Does Alzheimer’s Disease and Related Dementias Modify Delirium Severity and Hospital Outcomes?

Abstract

Background/Objectives:

We examined the association between delirium severity and outcomes of delirium among persons with and without Alzheimer’s Disease and Related Dementias (ADRD).

Design, Setting, Participants:

Prospective cohort study at of medical and surgical patients at an academic tertiary medical center (N=352).

Measurements:

Delirium incidence and severity were rated daily using the Confusion Assessment Method (CAM) and CAM-severity (CAM-S) score during hospitalization. Severe delirium was defined as a CAM-S short form score in the highest tertile (3-7 points out of 7). ADRD status was determined by a clinical consensus process. Clinical outcomes included prolonged length of stay (>6 days), discharge to post-acute nursing facility, any decline in Activities of Daily Living (ADLs) at one month from pre-hospital baseline, ongoing nursing facility stay and mortality.

Results:

Patients with ADRD (n=85, 24%) had significantly higher relative risk (RR) for incident delirium, RR=2.31, 95% Confidence Interval (CI) 1.64-3.28 and higher peak CAM-S scores, mean difference=1.24 points, CI 0.83-1.65 (p<0.001). Among patients with ADRD, severe delirium significantly increased the relative risk for nursing facility stay (RR=2.22, CI 1.05-4.69), p=0.04 and increased the relative risk for mortality (RR=2.10, CI 0.89-4.98), p=0.09. Among patients without ADRD, severe delirium was associated with significantly increased risk for all poor outcomes except mortality – including prolonged length of stay in hospital (RR=1.47, CI 1.18-1.82) and discharge to post-acute nursing facility (RR=2.17, CI 1.58-2.98) plus decline in ADLs (RR=1.30, CI 1.05-1.60) and nursing facility stay at one-month (RR=1.93, CI 1.31-2.83).

Conclusion:

Severe delirium is associated with increased risk for poor clinical outcomes in patients with and without ADRD. In both groups, severe delirium increased risk of nursing home placement. In patients with ADRD, delirium was more severe and associated with a trend towards increased mortality at one month. While the increased risk remains substantial by relative risk, the study had limited power to examine the rarer outcome of death.

Background

Dementia is a progressive neurodegenerative disease that causes decline in cognitive function, reasoning and memory.¹ By 2050, 13.8 million older persons in the United States are expected to have cognitive impairment due to Alzheimer’s Disease and Related Dementias (ADRD).² Alzheimer’s Disease is the most common form of dementia but there are a variety of related dementias including Lewy Body dementia, frontotemporal dementia and vascular dementia – all with distressing consequences and burden to patients and families. Since no cure has yet been identified for ADRD, prevention of reversible contributors has assumed a prominent role to slow the onset and progression of ADRD. Hospitalization often represents a pivotal event for older patients, particularly for ADRD patients, with potentially life-threatening outcomes, including delirium, decline in function, nursing facility stay and mortality.^3-7

Delirium, an acute decline in attention and cognitive functioning,⁸ has been associated with worsened prognosis,^9,10 with accelerated rates of cognitive and functional decline,¹¹ increased lengths of hospital stay,⁴ higher rates of re-hospitalization,¹¹ institutionalization¹² and mortality.^10,12 Approximately 1 in 8 hospitalized ADRD patients who develop delirium will have at least one adverse outcome, including cognitive decline, institutionalization and death.⁹ The estimated healthcare costs of delirium are over $190 billion per year (2018 U.S. dollars) for hospital and post-hospital costs.^13,14 Significantly, 30-40% of delirium cases are preventable.¹⁵ However, the contribution of delirium severity to hospital outcomes in older persons with and without ADRD has not been previously examined.

There has been growing interest in delirium severity, which can help to capture changes over time, monitor treatment response, track clinical course and prognosis, and correlate with potential pathophysiologic biomarkers.^15,16 Since delirium of mild-moderate degree is most amenable to intervention, understanding the relationship of delirium severity to outcomes may help advance risk stratification, treatment approaches, and pathophysiologic understanding. Thus, the present study extends previous work^9,17 by examining the unique impact of delirium severity on hospital outcomes in ADRD patients compared to those without ADRD.

For the present study, we examined the association of delirium severity measured by the Confusion Assessment Method-Severity (CAM-S) score with important clinical outcomes at hospital discharge and at one-month follow-up in ADRD patients, compared to those without ADRD.

