Efficacy of Non-Pharmacological Interventions to Prevent and Treat Delirium in Older Patients: A Systematic Overview. The SENATOR project ONTOP Series

1The healthcare system is increasingly demanding rapid access to current research to ensure evidence-based informed decision making and practice. Previously, guideline developers and decision makers were overwhelmed by the number of primary studies; they currently contend with an excess of reviews. The number and variety of systematic reviews (SRs) is rapidly growing. Various sources report that for a single topic several systematic reviews can often be identified. Furthermore, there is a tendency to perform systematic reviews of reviews in order to provide clinical decision makers with the evidence they need.

2There are several reviews of non-pharmacological interventions for the prevention and treatment of delirium in older patients. Delirium is the most common complication of hospital admission in older patients with an incidence rate that varies between 11% and 42% among patients in medical wards and is as high as 80% in some surgical conditions in the post-operative phase. Delirium is associated with increased morbidity, mortality and length of hospital stay as well as increased use of healthcare services and costs. Since there is no evidence that pharmacological prevention or treatment of delirium is effective, a great deal of attention has been devoted to non-pharmacological interventions.

3The non-pharmacological interventions to prevent or treat delirium are quite diverse, ranging from simple single component interventions (e.g., music therapy) to complex multicomponent interventions.

4This paper describes the methods used to identify all published systematic reviews concerning non-pharmacological interventions for delirium; identifies and critically appraises the primary studies included in the SRs; lists the elements that compose the multicomponent interventions and, based on the components shared among the studies, presents the meta-analyses, critically summarizes the evidence, discusses the limitations and proposes research priorities for future studies.

5This work is part of the ONTOP (Optimal Evidence-Based Non-drug Therapies in Older People) project, a workpackage of a European Union funded FP 7 research named SENATOR (Software ENgine for the Assessment & Optimization of drug and non-drug Therapy in Older peRsons). The ONTOP aim is to undertake a literature search of systematic reviews concerning evidence-based non pharmacological treatments of 15 prevalent medical conditions affecting older people, including delirium.

6To gather the evidence about non-pharmacological intervention to prevent or treat delirium, the ONTOP Evidence Group was established and took responsibility for defining the clinical questions. To define appropriate clinical question the Group identified a list of outcomes and a list of non-pharmacological interventions deemed relevant independent from the evidence according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE). After three rounds of consultations, important and critical outcomes were identified. For delirium prevention the outcome delirium incidence was considered critical whereas for delirium treatment the outcomes delirium improvement (intended as either a delirium resolution and a reduction in its severity) and functional status (intended as the degree of functional autonomy of the patient) were considered critical. In the present paper only the results of the critical outcomes will be presented. For details see S1 Box, S1 and S2 Tables. For results of secondary outcome see S1 File.

7The search sources included Cochrane Database of Systematic Reviews, PubMed, PsychINFO, EMBASE and CINAHL (S2 Box).

8Two criteria were considered for further evaluation of an abstract: a) a paper defined as a review or meta-analysis, b) the mention of any non-pharmacological intervention for delirium.

9Subsequently, full-texts of relevant abstracts were obtained and screened to identify SRs of interest based on: the use of at least one medical literature database;the inclusion of at least one primary study; andthe use of at least one non-pharmacological intervention for delirium prevention or treatment for patients of 60+ years of age.

10the use of at least one medical literature database;

11the inclusion of at least one primary study; and

12the use of at least one non-pharmacological intervention for delirium prevention or treatment for patients of 60+ years of age.

13Only studies written in English, Italian or Spanish were considered.

14We assessed the methodological quality of each systematic review using the AMSTAR (A Measurement Tool to Assess Reviews) instrument that contains 11-items to appraise the quality. Two reviewers independently assessed the quality of the SRs and disagreement was resolved by consensus.

15From the included SRs we identified any experimental comparative study, either randomized or nonrandomized, that investigated any non-pharmacological intervention to prevent or treat delirium in older patients. Primary studies were excluded if they were observational studies or before-after studies with historical controls.

16Extracted data were transferred onto data extraction forms. Information collected included trial characteristics, patients characteristics, intervention and comparator characteristics, and outcome measures.

17Pairs of reviewers independently screened titles, abstracts and full-texts of articles. Disagreement was resolved by discussion.

18To better understand the characteristics of the multicomponent interventions, we itemized each element of the multicomponent intervention. This process helped to decide whether or not it was appropriate to perform a formal meta-analysis of different studies.

