Rapid screening for neglect following stroke: A systematic search and European Academy of Neurology recommendations

Abstract Background and purpose Unilateral neglect is a common cognitive disorder following stroke. Neglect has a significant impact on functional outcomes, so it is important to detect. However, there is no consensus on which are the best screening tests to administer to detect neglect in time‐limited clinical environments. Methods Members of the European Academy of Neurology Scientific Panel on Higher Cortical Functions, neuropsychologists, occupational therapists, and researchers produced recommendations for primary and secondary tests for bedside neglect testing based on a rigorous literature review, data extraction, online consensus meeting, and subsequent iterations. Results A total of 512 articles were screened, and 42 were included. These reported data from 3367 stroke survivors assessed using 62 neglect screens. Tests were grouped into cancellation, line bisection, copying, reading/writing, and behavioral. Cancellation tasks were most frequently used (97.6% of studies), followed by bisection, copying, behavioral, and reading/writing assessments. The panel recommended a cancellation test as the primary screening test if there is time to administer only one test. One of several cancellation tests might be used, depending on availability. If time permits, one or more of line bisection, figure copying, and baking tray task were recommended as secondary tests. Finally, if a functional and ecological test is feasible, the Catherine Bergego Scale was recommended. Overall, the literature suggests that no single test on its own is sufficient to exclude a diagnosis of neglect. Therefore, the panel recommended that multiple neglect tests should be used whenever possible. Conclusions This study provides consensus recommendations for rapid bedside detection of neglect in real‐world, clinical environments.


INTRODUC TI ON
Unilateral neglect is a common poststroke cognitive impairment characterized by consistently lateralized spatial attentional deficits [1,2]. The occurrence of neglect acts as a key predictor of poor longterm recovery following stroke, with neglect patients experiencing lower quality of life and demonstrating reduced motor/functional abilities as well as higher levels of mood disorders than patients without neglect [3][4][5][6]. It is therefore critically important to detect neglect impairment to provide important prognostic indicators and to facilitate targeted rehabilitation approaches.
Currently, a wide range of methods are employed to screen for neglect within clinical environments [7]. Checketts et al. [8] conducted a large-scale, international survey aiming to identify common screening methods in clinical practice. Cognitive tasks were found to be the most popular form of neglect assessment (used by 82% of those responding to the survey), followed by functional assessments (used by 80%) [8]. A similar, Danish nationwide study conducted by Evald et al. [9] reported that subjective clinical observations were the most common assessment method, used by 90% of those surveyed, whereas pen-and-paper cognitive tasks were used by 49%. However, a wide range of individual tests were included within each of these reported assessment type categories. For example, Checketts et al. [8] reported on 14 different neuropsychological tests, including line bisection, copying, and cancellation tasks, within the "cognitive assessments" category. Similarly, 11 different screening methods were included within the "functional assessments" category, ranging from unstructured observations to standardized functional assessment tools [8]. Given this variation, it is important for clinicians to have access to recommendations for methods to detect neglect.

Previous investigations have come to varying conclusions on
whether it is better to use observational or pen-and-paper neglect screening methods [10][11][12], whether specific pen-and-paper tasks represent valid methods for detecting impairment [13,14], and what is the single best method for detecting neglect in clinical environments [15][16][17]. Overall, the existing literature has strongly suggested that, ideally, neglect should be screened for by comparing performance across a battery of independent and multimodal neglect assessments [15,[18][19][20][21]. However, given the time and resource constraints associated with real-world clinical environments, this practice is generally not feasible. It is therefore crucial to determine which neglect screening methods should be used in cases where real-world time and resource constraints allow for only one or a few screening tests.
An important issue for any recommendations regarding best tests to use for screening is that there is no objective gold standard against which tests can be compared. Furthermore, because there can be dissociations in the nature of neglect (e.g., egocentric vs. allocentric [22,23] or personal vs. extrapersonal [14,24,25]), some tests might, in theory, be able to detect only certain forms of neglect. It is important to detect different neglect subtypes, as previous research has demonstrated that these subtypes are dissociable and differentially associated with long-term recovery outcomes [5,15,16]. In addition, some patients show neglect during everyday functional tasks but perform normally on pen-and-paper tests of neglect, particularly due to testing/practice effects that can accompany repeated assessment. Thus, there are many factors in addition to reported number of neglect cases that must be considered to identify the best assessment methods. Moreover, neglect screening methods must be practical, inclusive, time-efficient, and easy to administer without specialist equipment or training. Given the diverse factors that must be considered when assessing the practicality of any single neglect test, there is a clear need for the existing literature to be systematically analyzed to identify the individual assessment methods that are most strongly supported by evidence.
Results: A total of 512 articles were screened, and 42 were included. These reported data from 3367 stroke survivors assessed using 62 neglect screens. Tests were grouped into cancellation, line bisection, copying, reading/writing, and behavioral. Cancellation tasks were most frequently used (97.6% of studies), followed by bisection, copying, behavioral, and reading/writing assessments. The panel recommended a cancellation test as the primary screening test if there is time to administer only one test. One of several cancellation tests might be used, depending on availability. If time permits, one or more of line bisection, figure copying, and baking tray task were recommended as secondary tests. Finally, if a functional and ecological test is feasible, the Catherine Bergego Scale was recommended. Overall, the literature suggests that no single test on its own is sufficient to exclude a diagnosis of neglect. Therefore, the panel recommended that multiple neglect tests should be used whenever possible.