Methods

Setting and participants

The Better Assessment of Illness (BASIL) study is an ongoing prospective cohort study of hospitalized older adults, designed to examine delirium severity and related outcomes. A complete description of the BASIL study, recruitment process, study procedure and data collection has been published previously.¹⁸ In brief, 352 patients were enrolled between October 20, 2015 and March 15, 2017. Patients age 70 years or older, who were English speaking, admitted or transferred as either emergency or elective admissions and residing within 40 miles of the Beth Israel Deaconess Medical Center (BIDMC) in Boston were eligible. BIDMC is an academic tertiary medical center with 673 beds, over 40,000 admissions and 10,000 operations per year. Exclusion criteria included legal blindness or severe deafness, active alcohol abuse, alcohol withdrawal within the last 6 months, diagnosis of schizophrenia or active psychosis, nonverbal condition, immediate discharge plans or imminent death. All study procedures were approved by the Institutional Review Boards of BIDMC and Hebrew SeniorLife, the study coordinating center.

Study Procedures

Trained interviewers conducted initial patient evaluations within 48 hours of hospital admission, followed by daily assessments during hospitalization, telephone interviews 5-10 days after hospital discharge, and in-person interviews at one month after discharge.¹⁸ A comprehensive, standardized medical record review was conducted after hospital discharge by an experienced research physician to abstract demographic characteristics and clinical information, including admission diagnoses, medical comorbidities, surgical type if any, length of stay and death. Based on the medical record information, the Charlson Comorbidity Index was calculated at initial assessment.¹⁹

Participants were also asked to provide the name and contact information of a family member or caregiver who was familiar with their functional status. During the participant’s hospitalization, and at the 1-month and 12-months follow-up assessments, these family members or caregivers (proxy reporters) underwent 10-minute interviews in-person or by telephone to determine participant’s mental and functional status, and any interval change.

Study Variables

Assignment of ADRD Status.

All participants from the full BASIL cohort (N=352) were categorized from chart review as either ADRD or no ADRD through a clinical consensus process that has been published previously.²⁰ Since chart documentation often includes heterogeneous etiologies for cognitive impairment, including non-specified dementia, Alzheimer’s disease, frontotemporal dementia, diffuse Lewy body disease, amnestic and non-amnestic Mild Cognitive Impairment (MCI), or other, for the purposes of this study, we included all etiologies of cognitive impairment under the category of Alzheimer’s Disease and Related Disorders (ADRD).²¹

Participants were selected for detailed chart review (n=149) by one of two expert clinicians (TTH, TGF) if any of the following criteria were present: (1) Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) ≥3.2;^22,23 or (2) documentation of dementia by International Classification of Diseases, Tenth Revision (ICD-10) code in the medical records; or impairment on each of three cognitive tests as defined by: (1) Montreal Cognitive Assessment <26^24,25; or (2) brief cognitive screen <14/16^18,26; or (3) MoCA delayed recall of 2 or less. The ICD-10 codes used for documentation of ADRD were Alzheimer’s disease (G30.9); other Alzheimer’s disease (G30.8); other frontotemporal dementia (G31.09); dementia with Lewy bodies (G31.83); vascular dementia (F01.50); unspecified dementia (F03.90); and dementia in other diseases (F02.80). Patients without ICD-10 diagnosis or data from IQCODE or cognitive tests were also included in the chart review (n=48). The remaining patients – not meeting any of the above criteria – were categorized as no ADRD. Thus, a total of n=197 charts were reviewed.

The ADRD chart review process abstracted the following data: relevant medications (cholinesterase inhibitors, memantine, vitamin E), neuroimaging studies, provider notes (inpatient and outpatient) documenting a diagnostic evaluation for ADRD or memory loss, or formal neuropsychological impression. Based on these data, the participants were classified using clinical criteria as having either ADRD (includes MCI or dementia) or No ADRD. Chart evidence of dementia was defined as documentation of cognitive impairment from at least two separate sources (i.e., primary care outpatient notes and inpatient admission note); for cases with limited chart documentation (i.e., only one source, n=24), available data was adjudicated by an expert panel of 4 interdisciplinary physicians with clinical expertise in dementia (ERM, TGF, TTH, SKI) until a consensus assignment as ADRD or no ADRD was reached.

Delirium and delirium severity assessment.

