Evaluation of Bedside Tests of Attention and Arousal Assessing Delirium in Parkinson’s Disease,Dementia,and Older Adults

INTRODUCTION

Delirium is an acute, neuropsychiatric disorder affecting approximately 15%of older people in hospital [1]. It is characterised by inattention, deranged level of arousal and fluctuating cognitive impairments [2]. It is highly distressing, expensive and associated with poor outcomes, including increased mortality, institutionalisation and cognitive decline [3–5]. However, delirium remains under-diagnosed and poorly documented, resulting in worse outcomes [1, 7].

People with neurodegenerative disorders, such as Parkinson’s disease (PD) [8] and dementia [9] may be at increased risk of delirium. The prevalence of delirium in those with PD admitted to hospital is high at 34%[10], and it is thought that PD with and without dementia confers a vulnerability to delirium [11]. Dementia is a syndrome characterised by cognitive decline and impaired functional independence and activities of daily living [2]. Dementia is a major risk factor for delirium [9]. The prevalence of delirium superimposed on dementia (DSD) is estimated to be 22–89%depending on the setting (e.g., in inpatients across specialties, in nursing homes or in the community) [12], population [9], and outcomes following DSD are worse than in delirium alone [12–14].

It is vital to differentiate delirium from underlying PD or dementia, as delirium may be the first or only symptom of a potentially life threatening, but treatable, underlying cause. Additionally, delirium itself has been shown to be preventable in some cases [15]. However, there is considerable uncertainty, regarding how best to diagnose DSD [16]. Currently, DSD is evaluated using tools designed for delirium alone. No studies have evaluated the efficacy of delirium diagnostic tools in PD or PD with dementia (PDD) [8]. This is problematic given that delirium and PD share many overlapping features, including sleep disturbance, daytime somnolence, impaired attention, hallucinations, and delusions [8, 17], making delirium in PD difficult to identify. Additionally, many of the tools used test multiple cognitive domains and so may be abnormal in someone with PD or dementia, even when delirium is not present. This means that a delirium diagnosis may be missed or incorrectly diagnosed in people with PD and dementia [8].

Demonstrating impaired attention and altered levels of arousal are essential components of Diagnostic and Statistical Manual of Mental Disorders fifth edition (DSM-5) criteria [2], required for a diagnosis of delirium, and both have been shown to be relatively specific to delirium [18–20]. Recent evidence suggested that altered arousal and inattention may be associated with increased mortality [21]. There is consensus that they should be considered together, with the inability to engage in cognitive testing due to abnormal arousal (above the level of coma) considered to be severe inattention for the purposes of delirium diagnosis [22]. A combined assessment of attention (letter vigilance “S-A-V-E-A-H-E-A-R-T” test) and arousal (Observational Scale of Level of Arousal, OSLA) has previously been shown to accurately identify DSD and provides a simple, brief bedside test for delirium, which can be administered with minimal training [17]. However, few studies have evaluated the performance of these tests in people with other neurodegenerative conditions, and none in PD [8]. PD, PDD, and dementia with Lewy bodies (DLB) have a number of intrinsic non-motor, neuropsychiatric manifestations, including inattention and fluctuating arousal [8, 24]. Given the considerable overlap in their clinical phenotypes, it is not known whether common bedside tests for attention and arousal will correctly identify delirium in patients with PD with or without dementia.

We aimed to evaluate the ability of simple bedside tests used by non-specialists in hospital settings for attention and arousal to detect delirium in those with and without PD and dementia. We also tested for the optimal cut-offs for each test for the different neurodegenerative disease groups. We hypothesized that simple bedside tests would be sensitive to identifying delirium in PD and dementia, but that different cut-offs would be required compared to older adults without underlying neurodegenerative diseases.

METHODS

RESULTS

In total, 249 participants accounting for 371 admissions were included in this analysis (Fig. 1). Fifty-four (21.7%) participants had a Lewy body disease comprising 30 (12.0%) participants with PD without cognitive impairment, 24 (9.6%) with LB-CI (PD-MCI, PDD, or DLB) and 16 (6.4%) with another dementia. Of these, 21.7%(n = 54) had prevalent delirium (delirium present on first assessment) on their first admission. Participants in the DETERMINE-PD study were significantly younger, had completed more years of education, were frailer and had a higher proportion of male participants compared to DECIDE (p < 0.05 for all, Supplementary Table 2).

