abstract only). There were no restrictions on study design. Studies solely investigating delirium-free-coma-free days were excluded since it is not possible to review as a delirium-specific result. One multi-center study was excluded as both the frequency and method of assessment for delirium were not specified for all study centers, making it difficult to reliably compare the results with other trials 22 .
Another study was excluded because neither the screening process nor the cohort were described other than total number of

Amendments from Version 2
This revision addressed additional limitations and future directions of our review topic (non-pharmacologic delirium strategies in critically ill adult patients). Segments were added to the discussion section regarding: the distinction between different etiologies of ICU delirium, expanding patient populations, and underlying differences among sub-populations of patients with ICU delirium. We also updated author affiliations and funding to their current status.
patients enrolled, and there were no exclusion criteria noted to infer any characteristics of the selected population 23 .

Risk of bias assessment
In addition to data extraction using the Cochrane Review Group's Data Extraction Form, a risk of bias assessment was performed by all authors on all included randomized controlled trials (RCTs) and randomized pilot studies. A risk of bias worksheet was developed by modifying Cochrane's Risk of Bias Tool and articles were ranked as having a low, high, or unclear risk of bias 21 . Disagreements were settled by discussion between the authors. A total of eleven included studies underwent this assessment. Details of the risk of bias assessment categories can be found in the Extended data, Supplementary Table 2 20 .

Results
After searching the literature, 27 articles are included in our systematic review 24-50 ( Figure 1). Study details of each reviewed trial are located in Table 1. The 27 included studies provide results on many distinct outcomes; however, only the deliriumrelated outcomes of incidence, prevalence, duration, and severity were reviewed (Table 2-Table 5). Outcomes combining delirium and coma into the same statistic were excluded, as no delirium-specific results could be assessed outright. An overall summary of delirium outcomes can be found in Table 2.

Bundled interventions Incidence Prevalence Duration Severity
Measurements for incidence, prevalence, and duration were based upon multiple methods of delirium screening, including the Confusion Assessment Method (CAM), CAM-ICU, Intensive Care Delirium Screening Checklist (ICDSC), and Neelon and Champagne (NEECHAM) scales. Incidence and prevalence were similarly defined in all studies except for one, and are recorded separately in Table 3; only one study looked at both incidence and prevalence 39 . Severity was assessed by using the Delirium Index (DI), the Delirium Rating Scale (DRS), the Revised Delirium Rating Scale (DRS-R-98), and NEECHAM scale (Table 5).   Legend: * = significant, p<0.05; † = 95% confidence interval. The remaining six were a case-matched control study, an evidence based protocol, a mixed-methods pilot study, a prospective multicenter cohort study, a retrospective cohort study, and an action research study 27,29,33,37,41,45 .

Risk of bias assessment
The ten RCTs and the randomized pilot study underwent a risk of bias assessment performed by all authors. Risks of bias fell into five major groups (selection bias, performance bias, detection bias, attrition bias, and reporting bias), and based on a study's scores in each of these groups it was labeled as having an overall high, low, or unclear risk of bias. Four were considered low risk of bias 24,34,42,48 , two had a high risk of bias 32,44 , and five had an unclear risk of bias 31,35,36,38,43 . The most common source of bias was performance bias due to the impossibility of blinding participants or personnel to certain treatments. Common sources of unclear and high risk of bias included the methods of randomization and allocation concealment, as well as how missing data was handled.
Individual interventions Early mobility. The effect of early mobility protocols on delirium was the most commonly studied individual intervention. Four of the studies included in our review individually assessed the efficacy of early mobility 24,29,38,45 in treating and preventing delirium; of these, two were RCTs 24,38 , one was an evidencebased project 29 , and one was a case-matched control study 45 .
The pilot RCT performed by Álvarez et al. investigated the effect of early mobilization through early and intensive occupational therapy (OT), including polysensory stimulation, body positioning, cognitive stimulation exercises, basic activities of daily living, upper extremity motor exercises, and family involvement, on non-intubated, elderly patients (≥ 60) in addition to the study center's standard, non-pharmacological delirium prevention care 24 . Delirium associated outcomes included incidence, duration, and severity; they found significant differences in incidence and duration of delirium, with both p-values ≤ 0.001, but no significant difference in severity (Table 3-Table 5).
Another RCT by Karadas and Ozdemir assessed the effect of range of motion (ROM) exercises on delirium in elderly ICU patients (≥ 65 years) 38 . Interventional care included ROM exercises for 30 minutes daily after establishing the patient's ability to complete 10 repetitions on each of the four extremities while lying in bed. They reported no statistically significant differences between cohorts for delirium associated outcomes (Table 3 & Table 4).
Campbell addressed early mobilization in mechanically ventilated ICU patients with an evidence-based project 29 . They measured the effect of a tiered protocol of ROM exercises, bed mobility exercises, seated balance activities, transfer activities (such as bed to chair), standing exercises, and ambulation on delirium incidence and duration but found neither to be significant (Table 3 & Table 4).
The effectiveness of functional electrical stimulation (FES) to promote mobility and recovery in mechanically ventilated patients with sepsis was evaluated by Parry et al. in a case-matched control study 45 . The intervention included use of a motorized cycle ergometer to directly stimulate four major lower limb muscles (quadriceps, hamstrings, gluteals, and calves) five times weekly for 20-60 minutes a session dependent on the individual patient's tolerance. While delirium incidence was not significantly affected (Table 3), the median days of delirium differed between arms (6.0 in control and 0.0 in intervention) ( Table 4).