The Confusion Assessment Method (CAM) and CAM-Severity (CAM-S) was used for rating the presence of delirium and its severity. Brief cognitive testing using the MoCA, combined with supplemental attention tasks (days of the week and months of the year backwards) was used to score the CAM. The CAM diagnostic algorithm for delirium requires the presence of both 1) acute onset and fluctuation and 2) inattention, and either 3) disorganized thinking or 4) altered level of consciousness.²⁷ The CAM has been demonstrated to have a sensitivity of 94% (95% confidence interval [CI] 91-97%), specificity of 89% (95% CI 85-94%) and inter-rater reliability of 0.70 to 1.00 in at least seven high-quality studies involving over 1,070 participants.²⁸ Diagnosis of delirium was made using the CAM diagnostic algorithm. Delirium severity was measured with the CAM-S long form (10 items, scored 0-19 with 19 worst) and CAM-S form (4 items, scored 0-7 with 7 worst).²⁹ In the analysis for this study, the CAM-S short form was primarily used. For the purposes of clinical relevance, we defined severe delirium as peak CAM-S short form score in the highest tertile (3-7 points out of 7); not severe delirium was defined as peak CAM-S short form score in the low-middle tertiles (0-2 points).

Clinical Outcomes

Pertinent outcomes evaluated at hospital discharge included length of stay >6 days (median) and discharge to a nursing facility. One-month outcomes included any decline in Activities of Daily Living (ADLs), new or ongoing nursing facility stay, and mortality. ADL decline was collected by patient interview, with self-report on ability to perform seven basic care skills (bathing, dressing, grooming, feeding, using the toilet, transferring, and walking), scored 0-14 with 14=completely dependent.³⁰ Decline in ADLs was defined as any decline in ADLs score between pre-hospital baseline and the one-month follow-up. All deaths were confirmed by at least two sources of information (i.e., family interview, medical record review, published obituary, or state death certificate).

Statistical Analysis

Comparisons between those with and without ADRD at baseline were conducted by one-way analysis of variance (ANOVA) for continuous variables or appropriate statistics for categorical variables. Generalized linear logistic models were used to determine relative risk (RR) for clinical outcomes given ADRD status or delirium incidence, stratified by ADRD status. Generalized linear models were used to determine mean differences in CAM-S peak (short and long) scores given ADRD status and to estimate the RR of clinical outcomes for each standard deviation increase in peak CAM-S score. Relevant covariates included age (continuous), sex (categorical), and Charlson Comorbidity Index score ≥ 2 (categorical). Most variables had no missing data with the following exceptions: at baseline, education (missing n=6, 2%); marital status (missing n=2, 1%), living alone (missing n=2, 1%), lives in a nursing home (missing n=2, 1%), ADL score (missing n=10, 3%); at one-month follow-up, re-hospitalization (missing n=5, 2%), and nursing facility stay (missing n=4, 1%). Mortality was ascertained in full, without missing data. At the one-month time point, there were 48 deaths and 26 patients lost to follow-up; therefore, nursing facility and decline in ADLs were not assessed for these individuals.

All analyses were performed using Stata, version 13 (StataCorp, College Station, Texas. All statistical tests were two-tailed and a P-value of less than 0.05 was considered statistically significant.

Results

On average, the 352 BASIL participants were 80.3 years old (standard deviation, SD 6.8); 148 (42%) were male, and 52 (15%) were non-white. The mean educational level was 14.5 (SD 3.0) years and 139 (40%) patients were married; 135 (39%) were living alone and 13 (4%) were living in nursing facilities. Most patients had comorbid conditions with Charlson Comorbidity Index Scores ≥ 2 (201 patients, 57%) and 272 (80%) patients had at least one ADL impairment at baseline. Eighty-five participants (24%) were categorized as having ADRD at the initial assessment by our clinical consensus approach. Patients with ADRD were older (82.0 years old, SD 6.5) than their non-ADRD counterparts (79.8 years old, SD 6.8, p=0.01) but otherwise did not differ significantly in their baseline characteristics (Table 1). There were 89 (25%) deaths, including two patients who died after enrollment but while in the hospital. The study continued for 12 months, where only 8 (2%) patients were completely lost to follow-up. The reasons for study dropouts were lack of time, declining health or memory, or family member requests. The enrollment flow into this study is delineated in Figure 1.

Table 1.