19Assessment of risk of bias for the included primary studies was carried out using criteria from the Cochrane Collaboration. The domains considered were random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other potential biases (e.g., balance in baseline characteristics). We assigned a risk of bias to one of three categories: low risk, high risk and unclear risk. Two reviewers independently assessed the risk of bias of individual studies and any differences in quality assessment results between raters was resolved through consensus.

20The quality of evidence was assessed with GRADE (Grading of Recommendations, Assessment, Development and Evaluation) methodology by a panel of reviewers with experience in geriatrics, internal medicine and methodology.

21GRADE assessment took into consideration the risk of bias, consistency of results across the available studies, precision of the results, directness, and likelihood of publication bias. Randomized trials were privileged for the GRADE evaluation. Where evidence is without randomized trials the GRADE evaluation is performed for other study designs (e.g., controlled clinical trials). The quality of the evidence was categorized as high, moderate, low, or very low based on the judgments for the primary outcome (delirium incidence for prevention; delirium improvement and functional status for treatment). Summary tables were constructed using GRADEpro version 3.6.

22Where a meta-analysis was possible with at least two studies, DerSimonian and Laird random-effects models were used to pool risk ratios of delirium incidence (Review Manager software version 5.3). We used the Chi2 test and I2 statistic to assess heterogeneity. We considered heterogeneity to be statistically significant if the P value is less than 0.1. Publication bias was assessed by visual judgment of a funnel plot and by Egger’s regression method.

23Our search identified 3329 abstracts after excluding 295 duplicates. Among the 80 potentially relevant publications, 26 were considered relevant for inclusion and 54 were excluded (see Fig 1 for study screening process and S3 Table for list of excluded reviews with reasons). The publication year ranged from 1996 to 2014 and 5 reviews were published in 2014.

24While all articles used PubMed to search for primary studies, 13 papers also employed CINAHL, 11articles also used the Cochrane Library, 10 studies also accessed Embase, and 3 studies also searched PsychINFO.

25None of the SRs was sponsored by a company. Six studies were funded by a governmental institution or a non-profit organization.

26The reviews were heterogeneous. Some reviews, in addition to examining interventions to treat or prevent delirium, also evaluated the pathogenesis of delirium, examined the role of sitters, or studied approaches to diagnose delirium. Only two reviews assessed single component interventions (music therapy and earplugs). The remaining SRs appraised multicomponent interventions.

27Three SRs ranked as being of high quality (scoring 8–11), 12 of medium quality (scoring 4–7), and 11 of low quality (scoring 0–3) (S4 Table).

28Table 1 summarizes the basic characteristics of the SRs.

29Overall the 26 SRs yielded 78 primary studies of which 31 satisfied our inclusion criteria. (see Fig 1 and S5 Table for excluded primary studies).

30The identified primary studies are described based on the type of the intervention (single or multicomponent), the aim of the intervention (prevention or treatment), the setting (e.g., surgical), and the study design [randomized controlled trial (RCTs), Controlled Clinical Trials (CCT), and Before After (BA) study].

31Nine studies evaluated the efficacy of multicomponent non-pharmacological interventions to prevent delirium in surgical wards. Two studies were randomized trials, one was a CCT and six were BA studies. Except for Chen et al. all studies assessed patients with hip fractures. The study characteristics are described in Table 2. All studies considered the incidence of delirium as primary outcome except Chen et al, who investigated functional and cognitive function as the primary outcomes.

32The components of the non-pharmacological intervention that were common among the studies are shown in Table 3.

33Given the nature of the intervention, all the comparative studies suffered from performance bias because blinding of patients and personnel could not be carried out. However, all the three trials were immune from detection bias. Of the two RCTs, only Lundstrom et al. explicitly reported the method of allocation concealment. The non-randomized studies (CCTs and BA studies) were, by their nature, at risk of selection bias. In two of the four BA studies, the outcome assessor could not be blinded, thus raising the possibility of detection bias. Fig 2 summarizes the risk of bias in each study.

34Two RCTs presented data that could be combined, given the similarity between the population samples and the items of the non-pharmacological interventions. In fact these two trials had the following interventions in common: comprehensive geriatric assessment, management and rehabilitation, prevention, early detection and treatment of major postoperative complications, oxygen therapy, regulation of bowel/bladder function, nutrition and hydration (Table 3).

35The study by Marcantonio et al. reported the cumulative incidence of delirium during hospitalization and used the Delirium Symptom Interview (DSI), the Memorial Delirium Assessment Scale (MDAS), and the Confusion Assessment Method (CAM) to assess delirium. Lundstrom et al.measured delirium based on nurses’ interviews and the modified Organic Brain Syndrome Scale, which was applied at 3–5 days after hospital admission. The pooled results showed that the multicomponent intervention significantly reduced the incidence of delirium by 29% [RR 0.71 (95% CI, 0.59 to 0.86); I20%]. The overall GRADE quality of evidence was judged to be moderate (Table 4).