Family involvement.
Of the 12 studies in our review which focused on individual interventions, three studied the effect of family involvement on delirium in adult ICU patients. One was a randomized pilot study 42 , one an RCT 32 , and one was a pre-post study 47 . All three studies utilized CAM-ICU in their assessment of delirium.
Mailhot et al. constructed a randomized pilot study to explore the effect of a family caregiver (FC) assisting with delirium management after being 'mentored' (MENTOR_D) by nurses in the 'MENTOR_D' intervention 42 . They assessed the efficacy of this intervention on all delirious, adult coronary artery bypass graft (CABG) patients admitted to the surgical ICU by measuring the outcomes of duration, occurrence, and severity of delirium over three days 42 . This intervention enrolled 14 patient-nurse (control) care dyads and 16 patient-FC care dyads, which had the FC apply bedside strategies to aid the patient in reorientation. In addition to reorientation, the FC was asked to observe and communicate signs of delirium with nursing staff, present family memories, and speak clearly and simply. Delirium duration and occurrence on post-operative Day 2 improved clinically between groups (duration, mean days from 4.14 to 1.94; occurrence, from 71.40% to 43.80%); however, this result was not assessed for statistical significance and the severity result was not found to be significant (Table 3-Table 5).
The RCT performed by Eghbali-Babadi et al. investigated a modified family visitation policy, implementing an additional 30-40 minute special visit by an approved family member, and its effect on delirium incidence in non-intubated adults aged 18-70 after elective open heart surgery 32 . They found a statistically significant reduction in delirium incidence in the intervention group with a p-values of 0.04, <0.05, and 0.03 at three different time points (Table 3).
Rosa et al. also measured the effect of a modified family visitation policy on delirium incidence, although their population was less restrictive and included any adult ICU patient 47 . Their pre-post study included the extension of visitation hours from 4.5 hours per day over three visitation blocks to 12 hours per day between 09:00-21:00. This resulted in a statistically significant difference in delirium incidence, improving from 20.5% to 9.6% (Table 3).

Environmental approaches (lighting, music therapy, automated reorientation).
Three RCTs assessed the impact of environmental factors on delirium in the ICU, assessed by CAM-ICU, through manipulation of light 48 , music therapy 31 , or automated reorientation 44 .
In Simons et al.'s dynamic lighting application RCT, adult ICU patients were exposed to variations in high intensity, blueish-white lighting while delirium incidence and duration were measured 48 . The intervention group was exposed to a peak of 1700 lux (brightness)/4300 K (color temperature) from 09:00-11:30 and 13:30-16:00, and a daytime minimum of 300 lux/3000 K from 11:30-13:30; the control group was exposed solely to 300 lux/3000 K (Table 1). Neither the cumulative incidence of ICU-acquired delirium nor the duration were significantly affected, and the trial was ended early after the intervention was deemed futile (Table 3 & Table 4).
In another RCT, Damshens et al. introduced therapeutic music selected by a music expert, twice a day for 45 minutes to assess the effect on delirium incidence in adults admitted to ICU trauma service 31 . Patients in the control group received conventional care for the duration of their admission. There was no resultant change to delirium incidence between the two groups ( Table 3).
Munro et al. developed a novel patient reorientation strategy in an RCT, which utilized bilingual (Spanish or English) messages pre-recorded by either family members or females unknown to the adult ICU subjects 44 . The recordings included an introduction with the patient's name and location, with several additional randomly ordered statements in order to reorient the patient to their unfamiliar surroundings and reason for hospitalization. All three arms (two intervention groups and one control group) were compared and it was found that the family voice group had a significant improvement in delirium free days (p= 0.0437) but not mean days of delirium (Table 4).

