Identifying spatial neglect – an updated systematic review of the psychometric properties of assessment tools in adults post-stroke

ABSTRACT

Spatial neglect commonly occurs after a stroke, resulting in diverse impacts depending on the type and severity. There are almost 300 tools for assessing neglect, yet there is a lack of knowledge on the psychometric properties of these tools. The objective of this systematic review, registered on Prospero (CRD42021271779), was to determine the quality of the evidence for assessing spatial neglect, categorized by neglect subtype. The following databases were searched on 3rd May 2022 from database inception: Ovid Emcare, Embase, Ovid MEDLINE, APA PsycINFO, Web of Science (SCI-EXPANDED; SSCI; A&HCI; ESCI) and Scopus. All primary peer-reviewed studies (>5 participants) of adults post stroke, reporting any psychometric property of 33 commonly used neglect assessment tools were included. The COSMIN (COnsensus-based Standards for the selection of health Measurement INstruments) risk of bias tool was used to assess the methodological quality of the studies and summarize the psychometric properties of each tool. 164 articles were included, with a total of 12,463 people with stroke. The general quality of the evidence was poor and no one tool had high-quality evidence of both validity and reliability. Eleven tools show some promise as they meet the minimum criteria for good measurement properties for both validity and reliability.

KEYWORDS:

• Spatial neglect
• Psychometrics
• Stroke
• Patient outcome assessment
• Neuropsychological tests

Introduction

Spatial neglect, experienced by approximately 50% of all stroke survivors (Esposito et al., 2021; Pedroli et al., 2015), is characterized by difficulties with reporting, responding or orientating to stimuli from the side opposite the brain lesion, and this deficit is not attributable to a primary motor or sensory impairment (Heilman et al., 2003). Subtypes of neglect have been identified to capture the range of neglect manifestations observed in different domains such as: visual, tactile, auditory, motor, representation (mental imagery); and the spatial regions of personal (on the body), peripersonal (near space) and extrapersonal (far space), see (Williams et al., 2021) for a summary. Spatial neglect can also be categorized as either egocentric (body-centered) and/or allocentric (object-centered) as it relates to the frame of reference.

There are almost 300 tools for the screening or diagnosis of neglect (Williams et al., 2021), however there is a lack of knowledge of the psychometric properties of many of these tools. Many assessment tools underdiagnose neglect as they lack the sensitivity to detect mild or moderate neglect (Bonato et al., 2013; Pflugshaupt et al., 2004; Puig-Pijoan et al., 2018). A recent scoping review provided an overview of all the neglect tools and identified many tools that do not specify the neglect subtype assessed (Williams et al., 2021). This review analyzed if the scoring differentiated behaviours for specific neglect subtypes, however, did not review the psychometric properties. One systematic review has summarized the psychometric properties of 19 spatial neglect assessment tools (Menon & Korner-Bitensky, 2004). However, this previous review is outdated as many more tools and studies of psychometrics have been published in the subsequent 20 years. Additionally, this earlier review only included egocentric and perceptual components of neglect and did not include tools for representational, motor, allocentric, tactile, or auditory neglect (Menon & Korner-Bitensky, 2004). A recent international survey of tools used in current practice has recently been completed (Checketts et al., 2021). However, to our knowledge, no recent review has systematically compared relevant psychometrics such as validity and reliability across the varied tools used in clinical practice today.

Assessment tools with good psychometric properties are crucial for accurate and reliable measurement of spatial neglect. The psychometric properties of tools need to be rigorously tested to show that they measure what they intend to measure (content and construct validity), they can do this accurately (criterion validity, sensitivity and specificity), and consistently (reliability), and can detect changes over time (responsiveness) (Mokkink, Prinsen, et al., 2018). Using tools that do not have adequate validity and/or reliability, introduces randomness and imprecision. This can have detrimental consequences such as improper diagnosis, undetected neglect, or inappropriate interventions, thus negatively impacting rehabilitation outcomes (Kerkhoff & Schenk, 2012).

Tools with high test-retest reliability are essential to track change in neglect signs over time (Aldridge et al., 2017). However, measuring the reliability of neglect assessment tools is challenging, as reliability inherently refers to the ability of a tool to measure a stable construct over time (Mokkink et al., 2020). However, spatial neglect is not stable over time, as the severity of neglect signs can vary from day-to-day and even over the course of a therapy session (Anderson et al., 2000; Woods et al., 2017), and often improves over time. Therefore, studies exploring the reliability of tools to consistently identify neglect need to be carefully planned and implemented, taking into account the varying nature of neglect. A low value for a test-retest correlation may be due to the varying nature of neglect itself, rather than the variability in how it is measured.

This systematic review provides an update on the state of the evidence by critically appraising the psychometric properties of commonly used assessment tools for identifying spatial neglect in clinical practice. This review applied the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) methodology for systematic reviews (Mokkink et al., 2020; Mokkink, de Vet, et al., 2018). A recent international survey for screening and assessment of neglect was used to identify the tools most commonly used in clinical practice (Checketts et al., 2021) for inclusion in this review.

The questions that this systematic review answered are:

1. What are the psychometric properties (validity, reliability, responsiveness) of spatial neglect tools that are commonly used in clinical practice, categorized by neglect subtype?

2. What is the feasibility of the spatial neglect tools? (Feasibility defined as the ease of administration and process, such as the time it takes to complete, resources and training required)

3. What is the utility of the spatial neglect tools? (Does the tool provide useful and relevant information for clinicians, such as whether the tool specifies the neglect severity and what types of neglect can be identified?)

