Delirium Prevention, Detection, and Treatment in Emergency Medicine Settings: A Geriatric Emergency Care Applied Research (GEAR) Network Scoping Review and Consensus Statement

Abstract

Background:

Older adult delirium is often unrecognized in the emergency department (ED), yet the most compelling research questions to overcome knowledge-to-practice deficits remain undefined. The Geriatric Emergency care Applied Research (GEAR) Network was organized to identify and prioritize delirium clinical questions.

Methods:

GEAR identified and engaged 49 transdisciplinary stakeholders including emergency physicians, geriatricians, nurses, social workers, pharmacists, and patient advocates. Adhering to Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Reviews, clinical questions were derived, medical librarian electronic searches were conducted, and applicable research evidence was synthesized for ED delirium detection, prevention, and management. The scoping review served as the foundation for a consensus conference to identify the highest priority research foci.

Results:

In the scoping review, 27 delirium detection “instruments” were described in 48 ED studies and used variable criterion standards with the result of delirium prevalence ranging from 6% to 38%. Clinician gestalt was the most common “instrument” evaluated with sensitivity ranging from 0% to 81% and specificity from 65% to 100%. For delirium management, 15 relevant studies were identified, including one randomized controlled trial. Some intervention studies targeted clinicians via education and others used clinical pathways. Three medications were evaluated to reduce or prevent ED delirium. No intervention consistently prevented or treated delirium. After reviewing the scoping review results, the GEAR stakeholders identified ED delirium prevention interventions not reliant on additional nurse or physician effort as the highest priority research.

Conclusions:

Transdisciplinary stakeholders prioritize ED delirium prevention studies that are not reliant on health care worker tasks instead of alternative research directions such as defining etiologic delirium phenotypes to target prevention or intervention strategies.

Delirium occurs in 8% to 10% of older adults in the emergency department (ED) and is the underlying reason for about 1.5 million ED visits annually in the United States.^1–4 Delirium manifests clinically as sudden onset neuropsychiatric dysfunction with a waxing and waning course characterized by inattention, disorganized thinking, and changes in level of consciousness. Delirium is not attributable to an established neurocognitive disorder, but is a direct consequence of one or more medical conditions.⁵ Older adults are susceptible to developing delirium while in the ED.^6,7 Delirium is often missed in the ED, because emergency staff are only 35% sensitive in detecting delirium.^1,8–12 Even when recognized, delirium may lead to prolonged hospital stay, functional decline, accelerated cognitive decline, and postdischarge depression.^13–16 When unrecognized in the ED, delirium is associated with increased 6-month mortality.⁸

In 2007, the Society for Academic Emergency Medicine (SAEM) Geriatric Task Force identified cognitive impairment (including delirium) as one of three conditions with substantial quality gaps for geriatric ED patients.¹⁷ In 2010, this SAEM Task Force identified four high-priority delirium research foci: 1) optimal methods of screening for and diagnosing delirium in the ED, 2) risk factors for delirium in the ED and interventions for prevention and moderation, 3) criteria for safe discharge for older ED patients with delirium, and 4) interventions to improve outcomes for older patients with delirium in the ED.¹⁸ In response, efforts to improve the timely recognition and management of older ED patient delirium have developed. These ED-focused responses include resident core competencies,¹⁹ multiorganizational operational guidelines,²⁰ and quality indicators.^21,22 Despite these efforts, the best practices for preventing, identifying, and managing delirium in the ED are still unknown.

The Geriatric Emergency care Applied Research (GEAR) Network is an interdisciplinary group of clinicians and researchers focused on the identification and execution of high-impact research questions for geriatric emergency medicine domains with substantial and clinically important knowledge gaps. GEAR has addressed questions in medication safety, elder abuse, falls, and care transitions using a framework of patient–intervention–control–outcome (PICO questions), scoping literature review, and building consensus on best practices and areas of future research.²³ The primary objective was to prioritize high-yield older adult delirium research questions for the ED setting through an evidence-based consensus statement. A secondary objective was to report ED best practices for delirium prevention, recognition, and treatment based on a scoping review of published emergency medicine research.

METHODS

Study Design

Participating GEAR researchers were identified by their active membership in geriatric emergency medicine interest groups (within the SAEM, American Geriatrics Society, and Gerontological Society of America) and prior pertinent published research. GEAR researchers chose to participate in one of five subgroups as determined by the GEAR core team: cognitive impairment, medication safety, falls, geriatric abuse, and care transitions. The GEAR Cognitive Impairment (GEAR CI) subgroup included three geriatricians (UO, WWH, LAL); four emergency physicians (CRC, SMD, MS, UH); three research assistants (NH, EAL, DW); and one PhD researcher (KJK) with expertise in geriatrics, gerontology, and health services research.