Self-involvement approaches (mirror usage, mindfulness exercises).
The remaining two studies on the effect of individual interventions assessed the impact of self-involvement approaches, including mirror usage 34 and mindfulness exercises 41 , on ICU delirium measured by CAM-ICU. One study was a pilot RCT 34 , while the other was a mixed-methods pilot study 41 .
In a pilot time-cluster RCT, Giraud et al. tested the effect of introducing structured mirror usage into post-operative recovery in elderly ICU patients (≥70 years) after cardiac surgery 34 . Mirror usage was standardized by developing a protocol for nurses and physiotherapists, aiming to use both small, personal mirrors as well as larger posture mirrors in order to help the patient with reorientation and self-awareness, enhance multisensory feedback on minor procedures, and augment passive and active physical therapies. The control cohort received usual care, including allowing control patients who brought a mirror from home to use it per their normal habits. After comparing the usual care group with the mirrors group, no significant improvement was found in delirium incidence, ICU days with delirium, or the proportion of the total ICU length of stay that the patient spent delirious (Table 3 & Table 4).
The mixed-methods study by Lisann-Goldman et al. had subjects who were 40 years of age or older participate in Langerian mindfulness discussion exercises both prior to and after elective cardiac surgery with cardiopulmonary bypass 41 . In addition to discussion exercises, patients listened to an audio file before surgery. This audio file walked them through techniques on how to re-assess one's situation and improve their outlook by taking emotional control of the situation, encouraging the patients to focus on the process of change and allowing oneself to accept new ideas and remain confident about the unknown. The discussion exercises continued post-operatively twice daily. In contrast, the 'informational control' group went through normal pre-operative discussions followed by an audio file describing the process of cardiac surgery. They found that no subject developed delirium in either the interventional or the 'informational control' group so the effectiveness of the treatment could not be assessed.

Bundled protocols 'Wake Up and Breathe' protocol. Khan et al. designed a 'Wake
Up and Breathe' protocol in a pre-post interventional study to assess for any change in delirium and sedation in mechanically ventilated, adult ICU patients 39 . They modified elements of the Awakening and Breathing Controlled trial (ABC) to implement a spontaneous awakening trial and daily sedation vacation followed by a spontaneous breathing trial, depending on the patient's response 51 . Delirium was assessed by CAM-ICU, and both the incidence and prevalence of delirium were analyzed, with the study finding no significant change in either measured outcome (Table 3).