Methods

Protocol and registration

This systematic review followed the PRISMA 2020 statement: An updated guideline for reporting systematic reviews (Page et al., 2021), and the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) methodology for systematic reviews (Mokkink, de Vet, et al., 2018; Mokkink et al., 2020). The review was registered on Prospero, the international prospective register of systematic reviews (registration number: CRD42021271779) and is available at: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021271779. This review was registered prior to the final selection of tools to be included in this review.

Inclusion of tools in this review:

Initially all spatial neglect screening and assessment tools were to be included. However, after almost 300 tools for the assessment of spatial neglect were determined in a recent scoping review (Williams et al., 2021), the criteria for choosing included tools were refined. The updated criteria included tools: (1) being used in clinical practice (identified from an international survey on current clinical practice (Checketts et al., 2021), or suggested for inclusion from a multi-disciplinary panel of experts International group for Spatial Attention and Neglect Disorders (I-SAND) (www.isand.org)); (2) designed to detect neglect of any type; (3) with evidence of standardization (clear documented protocols that are replicable); (4) some form of validity and reliability reported (determined through a preliminary search of the literature which included screening previous systematic reviews, review of the stroke engine website and Google scholar search of tool name until two full pages of irrelevant hits); (5) are transdisciplinary (Tools were excluded if they were specifically designed to be used by only one discipline. For example, the “Arnadottir OT-ADL Neurobehavioural Evaluation (A-ONE) Assessment is designed specifically for Occupational Therapists and as such the administration and scoring of the tool assumes prior knowledge of occupational therapy theory); (6) tools available in English. Refer to Online Resource 1 for the full list of all 33 included assessment tools, and the screening flowchart which outlines the inclusion of tools and the list of exclusion reasons.

Search strategy

A search strategy was developed based on the scoping review of spatial neglect subtypes, definitions and assessment tools (Williams et al., 2021), which was modified from a Cochrane review of spatial neglect (Bowen et al., 2013) by removing search strings aimed to find RCTs. The names and abbreviations of the included tools were added to the search strategy, along with a validated search filter for finding studies on measurement properties (Terwee et al., 2009). The search strategy was then reviewed by an academic librarian to ensure all relevant studies were likely to be found. The final search was completed on the following databases on 3rd May 2022, from database inception: Ovid Emcare, Embase, Ovid MEDLINE, APA PsycINFO, Web of Science (SCI-EXPANDED; SSCI; A&HCI; ESCI) and Scopus. Refer to Online Resource 2 for the full search strategy.

Eligibility criteria

This systematic review included primary studies reporting any psychometric property (defined by COSMIN) of the 33 included spatial neglect tools in adults’ post-stroke (with more than 85% of population with a diagnosis of stroke), and no restriction to setting or recovery stage post stroke. Studies were excluded if not peer reviewed (e.g., book chapters); were not written in English; had five participants or fewer; or if the sample included neglect resulting from a traumatic brain injury or tumour. Studies including children, or only healthy controls were also excluded. Two reviewers independently screened the titles and abstracts in Covidence for eligibility for inclusion. The full text of potentially eligible studies were retrieved and assessed for full eligibility independently by two researchers from the authorship team. Any discrepancies over the eligibility of a particular study were resolved by a third researcher or a team discussion.

Data extraction

A data extraction form was developed based on the COSMIN data extraction templates for systematic reviews (Terwee and Mokkink). Data were extracted from the full texts of potentially eligible studies by the first author. A selection of 46 studies were randomly selected to have the data extraction checked by an additional researcher; with 30 studies independently extracted, and 16 studies double-checked to ensure completeness and accuracy of extracted data. The agreement between the data extracted for the randomly selected 46 studies was 97.9% agreement. The following data items were extracted from each study (1) characteristics of the included patient sample; (2) characteristics of each tool, such as a description of each test/subtest, administration procedures, scoring system, utility and feasibility; and (3) measurement properties for each tool defined as per COSMIN (Mokkink et al., 2020; Mokkink, de Vet, et al., 2018). Refer to Online Resource 3 for the full list of data items extracted.

Evaluation of the content validity as per the COSMIN guidelines was considered not appropriate as this section of the guidelines is very specific to patient reported measures (Mokkink, Prinsen, et al., 2018), whereas the tools included in this review were all clinician reported or performance based measures. Additionally, it was considered not necessary to evaluate content validity as the determination of the constructs (neglect subtypes) assessed by each tool was completed in a recent scoping review (Williams et al., 2021). The descriptions of each neglect tool and the documented protocols were systematically analyzed by two authors of the scoping review to determine (1) the main neglect subtypes that are specifically identified by each tool, and (2) other subtypes that deficient performance may be attributed to, but the scoring does not allow for them to be differentiated (Williams et al., 2021). Refer to the scoping review for a full description of the methodology (Williams et al., 2021). A summary of the neglect subtypes assessed by each tool (from the scoping review) was added to the utility section in this systematic review as this provides context to clinicians and researchers on the information that each assessment provides.

Risk of bias assessment

The methodological quality and the quality of the results were evaluated for each study as per the COSMIN guidelines (Mokkink, de Vet, et al., 2018; Mokkink et al., 2020). The risk of bias checklist included an evaluation of methodological design, statistical methods and bias relating to other flaws relevant to each psychometric property. The risk of bias checklist was completed several times for each article if it included data for multiple psychometric properties, such as internal consistency, structural validity, reliability, and construct validity. Each standard in the COSMIN risk of bias checklist was rated on a four-point scale as either “very good”, “adequate”, “doubtful” or inadequate”. For a detailed explanation of the individual standards and guidelines, refer to pages 23–37 of the COSMIN risk of bias checklist (Mokkink et al., 2018), and Mokkink et al., for the updated standards of assessing reliability and measurement error (2020). The overall methodological quality of each psychometric property reported in each study was then rated according to the “worst-score” counts method according to the COSMIN guidelines for the simple reason that poor methodological aspects cannot be overcome by good aspects (Mokkink, de Vet, et al., 2018). The risk of bias evaluation was completed by the first author for all studies, with 101 studies rated by another member of the research team, achieving 89.23% agreement. Any discrepancies were resolved by discussion until a consensus was reached.