The GEAR CI subgroup performed a scoping review that adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) reporting guidelines.²⁴ The 14 GEAR CI members participated in monthly teleconferences to derive pertinent PICO questions, prioritize the PICO questions, derive a reproducible search strategy, independently filter the results of the electronic search, abstract key study data from the studies that met inclusion criteria, and synthesize the research findings into best practices based on currently available research.

PICO Questions

The GEAR CI subgroup derived and refined the following PICO questions:

PICO-1

• Population: ≥65-year-old ED patients from any pre-ED setting;

• Intervention: Risk stratification for delirium during ED episode of care using feasible and validated instrument;

• Comparison: Clinician gestalt for delirium;

• Outcome: Sensitivity, specificity, likelihood ratio, receiver operating characteristic area under the curve (ROC AUC), association of delirium identification with ED revisits and mortality, comparison of claims data delirium incidence with prospective ED study data.

PICO-2

• Population: ≥65-year-old ED patients identified as high-risk for delirium during ED episode of care;

• Intervention: Action to reduce the occurrence (i.e. prevent), duration, or severity of prevalent or incident delirium;

• Comparison: Standard care;

• Outcome: ED length of stay, hospital length of stay, ED returns, quality of life, mortality at 1 year.

PICO-3

• Population: ≥65-year-old ED patients from any pre-ED setting;

• Intervention: Risk stratification for dementia during ED episode of care using feasible and validated instrument;

• Comparison: Clinician gestalt for dementia;

• Outcome: Sensitivity, specificity, likelihood ratio, ROC AUC, association of dementia with ED revisit rates, comparison of claims data dementia incidence with prospective ED study data.

PICO-4

• Population: ≥65-year-old ED patients identified as high risk for dementia during ED episode of care;

• Intervention: Referral for definitive dementia diagnostic testing;

• Comparison: Standard care;

• Outcome: ED length of stay, ED returns, quality of life, institutionalization at 1 year.

The GEAR CI subgroup identified exemplar articles for each PICO question and surveyed 33 GEAR investigators from all five cores to prioritize the PICO questions via teleconference discussions and online voting. PICO questions 1 and 2 were selected as highest priority.

Search Strategy

Based on PICO questions 1 and 2 as well as exemplar articles that a search should identify, a medical librarian (MD) created electronic search strategies for OVID MEDLINE, Embase, CINAHL, and CENTRAL from inception through November 2019. Full details of the search terms are provided in Data Supplement S1, Appendix S1 (available as supporting information in the online version of this paper, which is available at http://onlinelibrary.wiley.com/doi/10.1111/acem.14166/full).

Study Selection and Data Abstraction

Two authors independently reviewed the titles and abstracts for the search results of each PICO question (PICO-1 = CRC, KJK; PICO-2 = CRC, UO). Inclusion criteria for PICO-1 were adults over age 65 with quantitative evaluation of one or more delirium identification methods in ED settings. Inclusion criteria for PICO-2 studies were one or more intervention(s) intended to improve the identification or outcomes of delirium in ED patients over age 65. No publication date or language exclusion criteria were applied. Unweighted Cohen’s kappa (κ) was used to quantify interrater agreement. An adjudicator (UH) was named a priori to resolve continued interrater discrepancies.

Two authors (PICO-1 = CRC, NH; PICO-2 = CRC, EAL) reviewed the selected articles and conference abstracts independently collecting key elements using a predesigned template. Information abstracted included the study setting, inclusion/exclusion criteria, study design, comparator reference standard or control group, and primary/secondary outcomes. The first authors of unpublished abstracts were contacted to determine whether that research has since been published. Manuscript and abstract authors were also contacted for additional study details if key elements could not be abstracted from the publication. Studies reporting none of the PICO outcomes were excluded. Data abstractors identified additional studies for potential inclusion by reviewing full-text articles’ references.

Synthesis of Best Practices and Building Consensus

Abstracted data were independently reviewed and summarized by three authors (UH, UO, SMD). We reported preliminary results to the CI subgroup, refined concepts and recommendations via e-mail and teleconference, and presented at an in-person consensus conference in October 2019 that included the entire GEAR network. The GEAR CI subgroup presented recommendations for prioritizing ED delirium research and GEAR Consensus Conference attendees voted on the final rank order of research priorities. Two non-emergency medicine delirium experts subsequently reviewed, edited, and endorsed a statement regarding the impact of the scoping review by the CI subgroup. Two patient, family, and caregiver representatives (LH, LN) reviewed this scoping review and PowerPoint presentations from the GEAR Consensus Conference and then provided their perspectives via e-mail and teleconference calls.

RESULTS

Evidence Synthesis

PICO-1: Delirium Risk Stratification.