36Combining the former results with the single CCT, which had similar characteristics, yielded pooled results which remained statistically significant with no change in heterogeneity [RR 0.71 (95%CI, 0.60 to 0.84)] (Fig 3).

37In three of the 5 BA studies the CAM was employed to assess delirium, while a clinical evaluation was used by Williams et al.and SPMSQ and OBS scale was used by Bjorkelund. Moreover, the assessment time points varied considerably; Wong et al. measured the incidence of delirium every month, Williams et al.measured it within 5 postoperative days, while the remaining studies did not provide the time frame. However, an attempt to pool the data across the studies with patients that received orthopedic surgery, in a meta-analysis, yielded a statistically significant result in favor of the multicomponent interventions [RR 0.57 (95% CI 0.39 to 0.85): I2 27%, P = 0.25].

38Seven studies evaluated the efficacy of multicomponent interventions to prevent delirium in older patients hospitalized in medical departments. Two studies were RCTs, three were CCTs and two were BA studies(Table 5).

39Yoo et al. evaluated, in a CCT, the efficacy of interdisciplinary intervention by non-geriatricians to prevent delirium. The components of the non-pharmacological intervention in the remaining studies are shown in Table 3. Delirium incidence was a primary outcome in every study except in Asplund et al..

40Of the two RCTs, Asplund et al. did not report the exact method of random sequence generation and the personnel could have been aware of the group allocation with block randomization, therefore the study was judged to be at high risk of selection bias. In addition, Asplund et al. was at high risk of attrition bias, because the data were analyzed per protocol.

41In the remaining non-randomized studies despite selection bias is a threat to these studies, the baseline characteristics of the population were well balanced. Fig 2 describes the risk of bias table.

42The two RCTs differed substantially in the patient population evaluated. In Asplund et al. where delirium was not a primary outcome and patients were at low risk of developing delirium, non-pharmacological interventions were not able to prevent delirium [RR 1.75 (95% CI 0.50 to 6.10), GRADE quality of evidence was very low]. Conversely, in Martinez et al. patients were at high risk of delirium (age >70 years, a documented cognitive impairment, alcoholism and metabolic imbalances) and the multicomponent intervention, which was performed by family members, was able to reduce the incidence of delirium (evaluated daily with the CAM) by 58% [RR 0.42 (95% CI 0.19 to 0.92)GRADE quality of evidence was low].

43Two of CCTs investigated very similar patient populations (patients at intermediate/high risk of delirium), and type of multicomponent interventions (Table 3), targeting the same risk factors. In both studies, delirium was evaluated daily. The meta-analysis demonstrated a significant risk reduction of 35% of delirium incidence [RR 0.65 (95% CI 0.49 to 0.86); I2 0%] (Fig 4). The overall GRADE quality of evidence was judged to be moderate.

44The third CCT evaluated an intervention carried out by non-geriatricians which was not effective in preventing delirium.

45Before-after studies

46The two BA studies were different in the patient populations, settings and interventions (Table 5) and the interventions in the two studies were not effective in preventing delirium.

47Four RCTs evaluated the efficacy of non-pharmacological interventions to treat delirium in older patients hospitalized with acute illness in a medical ward trials and their characteristics are described in Table 6.

48The components of the non-pharmacological interventions differed between the studies, with only the individual care planning being a common element to all four studies (Table 3). Three studies measured the mortality rate as a primary outcome, while Cole et al. assessed cognitive improvement as a single primary outcome. The length of hospital stay was evaluated as a primary outcome in two studies.

49Two of the RCTs had an adequate method of randomization. In the study by Lundstrom et al., the randomization method was unclear, in addition to having a high risk of bias in the allocation concealment (their allocation method depended on the availability of a free bed). The allocation concealment in the study by Cole et al. was unclear. In terms of detection bias three studies displayed low risk of bias and one study had an unclear risk (Fig 2).

50The differences in the components of the interventions and the methods used to assess the outcomes precluded a meta-analysis. In the study by Lundstrom et al., delirium was assessed at day 1, 3 and 7 of hospitalization. Complete remission rate at day 7 was significantly higher in patients that received the non-pharmacological intervention [RR 1.58 (95%CI 1.15 to 2.17)]. Pitkala et al. assessed delirium severity using the MDAS scale. Despite the significant statistical difference in favor of the experimental group, the number of incident cases of patients who improved was not reported. Cole et al. used the CAM to assess delirium, but not to determine delirium improvement. In a subsequent publication in 2002, the authors assessed the time and rate of improvement of Delirium Index Score, without finding a significant effect of the intervention. The overall GRADE quality of evidence was judged to be very low (Table 4).