Appraisal of measurement properties

Refer to Table 1 for the definitions of the psychometric properties, according to COSMIN, that were appraised for all the included tools (Mokkink et al., 2018). The psychometric properties reported in the individual studies were compared against COSMIN criteria for good measurement properties, and rated as either sufficient, insufficient, or indeterminate (Refer to COSMIN manual for detailed criteria) (Mokkink, Prinsen, et al., 2018, pp. 28–29). The first author appraised the psychometric properties in all the studies, with 101 studies appraised by another member of the research team with 91.7% agreement. Any discrepancies were resolved by discussion until a consensus was reached.

Table 1. Definitions of the psychometric properties that were appraised for all the included tools, as per COSMIN guidelines.

Structural Validity:	The degree to which the scores of all the items in a tool are an adequate reflection of the dimensionality of the construct to be measured. Evidence for structural validity is a prerequisite for evaluating internal consistency. If a construct is seen as a single domain, a factor analysis will produce a single factor and all items will be assumed to be related.
Internal Consistency:	The degree of interrelatedness among the items in a tool (based on the assumption that all items should be related). An aspect of reliability that examines the maintenance of a consistent score across all subtests for the same patient.
Reliability: (test-retest, inter rater and intra – rater)	The extent to which scores are the same for repeated measures in a person that has not changed. This includes measures over time (test-retest); by different persons on the same occasion (inter rater); or by the same persons on different occasions (intra-rater).
Measurement Error:	The systematic and random error of a patient’s score that is not attributed to true changes in the construct to be measured. A tool with acceptable measurement error has a smallest detectable change (SDC) that is smaller than the Minimal Important Change (MIC).
Construct Validity:	The degree to which the scores of a tool are consistent with known hypotheses – such as relationships to other tools, or differences between groups. This assumes that the tool validly measures the construct to be measured.
Criterion Validity:	The degree to which the scores of a tool are an adequate reflection of a “gold standard”.
Responsiveness:	The ability of a tool to detect change over time in the construct to be measured.

As there is no gold standard tool for identifying spatial neglect, sensitivity analyses were unable to be compared or synthesized across the studies. COSMIN recommends that, in this case, sensitivity analyses be evaluated as evidence of construct validity rather than criterion validity where appropriate (Mokkink, de Vet, et al., 2018). Criterion validity was only considered when the sensitivity/specificity of a modified or simplified version of a tool was compared to the original full version.

Structural validity is only relevant for tools with more than one subtest, that are based on a reflective model, which assumes all items and subtests of the tool are manifestations of one underlying construct, and thus expected to be correlated (Mokkink, de Vet, et al., 2018). For this review, all tools with subtests were assumed to be based on a reflective model (if not otherwise stated), so structural validity was examined. The review team also decided that neglect tools would expect to be correlated as evidence for construct validity. However, as neglect is not considered to be a unitary syndrome due to many documented double dissociations between different neglect subtypes (Demeyere & Gillebert, 2019; Ortigue et al., 2006), it was expected that the correlations between tools assessing different neglect subtypes would not necessarily be ≥0.5, as COSMIN recommends for similar constructs. Therefore the criteria for sufficient construct validity were any significant correlations >0.3 with another neglect tool, as per the COSMIN recommendations for “related but dissimilar constructs’ that should have correlations between 0.3 and 0.5 (de Vet et al., 2011; Mokkink, Prinsen, et al., 2018).

Data synthesis

The results of available studies for each psychometric property were pooled or summarized for each tool to determine the overall ratings as per COSMIN (Mokkink, de Vet, et al., 2018). If the results of all the studies were consistent, then the results were pooled to give an overall rating of either “sufficient” (+) or “insufficient” (-). If the results were inconsistent (for example, both sufficient and insufficient results were found), then three strategies were used: (1) the conclusions were based on the majority of consistent results (if >75% were consistent); (2) explanations for the inconsistencies were explored and if found, were summarized per subgroup, or; (3) an overall rating of “inconsistent” was given. Refer to Figure 1 for a flow chart of how the ratings were assigned for each psychometric property of a tool.

Figure 1. Decision flowchart for rating the overall psychometric property for each assessment tool, according to COSMIN.

Decision flowchart for rating overall psychometric properties for each tool as either sufficient, insufficient, or inconsistent.

The overall quality of evidence for each measurement property of a tool was then collated across the studies and summarized according to the modified GRADE approach as either (Mokkink, de Vet, et al., 2018):

• High – confident that the measurement property lies close to the estimate (pooled or summarized result).

• Moderate – moderate confidence in the measurement property estimate. It is likely to be close to the estimate (pooled or summarized result), but there is a possibility that it is substantially different.

• Low – limited confidence in the measurement property estimate. The true measurement property may be substantially different from the estimate.

• Very Low – very little confidence in the measurement property estimate. The true measurement property is likely to be substantially different from the estimate.

As uni-dimensionality is a prerequisite for internal consistency analyses, the quality of the evidence for structural validity was assessed first and taken as a starting point for interpreting the quality of the evidence for internal consistency (Mokkink, Prinsen, et al., 2018). For example, if the structural validity for a tool was rated “moderate”, then this was the starting point for rating internal consistency of that tool (rather than starting from “high”) and the quality was downgraded accordingly.

The risk of bias ratings and ratings for each measurement property were charted using the COSMIN online tools (see https://www.cosmin.nl/tools/guideline-conducting-systematic-review-outcome-measures/) (Terwee and Mokkink).

