Delirium Among Critically Ill Patients With Stroke: Prevalence, Severity, and Outcomes Free

Stroke is a common and often life-altering medical condition that results in a significant public health burden. Stroke is the third leading cause of death globally,¹  accounting for 6.3 million annual deaths worldwide, and rates are anticipated to increase in the next decade.²  Intensive care unit (ICU) level of care is indicated for approximately 20% to 50% of patients experiencing acute ischemic stroke (AIS), and rates of intensive care are much higher for patients with intracerebral hemorrhage (ICH) or aneurysmal subarachnoid hemorrhage (aSAH).^3–5 

Delirium is a highly prevalent syndrome among patients with acute stroke and is associated with poor outcomes such as impaired physical, cognitive, and psychological function and worse survival.^6,7  Considerable research and clinical interest has been devoted to delirium occurring in critically ill adults, but much of this research has excluded patients with neurologic disease, including stroke.⁸  As such, the incidence and prevalence of delirium among critically ill patients who have had a stroke is unclear. Previous estimates of incidence of delirium among critically ill patients after stroke vary widely, ranging from 12% to 75%, most likely because the studies included patients with mixed acuity, including patients who were not critically ill.⁹  Because patients who have had a stroke are classified as a vulnerable population of interest in the 2017 ICU delirium research agenda, further research into delirium in critically ill patients after stroke is warranted.¹⁰ 

The extant literature leaves uncertainty regarding how many critically ill stroke patients experience delirium during an ICU stay, which factors increase the risk for delirium, and the effect delirium has on important patient- and family-centered outcomes. These gaps in knowledge prevent appropriate targeting of preventive and treatment interventions. Therefore, this study aimed to determine the (1) prevalence, (2) severity, and (3) relationship between delirium and short-term outcomes among adults admitted to a neurocritical care unit (NCCU) with AIS, ICH, and aSAH.

We conducted a prospective, observational, cohort study.

This single-center study was conducted in the NCCU at The Ohio State University Wexner Medical Center, a tertiary center with a large catchment area for neurovascular patients. We included those who were (1) admitted to the NCCU with a diagnosis of stroke (AIS, ICH, or aSAH); (2) at least 18 years old; and (3) admitted to the NCCU within 48 hours of stroke onset. We excluded those who (1) had an ICH due to traumatic brain injury; (2) were prisoners; (3) were both blind and deaf; (4) were non-English-speaking; or (5) had imminent plans to withdraw life support. We screened the NCCU list of patients for eligible participants daily and after a brief medical record review to determine eligibility approached the participants or their legally authorized representative to request informed consent. The number of cases in the area during the study period determined the sample size. Ethical approval was obtained from The Ohio State University institutional review board (IRB) before initiating recruitment (title: Delirium Among Critically Ill Stroke Patients; approval date: 11/12/2021; IRB number: 2021H0325). We followed procedures in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration.

We collected demographic and clinical characteristics, including stroke type and laterality, severity of illness, and comorbidities, via medical record review.

After participants were enrolled in the study, trained study staff assessed their level of arousal, delirium, and delirium severity daily for up to 7 days while patients were in the NCCU. Study staff used the Richmond Agitation-Sedation Scale (RASS) to quantify level of arousal.^11,12  Unarousable patients with RASS scores of less than −3 were considered untestable for delirium and categorized as having coma. Delirium was measured with the Confusion Assessment Method for the ICU (CAM-ICU).¹³  The poststroke neurologic examination was considered as the baseline for feature 1 of the CAM-ICU (acute onset or fluctuating course of mental status).¹⁴  The CAM-ICU-7 is a delirium severity tool calculated from data derived from the CAM-ICU, resulting in a score of 0 to 7, with scores of 0 to 2 indicating no delirium, scores of 3 to 5 indicating mild-to-moderate delirium, and scores of 6 to 7 indicating more severe delirium symptoms.¹⁵  Every assessment generates a CAM-ICU-7 score, with 0 to 2 points assigned per feature; thus, even patients without delirium may accrue points. We defined the delirium group as patients ever having had delirium for at least 1 day during the ICU stay and the no delirium group as patients having no delirium at any time during the ICU stay. Patients with missing assessments were regarded as ever having had delirium only if they had at least one positive assessment.