Conclusions:
This study provides consensus recommendations for rapid bedside detection of neglect in real-world, clinical environments.

K E Y W O R D S
cognitive impairments, diagnostic screening programs, hemispatial neglect, stroke  The recommended tests were divided into primary and secondary categories. Primary tests represent assessments that were unanimously agreed to represent the best options for a timeefficient neglect screening assessment within a clinical environment.
Secondary tests include assessments that can be administered in addition to the recommended primary tests to provide additional details pertaining to the type, severity, and potential impact of neglect impairment.

Systematic search results
Systematic literature review yielded a total of 42 articles meeting all inclusion criteria ( Table 1). The process as well as the number of articles excluded at each stage is presented in Figure 1. Of the in- Cumulatively, these studies report data from 3367 stroke survivors assessed using 62 different neglect screening tools.
These screening tools can be grouped into cancellation, line bisection, copying, reading/writing, and behavioral test categories.
Cancellation tasks were found to be the single most frequently used assessment class (used in 97.6% of included studies), followed by bisection (used in 66.7%), copying tasks (used in 60%), behavioral tasks (used in 45.3%), and reading/writing assessments (used in 14.3%).
Overall, cancellation tasks most frequently resulted in the highest positive screening rates (most cases reported within 59.5% of studies), followed by behavioral (reported in 19.0%), bisection (reported in 14.2%), copying (reported in 4.8%), and reading/writing tests (reported in 2.4%; Table 1).
Within the 20 studies that conducted comparisons across several different cancellation tasks, the Star Cancellation from the Behavioral Inattention Test (BIT) [36] was most frequently found to be the best cancellation task (12/20), followed by the Bells Test [27] TA B L E 1 Summary of analyses conducted in studies identified within the literature review  [29]), so these findings may partially be explained by the test's popularity and history rather than its underlying sensitivity. For this reason, further direct, head-to-head studies are needed to evaluate assessment quality.  [5,23,79,80]. For this reason, cancellation tests that can distinguish between egocentric and allocentric neglect [29] are useful, if these are accessible.

Primary consensus recommendations
Reading-and writing-based neglect assessments were not recommended as primary neglect assessments, because assessment of function with these tasks might be precluded by comorbid language

TA B L E 1 (Continued)
and fine-motor deficits in a substantial portion of the stroke population [81][82][83][84]. Neglect assessments based on behavioral observations were not recommended for primary use due to the documented susceptibility to expectation biases due to lesion location [11].
Additionally, behavioral observation is not ideal for rapid, first-line assessment due to the potentially time-consuming need to observe patients interacting with real-world environments (as in the Catherine Bergego Scale, discussed below) [10].
Although line bisection tasks are commonly employed to quantify neglect, the results of some studies suggest that these tasks do not represent a valid method for detecting neglect impairment [14]. Bisection tests may measure a different behavioral construct than cancellation and copying tests [14] and yield finegrained continuous behavioral metrics that are vulnerable to confounding bias from comorbid fine-motor impairments, hemianopia, and optic ataxia [19]. For these reasons, line bisection tasks were not recommended for neglect screening if only one test can be used. Copying tests were also not considered to be suitable as the primary test for neglect screening due to potential interference from comorbid motor and cognitive deficits [85,86] as well as comparative difficulty in calculating quantitative neglect impairment scores [86]. Finally, the National Institutes of Health Stroke Scale (NIHSS) is commonly used as a first-line neglect screen in clinical environments [1]. However, previous literature has demonstrated that this screen is <30% sensitive compared to cancellation tasks, is highly susceptible to clinician expectation biases, and commonly misdiagnoses visual field impairment as neglect [1,87]. For these reasons, the NIHSS was not recommended to be used for neglect assessment.
Overall, the expert panel unanimously agreed that a form of cancellation task should be the first choice of neglect assessment if there is time for only one test ( Table 2). If available, established measures such as the BIT Sar Cancellation [36] and Bells Test [27] can be used. The OCS Hearts Cancellation Test [29] and the BCoS Apples Cancellation Task [16,69] are also recommended as primary neglect assessment methods within clinical environments. The latter tests are openly available, and also provide a potential means to distinguish between egocentric and allocentric neglect.