51Only two studies evaluated the functional status using the Barthel Index score and results were not statistically significant.

52A cluster RCT evaluated the efficacy of a nurse-led delirium abatement program for post-acute care (PAC) in skilled nursing facilities. The primary outcome was the delirium persistence at 2 weeks and 1 month after post-acute care unit admission.

53Methodological limitations of the trial are synthesized in Fig 2.

54The intervention allowed a better identification of delirium but was ineffective at reducing delirium.

55Nine studies evaluated the efficacy of single-component non-pharmacological interventions to prevent delirium in acute medical wards (Table 7).

56The risk of bias is summarized in Fig 5.

57Two RCTs evaluated the efficacy of Bright Light Therapy in an intensive care unit. The intervention reduced the incidence of delirium but without statistically significance[RR 0.29 (95% CI, 0.07 to 1.25)].

58Van Rompaey’s randomized trial studied the efficacy of using Ear Plugs to prevent delirium and found a hazard ratio of documenting any benefit for delirium prevention [RR 1.05 (95%CI 0.53 to 2.06)].

59In a large cluster-RCT Lapane et al. a Geriatric Risk Assessment MedGuide software was used to identify resident-specific medications that may contribute to delirium. The intervention significantly reduced the incidence of delirium in newly admitted residents in the intervention homes than those in usual care homes (hazard ratio 0.42, 95%CI 0.35 to 0.52).

60In a cluster RCT, Culp et al. evaluated the efficacy of Hydration management in 98 residents of 7 care homes without documenting any efficacy (RR 0.85, 95% CI 0.18 to 4.00).

61McCaffrey et al. studied Music Therapy in three RCTs but results of delirium incidence were not clearly reported.

62In a CCT, Tabet et al.evaluated the efficacy of Staff Education to prevent delirium in a medical ward. The incidence of delirium was significantly lower on the intervention ward despite a wide confidence interval [RR 0.50 (95%CI 0.26 to 0.96)].

63In a controlled BA study, Colombo et al. evaluated the efficacy of a Reorientation protocol in 314 critically-ill patients admitted to an ICU. Delirium occurrence was significantly lower in the experimental group [RR 0.63 (95%CI 0.44 to 0.91)].

64Table 8 displays the summary of finding with GRADE evidence profile for each single component.

65This systematic review was aimed at identifying systematic reviews and meta-analyses of non-pharmacological interventions used to prevent or treat delirium in patients aged 60 years or older to provide a summary for decision makers and guideline developers.

66From 26 SRs or meta-analyses meeting our inclusion criteria, we analyzed data from 31 primary studies published in the last 20 years.

67We found evidence of moderate quality supporting the efficacy of multicomponent non-pharmacological interventions to prevent delirium in older patients acutely admitted to a surgical or a medical ward. It must be emphasized, however, that these interventions are effective when administered to patient at intermediate/high-risk of developing delirium. When single component interventions were considered staff-education, reorientation protocol and the Geriatric Risk Assessment MedGuide software were the only interventions that produced significant reduction in delirium prevention albeit a wide confidence interval, likely due to the fact that only one study has been performed for each of them. However, staff-education, reorientation protocol and drug review were present in trials where the multicomponent interventions resulted effective in reducing the delirium incidence.

68On the contrary, evidence was insufficient to determine the benefit of multicomponent interventions in the prevention of delirium in other care settings (i.e., nursing homes). Finally, conflicting but mainly negative evidence was found concerning the utility of multicomponent non-pharmacological interventions to treat delirium in older medical patients.

69This is the first systematic review to gather evidence from different SRs and meta-analyses concerning non-pharmacological treatments for delirium.

70Firstly, most of the reviews included in our analysis had different aims and, consequently, the primary studies were distributed among the different reviews requiring the interested reader to consult all of them. For example, the review of Mak et al., focused on prevention of delirium in patients with hip fracture. Skingleyet al. were interested in music as a single component intervention to prevent delirium; Holroyd-Leduc et al.considered both pharmacological and non-pharmacological interventions without examining the single component intervention;the review by Hempenius et al. was limited only to delirium prevention (S6 Table shows how the primary studies were distributed among the reviews). Thus, our systematic overview provides a unique tool that synthesizes the evidence on non-pharmacological intervention.

