Cognitive Predictors of Delirium on Long-Term Follow-Up after TIA and Stroke: Population-Based Cohort Study

Introduction: TIA and stroke cause cognitive impairment with a typical “vascular” pattern, including prominent frontal/executive deficits. Cognitive impairment is associated with increased delirium risk and the few available data suggest that executive dysfunction is important. We therefore determined the predictive value of both severity and pattern of cognitive deficits for delirium on long-term follow-up after TIA/stroke. Methods: Surviving TIA/stroke participants on October 1, 2013, in the Oxford Vascular Study (OXVASC) were assessed prospectively for delirium during all hospitalizations over the subsequent 6 months. Associations between OXVASC pre-admission mini-mental state examination (MMSE) and Montreal Cognitive Assessment (MoCA) scores, and delirium during hospitalizations on follow-up were determined using logistic regression adjusted for covariates, including demographic factors, history of depression, baseline stroke severity, and admission illness severity. Results: Among 1,565 TIA/stroke survivors, 158 patients (mean/SD age = 79.2/11.5 years) had ≥1 admission and 59 (37%) had ≥1 delirium episode. Mean/SD time between baseline TIA/stroke and admission was 4.7/3.6 years and between most recent OXVASC cognitive testing and admission was 1.7/1.8 years. MMSE and MoCA scores were associated with delirium: odds ratio (OR) = 1.16 (95% CI 1.07–1.27, p < 0.0001 per point decrease in MMSE) and OR = 1.20 (1.11–1.30, p < 0.0001 MoCA) and associations were robust to adjustment for all covariates, including stroke severity: OR = 1.11 (1.01–1.22, p = 0.03, MMSE) and OR = 1.15 (1.05–1.25, p = 0.003, MoCA). All 10 subtests on the MoCA and 4/11 on the MMSE were significantly associated with delirium with highest predictive value for frontal/executive and recall domains. Conclusions: Cognitive impairment of increasing severity after TIA/stroke predisposed to delirium particularly deficits in frontal/executive domains and recall. Long-term risk of delirium should be considered as part of the overall cerebrovascular disease burden.


Introduction
Pre-existing cognitive impairment predisposes to delirium [1,2]. Stroke causes cognitive decline and dementia in proportion to its severity which may explain why a history of stroke is associated with increased delirium risk in general hospital cohorts [1][2][3]. The pattern of neurocognitive deficits may also be important: executive dysfunction has been postulated to increase delirium possibly through altered brain networks and reduced ability to integrate sensory inputs [4,5]. Although stroke-associated cognitive impairment is heterogeneous depending on stroke location and the underlying neuropathology, a characteristic "vascular pattern" of deficits is seen, including relatively prominent frontal/executive dysfunction [6][7][8]. This cognitive impairment profile might therefore also increase the likelihood of delirium after stroke.
Delirium in the acute phase of stroke has shown to be associated with pre-existing cognitive impairment, as well as with stroke severity and infection [9], but there are few studies of delirium in the longer term after cerebrovascular events. In a population-based study, we determined associations between the severity and pattern of cognitive domain deficits occurring after TIA and stroke, and future delirium occurring during hospitalization on long-term follow-up.

Oxford Vascular Study
The patients in the current study were participants with TIA or stroke previously recruited since 1 April 2002 in the Oxford Vascular Study (OXVASC), an ongoing longitudinal populationbased cohort study within a population of >92,700 covered by 9 primary care practices (∼100 primary care practitioners) in Oxfordshire, UK [10]. The study was approved by the local research Ethics Committee. Informed written consent (or assent from relatives) was obtained for study interview and face-to-face and telephone follow-up as well as indirect follow-up using medical records.
TIA and stroke were defined clinically by WHO criteria and index cerebrovascular event severity was measured using the National Institutes of Health Stroke Scale (NIHSS) [10]. Patient data were collected by interview using a standardized form and from general practitioner records. Follow-up interviews were done by clinical fellows and trained research nurses at 1 and 6 months and 1, 5, and 10 years either in the out-patient clinic or by home visit. Cognitive testing was conducted at baseline and all follow-ups ( Fig. 1) using one or more of mini-mental state examination (MMSE) [11] and Montreal Cognitive Assessment (MoCA) [12] validated for use in TIA/stroke [8].