Characteristics of study participants on enrollment

Characteristics	Full Sample (N=352)	ADRD (N=85)	No ADRD (N=267)	P-value
Age, mean years (SD)	80.3 (6.8)	82.0 (6.5)	79.8 (6.8)	0.01
Male sex, self-reported, n (%)	148 (42)	41 (48)	107 (40)	0.18
Non-white race, n (%)	52 (15)	11 (13)	41 (15)	0.59
Education, mean years (SD)	14.5 (3.0)	14.4 (3.1)	14.6 (3.0)	0.59
Married (vs. unmarried), n (%)	139 (40)	34 (41)	105 (39)	0.79
Lives alone (vs. with other), n (%)	135 (39)	27 (33)	108 (40)	0.20
Lives in nursing home (vs. not), n (%)	13 (4)	6 (7)	7 (3)	0.05
Admission diagnoses (top 10), n (%)
Fracture	32 (9)	6 (7)	26 (10)	0.45
Malignant neoplasm	26 (7)	7 (8)	19 (7)	0.73
Sepsis	19 (5)	5 (6)	14 (5)	0.82
Gall bladder infection, biliary infection, obstruction	18 (5)	2 (2) 6 (7)	16 (6)	0.19
Gastrointestinal bleed	18 (5)		12 (4)	0.35
Joint, orthopedic surgery infection	17 (5)	3 (4)	14 (5)	0.52
Pneumonia	12 (3)	5 (6)	7 (3)	0.15
Spinal stenosis/spondylosis	12 (3)	0 (0)	12 (4)	0.05
Encephalopathy	9 (3)	5 (6)	4 (1)	0.03
Congestive heart failure	9 (3)	1 (1)	8 (3)	0.36
Charlson Comorbidity Score ≥ 2, n (%)*	201 (57)	54 (64)	147 (55)	0.17
ADL score at baseline, mean (SD)	3.1 (2.3)	3.1 (2.2)	3.1 (2.4)	0.98
Any ADL impairment at baseline (%)	272 (80)	66 (81)	206 (79)	0.62

ADL = Activities of Daily Living; ADRD = Alzheimer’s Disease and Related Dementias; SD = standard deviation.

All variables had no missing data with the following exceptions: Education (missing n=6, 2%); married (missing n=2, 1%), living alone (missing n=2, 1%); lives in a nursing home (missing n=2, 1%).

^*Charlson Comorbidity Score was calculated based on diagnoses abstracted from medical record review, scored from 0-35 with higher scores indicating more comorbidity.

Figure 1.

Enrollment Flow into BASIL Study

Overall, incident delirium occurred in 25% (88/352) participants overall. Among patients with ADRD, incident delirium occurred in 45% (38/85), compared with 19% (50/267) in those without ADRD, for an increased relative risk (RR) of 2.31, 95% confidence interval (CI) 1.64-3.28 (Table 2). Peak CAM-S short-form score was significantly higher among patients with ADRD, 2.8 (SD 1.8) vs. without ADRD, 1.5 (SD 1.5); the mean difference was 1.24, CI 0.83-1.65 (p<0.001). We also examined peak CAM-S scores among delirious patients only, to ensure results were not skewed by non-delirious, non-ADRD patients. This demonstrated the same trend: delirious patients with ADRD had a peak CAM-S short form score of 4.4 (SD 0.9) vs. delirious patients without ADRD 4.0 (SD 0.9); the mean difference was 0.5 (CI 0.1-0.9, p=0.01) (Table 2). Regardless of delirium status, patients with ADRD were more likely to be discharged to post-acute nursing facilities, RR=1.33, CI 1.00-1.79 or to die within one-month after discharge, RR=2.02, CI 1.20-3.40.

Table 2.

Clinical outcomes by ADRD Status

Outcomes	ADRD n=85	No ADRD n=267	Relative Risk or mean difference (95% CI) §	p-value
Delirium Outcomes	n/N (%) *	n/N (%) *
Delirium Incidence	38/85 (45%)	50/267 (19%)	2.31 (1.64-3.28)	< 0.001
Peak CAM-S, short; mean (SD)†	2.8 (1.8)	1.5 (1.5)	1.24 (0.83-1.65)	< 0.001
Among delirious patients only	4.4 (0.9)	4.0 (0.9)	0.50 (0.10-0.90)	0.01
Hospital Outcomes
Length of stay > 6 days	45/85 (53%)	143/267 (54%)	0.96 (0.76-1.20)	0.71
Discharge to post-acute nursing facility	39/85 (46%)	89/267 (33%)	1.33 (1.00-1.79)	0.05
One-month Outcomes
Decline in ADLs ‡	37/54 (69%)	131/213 (62%)	1.15 (0.93-1.42)	0.21
Nursing facility stay (new or ongoing)	19/57 (33%)	66/217 (30%)	1.02 (0.66-1.58)	0.91
Mortality	20/85 (24%)	28/267 (10%)	2.02 (1.20-3.40)	0.008