Results

This systematic review included 165 studies after the systematic search and screening process. For details refer to Figure 2 for the search and screening flowchart. A total of 12,974 people with stroke (Mean: 79; SD: 102), and 381 healthy controls were recruited to the included studies, from 29 different countries. The mean age of participants ranged from 55.1–77 years. Table 2 presents a description of the 33 assessment tools, alternative versions and scoring methods.

Figure 2. Search and screening flowchart.

Flowchart depicting how many records were identified from each database and pearling, number of duplicates removed, number of abstracts and full texts screened for eligibility, and the number of articles included and excluded at each stage.

Table 2. Description of the assessment tools, alternative versions and methods of scoring.

Tool name (reference)	Description of test components/subtests	Scoring	Modified versions/scoring	Neglect subtypes	Feasibility
Behavioural Inattention Test (BIT) (Wilson et al., 1987)	Conventional (BIT-c) subtest 6 items (line cancellation, letter cancellation, star cancellation, line bisection, figure copying, drawing) Behavioural (BIT-b) subtest 9 items (menu and article reading, phone dialling, time setting, coins and coin sorting, sentence copying, map navigation, scanning 3 images)	Scores of each subtest are summed. Total score cut-off 196 (Max 227). Conventional subtest cut-off 129 (Max 146). Behavioural subtest cut-off 67 (Max 81) Starting points of cancellation tasks are measured from centre of page.	BIT-10–10 item version mBIT – 8 item version sBIT – 3 item version Modified scoring: Index of laterality (Halligan et al., 1991) Proportional severity = [max score minus score obtained] /max score x 100 (Luukkainen –Markkula, 2011)	Egocentric neglect in peripersonal space Cannot differentiate visual & motor neglect.	Time: 30–40 min To see the tool: Pearsonassessments.com (at a cost)
mBIT (Stone et al., 1991)	8 item version of the BIT: (1) Pointing to objects about the ward; (2) Food on a plate; (3) Reading menu; (4) Reading newspaper; (5) line cancellation; (6) star cancellation; (7) coin selection and (8) figure copying.	Percentage of omissions for each test is calculated, and starting point is recorded.	No alternative versions	Egocentric neglect in peripersonal and extrapersonal space. Cannot differentiate visual & motor neglect.	Time: not stated To see the tool: Stone et al. (1991)
Sunnybrook Neglect Assessment Procedure (SNAP) (Leibovitch et al., 2012)	Drawing a clock and daisy Line cancellation Line bisection Copying of clock and daisy Mesulam Shape cancellation.	A weighted sum of the four SNAP subtests, a total neglect score out of 100. Score < 5 = within normal limits, 5–40 = mild-mod neglect, > 40 = severe neglect. (OR score ≥1 on line cancel or drawing indicates neglect)	No alternative versions	Egocentric neglect in peripersonal space Cannot differentiate visual & motor neglect.	Time: not stated To see the tool: Leibovitch et al. (2012)
BIT line bisection (Wilson et al., 1987)	Three horizontal, 8-inch (204 mm) black lines (l mm thick) displayed in a staircase fashion across the page. The extent of each line is clearly pointed out to the patient who is then instructed to mark the centre of each line.	Deviations from the midpoint of each line are measured. Scores range 0–3 for each line from the template. Smaller score indicates greater deviation from midline. Max score of 9. Scores <6.5 indicate neglect.	No alternative versions	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Pearsonassessments.com (at a cost, via BIT test kit)
Line bisection (Schenkenberg et al., 1980)	20 horizontal lines on A4 sheet of paper. Eighteen lines organized in three sets of six (centre, right and left). Each set contains lines of 10, 12, 14, 16, 18 and 20 cm. Two 15 cm lines at the top and bottom (centre) of the page are used when giving instructions.	A cancellation score (max 18). A measured bisection score includes the percent deviation for each line and the mean percent deviation for each set of lines at each position on the page, and for the whole page combined.	An alternative form is obtained by rotating the paper 180”. Alternate scoring: % deviation = (measured right part) ÷ (measured right part + measured left part) x 100.	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Schenkenberg et al. (1980)
Line cancellation	Includes 40 lines each 2.5 cm in length drawn in apparently random manner on an A4 sized landscape sheet (18 lines on each half of the page arranged in three columns of 6 lines each) and four central lines.	Number of omissions on L/R sides and asymmetry score. Starting point – distance from median line. (Range −13.4–13.2 cm) (Kettunen et al., 2012; Nurmi et al., 2010).	Alberts original version presented on a slightly smaller sheet of paper (11x 8.5-inch) with 41 lines 2 cm in length each.	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Pearsonassessments.com (at a cost, via BIT test kit)
Letter cancellation task (Diller & Weinberg, 1977)	Includes 104 target letters (uppercase H’s) interspersed among 386 distractor letters arranged in six horizontal lines on a horizontal A3 sheet of paper.	Cancellation score (ranging from 0 to 104). Cutoff ≥ 4 omissions on either side of page, or > 2 difference in omissions between the left and right sides.	No alternative versions	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Diller and Weinberg (1977)
Bells Test (Gauthier et al., 1989)	Cancellation test on a landscape sheet of A4 paper, including 35 small bells dispersed among 280 distractors in apparently random array, with targets equally distributed in 7 columns (with 5 targets and 40 distractors in each) Three columns on the Left side, three on the right and one in the middle.	The total number of omissions (/35). Cut-off >3 omissions for “attentional deficit”, or >5 on contralesional side for visual neglect. Starting point is recorded.	Alternative scoring/cut-offs: • Lateralised score = difference between L/R omissions (Azouvi et al., 2002). • Cut-off >2 difference between R/L omissions (Saj et al., 2012) • Centre of cancellation (Rorden & Karnath, 2010)	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: (Gauthier et al., 1989) or strokengine.ca