After duplicates were removed, the literature search identified 814 potential studies. One study was identified when one author served as an invited peer reviewer for a journal submission. The initial independent review demonstrated moderate agreement (κ = 0.50, 95% confidence interval [CI] = 0.41 to 0.59).²⁵ After full-text review, adjudication, and abstraction, 48 articles were included in the scoping review (see Figure 1 for details of the inclusion and exclusion of these studies).²⁴

Figure 1.

Summary of scoping review results for delirium screening for older adults in emergency settings.

Study Characteristics.

Of the 48 selected studies, eight were reported only as conference abstracts.^26–33 These studies analyzed 27 different instruments’ accuracy to risk stratify for delirium. The most commonly evaluated stratification tool was clinician (nurse or physician) judgment (“gestalt”; studied in 12 papers).^{11,12,27,28,30,32,34–39} The next most commonly evaluated instruments were the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU; five),^39–43 Confusion Assessment Method (CAM; four),^27,29,44,45 Brief Confusion Assessment Method (bCAM; four),^46–49 Richmond Agitation Sedation Scale (RASS; four),^50–53 4AT (three),^29,54,55 and the tablet “serious game” (three).^56–58

Participant Characteristics.

The sum of study populations totaled 15,860 patients and two systematic reviews of 12 ED-based studies.^45,59 One group published seven manuscripts reporting diagnostic accuracy of different delirium screening instruments on the same population, so these patients are only counted once.^{42,47,48,51,52,60,61} Patients were recruited from 1992 to 2016 and the manuscripts were published between 1995 and 2019. The majority of studies were prospective observational (35),^{11,12,26,27,30,32,34–36,38,40–43, 46–52,54–58,60–68} followed by chart review (three),^37,39,69 before/after (two),^31,70 case–control (one),⁷¹ randomized controlled trial (one),²⁹ and a mixed prospective/retrospective study (one).⁵³ Three studies failed to report their research design.^28,33,44

Delirium Screening Instrument Characteristics.

Reported delirium prevalence in the ED ranged from 6% to 38%.^30,70 Sensitivity of the instruments ranged from 0% for clinician documentation²⁷ to 64% for 4AT.⁵³ Specificity of the instruments reported ranged from 27% for the Bergman-Paris question⁶⁷ to 100% for clinician gestalt.¹² The most commonly used reference standards were the CAM (18);^{12,26–28,30,32,33,36,38,54, 56–58,62,64,66,67,70} Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria applied by neurologist, geriatrician, or psychiatrist (16);^{29,39,42,44,46–48,50,51, 53,55,60,61,65,68,69} and the CAM-ICU (three).^43,52,63 Two original data studies did not report their reference standard^31,37 (for a summary table of each included study’s population, inclusion/exclusion criteria, study design, reference standard, and diagnostic accuracy outcomes, see Data Supplement S2, Appendix S2). The components and original derivation or validation study for each delirium instrument are provided in Data Supplement S3, Appendix S3 (reproduced with permission from their original sources), while Data Supplement S4, Appendix S4, provides pragmatic instrument-level details regarding diagnostic accuracy, time to administer, and potential advantages or disadvantages for ED use.

Clinician gestalt was the most commonly studied risk assessment method for delirium. However, clinician gestalt varies significantly depending on how “gestalt” is measured and which physician’s gestalt is measured. Sensitivity is as low as 0%^27,32 (when relying on a diagnosis recorded in the medical record) or as high as 81% when prospectively asking emergency physicians if they believe there is more than a 5% chance that their patient has delirium.³⁰ Specificity ranges from 65% (when prospectively asking an emergency physician if they believe there is more than a 5% chance their patient has delirium) to 100% when delirium is documented in the medical record.¹²

Time on task for formal screening tools range from 20 seconds (Delirium Triage Screen)⁴⁷ to 20 minutes (CAM).⁷² RASS (1 minute) and bCAM (2 minutes) demonstrated intermediate-range times on task. Interrater reliability was measured in 11 studies: CAM (κ = 0.7–0.91),⁷⁰ bCAM (κ = 0.87–0.88),^46–48 individual questions on the bCAM (κ = 0.25–0.50),⁴⁸ CAM-ICU (κ = 0.92),⁴² mCAM-ED (κ = 0.79),³⁵ “single question” (κ = 0.75),⁶¹ and RASS (weighted κ = 0.63).⁵¹ For the “serious game,” test–retest reliability was reported by Spearman’s rho and ranged 0.56 to 0.85.⁵⁷

PICO-2: Delirium Interventions to Reduce the Occurrence (i.e., Prevent), Duration, or Severity of Delirium.

The literature search identified 761 potential studies (see Figure 2 for details of the inclusion and exclusion of these studies).²⁴ The initial independent review of the raw search results demonstrated fair agreement (κ = 0.36, 95% CI = 0.22 to 0.50).²⁵ After full-text review, adjudication, and abstraction, 15 “intervention” studies were included in the scoping review.