ADLs = Activities of Daily Living; ADRD = Alzheimer’s Disease and Related Dementias; CAM-S = Confusion Assessment Method-Severity; CI = Confidence Interval; SD = Standard Deviation.

At the one-month time point, there were 48 deaths and 26 patients lost to follow-up; therefore, nursing facility and decline in ADLs were not assessed for these individuals. All variables had no missing data with the following exceptions: ADL score at baseline (missing n=10, 3%); re-hospitalization (missing n=5, 2%); nursing facility stay (missing n=4, 1%).

^*Outcome descriptives are n/N (%) unless otherwise indicated.

^†CAM-S severity scores range from 0-7 for the short form, with higher scores indicating greater delirium severity.

^‡Decline in ADLs is by patient report, and defined as any decline in ADL score between baseline and one-month time point.

^§All results are relative risks and 95% confidence intervals, except for peak CAM-S results, which are mean differences and standard deviations comparing ADRD and no ADRD groups. Relative risks with 95% CI utilized generalized linear logistic model analysis, adjusted for age, male sex, Charlson Comorbidity ≥ 2 (see text for details). Mean differences and standard deviations for peak CAM-S results utilized linear regression analysis

Incident delirium was associated with poor clinical outcomes at both hospital discharge and one-month after discharge in the full sample (Figure 2). When stratified by ADRD status, incident delirium was associated with significantly worse one-month outcomes among patients with ADRD, including higher relative risk for decline in ADLs, (RR=1.53, CI 1.08-2.17), nursing facility stay (RR=2.06, CI 1.00-4.25) and mortality (RR=2.37, CI 1.03-5.44). In patients without ADRD, incident delirium was associated with adverse outcomes at hospital discharge and some one-month outcomes, but not mortality. Specifically, the group without ADRD demonstrated an increased relative risk for prolonged length of stay (RR=1.57, CI 1.28-1.94) and discharge to a post-acute facility (RR=2.21, CI 1.62-3.03) at hospital discharge, and decline in ADLs (RR=1.30, CI 1.05-1.61) and nursing facility stay (RR=1.68, CI 1.11-2.52) at one-month after discharge (Figure 2).

Figure 2.

Incident delirium and clinical outcomes by ADRD status. * ADLs, activities of daily living; ADRD, Alzheimer's disease and related dementias; CI, confidence interval; RR, relative risk. At the 1-month follow-up, there were 48 deaths and 26 patients lost to follow-up. Sample size for patients with ADRD was reduced to 60; for patients with no ADRD it was reduced to 218 due to deaths and loss to follow-up. All variables had no missing data with these exceptions: ADL score at baseline (missing n = 10 [3%]); rehospitalization (missing n = 5 [2%]); and nursing facility stay (missing n = 4 [1%]). Decline in ADLs is by patient report and defined as any decline in ADL score between baseline and 1-month time point. *Generalized linear logistic models used, adjusted for age, male sex, and Charlson Comorbidity Index of 2 or higher (see text for details).