BIT Letter cancellation (Wilson et al., 1987)	Includes 40 target letters (E, R) among 130 non-target letters, arranged in five rows on an A4-sized landscape sheet (maximum score 40, cutoff point 32). Two additional letters (E & R) beneath the actual stimulus rows serve as examples and are not scored.	Scoring template divides the total array into four columns, two left and two right. The number and location of the omissions is noted.	N/A	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Pearsonassessments.com (at a cost, via BIT test kit)
BIT Star cancellation test (Wilson et al., 1987)	Cancellation test printed on horizontally presented A4 paper, consisting of a random array of 56 small stars (the targets), 52 larger stars, 13 letters, and 10 short words. The two small stars in the centre are used for demonstration.	Cut-off for neglect: • <44 (Halligan et al., 1992). • Laterality index = (Halligan et al., 1992) left cancellations divided by total cancelled. Range 0–1. L neglect = 0–4.6 R neglect = 0.54–1 • Starting point (Kettunen et al., 2012; Nurmi et al., 2010; Samuelsson et al., 1996)	Alternative scoring/cut-offs: • < 51 (Bailey et al., 2004; Civelek et al., 2015; Kim et al., 2021) • older adults <39 (Bickerton et al., 2011; Stone et al., 1991) Omissions on one side: • ≥3 omissions (Gerafi et al., 2017) • ≥5 omissions (Ye et al., 2022) Centre of cancellation (Rorden & Karnath, 2010)	Egocentric neglect in peripersonal space Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Pearsonassessments.com (at a cost, via BIT test kit)
Ogden Figure copy (Ogden 1985)	Required to copy four objects arranged horizontally on an A4 sheet of paper (pine tree, fence, house, and right tree).	Scored from 0 (no omissions) to 4 (omission of the left tree and one half of another element). Score +1 for omission of each object, +0.5 for objects with missing elements on one side and 0 for a complete figure. Conversely, scored −0.5 for right-sided object-centered neglect and −1 for right scene-based omissions (Lee et al., 2004).	Alternate scoring: Stein et al., 2022 - 36 total components (pen strokes), with 16 components on the left and 20 on the right. Each omitted component scored as an error; and misplaced or distorted components scored as half an error.	Neglect in peripersonal space, egocentric and allocentric neglect. Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Ogden (1985)
Clock drawing (15 pt scale) (E.g. Clox)	CLOX 1 – Required to draw a clock with the time set to 1:45. CLOX 2 – Observe the examiner sketch a clock within a pre-drawn circular stimulus and are invited to copy the figure.	A 15-point scoring system. Drawings scored according to the presence of critical features (e.g., resemblance to a clock and the absence of inappropriate details) (Delazer et al., 2018; Rengachary et al., 2009; Zuverza-Chavarria & Tsanadis, 2011).		Neglect in peripersonal space. Cannot differentiate visual and motor neglect, or egocentric and allocentric neglect.	Time: < 5 min
Catherine Bergego Scale (CBS) (Azouvi, 1996)	Standardized checklist to guide the observation of neglect behaviour in 10 everyday activities (grooming or shaving, dressing, eating, mouth cleaning, gaze orientation, knowledge of the left side of the body, auditory attention, collisions into objects on the left while moving, finding one’s way, and finding personal belongings). Can be used as a patient reported evaluation to determine anosognosia.	Each item score of 0 (no neglect) to 3 (severe neglect), with total score of 30. Cut-off: ≥ 1 Classification: 1–10 = mild neglect 11–20 = moderate neglect 21–30 = severe neglect	Alternative cut-offs: • ≥ 5 (De-Rosende-celeiro et al., 2021; Luukkainen-Markkula et al., 2011) • ≥ 6 (Nelemans et al., 2022; Rousseaux et al., 2015) Scoring neglect subtypes: Extrapersonal space = items 5, 7, 8, 9 Personal space = items 1, 2, 4, 6. (Cunningham et al., 2017)	Neglect in personal, peripersonal and extrapersonal space. Scoring does not differentiate visual, motor, representational or tactile neglect.	Time: 30–40 min To see the tool: Azouvi (1996)
CBS via KF-NAP (Kessler Foundation Neglect Assessment Process)	Defines and standardizes the administration of the 10 CBS items. All items are completed during one session and in a specific order.	Each item is scored 0–3, higher score indicating more severe symptoms. Total score of 0 = no neglect 1–10 = mild neglect 11–20 = moderate neglect 21–30 = severe neglect.	N/A	Egocentric neglect in personal, peripersonal, and extrapersonal space. Scoring does not differentiate visual, motor, representational or tactile neglect.	Time: 30–40 min To see the tool: https://www.kflearn.org/courses/KF-NAP
Mobility Assessment Course (Verlander et al., 2000)	Required to locate 24 laminated cards (each 10 × 10 cm) positioned on each side (12 left, 12 right) in a corridor of ∼1.5 m wide and ∼2 m high while mobilizing (independent or with walking frame or in wheelchair if needed) along a predefined route in the building.	(1) Total % of targets found (2) Asymmetry – difference between sides (3) number of omissions (4) number of collisions (Grech et al., 2017) Score of >2.3 = neglect	Modified protocol in attendant propelled wheelchair suitable for patients with reduced stamina/ strength (Ten Brink et al., 2018)	Egocentric neglect in peripersonal and extrapersonal space. Cannot differentiate visual and motor neglect.	Time: 2–9 min To see the tool: Ten Brink et al. (2018)