Figure 2.

Summary of scoping review results for delirium interventions for older adults in emergency settings.

Study Characteristics.

All included studies had interventions implemented in the ED from 1996 to 2015 and published between 2001 and 2018. Among the 15 studies, one was a randomized, placebo-controlled trial⁷³ while five were retrospective observational and pre-/post- studies.^74–78 Seven of the studies were prospective observational and pre-/post- studies,^1,34,79–83 while another was a case–control study⁸⁴ and another was a pre-/post- survey following a workshop⁸⁵ (for a summary table of each included study’s population, inclusion/exclusion criteria, study design, intervention, and outcomes see Data Supplement S5, Appendix S5).

Participant Characteristics.

The sum of study populations totaled 9,507 individuals. However, significant between-study differences in subjects recruited exist. For example, some interventions focused on older adults admitted for acute medical illnesses,⁸¹ whereas others were solely hip fracture patients^{77,80,82–84} and others were critically ill patients admitted to the ICU from the ED.⁷⁹ In other studies, the subjects were members of the health care team.⁸⁵ For these studies focusing on health care team subjects, the numbers and characteristics of ED nurses and physicians who participated in workshops and grand rounds were not reported.

Intervention Characteristics and Outcomes

Study intervention fell into three different categories. The first consisted of clinician-targeted interventions, ranging from educational interventions for nurses and physicians^81,82,85 to provision of reminders to providers.⁷⁸ Second were care process interventions consisting of implementation of emergency care bundles,⁸³ clinical pathways,^80,84 screening processes,³⁴ and provision of screening results to physicians.¹ Other care process interventions included patient evaluations in the ED by geriatricians and interdisciplinary team rounds^75,76 or early hospitalization or admission to the acute geriatric unit when delirium was recognized.⁷⁴ Finally, the third category was pharmacologic interventions which included an adapted pain management pathway⁸⁴ and use of antipsychotics (haloperidol and blonanserin).^73,79

Primary outcomes included detection and incidence of delirium in seven studies,^{34,73,75–77,83,84} change in patient’s care plan measured by alteration in disposition or additional diagnostic evaluation in two studies,^1,81 and repeat ED visits and hospital readmissions in one study.⁷⁴ Primary outcomes were not defined in five of the studies.^78–82 Secondary outcomes studied include severity and duration of delirium;⁷³ secondary complications such as falls, fractures, aspiration pneumonias, and mortality;³⁴ and health care utilization characteristics such as ED revisits, hospital readmissions and associated cost, length of stay, or placement in a long-term care facility.^1,34,73,74 Other secondary outcomes studied included reduction in pain⁸⁴ and referral for a geriatric assessment at home.⁸⁵ Eight of the studies did not report any secondary outcomes.

Clinician Targeted Interventions

Among the interventions, there were six provider-targeted interventions. Three of these interventions focused on provider education through workshops, grand rounds, and small group meetings.^81,82,85 Two of the educational interventions also included implementation of process changes.^34,80 The other three provider interventions focused on provider reminders for delirium screening and provision of screening results to the physicians.^1,78 Two before–after retrospective studies by the same investigators evaluated geriatrician care in the ED for Parkinson’s disease patients.^75,76 Among the educational interventions was a 1-day workshop for clinicians that was associated with an improvement in the outcomes of increased delirium screening (p = 0.006) and referral for home care and geriatrics evaluation.⁸⁵ Educational interventions were also significantly associated with reduced delirium duration from a median of 4 days to 1 day and reduced severity by 2.94 points using a modification of CAM,⁸² reduced use of benzodiazepines (p < 0.01) and antihistamines (p < 0.05), and reduced length of stay. Naughton et al.⁸¹ used small group meetings between the ED and an acute geriatrics unit with audit and feedback of nurses to reduce delirium and estimated that each case of delirium prevented saved a mean of 3.4 hospital days. An educational intervention paired with development and preferential admission to acute geriatric unit for patients with cognitive impairment or delirium was associated with a decrease in prevalent delirium on hospital day 4 from 41% to 19%.⁸¹ Among interventions focused on providing reminders to providers, no significant increase in proportion of patients diagnosed with delirium,¹ change in patient care plans by emergency physicians,¹ nor reduction in secondary complications of delirium postintervention were observed.³⁴ Using an electronic medical record reminder for ED triage nurses to obtain RASS and bCAM with an automatic physician alert when delirium screening was positive, Delaney et al.⁷⁸ demonstrated a reduction in 30-day ED return visits from 90 to 35 per month.