Among persons with ADRD, significantly higher RR were observed for mortality (RR=1.59, CI 1.15-2.19) and nursing facility stay (RR=1.43, CI 1.01-2.03) at one-month for each standard deviation (SD=1.6) increase in peak CAM-S score, short form (Supplemental Table S1). No significant differences were found between the more delirious patients with ADRD and the less delirious with respect to hospital discharge outcomes (length of stay, discharge to post-acute nursing facility), or for decline in ADLs at one-month. Interestingly, among persons without ADRD, increasing delirium severity was associated with significantly increased risk for all poor outcomes, except mortality. This included poor outcomes at hospital discharge (length of stay, discharge to post-acute nursing facility) and at one-month post-discharge (decline in ADLs, nursing facility stay) (Supplemental Table S1). Given the possibility that patients without ADRD who became delirious were more likely to have poorer outcomes, we examined patients by tertiles of delirium severity to evaluate for trends. This also allowed us to examine delirium severity among patients with and without ADRD in a clinically relevant manner. We found the same associations between severe delirium and clinical outcomes by ADRD status with this approach. Among persons with ADRD, severe delirium increased the relative risk for mortality (RR=2.10, CI 0.89-4.98) and nursing facility stay (RR=2.22, CI 1.05-4.69) (Table 3). No significant differences were found between the severely delirious patients with ADRD and the less/not delirious for hospital discharge outcomes (length of stay, discharge to post-acute nursing facility) and for decline in ADLs at one-month. Again, interestingly, among persons without ADRD, severe delirium was associated with significantly increased risk for all poor outcomes, except mortality. This included poor outcomes at hospital discharge with prolonged length of stay (RR=1.47, CI 1.18-1.82) and discharge to post-acute nursing facility (RR=2.17, CI 1.58-2.98) as well as poor outcomes at one-month post-discharge with decline in ADLs (RR=1.30, CI 1.05-1.60) and nursing facility stay (RR=1.93, CI 1.31-2.83) (Table 3). We conducted detailed sensitivity analyses including and excluding the 38 patients with MCI from the analyses. We found that excluding these patients from analyses did not substantially alter the results or conclusions of the study (see Supplemental Tables S4-S5).

Table 3.

Independent contribution of delirium severity to clinical outcomes by ADRD status

Peak CAM-S, Short Form
Outcomes	Relative risk (RR) for outcome given severe delirium vs. not †
	ADRD (n=85)		No ADRD (n=267)
Hospital Outcomes	RR (95% CI) *	p-value	RR (95% CI) *	p-value
Length of stay > 6 days	1.18 (0.77-1.80)	0.45	1.47 (1.18-1.82)	< 0.001
Discharge to post-acute nursing facility	1.16 (0.72-1.87)	0.55	2.17 (1.58-2.98)	< 0.001
One-month Outcomes
Decline in ADLs ‡ §	1.41 (0.97-2.05)	0.07	1.30 (1.05-1.60)	0.01
Nursing facility stay (new or ongoing) §	2.22 (1.05-4.69)	0.04	1.93 (1.31-2.83)	0.001
Mortality	2.10 (0.89-4.98)	0.09	0.87 (0.37-2.08)	0.76

ADLs = Activities of Daily Living; ADRD = Alzheimer’s Disease and Related Dementias; CAM-S = Confusion Assessment Method-Severity (standardized); CI = Confidence Interval; SD = Standard Deviation; RR = Relative Risk.

At the one-month time point, there were 48 deaths and 26 patients lost to follow-up. All variables had no missing data with the following exceptions: ADL score at baseline (missing n=10, 3%); re-hospitalization (missing n=5, 2%); nursing facility stay (missing n=4, 1%).

^*Generalized linear logistic model analysis utilized, adjusted for age, male sex, Charlson Comorbidity ≥ 2 (see text for details).

^†Severe delirium defined as peak CAM-S score was in the highest tertile (3-7 out of 7); not severe delirium defined as peak CAM-S score was low-middle tertiles (0-2)

^‡Decline in ADLs is by patient report, and defined as any decline in ADL score between baseline and one-month time point.

^§Sample size for patients with ADRD reduced to 60; for patients with no ADRD n=218 due to deaths and loss to follow-up.

We also examined the association between delirium severity and ADRD, and their interaction with clinical outcomes to better clarify the relationship between these two variables. Delirium severity and ADRD status demonstrated a significant interaction for one-month outcomes, indicating significant interacting associations for increased risk of nursing facility stay, RR=1.14, CI 1.00-1.31 and mortality, RR=1.19, CI 1.08-1.30 for each standard deviation (SD=3.0) increase in peak CAM-S score, long form (Supplemental Table S2). Similarly, delirium severity by CAM-S, short form and ADRD status demonstrated a significant interaction for the one-month outcomes of increased risk of nursing facility stay, RR=1.26, CI 1.03-1.55 and mortality, RR=1.34 (CI 1.09-1.64) for each standard deviation (SD=1.6) increase in peak CAM-S score, short form (Supplemental Table S3). Thus, delirium severity demonstrates significantly different effects across the ADRD and non-ADRD strata for these outcomes.