Secondary consensus recommendations
Due to its heterogeneity, previous research has shown that neglect should be screened for by comparing performance across several, independent neglect assessments [18,19,21,26]. Where time allows for more than one test, clinicians should conduct additional neglect assessments. Therefore, the included literature was analyzed by the expert panel to provide secondary recommendations for additional neglect assessments.
Three types of test were recommended by the panel as adjuncts to a cancellation test. Despite discussed limitations associated with using manual line bisection tasks to assess visuospatial neglect impairment, some previous studies have identified patients demonstrating neglect on bisection, but not cancellation tasks [64,65,72].
Prior research has suggested that manual line bisection tasks may be most appropriate for detecting co-occurrence between visual field deficits and egocentric neglect [88]. In cases where bisection tasks are used, clinicians should aim to employ standardized manual bisection tasks with published normative performance thresholds (e.g., Wilson et al. [36]) rather than improvised, original tasks. This use of normative data is critically important, as controls have also been found to exhibit small biases in line bisection tests [76,77]. Given some of the limitations associated with using bisection tests to quantify neglect [14], the panel recommended that they be used as secondary assessments, but biased performance on bisection tests alone should not be considered sufficient evidence to detect neglect impairment.
Next, figure copy tasks [89] were also recommended for secondary neglect assessment. These are easy to administer and improvise within clinical environments. Copying and drawing tests may help provide insight into some components of neglect not clearly assessed by standard cancellation tests (e.g., drawing from memory for representational neglect) [90]. However, past research has demonstrated that these tasks might detect a lower frequency of neglect than cancellation tasks and are reliant on subjective interpretations F I G U R E 1 Visualization of systematic literature search and exclusions at each stage of impairment rather than quantitative scoring systems [90]. As in cancellation tests, a wide range of copying-based neglect assessments are in use. In general, copying tasks that display multiple stimuli on the horizonal axis and are able to distinguish between egocentric and allocentric neglect (e.g., scene copying tasks) are more informative than those that employ simpler stimuli (e.g., daisy copying) [90,91].
Finally, baking tray tasks [92] were recommended as a secondary neglect assessment method. In this task, patients are asked to arrange items evenly across a tray as if they were "buns on a baking tray" [92]. Patients with egocentric neglect have been found to demonstrate a clear spatial bias on this task, crowding all items onto one side of the tray area [15,92]. Baking tray tasks are easy to improvise within clinical environments by making a "tray" and "items" with standard, normed dimensions (e.g. Facchin et al. [93]). This task has been demonstrated to be highly sensitive to neglect if it is possible to perform it in a clinical environment [15,28,29].

Functional/ecological evaluation of neglect recommendation
The panel also acknowledged that if time is available and if the patient's condition allows, functional/ecological tests should be performed. If feasible, the Catherine Bergego Scale [10] was recommended. This is a functional observation checklist that provides a naturalistic assessment of how neglect impairment manifests in realworld activities, such as grooming and navigation. A standardized protocol for administering this assessment has been developed [94].
It can outperform many pen-and-paper assessments in detection of neglect [94. However, it is generally not feasible to observe all the behavior necessary to accurately complete the checklist within very brief initial clinical assessments, and the assessment requires experienced observers. Nevertheless, it can be a useful adjunct to rapid bedside assessments.
The Dublin Extrapersonal Neglect Assessment [45] also provides a highly naturalistic and informative assessment of how neglect impairment impacts on real-world function. In this test, patients are asked to navigate through a hallway and locate a series of signs placed by the examiner [45]. However, this requires patients to mobilize (or be assisted) down a hallway, which is often not possible, particularly in hyperacute stroke. For this reason, this might be better suited for use in a slightly later stage of the stroke pathway (e.g., occupational therapy assessment for discharge planning). Furthermore, because this test has not been extensively deployed, the panel did not recommend its routine use.

DISCUSS ION
It is critically important to screen for neglect in patients with stroke, as the occurrence of this cognitive deficit has been found to be a key predictor of recovery outcomes [3,4,5,6,16]

Conclusions
Overall, this study provides expert consensus recommendations on the best ways to detect neglect impairment within real-world clini- None of the other authors has any conflict of interest to disclose.

DATA AVA I LI B I LIT Y S TATEM ENT
All data associated with this project are openly available on the Open Science Framework (https://osf.io/fzmde/).