OPEN IN VIEWER

Fig. 1

Pooled participants from the DECIDE and DETERMINE-PD studies. PD, Parkinson’s disease; LB-CI, Lewy body cognitive impairment group; PD-MCI, Parkinson’s disease with mild cognitive impairment; PDD, Parkinson’s disease dementia; DLB, dementia with Lewy bodies, untestable, participants with abnormal arousal levels where it was not possible to complete tests of attention.

Demographics of participants from their first admission (Table 2) to hospital show that cases with prevalent delirium were significantly older and frailer (p < 0.05). Seventy-eight (31.3%) participants had ≥2 admission to hospital. Cases with prevalent delirium were more likely to have a dementia (23.1%vs. 6.5%, p < 0.01) or PD (29.5%vs. 14.3%, p < 0.01).

Attention and arousal accuracy

OPEN IN VIEWER

Fig. 2

ROC curves for (A) older adults without PD or dementia, (B) participants with PD/LB-CI, and (C) participants with dementia. PD, Parkinson’s disease; LB-CI, Lewy body cognitive impairment comprising PD-MCI, PDD, and DLB; PD-MCI, Parkinson’s disease with mild cognitive impairment; PDD, Parkinson’s disease dementia; DLB, Dementia with Lewy bodies; MDAS, Memorial Delirium Assessment Scale; GCS, Glasgow Coma Scale; OSLA, Observational Level of Arousal. To facilitate comparison, GCS scores were reversed.

The diagnostic test accuracy of each cognitive measure was assessed using ROC curves (Table 4) in PD and dementia-free older adults, in cases with PD (with and without cognitive impairment and DLB), and those with dementia (including LB-CI). Area under the curve ranged from 0.62-0.89. For measures of attention, only the MDAS attention scores significantly predicted delirium in all groups (AUC range 0.70–0.74, p < 0.001 for all). Forward digit span was significantly associated with delirium in PD and dementia-free older adults (AUC = 0.67, p = 0.01) but not in those with PD or dementia. Backwards digit span, MDAS digit span and MOTYB scores were significantly associated with delirium in PD and dementia-free older adults (AUC = 0.77, 0.77 and 0.74, respectively, p < 0.001) and PD (AUC = 0.70, 0.75 and 0.73, respectively), but not in dementia (p > 0.05). All tests of arousal were significantly associated with delirium in all admission groups (AUC = 0.61–0.89, p < 0.05 for all).

OPEN IN VIEWER

Table 4

Identifying cut-off points for identifying impaired tests in participants with delirium

			AUC	95% CI		p	Optimal cut-off	Sensitivity %	Specificity %	Youden index (J)
			Lower	Upper
Attention	Digit span – Forward	PD and dementia-free older adults	0.67	0.56	0.77	0.001	4/5	44.44	87.00	0.31
		PD/LB-CI	0.63	0.47	0.79	0.090
		Dementia (all)	0.60	0.42	0.77	0.270
	Digit span – Backwards	PD and dementia-free older adults	0.77	0.69	0.85	< 0.001	3/4	91.67	46.19	0.38
		PD/LB-CI	0.70	0.257	0.83	0.010	1/2	80.00	57.14	0.37
		Dementia (all)	0.62	0.46	0.79	0.160
	MDAS Digit span	PD and dementia-free older adults	0.77	0.69	0.85	< 0.001	1/2	61.11	79.37	0.40
		PD/LB-CI	0.75	0.62	0.88	< 0.001	1/2	80.00	59.52	0.40
		Dementia (all)	0.65	0.48	0.82	0.080
	MOTYB	PD and dementia-free older adults	0.74	0.65	0.83	< 0.001	0/1	73.68	73.54	0.47
		PD/LB-CI	0.73	0.59	0.87	< 0.001	0/1	71.43	69.05	0.40
		Dementia (all)	0.62	0.45	0.78	0.170
	MDAS attention	PD and dementia-free older adults	0.70	0.59	0.80	< 0.001	0/1	40.00	99.11	0.39
		PD/LB-CI	0.84	0.73	0.96	< 0.001	1/2	68.18	92.86	0.61
		Dementia (all)	0.78	0.65	0.92	< 0.001	1/2	68.18	83.87	0.52
Arousal	GCS	PD and dementia-free older adults	0.67	0.57	0.77	< 0.001	14/15	44.44	87.00	0.31
		PD/LB-CI	0.89	0.80	0.98	< 0.001	13/14	73.91	90.48	0.64
		Dementia (all)	0.88	0.78	0.98	< 0.001	14/15	88.00	75.76	0.64
	OSLA	PD and dementia-free older adults	0.74	0.64	0.84	< 0.001	0/1	52.27	94.02	0.46
		PD/LB-CI	0.85	0.75	0.95	< 0.001	3/4	86.96	73.81	0.61
		Dementia (all)	0.78	0.65	0.91	< 0.001	3/4	76.00	72.73	0.49
	MDAS Consciousness	PD and dementia-free older adults	0.61	0.51	0.72	0.017	0/1	22.73	100.00	0.23
		PD/LB-CI	0.87	0.77	0.98	< 0.001	1/2	65.22	97.62	0.63
		Dementia (all)	0.78	0.65	0.91	< 0.001	1/2	64.00	93.94	0.58
	m-RASS	PD and dementia-free older adults	0.69	0.59	0.79	< 0.001	±0/1	38.64	99.57	0.38
		PD/LB-CI	0.81	0.69	0.93	< 0.001	±1/2	56.52	95.24	0.52
		Dementia (all)	0.76	0.64	0.89	< 0.001	±1/2	52.00	90.91	0.43