Prospective Ascertainment of Delirium
The Oxford University Hospitals NHS Foundation Trust (OUHFT) provides acute services for the population of approximately 660,000 in Oxfordshire, including all patients in the OX-VASC study primary care practises. OXVASC patients requiring acute hospital care for any reason are admitted either to OUHFT or to Abingdon Emergency Medical Unit (EMU). To identify OX-VASC participants admitted to hospital over the study period, we conducted daily searches of electronic (OUHFT) and paper records (EMU) for all new admissions excluding day case procedures, and cross-checked these against the register of OXVASC participants.
All TIA and stroke participants in OXVASC surviving on October 1, 2013, were included in the current study. Subsequent hospital admissions for any reason were prospectively identified from October 2013-April 2014 at the OUHFT and the Abingdon EMU. Patients were assessed as soon as possible after admission by members of the OXVASC study team. Delirium was ascertained using gold standard clinical diagnosis rather than retrospective hospital administrative diagnostic (ICD-10)-coded data since the latter are insensitive [13].
Delirium diagnosis was made according to the DSM-IV [14] criteria by STP after discussion with the OXVASC study team supplemented by review of the medical notes, including the OUHFT cognitive screen as described previously [2,13] (see online suppl. methods; for all online suppl. material, see www.karger.com/ doi/10.1159/000519900). The OUHFT cognitive screen is mandated on all patients admitted to the OUHFT or EMU who are aged ≥70 years or <70 years with confusion/altered behaviour and includes the 10-point abbreviated mental test score (AMTS) [15] together with the Confusion Assessment Method (CAM) [16] for delirium. Illness severity was defined using the systemic inflammatory response syndrome (2 or more of heart rate >90 beats per minute, temperature <36 or >38°C, respiratory rate >20 breaths per minute, white blood cell count <4 × 10 9 or >12 × 10 9 cells per litre) [17].

Statistical Analysis
Baseline characteristics of patients with any admission with delirium versus those without delirium were compared using t test and ANOVA for continuous variables and χ 2 for categorical variables. MMSE and MoCA cognitive test scores were obtained from the most recent OXVASC study follow-up performed prior to the hospital admission episode (Fig. 1). Cognitive test scores were dichotomized into mild (MMSE <27, MoCA <26) and moderate/ severe (MMSE <24, MoCA <18) cognitive impairment using previously described cut-offs [8].
To determine the associations between pre-admission OX-VASC cognitive impairment (MMSE and MoCA scores) and delirium occurring during hospitalization, we used data from all admissions to calculate odds ratios (ORs). We adjusted for demographic factors and then for key covariates known to be associated with delirium in general cohorts that were available in our dataset [1,2]: age, sex, and education (model 1), age, sex, education, history of depression at baseline OXVASC assessment, and illness severity on admission (model 2), and then with the addition of baseline OXVASC stroke severity (model 3). Similarly, we examined associations between individual MMSE and MoCA subtest scores and delirium adjusted for age, sex, and education. We also performed the following sensitivity analyses:

Results
Among TIA and stroke patients recruited since April 2002 (mean age/SD = 68.9/13.3 years, 676 TIA) 1,565 were still alive on October 12, 2013. Over the subsequent study period, there was a total of 194 admissions (100 to OUHFT and 94 to EMU) in 158 OXVASC participants (mean/SD age at admission = 79.5/11.2 years, range 47-100 years). The majority (170 [88%]) of admissions were unplanned: 122 (72%) to acute general (internal) medicine; 15 other medical services; 23 surgery; 10 trauma. Mean/SD time between baseline OXVASC TIA/stroke event and first hospital admission was 4.7/3.6 years and between most recent cognitive tests at OXVASC followup and first hospital admission was 1.7/1.8 years (Table 1).
Associations between pre-admission cognitive impairment and delirium were stronger for the MoCA (OR = 1.20 95% CI 1.11-1.30, p < 0.0001 per point decrease) than the MMSE (OR = 1.16 1.07-1.27, p < 0.0001) and for more severe versus less severe impairment (Table 2). Associations were robust to adjustment for demographic factors (model 1) with OR = 1.14, 1.08-1.27 (p = 0.005) for MMSE and OR = 1.16, 1.06-1.25 (p < 0.0001) for MoCA and were largely unaffected by the addition of illness severity, depression history (model 2), and stroke severity (model 3, Table 2). Sensitivity analyses substituting post-stroke disability (modified Rankin score) or baseline stroke subtype for stroke severity in model 3 showed similar results as did further adjustment of model 3 for admission characteristics (elective vs. unplanned and treatment speciality, online suppl. Tables). Further   sensitivity analyses using only the first admission only gave similar results (data not shown). All 10 subtests on the MoCA and 4/11 on the MMSE were significantly associated with delirium (Table 3). Significant MMSE subtests were orientation to time (p = 0.02), calculation (p = 0.008), recall (p < 0.0001), and copying pentagons (p = 0.03, Table 3; Fig. 2). On the MoCA, the strongest associations were for sustained attention (tapping at the letter A, p = 0.001), verbal fluency (p = 0.001, total number of f words p < 0.0001), recall (p = 0.002), abstraction (p = 0.01), calculation (p = 0.01), visuo-executive function (p = 0.02), naming (p = 0.02), and orientation (p = 0.02, Table 3; Fig. 2). The time taken to perform the cognitive test was longer in those with versus without delirium for the MoCA (mean/SD = 11.8/4.7 vs. 9.6/3.3 min, p = 0.002) but not the MMSE (6.2/2.5 vs. 5.9/6.4 min, p = 0.54). Sensitivity analyses using only the first admission gave similar results (data not shown).