ABCDE(F) bundles.
Five of the 15 studies which examined delirium bundles studied the effectiveness of ABCDE(F) bundle protocols on reducing delirium. ABCDE(F) bundles have multiple components including: spontaneous awakening (A) and breathing (B) trials, interdisciplinary coordination of sedatives and medications (C), delirium screening and management (D), early mobilization (E), and family engagement and involvement (F) 52 . Of these five studies, two were pre-post studies 26,40 , one was a prospective multicenter cohort study 37 , one was a quasi-experimental quality improvement project 30 , and one was a retrospective cohort study 27 . Three measured delirium outcomes using CAM-ICU 26,30,40 , one utilized ICDSC 27 , and one multicenter study used either CAM-ICU or ICDSC 37 .
Balas et al. assessed the impact of an ABCDE bundle on adult ICU patients, evaluating the prevalence and duration of delirium in both total days and percent of ICU days spent delirious, with a pre-post study 26 . The prevalence and percent of ICU days spent delirious were improved in the post period with p-values of 0.03 and 0.003 respectively (Table 3 & Table 4). However, the overall duration of delirium was not significantly different ( Table 4).
The retrospective assessment of an ABCDE bundle by Bounds et al. evaluated its effect on delirium prevalence and duration in an adult ICU population 27 . Both the prevalence and duration were significantly decreased in the ABCDE bundle group (p= 0.01 and 0.001 respectively; (Table 3 & Table 4).
Kram et al. also looked at a similar patient cohort, all adult patients 18 or older admitted to the ICU, in a pre-post ABCDE bundle study with a smaller subject population (Kram, n=83; Balas, n=296; Bounds, n=159) 40 . They assessed the effectiveness of the ABCDE bundle on delirium by measuring delirium prevalence and comparing it to a control based on literature values. The measured delirium prevalence of 19% (Table 3)  Chai initiated an ABCDEF bundle in a mixed ICU setting and analyzed delirium incidence in the adult patients in a pre-post, quasi-experimental quality improvement project 30 . Delirium incidence was compared between morning and night occurrences (morning 08:00-11:00; night 20:00-23:00); both showed significant improvement in the intervention group with a p-value <0.001 for both morning and evening measurements (Table 3).
A prospective cohort study performed by Pun et al. through a national quality improvement initiative compared complete ABCDEF bundle performance with proportional ABCDEF bundle performance in adult ICU patients with an ICU stay of at least 48 consecutive hours 37 . Complete bundle performance was defined as a patient-day where 100% of the eligible bundle elements were performed, whereas proportional performance was anything less 37 . Their study was comprehensive, including 10,840 patients for delirium outcome analysis across 68 ICUs in the United States and Puerto Rico 37 . When comparing the incidence of delirium between patients with complete and proportional ABCDEF bundle performance, they found that patients with complete performance were significantly less likely to develop delirium (Table 3) 37 . In an additional analysis, Pun et al. found a dose-dependent reduction of delirium incidence when the more eligible ABCDEF bundle elements were performed (p < 0.0001) 37 . It is worth noting that this study had a high rate of 'missingness' for delirium data and the analysis team chose not to perform multiple imputations 37 .
A quasi-experimental study designed by Arbabi et al. developed a multi-component delirium management bundle comprised of staff education and environmental and non-pharmacologic care changes 25 . They measured the effectiveness of their bundle by assessing delirium incidence and duration in all adult patients admitted to the general ICU, finding a significant difference in both outcomes (p = 0.01 and 0.001 respectively; Table 3 & Table 4).
Bryczkowski et al. assessed the effectiveness of their bundle, which included a staff-patient-family education program, medication management strategies, and non-pharmacological sleep enhancement protocols, on delirium incidence and delirium free days in patients over the age of 50 years 28 . The research team found no significant improvement in delirium incidence (Table 3), although the average total number of delirium-free days out of 30 changed significantly from 24 to 27 between groups (p=0.002; Table 4).
Another bundle study developed by Fallahpoor et al. focused specifically on adults admitted to the ICU after elective CABG in an action research study. Their post-CABG delirium management bundle was assessed in an action research study and had three elements focusing on pre-, intra-, and post-operative methods to identify delirium risk factors, optimize time spent in surgery, and introduce staff education and post-operative environmental changes 33 . Delirium related outcomes included the incidence ratio and total number of recorded delirium events, with significant differences found in both (p=0.001 and 0.008 respectively; Table 3).
In the RCT conducted by Guo et al., the effect of a bundle consisting of cognitive prehabilitation, post-operative cognitive stimulation activities, environmental changes, music therapy, and non-pharmacologic care changes on delirium incidence and duration after oral tumor resection in patients aged 65-80 years was studied 35 . The incidence of delirium improved significantly overall, but was only significantly different on post-operative day one compared to days two and three (p=0.035, p=0.374, p=0.364 respectively; Table 3); the duration of delirium also differed significantly (p< 0.001; Table 4).
Hamzehpour et al. designed an RCT and implemented the Roy adaptation nursing model for all adult ICU patients, which focuses on balance of nutrition, electrolytes, and fluids while promoting activity, sleep hygiene, and monitoring of circulation and endocrine function 36 . Their primary delirium-specific outcomes were incidence and severity, and they analyzed both outcomes for two time points (morning & night) for seven days. Their research only showed significant improvements to incidence on day seven, both morning and night (p<0.008 and p<0.05; Table 3), but delirium severity, assessed with NEECHAM, improved through the morning of day four to the night of day seven at all measured time points (every time point, p≤0.028; Table 5).
Moon and Lee implemented a bundle which included early cognitive assessments and reorientation, sensory aids, environmental changes, consistent care staff and location, familiar items from home, nursing care changes, and early mobility as part of an RCT aimed at assessing delirium incidence in adult ICU patients with at least a 48 hour stay 43 . Their study did not show a significant difference between the intervention and the control group who received usual care (Table 3).
In a pre-post, observational quality improvement project, Rivosecchi et al. combined staff education with a non-pharmacologic bundle to look at incidence and duration of delirium in any adult patient aged 18 or older admitted to the medical ICU 46 . Their M.O.R.E. bundle included (M)usic, (O)pening blinds, (R)eorientation and cognitive stimulation, and (E)ye and ear care. Both delirium incidence and duration were significantly impacted, with incidence decreasing from 15.7% to 9.4% and a reduction in duration from 16.1% of the ICU stay to 9.6% (p = 0.04 and <0.001 respectively; Table 3, Table 4).
Sullinger et al. enrolled adult surgical-trauma ICU patients with acute delirium in a pre-post retrospective study, tailoring their bundle to incorporate staff education with sensory aids, healing arts techniques, mobility, environmental changes, and family presence 49 . Their bundle also included the initiation of anti-psychotic medications if non-pharmacologic tactics failed. The only specifically delirium related outcome was the number of days spent delirious, resulting in a significant decrease from 8.2 to 4.5 median days (Table 4).
Any patient 18 years or older admitted to a cardiothoracic ICU after CABG surgery was analyzed for incidence, duration, and severity of delirium by Zhang et al. in a prospective pre-post study 50 . Their delirium bundle targeted risk factor screening and modifications, including increased family visits, reorientation, and changes to nursing care. The only significant improvement was to incidence of delirium which dropped from 29.93% to 13.48% (Table 3), while the intervention had no impact on duration or severity (Table 4 & Table 5).