Care Process Interventions

These studies focused on establishing clinical pathways and implementation of emergency care bundles along with interdisciplinary team rounds and geriatrician ED visit.^{74–76,80,83} For example, the pathway described by Fleury et al.⁸⁰ for ED patients with suspected proximal femur fractures included imaging protocols and requisite labs followed by a geriatric consult “to detect, treat, or prevent delirium.” The description of components of emergency care bundles by Thomson et al.^75,76 are completely lacking and are only evaluated in individuals with Parkinson’s disease so may not be applicable to other populations. Emergency care bundles were associated with increased delirium detection rates from 22% to 29%, reductions in hospital length of stay from 16 to 11 days, and reduced mortality from 24% to 14% for patients with identified delirium.^75,76,80 Other care processes evaluated included maintaining oxygen saturation, intravenous fluids, preventing hypotension, and temperature changes.⁸³ A multifactorial protocol extending from the community prehospital through the ED to postoperative settings for patients with suspected hip fracture was associated with lower rates of postoperative delirium (33% to 21%) and more rapid ED-to-ward time (4.6 to 2.7 hours).⁸³ In one observational study using the National Health Insurance Research Database examining 2,780 ED visits for patients with delirium between 2000 and 2008, decisions to admit patients with delirium at the initial ED visit reduced ED returns and subsequent hospitalizations within 28 days when compared with discharge home during the initial ED evaluation.⁷⁴

Pharmacologic Interventions

Three medications were explored to reduce or prevent ED delirium: haloperidol,⁷³ blonanserin,⁷⁹ and regional anesthesia.⁸⁴ Neither haloperidol nor opioid analgesics demonstrated significant reductions in delirium incidence or severity. On the other hand, an open-label, noncontrolled, nonrandomized trial of blonanserin reduced hyperactive delirium by 33% and mixed delirium by 35%.⁷⁹ Patients who received scheduled prophylactic haloperidol had more severe and longer duration of delirium.⁷³ LeBlanc et al.⁸⁴ described no significant decrease in delirium for hip fracture patients receiving a nerve block compared with opioid therapy.

Consensus of Highest-yield ED Delirium Research Priorities

Based on the scoping review above and prior to the consensus conference the GEAR CI subgroup attained agreement on the following primary research objectives in order of highest to lowest priority with the objective to accelerate the identification and reduction of ED delirium:

1. Phenotyping: Define ED delirium phenotypes based on clinical features (presenting complaint, underlying cognitive frailty, and suspected delirium precipitant) and/or readily available delirium biomarkers.^86–88 Then, by phenotype and severity^89,90 stratify delirium identification strategy accuracy and intervention effectiveness.

2. Building for Implementation: Evaluate the pragmatic reproducibility, fidelity, adaptability, and sustainability⁹¹ of different ED delirium detection protocols ranging from nurse/physician education to trained volunteers, computer-based screening, and family report.

3. Phenotype-targeted Interventions: Rather than apply monomorphic intervention(s) to every delirium identified, base ED intervention studies on biologically plausible delirium precipitant informed by delirium phenotype research and ED-validated delirium identification strategies.

4. Transdisciplinary Expansion: Harmonize delirium identification and phenotype-guided intervention strategies across medical and surgical specialties to build the foundation for more transparent interdisciplinary comparative analyses and catalyze meaningful transdisciplinary research.^92,93

5. Organized Knowledge Acquisition: Develop a Pediatric Emergency Care Applied Research Network (PECARN)-like multi-institutional data repository of ED delirium cases identified using the same methods permitting stratification by delirium phenotype.⁹⁴ The data repository would include key prognostic variables (dementia,⁹⁵ frailty⁹⁶); pathophysiology-guided interventions; and both process-and patient-centered outcomes including the burden of delirium on caregivers.⁹⁷

During the consensus conference, attendees added to, reworded, and reranked the proposed research priorities. The GEAR work group included 49 members. Those who were able to attend the consensus conference in person voted on research priorities using polleverywhere.com. Those who did not attend completed their rankings through a RedCap Survey link that was e-mailed. There was 100% voting participation by all 49 members of the task force. Table 1 represents the final GEAR research recommendations for ED delirium care.

Table 1.

Consensus Conference Ranking of ED Delirium Research Priorities

Research Priority Rank—Descriptor

1—Prevention focus including developing screening instrument/risk score that does not entail additional nurse/physician workload

2—Testing delirium prevention strategies in ED once high-risk individuals identified

3—Identifying high-risk patients for cognitive impairment (lumping delirium and dementia together) and then unique care bundles depending on dementia, delirium, or delirium superimposed on dementia

4- Build a multi-institutional data repository and research infrastructure for delirium hypothesis testing similar to PECARN

5- Structured implementation science evaluation of ED delirium detection and intervention strategy pragmatic reproducibility, fidelity, adaptability, and sustainability