Wheelchair obstacle course (Whitehouse et al., 2019) modified from Webster et al. (1989)	Required to mobilize in a wheelchair twice along a course (once backward, once forward) of 150 feet long × 5 feet wide demarcated by rope, requiring three right and three left turns, with six obstacles (folding chairs) located on each side	Two types of errors investigated: Direct hits (direct frontal contact with an obstacle or border) and sideswipes (contact made with side plane of the chair). (Webster et al., 1989)	Moss-Magee Wheelchair Navigation Test (MMWNT) modified from Webster et al., (1989) is another alternative version (not included in this review).	Egocentric neglect in peripersonal and extrapersonal space. Cannot differentiate visual and motor neglect.	Time: not stated To see the tool: Whitehouse et al. (2019)
Semi-Structured Scale for the functional evaluation of personal neglect (Zoccolotti et al., 1992)	Required to demonstrate the use of common objects – comb, razor/compact, eyeglasses.	Score from 0 (no asymmetries) to 3 (reduced exploration of space on one side) for each object use. Total score of 9.	N/A	Egocentric neglect in personal space. Cannot differentiate visual and motor neglect.	Time: not stated To see the tool: Zoccolotti et al., (1992)
Comb and Razor Test (McIntosh et al., 2000)	Give comb and ask person to comb hair for 30 s. Categorize each stroke of the comb as either on the left or right of the head or ambiguous. Razor condition (used with men). Powder compact case (used with women). Same procedure used but with a brush.	Scoring calculated with formula: % left = (left strokes)/(left + right + ambiguous strokes).	Alternative scoring: % bias = (left – right stroke) / (left + ambiguous + right strokes). score range – 1 (total left neglect) to +1 (total right neglect) (McIntosh et al., 2000)	Egocentric neglect in personal space. Cannot differentiate visual and motor neglect.	Time: Not stated To see the tool: McIntosh et al., (2000)
Two Daisy figure (Marshall & Halligan, 1993)	Required to copy a drawing of two flowers in pot on A4 paper centred on the midsagittal plane, immediately above a blank A4 sheet.	Three objects are scored (left daisy, right daisy and pot). Omission of a petal or stalk on one side considered abnormal. Max score of 3. Score +1 for omission of each object, + 0.5 for objects with missing elements on one side and 0 for a complete figure (Lee et al., 2004).	N/A	Egocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Marshall and Halligan (1993)
Baking Tray Task (BTT) (Tham & Tegner, 1996)	Tray or board (100 × 75 cm). 16 (3.5 cm) cubes in a box placed directly in front. Required to “place the blocks as symmetrically as possible as if they were buns being placed on a baking tray to be put in the oven”. All 16 cubes need to be used and are reminded if any omitted.	Score number of cubes in each half of the tray. A cube placed on the midline, scores half for each side of the tray. The ratio, of cubes placed on left side of tray to the total of 16 is scored (Bailey et al., 2000) Cut-off = >2 difference between left/ right.	Smaller “Baking Tray” version: consisting of whiteboard (75 × 50 cm) (Bailey et al., 2000, 2004).	Egocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Tham and Tegner (1996)
Apples Test	Required to search for and cross out all the complete apples on a landscape A4 page containing 150 apples pseudo randomly scattered. Two thirds of the apples are distractors (half with an opening on the left side and half with an opening on the right side). Up to two practice trial can be presented before the test to ensure understanding of instructions. Max of 5 minutes is allowed.	Accuracy score = total number of targets cancelled (max. 50). Cut-off = >42/50 Egocentric asymmetry = difference between the right-side cancelled targets and left-side cancelled targets (max. 20). Cut-off = < – 2 or >2 Allocentric asymmetry = difference between the number of distractors cancelled with a left opening and the number cancelled with a right opening. Cut-off = <−1 or >1.	Restriction to only positive scores, excluding from VLSM analysis patients with negative scores (Spano et al., 2022)	Egocentric and allocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: Bickerton et al. (2011)
Broken Hearts cancellation (Oxford Cognitive Screen)	Cancellation task requiring to search for and cross out complete heart outlines while ignoring incomplete hearts.	Egocentric score = subtracting the number of targets identified on the left from the right side of the page. Cut-off >3. Allocentric score = subtracting the number of right-gap and left-gap hearts identified. Cut-off >1.	Divide score by 20 to convert to score between – 1 and +1 (cut-offs =  – .10 and .15)	Egocentric and allocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect.	Time: < 5 min To see the tool: http://www.ocs-test.org
Ota's cancellation task (Ota et al., 2001)	Cancellation task with 30 on an A4 page. Ten complete circles; 10 with gaps on the right; and 10 with gaps on the left. Fifteen circles on the right side of the page, and 15 on the left. Requires to circle all complete circles and cross out any circles with gaps.	Two types of errors recorded: (i) stimuli on the left of the page that was not cancelled (egocentric neglect) or (ii) stimuli with a left gap that was circled, i.e., marked incorrectly as a full circle (allocentric neglect) (Stein et al., 2022)	Tactile version – raised circles in identical layout. Required to feel board and determine if circles are complete or have gaps in them, with eyes closed.	Egocentric and allocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect.	Time: < 5 min