Summary of findings
Our review included 27 trials that evaluated the effect of various non-pharmacological treatment and management protocols on delirium in an ICU setting. Assessment of the efficacy of these protocols in the last five years was most commonly done by considering incidence and/or prevalence. A total of 25 studies assessed for the effects of these protocols on incidence and/or prevalence, with 11 studying individual approaches and 14 studying bundles. Of these 25 trials, 11 reported significant improvements overall [24][25][26][27]30,32,33,37,46,47,50 , nine found no significant improvement 28,29,31,34,38,39,43,45,48 , and two only found significant change at certain time points 35,36 ; the remaining three did not analyze for statistical significance of their results [40][41][42] . The 11 effective interventions for incidence and/or prevalence were primarily bundled protocols (eight trials) [25][26][27]30,33,37,46,50 , followed by family approaches (two trials) 32,47 , and early and intensive OT (one trial) 24 . The two studies with time point dependent changes were both bundles, one non-pharmacologic 35 and the other introduced the Roy adaptation nursing model 36 . The multicomponent non-pharmacologic bundle found improvements in incidence both overall and on day one, while the Roy adaptation trial only saw a change in incidence on day seven in both the morning and the evening. Three studies did not analyze for statistical significance [40][41][42] ; however, the family caregiver intervention saw an overall reduction in the percent of subjects who developed delirium from 71.40% to 43.80% 42 . The study on mindfulness exercises had no subjects in either investigational group develop delirium 41 , and an ABCDE bundle reported a post-bundle incidence of 19% but stated there was no pre-bundle data with which to compare 40 . No other studies that looked at delirium incidence were effective.
In addition to incidence and prevalence, another common outcome was a change in the duration of delirium. Sixteen of the reviewed studies evaluated the duration of delirium, eight focusing on individual interventions and eight introducing bundled protocols. Of these 16 studies, eight found significant changes overall 24,25,27,28,35,45,46,49 , two had significant improvements at select time points 26,44 , and five did not have significant results 29,34,38,48,50 ; the one remaining study did not assess for statistical significance 42 . Six of the eight successful trials were bundles and both of the two effective individual therapies were mobility-focused (early and intensive OT, and FES) 24,45 . Only four studies looked at delirium severity 24,36,42,50 with only one finding any significant results, and only finding them at select time-points (Roy Adaptation Model) 36 .
The pilot RCT performed by Alvarez et al. utilized a unique method for assessing the performance of their intervention. In addition to assessing delirium incidence, they measured the ratio of delirium duration to the amount of time exposed to the treatment (IRR) 24 . They found that IRR decreased as the time exposed to treatment increased to a significant degree (p= 0.000). This ratio could be explained in three ways. Either the duration of delirium stayed the same as the time exposed to treatment increased, the duration of delirium increased slower than the time exposed increased, or the duration of delirium decreased while the time of exposure increased. However, the last explanation is impossible, due to the duration of delirium being a sum overtime which could not decrease, such that the result must be explained by either a small increase or no increase in the duration of delirium. If the decrease in the IRR is explained by smaller and smaller increases to delirium duration it is likely that the IRR results from either the trend of patients slowly becoming healthier over time, or the conjunction of that with the intervention. However, if the IRR is explained by the delirium duration ceasing to increase, then once it stops the treatment may still be effective, but it is not becoming more effective over time and plateaus in effectiveness.