6—Identifying delirium phenotypes via pathophysiology, anticipated disease trajectory, anticipated response to interventions, and accuracy of different ED screening instruments

7—Targeting delirium prevention or treatment interventions based on delirium phenotypes based on biologic plausibility of effect and pathophysiologic etiology of episode delirium

DISCUSSION

This PRISMA-ScR adherent scoping review and GEAR consensus process provides a rigorous approach to defining the current state of ED delirium research, while outlining a high-yield path forward for multiple stakeholders. Delirium is prevalent, harmful, detectable, and potentially preventable in the ED, but will require a shift from tool-based research to pragmatic interventional trials.^98–100 Over two dozen ED delirium screening instruments exist of varying complexity and levels of validation. Unfortunately, many delirium-screening studies focus on diagnostic accuracy without measuring or contemplating health care personnel acceptance, incorporation into practice, or sustainability of use in the busy ED.¹⁰⁰ Although researchers from non-ED settings have evaluated psychometric properties such as tool development methods, reliability, validity, feasibility, and implementation, these delirium screening instrument properties are rarely evaluated in emergency medicine settings.¹⁰¹ The psychometric properties of delirium screening instruments from non-ED settings in non-English languages have also been studied.^102,103

Emergency department delirium instrument diagnostic accuracy studies also rarely cite adherence to Standards for Reporting Diagnostic Accuracy (STARD) guidelines calling into question the rigor of the studies,^104,105 notably the risk of common biases that skew observed sensitivity/specificity such as incorporation bias, partial verification bias, differential verification bias, imperfect criterion standard bias, or spectrum effect.¹⁰⁶ The relevance of these diagnostic biases for clinicians and researchers is that the true sensitivity and specificity of delirium screening instruments is unknown. If researchers minimized or eliminated these biases, sensitivity and specificity may be higher or lower than reported. In addition, delineating accuracy (the ability to distinguish those with delirium from those without) is the lowest tier of diagnostic research. More impactful research is required to determine whether using any of the ED delirium instruments identifies a subset of delirium patients not previously recognized or whether that newly identified subset of delirium patients can be positively impacted in any measurable way by any intervention.^107,108 Once a diagnostic screening instrument demonstrates accuracy, randomized controlled trials are helpful to evaluate patient-centric benefit or harm.¹⁰⁸ Thus far, these types of efficacy trials for ED delirium-screening instruments do not exist. As an example of this diagnostic impact research, brain natriuretic peptide (BNP) is undeniably accurate to distinguish congestive heart failure from other causes of acute decompensated dyspnea, but multiple trials randomizing ED physicians to be aware of the BNP result or to not be aware of it demonstrate no consistent differences in length of stay, admission rates, mortality, or readmission rates.¹⁰⁹

Gestalt is the most frequently evaluated approach to identify delirium and has the advantage of not adding burden to the clinician workload. However, significant interphysician variability exists so a role for an accurate and reliable delirium instrument exists. Among existing delirium screening instruments, CAM has been most extensively validated for accuracy in hospital settings with 94% sensitivity and 89% specificity,¹¹⁰ but far fewer ED studies exist.^27,29,44,71 The time and equipment (photos, training) required to administer CAM in real-world ED settings is an inconvenient but often ignored reality between inpatient and ED environments that limits feasibility and acceptance of CAM. ED health care teams, which have limited time and are often interrupted while performing tasks, judge usability on brevity and accuracy.¹¹¹ CAM requires more time and training to administer, so ED providers may sacrifice a degree of accuracy for feasibility.⁷² For example, bCAM requires 2 minutes to administer with a positive likelihood ratio of 10 to 20 that is quite accurate to rule-in delirium.^46,47 The Delirium Triage Screen requires 20 seconds to administer with a negative likelihood ratio of 0.04, which is quite accurate to rule out delirium.⁴⁷

Furthermore, ED delirium intervention studies depend on accurate ED delirium screening instruments, yet ED-specific validity of the tools to distinguish patients with delirium is rarely reported in intervention studies. Interventional trials to reduce the occurrence or duration of delirium during or immediately following an episode of ED care are limited in number and varied in interventional focus, populations evaluated, and outcomes assessed. The handful of ED delirium intervention studies explores a variety of unrelated interventions on heterogeneous patient populations with different outcomes assessed at variable time frames postintervention. Almost every interventional study is a nonrandomized trial design with significant risk of bias for any measured effects (or lack thereof). Clinician-targeted interventions are associated with increasing rates of delirium screening and referral to home care and geriatrics evaluation, but unclear impact on patient-oriented outcomes.