Discussion

Our study demonstrates the impact of delirium and delirium severity on clinical outcomes among patients with and without ADRD. As with prior studies, we found that persons with ADRD are at substantially increased risk for developing incident delirium and poor subsequent outcomes.^4,9-12 Extending our previous work, we found, on average, that persons with ADRD are at substantially increased risk for developing more severe delirium (Table 2). Incident delirium substantially increases the risk of mortality in persons with ADRD, beyond the risk observed for ADRD or delirium alone; it also increases the risk for other one-month outcomes, including functional decline and nursing facility stay (Figure 2). These longer term outcomes arguably have greater impact on patients’ overall life trajectory and play an important role in recovery beyond hospitalization.³¹

Patients with ADRD did not demonstrate any differences in clinical outcomes at hospital discharge, including prolonged length of stay and post-acute facility discharge. It is possible that for patients with ADRD, care has already been established (i.e., home care or nursing home placement) that can accommodate cognitive impairment, and thus, the development of new delirium may not impact on these discharge outcomes. It is also possible that outcomes at hospital discharge are more influenced by external factors – such as type of medical insurance, availability of caregivers or nursing facility beds – which are not influenced by delirium status in persons with ADRD. In persons without ADRD who have never before been confused, delirium may necessitate further evaluation and prolonged hospital stay to determine appropriate discharge planning with provision of adequate services. On a more mechanistic level, persons with underlying ADRD have increased brain vulnerability, and often develop delirium and more severe delirium with lesser or milder insults. Thus, the impact on adverse clinical outcomes may be reduced proportionately.

Among persons without ADRD, incident and severe delirium contribute to increased risk for poor outcomes both at the time of hospital discharge, as well as functional decline and nursing facility stay one month later. These findings highlight the pivotal role that delirium plays in outcomes of hospitalization for all older patients. Delirium results in downstream complications in older hospitalized adults, regardless of baseline cognitive status, and demonstrates an interactive, synergistic effect with cognitive impairment.

Delirium severity adds important prognostic information, highlighting the impact of delirium on persons with ADRD. Our study found an exposure-response relationship between delirium severity, ADRD and poor clinical outcomes. We found that with interval increase in severity of delirium (by both tertiles and standard deviation units), ADRD patients are at significantly increased risk for serious clinical consequences of delirium at one-month post-hospitalization, specifically increased nursing facility placement and death (Figure 2, Table 3, Supplemental Table S1). ADRD likely exacerbates vulnerability to delirium, resulting in more severe delirium and a stronger impact on outcomes at one-month following hospitalization. The lack of impact of delirium severity in ADRD at hospital discharge was not surprising, since many external factors beyond ADRD influence length of hospital stay and likelihood of nursing facility placement.

Our study is innovative in examining the impact of delirium severity and severe delirium on hospital outcomes, in an exposure-response relationship, stratified by ADRD status. Previous studies have examined the effects of incident delirium on cognition, function, and mortality.^9,11,32 Our study reiterates the morbidity and mortality that is associated with delirium, and uniquely demonstrates the cumulative effects of delirium severity. Our results underscore the importance of intervening to prevent severe delirium to improve adverse hospital outcomes.

Severe delirium among persons without ADRD affects outcomes at hospital discharge, including increased risk for prolonged hospital stay and discharge to post-acute nursing facility. These findings were likely due to the immediate effects of severe delirium on patients, necessitating increased nursing and medical care in the acute illness period. Severe delirium among persons without ADRD also results in adverse one-month outcomes, such as functional decline and increased nursing facility stay – but not mortality. This may suggest more severe underlying pathophysiological mechanisms at play when ADRD interacts with severe delirium, to result in death among these vulnerable patients.³³

Our study has some noteworthy strengths. It presents a novel and innovative examination of delirium severity using a well-validated tool on hospital outcomes for patients with and without ADRD. Our categorization of ADRD involved a rigorous, sequential approach, informed by chart review, cognitive testing, patient and family interviews, and expert panel consensus to categorize ADRD status for all patients. However, several important caveats are worthy of note. We lacked pre-admission baseline assessments of the study participants; thus, categorization of pre-baseline status (such as ADLs) was determined by proxy interview reporting pre-admission status but obtained during hospitalization, which may be influenced by recall bias. Our ADRD group included some patients with MCI (mild cognitive impairment). Thus, this represents a more heterogeneous group in terms of pre-hospitalization cognitive status. Another limitation is that this study was conducted at a single academic medical center on general medical and surgical services; thus, generalizability to other patient populations will need to be confirmed in future studies.