We provided study personnel with didactic training on administration of the RASS, CAM-ICU, and CAM-ICU-7 instruments with clinical-reasoning-based case scenarios.¹⁶  In addition to the scenarios, the principal investigator performed 2 to 3 co-assessments with the study personnel before data collection to verify their initial competency. Throughout the data collection period, additional co-assessments were performed randomly on 6% of the assessments to ensure reliability.

The presence of receptive aphasia was noted with each assessment. If a patient could participate in the examination and did not have complete receptive aphasia, defined as an inability to process any verbal input, the CAM-ICU was scored. If the patient had a component of receptive aphasia and radiographic rationale for the aphasia, such as a stroke affecting the Wernicke area, patients were marked as aphasic. Patients with complete receptive aphasia that precluded any participation in the assessment were considered untestable and were excluded from the analysis. Patients without a radiographic rationale for aphasia were considered testable for our purposes.

We collected data on mortality in the ICU and hospital, disability, dependance as measured by the modified Rankin Scale (mRS) at discharge (as rated by the clinical team),^17,18  tracheostomy, ventilator days, and discharge disposition.

Patient characteristics are expressed as mean (SD) or number (%). Characteristics between the delirium and no delirium groups were compared with χ², Fisher exact test, and 2-sample t tests as appropriate. A P value of .05 or less was considered statistically significant. The associations between ever having had delirium and outcomes were analyzed with univariate logistic regression for binary outcomes and linear regression for continuous outcomes. The maximum CAM-ICU-7 score for each patient in the subgroup with delirium was used as a predictor variable for outcomes in univariate linear and logistic regressions. A sensitivity analysis was performed for missing delirium assessments with best case, worst case, and last value carry forward. Delirium-and-coma-free days were calculated by subtracting the number of days with coma or delirium from the lesser of the 7-day data collection period or the ICU length of stay for each patient (in both patient groups) and then reported as mean days.

Between January and December 2022, we screened 269 patients. Of these, 53 were excluded, 211 were approached for consent, and 169 were included in the analysis (see Figure). The sample was predominantly female and White and had a mean age of 65.5 years. Demographics and baseline clinical characteristics stratified by delirium status are included in Supplemental Table 1 (available online only at ajcconline.org). The participants’ mean (SD) National Institutes of Health stroke score was 14.8 (9) and the distribution of stroke type was 54.7% (n = 93) AIS, 33.7% (n = 57) ICH, and 11.2% (n = 19) aSAH. Age, stroke type, and ICH laterality were significantly associated with ever having had delirium (P = .03, .02, and .02, respectively).

Of the 169 participants included in the analysis, 75 (44.4%) had ever had delirium (Table 1). Among stroke subtypes, the prevalence of ever having had delirium was 37.6% (n = 35/93), 59.6% (n = 34/57), and 31.6% (n = 6/19) among patients with AIS, ICH, and aSAH, respectively. Participants with delirium were more likely to be older (P = .03) with an ICH stroke subtype (P = .02). Across all stroke subtypes, the mean (SD) number of delirium-and-coma-free days was 3.2 (2) (Supplemental Table 2, Supplemental Figure [available online only at ajcconline.org]). In an analysis at the level of days, versus at the patient level, the Supplemental Figure illustrates the discrete and overlapping possibilities of how each of the days were or could be categorized.

The rate of missingness of the daily delirium assessments was 23.4%. Many of the missing assessments were on the final ICU day when the patient physically transferred off the unit before staff could perform an assessment. Other missed assessments occurred randomly as a result of staffing constraints or clinical work being prioritized. Our prevalence results of 44% were robust based on a sensitivity analysis with best case (43.5%), worst case (86.5%), and last value carry forward (50%). These analyses indicate that missing data had a limited impact on the overall prevalence findings.

The prevalence of coma at any time during the study was 32.5% (n = 55); 10 patients (5.9%) in the study sample were persistently comatose. Of the 159 participants who were not persistently comatose, 45 (28.3%) had aphasia at 1 or more assessments and 28 (17.6%) were persistently aphasic throughout the entire study period.