Implications of results and application to practice
The reviewed studies focused on individual interventions that had a wide range of limitations and were, on the whole, less effective than bundled protocols in the treatment and management of delirium. Many of these studies had limited reliability due to small or extremely small sample sizes 29,41,44,45 . Additionally, even when results were significant, they often had limited applications to practice due to the prevalence of restricted populations. Three of the individual intervention studies limited their study cohort to elderly adults 15,24,38 , a population which is at an increased risk of delirium. It is unclear whether these results would apply to younger patients. Studied populations were also commonly narrowed to either exclusively intubated patients 45 , non-intubated patients 24,32 or patients with a particular illness 34,42,45 . Another possible limitation was in the questionable reliability of delirium assessment criteria. This is mentioned by Campbell who stated that 35% of CAM-ICU were incorrectly labeled as 'unable to assess' 29 . The question of reliability was also raised in Lisann-Goldman et al.. This study could not assess the effectiveness of their intervention as no patients developed delirium 41 . However, this could be explained by the fact that fully sedated patients were considered 'not delirious' since CAM-ICU could not be performed. The authors also noted that since it often takes weeks or months to fully integrate new behavioral thought techniques, a study focused on changing thought patterns in days would not entirely reflect the full benefit if any were present 41 .
Eight of the individual intervention studies were RCTs. This type of study introduces the possibility of additional limitations due to the nature of its design. Two of these RCTs had a high risk of bias 32,44 due to a failure to blind patients and personnel, as well as blinding of the outcome assessment and improper allocation concealment. The question of blinding raises another possible limitation of many of these studies, namely the possibility of the Hawthorne effect in patients who knew that they were being observed and receiving an intervention for the treatment/prevention of delirium.
The 15 studies which investigated bundled protocols had, overall, larger sample sizes, fewer cohorts with limited populations, and indicated better reliability in their delirium assessment than the studies which focused on individual interventions. The smallest sample size was 83 patients 40 ; however, this study did not split the sample into multiple cohorts and all patients received the intervention. One study had a sample size of 89 49 , and all other studies had a sample size of at least 100 patients. A total of five studies restricted their studied population beyond adult ICU patients 28,33,35,39,50 . Two of these limited their population by age 28,35 ; however, Bryczkowski et al., despite limiting their population by age, included any patients greater than 50 years old, younger than the age when delirium risk is noted to increase 53,54 . One study limited its population to mechanically ventilated patients 39 , two considered only patients undergoing CABG 33,50 , and one studied patients after oral tumor resection 35 . These population restrictions could limit the generalizability and applicability of the interventions; however, this risk is reduced since bundles were often investigated in multiple studies with similar results.
Only three of the bundle studies were RCTs 35,36,43 . Each of these RCTs had an unclear risk of bias with the most common risk being the inability to blind participants and personnel. The impossibility of blinding in delirium intervention studies makes RCTs a questionable approach. Eight of the bundle studies, recognizing blinding as an impossibility, chose to conduct pre-post prospective studies rather than RCTs 26,28,30,39,40,46,49,50 . These studies carried a lower risk of introducing bias to their studies and avoided crossover between arms. The pre-post study performed by Kram et al. had the major limitation of not including a pre cohort and only comparing the results of their intervention with literature values 40 . Additionally, while they found their measured delirium prevalence to fall outside their included literature values (19%), this prevalence falls within the values provided by the current literature (19-87%) 2 . Chai's pre-post study had a delirium assessment with questionable reliability. While all other studies assessed delirium whenever the Richmond Agitation-Sedation Scale (RASS) was ≥ -3, they reported that patients were unable to be assessed whenever RASS was < -2, resulting in a greater proportion of patients not assessed for delirium 30 .
In sum, multicomponent, bundled approaches were more successful at improving delirium outcomes compared to individual techniques; however, the effective individual tactic of family engagement was included as a component in the effective bundles. Although a majority of the reviewed bundles were effective, it is difficult to compare results as the trials had large differences in study design, enrollment numbers, and delirium assessment measures.
Given the findings of this systematic review, further research is warranted in order to confirm these results and apply them to other patient populations. For example, patients in non-ICU hospital wards would likely be more capable of mobility or visitation interventional components that in turn increase their sense of independence, wellbeing, and recovery support compared to critically ill persons. Due to the nature of these differences, it is worth considering tangential application of ICU delirium prevention and management interventions dependent on a patient's level of care and health stability. An additional consideration for future research is to ascertain the level of effectiveness of individual intervention components against differently aged patient subpopulations. It is possible that certain interventional elements may be more effective for older versus younger critically ill adults at risk for ICU delirium due to differences in underlying risk factors and etiologies.