Significant results are highlighted in bold. PD, Parkinson’s disease; LB-CI, Lewy body cognitive impairment comprising PD-MCI, PDD, and DLB; PD-MCI, Parkinson’s disease with mild cognitive impairment; PDD, Parkinson’s disease dementia; DLB, Dementia with Lewy bodies; MDAS, Memorial Delirium Assessment Scale; MOTYB, months of the year backwards; GCS, Glasgow Coma Scale; OSLA, Observational Level of Arousal; m-RASS, Modified Richmond Agitation Scale.

Discussion

This is the first study to evaluate simple bedside tests of attention and arousal in PD and dementia compared with PD and dementia-free older adults, and to identify optimal cut-offs. We found that all tests were sensitive to detecting delirium in older adults without PD or dementia during a single study visit. Although all arousal measures could accurately detect delirium in participants with PD and dementia, only selected attention measures were able to identify delirium in these groups. Furthermore, we identified that stricter cut-offs for attention measures may be needed to accurately detect delirium in those with PD or dementia. In PD and dementia-free older adults, impaired digit span and OSLA were the optimal tests to detect delirium, while in PD and dementia the optimal tests were MDAS attention and GCS. Consistent with findings by Richardson et al. [17], a single combined score summing the two tests (attention plus arousal) performed better than the two tests individually in all groups.

Consistent with previous studies in other populations, we found that brief measures of attention and arousal are able to detect delirium in PD and dementia-free older adults with high levels of sensitivity and specificity [17, 20]. Dysfunction in attention and arousal are essential criteria for the diagnosis of delirium, and in the absence of an existing biomarker, simple bedside diagnostic tests are used to identify delirium in clinical practice [31]. However, few studies have evaluated the psychometric properties of scales or measures used to determine their accuracy in identifying dysfunction. In addition, it is important to identify whether these tools differentiate delirium from important differential diagnoses with overlapping phenotypes, including PD [8] and dementia [17].

We found that the forward digit span test was only sensitive to detecting delirium in PD and dementia-free older adults, while backward digit span, MDAS digit span and MOTYB were sensitive to detecting delirium in PD and dementia-free older adults and PD and LB-CI, but not those with non-Lewy body dementia. Previous studies have reported that impaired MOTYB [32–34], letter vigilance test [17, 33] and digit span tests [33] were significantly associated with delirium in older adults and dementia participants, although accuracy differs between studies, and compromise between sensitivity and specificity decrease with existing cognitive impairment [35]. Performance on the forwards and backwards digit span tests rely on different cognitive processes [36]; forward digit span relies on attention while backwards digit span relies on executive function. It has been suggested, therefore, that these tests should not be combined into a single score, as per the MDAS digit span item [37]. Our results highlight that forward and backward digit span tests should not be used interchangeably, particularly in groups with underlying neurocognitive disorders, such as those with Lewy body disease, where impaired attention and executive function are common in the absence of delirium [38].

Although MOTYB backwards has been used as part of screening tools for cognitive impairment and dementia in PD [23, 35], to our knowledge this is the first study to show it may have clinical utility in detecting delirium in PD and LB-CI patients. Only MDAS attention was significantly associated with detecting delirium in all three groups (PD/LB-CI, other dementia and PD and dementia-free older adults), with stricter cut-offs being applied in those with PD/LB-CI or other dementia. This may be due to a higher proportion of participants with PD and/or dementia being unable to complete attention tests due to underlying cognitive impairment or due to low arousal levels, which is a limitation of most tests of attention [17]. The MDAS attention item, however, is a subjective rating based on the assessor’s evaluation of performance, e.g., whether the participant was distractible or loses track of the conversation, whether questions had to be repeated and how much this disrupted the interview. In the DECIDE and DETERMINE-PD studies, this scale was completed at the end of the assessment, which included several tests of attention. Therefore, the assessor’s subjective impression may be influenced by impairment in specific tests. Thus, the MDAS attention item in isolation without the other attention tests (e.g., digit span) should be used with caution as it may not have the same level of accuracy.