Discussion
In our study of TIA/stroke patients on long-term follow-up, rates of delirium associated with hospitalization were high and were broadly similar to those reported in general hospital cohorts of older patients [1,2]. Pre-existing cognitive impairment of increasing severity, especially in frontal/executive and recall domains, predicted delirium in the long-term after TIA or stroke. Associations were robust to adjustment for other important factors including demographic characteristics, history of depression, and illness severity at the time of hospital admission and baseline stroke severity.
Studies in non-stroke cohorts have shown that delirium is more likely in patients with pre-existing cognitive impairment of increasing severity [1,2]. Since stroke-associated cognitive impairment is linked in a stepwise fashion to the severity of the stroke [3], we might have expected less strong associations between preadmission cognition and delirium after adjustment for baseline stroke severity. In fact, this had little effect probably because the majority of survivors in our cohort had relatively minor strokes or TIA. Although the link between pre-existing cognitive impairment and delirium is wellestablished, there are few data on whether the underlying subtype of neuropathology is important. One small autopsy study failed to show a relationship between any specific neuropathological marker and delirium [18]. In contrast, patients with vascular versus Alzheimer's dementia appear more likely to get delirium [19] and are over-represented in hospital cohorts [13]. In addition, neuroimaging studies suggest a stronger role for white matter disease than neurodegeneration in mediating delirium risk [20].
In our study, we saw strong associations between deficits in frontal/executive domains and memory and delirium risk. The MoCA has broader coverage of frontal and attentional cognitive domains than the MMSE and more difficult memory and language items. Consequently, impairment in nearly all the MoCA subtests predicted future delirium compared to only 4 of the MMSE subtests. The typical "vascular pattern" cognitive impairment profile after stroke with frontal/executive and recall deficits contrasts with that of early Alzheimer's disease in which short-term recall is most affected [6,7]. Previous data on cognitive domain impairment profile and delirium risk are largely from highly selected elective surgical populations. Almost all studies are concerned with shortterm delirium risk. Available data suggest associations with executive dysfunction: in 1 inclusive study of longterm (up to 1 year) prediction of delirium in older people living in residential care, strong associations with baseline Stroop test were found [21], although other studies have shown associations also with memory [22].
The mechanisms underlying the association between the cognitive domain impairment profile and delirium are unclear. Altered brain networks have been shown to underpin executive/attentional as well as memory deficits, and occur in stroke and neurodegenerative disease [5,22]. These network changes may predispose to network disintegration and the delirium syndrome in the face of precipitants including acute illness or environmental changes [23,24]. Alternatively, deficits in executive function and recall may associate with delirium because they are a clinical manifestation of the underlying cerebral pathology, including cerebrovascular disease burden, and associated blood-brain barrier permeability. Damage to the blood-brain barrier is thought to be an important mechanism in delirium through enabling inflammatory mediators to reach the brain [25].
Strengths of our study include the nesting within an on-going inclusive longitudinal population-based cohort study enriched for cerebrovascular disease with prospective patient evaluation for delirium. The study design allowed prospective assessment of pre-admission cognitive function using validated tests and robust adjustment for confounders often not considered in previous studies. Limitations include the use of cognitive screening tests rather than a comprehensive neuropsychological battery, although this enabled inclusion of older more impaired patients at greater risk of delirium. In-depth neuropsychological assessment may have provided more detail on the pattern of cognitive deficits contributing to delirium vulnerability. In addition, we did not ascertain episodes of delirium occurring in non-hospitalized patients in the community. There are few existing data but community prevalence of <0.5% has been reported in older people without dementia [26]. Since delirium is a powerful driver of hospital admission even in those in long-term care [1,26,27], it is unlikely that there were large numbers of delirium cases in the community without admission. Also, we did not examine the precipitating factors for delirium in detail but the breakdown by admission type suggested that acute medical emergencies were important. Finally, we were only able to prospectively collect data on hospital admissions over a 6 month snapshot period rather than over the entire OXVASC study timeframe which would have enabled us to undertake a linear mixed-effects analysis increasing power and fully exploiting the longitudinal nature of OXVASC. The move to electronic patient records and better routine documentation of delirium may make this possible in future studies.

Conclusion
TIA/stroke-associated cognitive impairment increases the risk of future delirium with particular importance of deficits in frontal/executive function and recall. Further work is needed to establish the underlying mechanisms and whether inclusion of cognitive impairment profile improves the accuracy of delirium risk models [28]. Future large studies are required to enable mixedeffects modelling to determine the impact of delirium occurrence on future dementia risk. Long-term delirium risk and acute delirium at the time of stroke should be considered as part of the overall burden of TIA and stroke.