Of those participants who experienced delirium (n = 75), the mean (SD) CAM-ICU-7 delirium severity score was 5.64 (1.4 [range, 4-7]). Delirium severity as a predictor of short-term outcomes is discussed below.

Participants who experienced delirium had more days receiving mechanical ventilation (point estimate, 2.59 [95% CI, 0.73-4.44]), had longer ICU (point estimate, 3.33 [95% CI, 1.36-5.31]) and hospital (point estimate, 6.76 [95% CI, 3.43-10.09]) lengths of stay, were less likely to be discharged to home (versus long-term acute-care hospital, skilled nursing facility, inpatient rehabilitation, or hospice; odds ratio [OR], 0.42 [95% CI, 0.19-0.94]), and had worse dichotomized discharge mRS scores using a dichotomization cut point of 3 (OR, 2.58 [95% CI, 1.04-6.36]).

Mortality rates, both ICU and hospital, were not significantly different between those who had ever had delirium and those who had not (OR, 0.67 [95% CI, 0.42-1.05]; OR, 0.94 [95% CI, 0.61-1.47], respectively). Tracheostomy outcomes associated with delirium severity could not be estimated because of quasi-complete separation; all 6 tracheostomy patients had a maximum CAM-ICU-7 score of 7. Descriptive statistics, stratified by ever having had delirium, and a summary of the results of regression models of the associations between ever having had delirium and short-term outcomes are presented in Table 2. Increasing delirium severity was also associated with worse short-term outcomes including more ventilator days (point estimate, 1.93 [95% CI, 0.37-3.49]), longer ICU stay (point estimate, 1.81 [95% CI, 0.27-3.35]), and discharge mRS scores of 3 or higher (OR, 2.35 [95% CI, 1.09-5.09]) (Table 3).

This prospective observational cohort study examining the epidemiology of delirium in patients who have had a stroke is among the first to do so with an exclusively critically ill cohort. Our findings contribute important information about prevalence, severity, how coma and delirium interact, and adverse outcomes among patients who experienced delirium. First, we report a 44.4% prevalence among critically ill stroke patients and a significantly higher rate among our ICH subgroup. Second, we found that ever having had delirium was associated with more days receiving mechanical ventilation, longer length of stay, lower likelihood of discharge to home, and higher odds of an mRS score of 3 or higher at discharge, indicating an unfavorable outcome. Finally, increasing delirium severity was associated with more ventilator days, longer ICU stay, and higher odds of an mRS score of 3 or higher at discharge.

Our prevalence study complements prior work reporting incidence, reflecting the overall burden of delirium in this population. Other studies have primarily reported incidence among mixed-acuity populations (ie, patients in both lower-acuity units and ICUs who have had a stroke) or among mixed-diagnosis populations (ie, an ICU including a sub-population of stroke patients).⁹  The prevalence of delirium varied significantly with stroke type. Our findings of 58.6% prevalence in ICH patients exceeded the 27% to 30% incidence among ICH patients reported in 2 prior studies,^19,20  but was closer to the 50% prevalence reported for a critically ill stroke cohort by Haight and Marsh.²¹  Our prevalence findings of 37.6% among AIS patients exceeded the incidence of 22% in a mixed ICU population including stroke and transient ischemic attack reported by Limpawattana et al,²²  but was comparable to the incidence of 39% in a mixed-acuity study of stroke patients reported by Zisper et al.²³  In a large retrospective cohort, stroke patients had higher odds of positive CAM-ICU assessments (OR, 4.2 [95% CI, 3.3-5.5]) and “unable to assess” CAM-ICU assessments than did a sepsis cohort.²⁴  If the RASS variability described by Reznik et al²⁴  truly indicates that the CAM-ICU underestimates delirium in stroke patients, then the prevalence of delirium in our sample, in which we used the CAM-ICU, is most likely underestimated.