Strengths and limitations
The strengths of our systematic review include thorough search terms and the methodology to assess a vast majority of recent literature in this field.
One limitation of this systematic review is that we only focused on trials within the past five years which excluded some well-cited early studies on delirium. We also did not evaluate other listed outcomes which could provide additional insight into any change in delirium status. Since the condition can be transient and delirium screenings are not performed as frequently throughout the ICU day as other measurements, outcomes such as restraint use or amount of prescribed sedatives or anti-psychotic medications would be beneficial to assess in this setting. While the decision to omit exclusion criteria on study design allowed for assessment of a broader range of trials, it was difficult to compare outcomes when multiple differing designs and measurement tools were used. Although the CAM-ICU was widely used, some studies used alternative tools and there was no standardized way of defining or measuring delirium duration or severity. A different measurement tool was used to evaluate severity in each of the four studies reviewing this outcome, and duration was defined in a multitude of fashions. Combining this realization with the fact that some studies focused on highly specific subpopulations suggests that some trials may need to be replicated in a standardized fashion to account for any differences in methodology or subjective assessments.
Additionally, this review did not clearly delineate between patients with post-operative ICU delirium and those who developed nonsurgical-related ICU delirium. This distinction is imperative, as the risk factors and appropriate treatment approaches may differ for these patient groups. It would be of future benefit to determine such differences, and which interventional components or bundles are best designed for each etiology.

Conclusions
Many ICU delirium treatment and management protocols were developed and tested within the last five years in a variety of study designs. Few trials on individual interventions had positive effects on delirium incidence and duration, but multicomponent bundles were found to be more effective overall while incorporating the effective individual intervention of family engagement. Based on the results of bundle studies, the implementation of multi-component protocols in ICUs can reduce ICU delirium, thereby reducing cost of care, improving overall outcomes, and limiting time spent mechanically ventilated, medicated, or admitted. Despite these results, further research is needed on individual interventions in order to improve specific elements of multicomponent bundles by adding or removing ineffective therapies. Additional research is also warranted to evaluate for any positive effects in more generalized hospital populations.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required. (Jan 1, 2014 to May 15, 2019) as stated in the manuscript? Or was it 10 years, as is listed in the search terms of the supplementary appendix 2? Thank you so much for this comment so that we can clarify. The searches for each database listed in Supplementary Appendix 2 were initially performed for 10 years from the date of the search. During the title and abstract screening process, we ultimately decided to narrow the results further to within 5 years of the dated search. This step can be seen in Fig 1 PRISMA Record Screening Flow Chart, where we removed 3039 records prior to 2014. We added a statement to our methods section to further clarify this aspect.

Extended data
Did all the authors screen the titles and abstracts? And did all the authors independently review the full text of the 113 eligible articles, or only some of the authors? Thank you for informing us of this insufficiency. Four authors (JSC, HH, MMR, RB) screened the titles and abstracts, and three authors independently reviewed the full text (JSC, HH, JL). We added a statement to our methods section clarifying this point.

2.
For Table 2, I found it confusing using x's, y's and o's. It may be better to title the Table "Summary of delirium outcomes by statistical significance (p<0.05)," and then spell out "yes" instead of y, and either "no" or "not significant" instead of x. For the outcomes that were measured but not analyzed, it may be clearer to write "not analyzed" or "NA" rather than "o." Thank you for the feedback about this table. This change will definitely make our table more clear to the reader, and Table 2 has been updated to reflect this amendment.

3.
Competing Interests: No competing interests were disclosed.
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias • You can publish traditional articles, null/negative results, case reports, data notes and more • The peer review process is transparent and collaborative • Your article is indexed in PubMed after passing peer review • Dedicated customer support at every stage • For pre-submission enquiries, contact research@f1000.com