VR-DiSTRO (Fordell et al., 2011)	Subtests on computer with headphones, robotic pen, keyboard, monitor and shutter glasses for stereoscopic vision. (1) VR star cancellation test (VR-SCT) (2) VR-line bisection (VR-LB) (3) VR-Visual extinction (VR-EXT) (4) VR Baking Tray task (VR-BTT).	Neglect if at least one of the four subtests has asymmetrical score for right and left sides.		Egocentric neglect in peripersonal space. Cannot differentiate visual and motor neglect	Time: Approx. 15 min To see the tool: Fordell et al. (2011) Requires: computer program, shutter glasses for stereoscopic vision and robotic pen
Indented paragraph reading test	Requires reading a paragraph on A4 sheet of paper. Thirty lines of text taken from the Reader’s Digest variably indented O–25 spaces from left margin. The right margin is indented O–6 spaces determined by length of the final word on the line.	Score total number of omitted words on the page. Criterion for neglect is omission of one or more words.		Visual neglect in peripersonal space.	Time: not stated To see the tool: Caplan (1987)
Tool name (reference)	Components/subtests – a description of the test	Scoring	Modified versions/scoring	Neglect subtypes	Feasibility
Fluff test (Cocchini et al., 2001).	Test of personal neglect. Examiner attaches targets to a blindfolded patients left arm, leg and trunk. Subjects distracted by being engaged in conversation during application. Patients were to remove the targets with their right hand. No time restriction	Variation on scoring. Can either: (1) Record amount of missed targets, or (2) Percentage of targets detached from each side.	Modified versions with no leg targets require less trunk control. • 6 targets on left side (Hamilton et al., 2008) • 8 targets (Klinke et al., 2018)	Egocentric neglect in personal space. Cannot differentiate motor and tactile neglect.	Time: Not stated To see the tool: Cocchini and Beschin (2022)
Video-oculography with Free visual exploration (Kaufmann et al., 2020; Paladini et al., 2019).	A free visual exploration task of 48 images on computer, comprising 24 photographs of natural/urban scenes, and their 24 mirrored versions. Required to freely explore each image for 7 s.	Eye movements recorded with infrared, video-eye-tracking system (Eyelink 1000 Plus System). Fixations shorter than 100 ms excluded from analysis. Neglect defined by the spatial distribution of fixations.	Two different test sets (A and B) (Kaufmann et al., 2020).	Egocentric visual neglect in peripersonal space.	Time: < 10 min To see the tool: Kaufmann et al. (2020)
MonAmour Test (Montedoro et al., 2019)	Requires finding grey targets randomly amongst black distractors. Targets are complete (full conductor) requiring a push response or incomplete (missing part of left/right arm) requiring a pull response, for egocentric and allocentric scoring. Includes 100 trials, each with a single target and 119 distractors.	Egocentric score: percentage of omissions, asymmetry score, mean reaction time asymmetry. Allocentric score: false positive response, asymmetry of false positive responses and total mean reaction time.	No alternative versions	Egocentric and allocentric neglect in peripersonal space.	Time: not stated To see the tool: Montedoro et al. (2019) Requires: Specialized equipment – REAplan® robot
NIHSS – Neglect/extinction item	In addition to general observation, the patient is tested with visual double simultaneous stimulation (performed during the visual field item of the NIHSS) and double simultaneous cutaneous stimulation of both upper limbs and lower limbs	Scoring: 0 = no abnormality, 1 for visual, tactile, auditory, spatial, or personal inattention or extinction to bilateral simultaneous stimulation in one sensory modality, and 2 for severe neglect or extinction to more than one modality.	Modified version – video conferencing on iOs using facetime for telestroke consultations – (Demaerschalk et al., 2012)	Tactile, auditory and visual extinction. Cannot differentiate motor and visual neglect with pointing response.	Time: < 10 min To see the tool: http://www.nihstrokescale.org/ Training and certification required to administer.
DSS confrontation test for extinction (Visual, tactile, auditory)	Visual: Examiner wiggles index fingers in the left and right field of view. Required to indicate when movement detected in either or both sides (5 x left, 5 x right, 5 x both sides). Auditory: Examiner rustles small piece of paper behind the left, right or both ears. Required to indicate location of stimuli (5 x left, 5 x right, 5 x both sides). Tactile: With eyes closed and palms face down. Examiner touches cotton wool to dorsal surface of each hand. Required to indicate location of stimuli (5 x left, 5 x right, 5 x both).	Extinction classification requires: • At least 80% (≥4) correct answers for single contralesional stimuli • Less than 80% (≤3) correct answers for bilateral stimuli (Kamtchum-Tatuene et al., 2017)	Alternate version: Three trials each of left, right and bilateral presentations.	Tactile, auditory and visual extinction.	Time: < 5 mins
Halifax Visual Scanning Test (Whitehouse et al., 2019)	Requires reading as many words (interspersed amongst distractors) in 2 mins at 3 distances: (1) Shirt subtest – stimuli printed on a shirt in personal space. (2) Table subtest – stimuli printed on sheet of paper in peripersonal space. (3) Wall subtest – on a wall at 250 cm in extrapersonal space	Total score = number of words read aloud (maximum  = 24 words in each space).	No alternate versions	Visual neglect in personal, peripersonal and extrapersonal space.	Time: Approx. 10 min To see the tool: Whitehouse et al. (2019)
Landmark test (verbal and motor response) (Bisiach et al., 1998)	LANDMARK-V: Pre-bisected 180 mm lines with black and red segments requiring a verbal-response to name the colour of the shorter or longer segment. LANDMARK-M – requires motor response to point to the shorter or longer segment of pre-bisected black lines, presented horizontally in centre of A4 page.	Percentage of correct responses for both LANDMARK – V and LANDMARK-M tasks.		Visual and motor neglect in peripersonal space.	Time: Not stated. To see the tool: Bisiach et al. (1998)