There is a paucity of evidence for effective ED-based prevention and treatment strategies for delirium. This lack of evidence stands in stark contrast to evidence on delirium prevention and management for ICU, ward, and postoperative settings.^112–116 The ED is often the first point of entry to the hospital and presents an upstream opportunity to improve older adult delirium outcomes across the continuum of care. Identifying patients with delirium or those at risk for delirium is particularly important in the ED since delirium commonly occurs with acute illness or injury. Early identification of delirium (or risk for delirium) may catalyze more effective prevention and treatment protocols. The challenges in studying and adopting delirium screening and intervention practices in the ED setting pertain largely to time constraints. Pragmatically, less time is available to assess older adults in the ED, collateral information is often limited, and quicker decision making is expected. Optimizing the environment and resources used for prevention or treatment of delirium mandates adaptation to the ED setting.

The available single-center observational studies of changes in the processes of care for patients with possible delirium report reductions in delirium prevalence and duration and hospital length of stay.^{74–76,80,83} Unfortunately, these process intervention studies inadequately report the components of care process changes and evaluate heterogeneous outcomes including delirium detection rates, ED and hospital lengths of stay, mortality, postoperative delirium incidence, and incidence of hypoxia. Because of the observational study designs, variability of care process components implemented, study environments, and outcomes assessed, additional studies are required to delineate the specific care processes or combinations that positively or negatively influence delirium outcomes.

More ED-based delirium research is essential to guide clinicians on best practices for delirium detection in the ED setting, along with interventions to reduce incident delirium and to treat prevalent delirium. Compared with operative, ICU, and ward settings, the ED environment is uniquely challenging for the identification, treatment, and investigation of delirium, but extrapolating delirium research from dissimilar settings is imperfect and possibly harmful. Interventions to identify, prevent, or treat delirium need to consider aspects of feasibility, acceptability, and effectiveness when adopted in the ED setting, ideally utilizing implementation science and human factors engineering approaches.¹¹⁷ Because pharmacologic studies have demonstrated disappointing results on the prevention or treatment of delirium in the ED setting, future ED delirium studies might evaluate the pragmatic application of nonpharmacologic interventions currently demonstrating effectiveness in hospital settings.¹¹² Potential intervention strategies include targeting delirium risk factors such as immobility, vision or hearing impairment, functional decline, dehydration, and sleep deprivation as well as pharmacologic measures of avoiding or reserving the use of antipsychotics or sedation only in patients with severe agitation posing risk to patient or staff safety.¹¹⁸

Existing ED delirium screening and intervention research conceptualizes delirium as a monomorphic entity without contemplating potentially unique patient phenotypes.^119,120 Do medical, surgical, or traumatic emergencies lead to different pathophysiology of delirium? Does delirium in the cognitively frail individual differ in neurobiologic etiology from those without underlying cognitive deficit? Does the delirium precipitant (e.g., medications, infections, pain, or other physiologic precipitant) affect the diagnostic accuracy of ED delirium screening instruments or the observed effectiveness of delirium interventions? ED delirium interventions require extensive funding to facilitate research and perhaps more importantly time to fully understand the pathophysiology of delirium as these frail individuals enter the front door of the hospital. In addition, the timing of delirium identification and interventions might be associated with effectiveness, so ED researchers could explore collaborative approaches with long-term care facilities and/or ambulance services. For example, long-term care facilities could provide a video depicting an individual patient’s baseline mental status for emergency medical services or ED personnel to review when assessing for delirium. Much in the same way that early sepsis treatment has been effective even though similar interventions had not worked later in the hospital course for critically ill patients, delirium may be best treated at the earliest possible moment.^121–123 Interventions that have been ineffective in the hospital may be more beneficial when delivered earlier in the course of delirium and deserve to be evaluated in the ED with a better understanding of delirium phenotypes and stratified analyses of those subtypes, to understand whether some subtypes may be more responsive to different interventions.

Through the GEAR consensus conference process, two new research priorities were added in addition to the five presented by the GEAR CI subgroup (Table 1). These two priorities (1—prevention focused research, including developing a screening instrument or risk score that does not entail additional nurse or physician workload and 2—testing delirium prevention strategies in the ED once high-risk individuals are identified) were subsequently selected as the top research priorities by the GEAR Consensus Conference participants after some debate. GEAR members who were not involved in the scoping review, but who were the first to hear its results, added these two priorities. These recommendations are the most basic of the seven research priorities ultimately identified. Other research priorities representing broader tiers of inquisition included: 3—providing unique care bundles, 4—building a multi-institutional data repository and research infrastructure, 5—a structured implementation science evaluation of delirium, 6—identifying delirium phenotypes, and 7—targeting delirium prevention or treatment based on delirium phenotypes. Although GEAR Consensus Conference attendees did not provide rationale for their votes, their votes seem to reflect the paucity of data for ED-based delirium risk-stratification and interventions, thereby favoring a back-to-basics approach.