This study holds important implications for the need to prevent delirium and particularly severe delirium in patients both with and without ADRD. Targeted multi-component, non-pharmacologic strategies such as the Hospital Elder Life Program have demonstrated significant impact on preventing delirium.^34,35 These “low-tech, high-touch” interventions – including frequent orientation, sleep-wake cycle preservation, early mobility, hydration and correcting visual/hearing impairments – have been shown to successfully prevent incident delirium and subsequent cognitive and functional decline in older patients both with and without ADRD.³⁴ The HELP protocols include adaptations for patients with ADRD. Our study suggests that a concerted effort to prevent delirium and attenuate severe delirium will help to prevent morbidity related to hospitalizations in those with ADRD. The current recommendation for clinicians when admitting patients with ADRD is to implement multi-component, non-pharmacologic interventions early and with good adherence to prevent delirium and the downstream cascade of complications and consequences. Future intervention trials focusing on prevention of delirium and severe delirium specifically in those with ADRD are greatly needed.

Supplementary Material

Supplementary Materials

Supplemental Table S1. Independent contribution of delirium severity (standardized peak CAM-S) to clinical outcomes by ADRD status

Supplemental Table S2. Association of delirium severity, Alzheimer’s Disease and Related Dementias status, and their interaction with clinical outcomes

Supplemental Table S3. Association of delirium severity, Alzheimer’s Disease and Related Dementias status, and their interaction with clinical outcomes

Supplemental Table S4. Clinical Outcomes by ADRD status excluding MCI cases

Supplemental Table S5. Incident delirium and clinical outcomes by ADRD status excluding MCI cases

BASIL Study Group

[Presented in alphabetical order; individuals may be part of multiple groups, but are listed only once under major activity].

Overall Principal Investigators (Multi PIs): Sharon K. Inouye, MD, MPH (Overall PI, HSL, BIDMC, HMS); Richard N Jones, ScD (BRN)

Co-Investigators: Tamara Fong, MD, PhD (HSL, BIDMC, HMS); Tammy Hshieh, MD (BWH); Edward R. Marcantonio, MD, SM (BIDMC, HMS); Annie Racine, PhD (HSL, HMS); Eva M. Schmitt, PhD (HSL); Dena Schulman-Green, PhD (Yale University); Patricia A.Tabloski, Ph.D,GNP-BC, FGSA, FAAN (Boston College); Thomas Travison, PhD (HSL, HMS).

Field Team: Tatiana Abrantes, BS (HSL); Brett Armstrong, MPH (BIDMC); Sylvie Bertrand, BA (HSL); Angelee Butters, MA (BIDMC); Madeline D’Aquila, BS (HSL); Jacqueline Gallagher, MS (BIDMC); Jennifer Kettell, BS (HSL); Jacqueline Nee, BA (HSL); Katelyn Parisi, BA, (HSL); Margaret Vella, BS (HSL); Guoquan Xu, MD, PhD (HSL); Lauren Weiner, MA (BIDMC).

Data Management and Statistical Analysis Team: Yun Gou, MA (HSL); Douglas Tommet, MPH (BRN).

Expert Review Panel¹: Charles H. Brown, M.D. (Johns Hopkins);^a,b Sevdenur Cizginer, M.D. (BRN);^a Diane Clark, PT, DScPT, MBA (University of Alabama);^a Joseph H. Flaherty, MD (St. Louis University);^a Anne Gleason, B.S. (HSL);^a,b Ann M. Kolanowski, Ph.D., RN (Penn State);^a,b Karen J. Neufeld, MD, MPH (Johns Hopkins University);^a Margaret G. O’Connor, PhD (BIDMC);^a Margaret A. Pisani, M.D., MPH (Yale);^a,b Thomas Robinson, M.D. (University of Colorado);^a,b Joe Verghese, M.B., B.S. (Albert Einstein);^a,b Heidi Wald, M.D., MPH (University of Colorado);^a,b Sharon M. Gordon, Psy.D. (Vanderbilt)^b

a) participated in expert panel to identify delirium severity items; b) participated in expert panel to identify delirium burden items. Abbreviations: BIDMC, Beth Israel Deaconess Medical Center; BWH, Brigham and Women’s Hospital; BRN, Brown University; HMS, Harvard Medical School; HSL, Hebrew SeniorLife; PI, principal investigator