In participants with PD/LB-CI and dementia, level of arousal measures were more accurately able to detect delirium compared to attention tests. A recent systematic review of rapid delirium tests found that measures of level of arousal, specifically the OSLA and RASS, correctly identified the most cases of delirium in PD and dementia-free older adults [39]. As with the MDAS attention measure, the arousal measures are observational and so it is always possible to score in all participants [17]. Consistent with previous studies, the OSLA [17, 39] had good specificity and sensitivity for delirium, even in those with PD and/or dementia, while the m-RASS had good specificity in all groups [19, 39]. However, we found the GCS had overall better accuracy in PD and/or dementia participants. Previous studies have shown that abnormal levels of arousal are strongly associated with delirium [20]. However, patients with PD, LB-CI or dementia may have an overlapping phenotype [10, 16] and present with lower levels of arousal and fluctuations as part of their clinical phenotype (e.g., daytime somnolence) [40]. We found that in participants with PD and/or dementia, the optimal cut-offs for all measures of arousal indicated greater levels of impairment (i.e., poorer scores) in these groups compared to PD and dementia-free older adults. Using stricter cut-offs in these groups and repeated testing may, therefore, reduce the risk of false-positive results in these participants and increase overall accuracy when using these measures.

Similar to a previous study [17], we found that combined attention and arousal testing increased the diagnostic accuracy of identifying delirium across all groups. Combining tests has the advantage of being able to assess patients and research participants when cognitive testing is not possible and may be more appropriate in those with underlying cognitive impairment [39]. These findings have implications for clinical practice. We suggest that in older adults without an underlying neurodegenerative condition, the MDAS digit span (≥2) and OSLA (≥1) should be considered as brief assessments to identify delirium, while in participants with established PD and/or dementia, the MDAS attention measure (≥2) and GCS (≤13) may be more appropriate tests to screen for delirium.

Strengths of this study include a large number of participants from two studies assessed using robust standardised procedures based on the DSM-5 criteria. Although the largest cohort of its kind, the sample size is relatively small, with small numbers of PD, DLB and dementia participants relative to the older adult group. The small numbers in the subgroups may therefore lack statistical power to detect subtle differences in attention or arousal in some measures. There were also demographic differences between the cohorts in terms of age, sex, and years of education; future studies should use age, education, and sex-matched cohorts to compare test performance. We compared several tests of attention and arousal across groups that were common to both studies in this analysis but acknowledge that some commonly used bedside tests that were part of DECIDE (e.g., S-A-V-E-A-H-E-A-R-T and 20-to-1 tests) were not used in the DETERMINE-PD study and so could not be explored. Future studies should consider a wider range of attention and arousal tests. Finally, this analysis is cross-sectional and only includes participants with prevalent delirium on hospital admission and does not consider the fluctuating nature of delirium. Future studies should consider prospective longitudinal assessments to determine changes in attention and arousal measures in those with incident delirium and once delirium resolves.

We found that simple bedside tests of attention and arousal could accurately detect delirium in PD, dementia and PD and dementia-free older adults in hospitals; however, the optimal tests and cut-offs differ between groups. In inpatients with PD and LB-CI, backwards digit span, MOTYB and the MDAS attention item may be useful tests to identify impaired attention associated with delirium, while the GSC, OSLA, m-RASS and MDAS consciousness item may be useful tests to identify altered levels of arousal. Consistent with previous studies, a combined attention and arousal score performed better than the two tests individually. We propose that these tests have clinical utility to aid the identification of delirium in those with neurodegenerative disorders with overlapping presenting phenotypes in inpatients, specifically in PD and dementia. However, these tests should not be used to diagnose delirium in isolation as they do not capture the fluctuating course of delirium and should be used instead within the context of the DSM-5 criteria. Further work is needed to determine their utility in the community and by non-specialists in secondary care. Future studies should consider replicating these findings in a larger cohort across a wider range of neurodegenerative conditions, and evaluating a wider range of cognitive tests longitudinally to identify incident delirium, and developing robust composite tools to aid delirium diagnosis.