This study adds support to the extant ICU delirium literature that delirium is both common and a problem that leads to further suffering and cost, including institutionalization after acute care and longer stays. We found that delirium was associated with numerous adverse clinical outcomes. Similarly, in a review of outcomes among critically ill stroke patients, we found that those who experienced delirium had longer stays, higher mortality rates, poorer mRS scores, and diminished cognitive and executive function at various follow-up intervals.⁹  In a mixed-acuity stroke population, Zisper et al²³  found that patients who experienced poststroke delirium had longer stays (mean [SD] of 11.5 [4] days vs 9.3 [3.4] days, P < .001). In a study of non-ICU patients with acute stroke, Rollo et al⁶  found that those with delirium had higher mRS scores at 3 months. When studying aSAH patients with hyperactive delirium, Reimann et al²⁵  noted that delirium was not associated with worse outcomes, but with intermediate outcomes (mRS score = 1-3), whereas patients without delirium had more pronounced bifurcated outcomes (mRS scores of 0 and ≥4). This finding is likely a factor of the effect of the primary disease on consciousness, in which severe aSAH often produces a coma examination (neither testable nor delirious) and results in high mortality. Patients with poststroke delirium had higher 3-month mortality in both a mixed-acuity hospital-wide stroke cohort²³  and a non-ICU acute stroke cohort.⁶  Demonstrating this association between delirium and adverse outcomes among stroke patients lends support to clinicians’ advocating for improved delirium prevention and treatment measures.

A particularly novel finding is the effect of delirium severity on important outcome measures. Among patients who experienced delirium, higher delirium severity, as measured by the CAM-ICU-7, was associated with more ventilator days, longer stays, and worse discharge mRS scores. Similarly, prior research in a mixed-diagnosis ICU population reported an association between delirium severity and higher in-hospital mortality (OR, 1.47 [95% CI, 1.30-1.66]).¹⁵  However, those study authors published no subgroup analyses of stroke. Our study is among the first to evaluate the severity of delirium among critically ill patients who have had a stroke.

This study has several limitations. Our sample overrepresented women and was less racially diverse than the community. Our study was limited by the lack of a prospective power analysis to determine the sample size. However, a post hoc power analysis suggested that the study had adequate power (>80%) to detect clinically meaningful differences between the delirium and no delirium patients for most short-term outcomes (ventilator days, ICU length of stay, hospital length of stay, disposition to home, and poor mRS score). Because of the low number of events for ICU death, hospital death, and tracheostomy placed, the study was underpowered (<80%) to detect a between-group difference of less than 8% for ICU death and hospital death and a between-group difference of less than 12% for tracheotomy placed. The outcomes reported in our study are univariate predictors, without control for what is a complex and multifactorial situation. Thus, the results are subject to confounding and selection biases. These outcomes are short-term, but approximately 64% of stroke deaths occur outside an acute-care hospital, presumably due to poststroke complications²⁶ ; thus, these mortality rates likely do not reflect the full mortality burden of delirium in this population. We obtained daily delirium assessments; because delirium fluctuates, more frequent assessment would have been preferred. Aphasia, particularly receptive aphasia, has a complex interaction with the clinical diagnosis and screening tools for delirium. The presence of receptive aphasia was collected alongside daily delirium assessments. Because of some ambiguity with this aspect of data collection, there were inconsistencies in the data and we could not make reliable inferences regarding aphasia. The aims of the study did not include screening tool development or validation; rather, we used the CAM-ICU as directed in the manual.¹⁴  The CAM-ICU has acceptable sensitivity and specificity (76% and 98%, respectively) in the stroke population.²⁷  While conducting this study, a novel tool, the Fluctuating Mental Status Evaluation, was published and retains high sensitivity and specificity in patients with aphasia in pilot testing among patients with ICH.²⁸  Incorporating this or similar assessment tools into future delirium studies including patients with aphasia may further enhance understanding of delirium in this unique population.

Delirium is a highly prevalent condition among critically ill stroke patients. In our cohort, delirium occurred at least once during the first week of the ICU stay in 44% of critically ill stroke patients and in 60% of ICH participants. Both the occurrence of delirium and higher delirium severity are associated with poor outcomes and longer stays. Given the association between delirium and adverse outcomes in this population, critical care providers should prioritize delirium prevention and treatment measures. This study contributes an important step toward a fuller appreciation of the epidemiology of delirium and its associated outcomes among critically ill stroke patients.

This work was performed at The Ohio State University, College of Nursing, and Wexner Medical Center.