71Secondly, unlike previously published reviews, we carefully examined the multicomponent intervention in order to identify the components that were common among different studies and used this information to decide whether it was feasible to perform a meta-analysis. In fact, in the assessment of non-pharmacological interventions for delirium prevention in older surgical patients, we were able to combine 2 RCTs together with the recently published CCT and from this combination of studies, we obtained a more precise estimate of the efficacy of the interventions. Similarly, in the setting of acute medical wards, we meta-analyzed Inouye and Vidan’s CCTs based on the fact that they had at least seven components in common.

72Third, we categorized the studies based on study design, the provision of intervention (for prevention or treatment), and the setting in which the intervention was provided as well as the risk of bias for each study. We believe that this classification approach will facilitate the formulation of clinical questions to assist clinicians to make decisions and to help guideline developers produce recommendations.

73Fourth, the strength of evidence is evaluated according to the GRADE items of risk of bias, inconsistency, indirectness, imprecision, and publication bias.

74We acknowledge that two studies were recently published that compares to our analysis. The first, was a meta-analysis that considered the efficacy of multicomponent interventions on, among other outcomes, delirium prevention. The study pooled the results of 11 studies including randomized, controlled clinical trials and non-randomized studies and found an OR 0.47 (95% CI, 0.38 to 0.58) in favor of non-pharmacological intervention. The second article is a practice statement from the American Geriatrics Society and it deals with the risk factors, diagnosis, and management of delirium including the use of non-pharmacological intervention for prevention and treatment of postoperative delirium. The authors identified 11 studies including randomized and non-randomized studies and state that the incidence of delirium was reduced but did not perform a meta-analysis. The conclusions of the two paper about multicomponent interventions to prevent delirium are similar to ours although our approach to meta-analyze data considered the setting in which the intervention is provided, the elements that composed the multicomponent interventions and the study design. In addition, we applied the GRADE approach and provide a summary of findings table that can provide the reader a complete account of the confidence of the evidence for the recommendation of multicomponent interventions.

75We acknowledge that our review has some limitation. Firstly, the arbitrary age cut-off that may limit the applicability of the evidence from the present overview to patients with less than 60 years of age. Secondly, the lack of assessment of cost-effectiveness reviews does not allow us to reach any conclusions regarding this topic. Thirdly, the studies examined were heterogeneous in terms of intervention, study design, population, outcome and instrument assessment. To address this issue we adopted the best available methodology, i.e. GRADE, to evaluate and synthesize the available evidence.

76Delirium is a geriatric syndrome and as such, it is expected to have a multifactorial etiology in the majority of patients. This implies that multicomponent interventions are those with the best chance of being effective. We did find that multicomponent interventions were able to prevent delirium in hospitalized older patients. However, the multicomponent interventions adopted in different studies were quite heterogeneous and therefore difficult to compare. We were able to identify some common elements among different interventions, but no evidence is available to allow either the identification of which multicomponent program is more effective than others, or the relative contribution of each constituent intervention to the positive results in each individual patient..

77The process of de-itemizing the multicomponent interventions used to provide the rational to pool meta-analyses may have another implication. For example, in delirium prevention in surgical setting, early mobilization, nutrition and hydration, regulation of bowel/bladder function, and early prevention of complications were the items that randomized studies had in common; in medical setting, in addition to early mobilization, nutrition and hydration, the interventions that the two controlled trials had in common were staff education, orientation protocol, avoidance of sensory deprivation. It is conceivable that these items may have more weight in determining the efficacy. Future studies may consider to design different multicomponent modalities in order to understand which items provide the most important contribution to the efficacy of the multicomponent intervention, how they are interrelated and the resources needed to implement the intervention.

78Another point worthy of comment is the fact that the multicomponent interventions work well to prevent delirium, but their efficacy to treat delirium is at best controversial, with more negative than positive results from the studies evaluated. There is no simple explanation for these findings. One possibility is that while the risk factors for delirium are well characterized and their management can effectively reduce the incidence of new episodes of this geriatric syndrome, the pathophysiology underlying the onset, development and persistence of confusion is not well understood. Several theories have been proposed, including impairment of cerebral metabolism (metabolic encephalopathy), intoxication by drugs, especially those with anticholinergic effects, inflammation, hypercortisolism, and a combination of the former, but no hypothesis has a strong experimental evidence in support of its validity.

79Therefore, multicomponent interventions investigated in delirium treatment studies are quite similar to those developed for delirium prevention, but they are applied in a very different clinical scenarios. Further research concerning the pathophysiology of delirium is clearly needed to provide the data to support the development of more effective interventions to treat older patients suffering from delirium.

80In older patients at intermediate/high risk of delirium multi-component non-pharmacological interventions as well as some single-components intervention (staff education, reorientation and drug review) reduce the incidence of delirium. Evidence for the role of non-pharmacological interventions in the treatment of delirium is inconclusive.