After the scoping review and consensus conference we obtained further stakeholder input.⁹³ Two experts in delirium (a geriatrician and an intensivist) endorsed a statement on the scoping review and original research priorities in Data Supplement S6, Appendix S6. In summary, they note that GEAR has revealed significant delirium knowledge gaps between management recommendations and objective evidence of benefit for delirium patients and their families. There is not a single controlled ED study on which to base actionable protocols of delirium prevention or delirium reduction interventions. GEAR’s priorities to refocus ED delirium detection research on motor and etiology-subtypes aligns with emerging delirium research priorities outside of emergency medicine. The collaborative research laboratory envisioned by GEAR is highly appealing beyond emergency medicine and provides the accelerant needed for improved delirium outcomes. It is notable that these experts highlight the more advanced and complex research priorities, which were high priority for the GEAR CI subgroup, but ultimately deprioritized in the GEAR Consensus Conference.

Patient, family, and caregiver stakeholders provided the summary of the scoping review and consensus conference in Data Supplement S7, Appendix S7. This summary notes that screening for delirium or risk of developing delirium is important to minimize the longer-term negative effects of delirium; however, screening is difficult because of delirium’s complexity. Several screening tools are available, but few have been tested in the unique environment of the ED. Also, timing of the screening is important and it appears from the reviewed research that early screening is preferable to prevent downstream complications. Important gaps in current knowledge that are priorities for future research projects include studying the effects of delirium on patient health outcomes and improving the pragmatic implementation of delirium screening in the ED. Future treatment or prevention trials should focus on patient-centered outcomes in quantifying potential benefits and harms of delirium interventions aligned with individualized risk factors and goals of care.

LIMITATIONS

This scoping review and evidence-based consensus statement has several limitations. PICO-1’s focus on risk stratification might be interpreted to mean research to evaluate the risk of developing delirium (prognostic, incident delirium) rather than identifying delirium at a point in time (diagnostic, prevalent, or incident delirium). The GEAR CI subgroup’s intended focus was the latter diagnostic accuracy priority. PICO-2’s intervention also broadly targeted primary delirium prevention with secondary prevention and prevalent delirium amelioration strategies. Each of these delirium interventions may require quite different strategies and lumping them together may be unhelpful. In addition, delirium can accelerate cognitive decline, but the well-recognized morbidity was not an outcome of either PICO-1 or PICO-2. Dementia is another subtype of cognitive impairment commonly encountered and often unrecognized in ED settings, yet GEAR stakeholders excluded PICO-3 and PICO-4, which pertained to dementia. The National Institute of Aging recently awarded investigators to develop and pilot test strategies to improve the process and outcomes of care for ED dementia patients and their caregivers, so these PICO questions will soon be explored.

We also report no quality assessment because this scoping review was not designed as a systematic review. Nonetheless, our scoping review demonstrates an overall paucity of evidence regardless of the quality, particularly in the area of delirium interventions. Therefore, current research cannot adequately inform effective ED-based delirium interventions. Significant effort was made to identify all studies in accordance with PRISMA-ScR guidelines;²⁴ however, it is possible that our search strategy missed relevant studies or that new studies have emerged since the scoping review was performed (November 2019). Finally, a diverse group of stakeholders, including clinicians of varying specialties, researchers, patients, and their advocates, attended the consensus conference. However, this group does not encompass all possible stakeholders. A significant divergence of opinion between the GEAR CI subgroup and the remainder of GEAR’s stakeholders occurred in that research priorities presented by the GEAR CI subgroup and endorsed by delirium experts in intensive care and geriatrics were ultimately rejected by the rest of GEAR. Finally, all participants in the scoping review and consensus conference are from North America. Additional perspectives are important to develop more globally relevant ED delirium research priorities.

CONCLUSIONS

The results of this scoping review demonstrate over two dozen ED delirium-screening instruments of varying complexity and degrees of validation. However, ED-based interventional studies to prevent or treat delirium are almost nonexistent. Top research priorities for cognitive impairment in the ED continue to focus on the basics of identification, prevention, and treatment of delirium in the ED including: 1) development of a screening instrument or risk score that does not entail additional clinician workload and 2) testing delirium prevention strategies in the ED once high-risk individuals are identified.

Supplementary Material

supinfo2

Data Supplement S2. Delirium screening study synopsis.

supinfo 3

Data Supplement S3. Appendix of ED delirium screening instruments.

supinfo 4

Data Supplement S4. Administration time, accuracy, and pragmatic features of delirium screening instruments.

supinfo 5

Data Supplement S5. ED delirium intervention study synopsis.

supinfo 6

Data Supplement S6. Transdisciplinary commentary of GEAR cognitive impairment.

supinfo 7

Data Supplement S7. Patient/Family caregiver perspective on GEAR CI recommendations.

supinfo 1

Data Supplement S1. Search strategy details.