365 psychometric properties were extracted from the 165 included studies. The risk of bias of all 365 psychometric properties was also evaluated according to the COSMIN guidelines (Mokkink, Prinsen, et al., 2018). The majority of the studies (n = 95) reported multiple psychometric properties (e.g., reliability, measurement error and known groups validity), and sometimes for multiple assessment tools. The most frequently reported type of psychometric property was hypothesis testing for construct validity, with 222 reported within the 164 studies.

A summary of the findings including the overall ratings for each measurement property for each tool and the overall quality of the evidence as per the modified GRADE approach is outlined in Table 3. No tools showed full evidence for all eight measurement properties. Twelve tools showed some evidence of sufficient reliability, according to COSMIN criteria. However, none of the 33 tools had high quality evidence of sufficient reliability, only one tool had moderate evidence (Star Cancellation), whereas the others had either low or very low evidence.

Table 3. Summary of findings: describes the overall good measurement properties ratings and modified GRADE criteria for each psychometric property of the spatial neglect tools, derived from the pooled results from each study as per Online Resource 4.

		Structural validity	Internal Consistency	Reliability	Measurement Error	Criterion Validity	Construct/ Known Groups Validity (hypotheses confirmed)
Behavioural Inattention Test	GRADE		Low				Low
	Rating (Summary)		(?) Evidence for sufficient structural validity not met				(+) 3 out of 3
BIT – Conventional subtests	GRADE	Moderate	Moderate	Very Low			High
	Rating (Summary)	(?) Unidimensional (single factor)	(?) Evidence for sufficient structural validity not met	(+) Inter-rater ICC = 0.99			(+) 30 out of 33 ^a
BIT – Behavioural subtests	GRADE			Very Low	Very Low		High
	Rating (Summary)			(?) Test-retest (15d): r = 0.97	(?) Inter-rater = 100%		(+) 18 out of 20 ^a
mBIT	GRADE				Very Low		Very Low
	Rating (Summary)				(?) Interrater = 100% (except coins, reading)		(−) 0 out of 1
Sunnybrook Neglect	GRADE	Low	Low	Very Low			Very Low
Assessment Procedure (SNAP)	Rating (Summary)	(?) R CVA: One factor L CVA: two factors	(?) Evidence for sufficient structural validity not met	(?)			(+) 1 out of 1
BIT Line bisection	GRADE			Moderate	Very Low	Low	High
	Rating (Summary)			(+) Inter-rater ICC = 0.99 Test-retest (1 h) ICC = 0.97	(?) 95% LoA = −2.7 to +3.0 (bias 0.13)	(?)	(+) 16 out of 19 ^a
Schenkenberg line bisection	GRADE			Low	Very Low		High
	Rating (Summary)			(+) Intra-rater: ICC = 0.72–1 Inter-rater: ICC = 0.82–1	(?) Quality too poor to summarize		(+) 11 out of 11^a
BIT Line cancellation	GRADE			Very Low		Low	High
	Rating (Summary)			Test-retest: (+) Acute(2 h) K = 0.82 (−) Subacute (75d) K = 0.58 Inter-rater ICC = 0.999		(?)	(+/−) 41 out of 59
		Structural validity	Internal Consistency	Reliability	Measurement Error	Criterion Validity	Construct/ Known Groups Validity (hypotheses confirmed)
Letter cancellation task (Diller & Weinberg, 1977)	GRADE						Moderate
	Rating (Summary)						(+/−) 4 out of 8
Bells Test	GRADE			Very Low	Very Low		High
	Rating (Summary)			(+) Test-retest (1d): ICC = 0.83–0.84	(?) SEM (Total) = 2.43 SEM (LI) = 0.17		(+) 20 out of 22^a
BIT Letter Cancellation	GRADE			Very Low		Very Low	Moderate
	Rating (Summary)			(+) Inter-rater ICC = 0.99		(?)	(+) 17 out of 21
BIT Star Cancellation	GRADE			Moderate	Very Low	Low	High
	Rating (Summary)			(+) Inter-rater: ICC = 0.98 Test-retest(1 h): ICC = 0.89	(?) 95% LoA = −15 to +11 (bias – 2.2) ^b	(?)	(+) 77 out of 82^a
Ogden Scene	GRADE			Low			Moderate
	Rating (Summary)			(?) Inter-rater r = 0.94			(+) 16 out of 19^a,c
Clock drawing	GRADE	Very Low	Very Low				Moderate
	Rating (Summary)	(+) Unidimensional 80.5–89.9%	(?) Person = 0.62–0.68 Item reliability 0.94–0.95				(+/−) 3 out of 4 (Can’t discriminate neglect/non-neglect groups.)
Catherine	GRADE	Moderate	Moderate	Low	Very Low		High
Bergego Scale	Rating (Summary)	(+/−) one / two factors	(?) Evidence for sufficient structural validity not met	(+) Inter-rater ICC = 0.85 wK range = 0.27–0.86	(+) Inter-rater: Mean diff = 0, LoA = ±2. (MIC = 4)		(+) 45 out of 49 ^a c
CBS Via KF-NAP	GRADE	Moderate	Moderate	Very Low			High
	Rating (Summary)	(?)	(?) Evidence for sufficient structural validity not met	(+) Inter-rater ICC = 0.92 wK = 0.35–1			(+) 15 out of 17 ^a
		Structural validity	Internal Consistency	Reliability	Measurement Error	Criterion Validity	Construct Validity / Known Groups (hypotheses confirmed)
Mobility Assessment Course	GRADE			Low
	Rating (Summary)			(+) Inter-rater: ICC =0.76–0.94
Wheelchair Obstacle Course	GRADE			Moderate	Very Low
	Rating (Summary)			(?) Inter-rater: r = 0.90–0.96	(?) Agreement range 80–100%
Semi-Structured Scale for the	GRADE		Very Low	Very Low			Very Low
functional evaluation of personal neglect	Rating (Summary)		(?) Evidence for sufficient structural validity not met	(?) Inter-rater: Kendall’s Tau = 0.88			(−) 0 out of 3
Comb and Razor	GRADE		Very Low	Very Low			Moderate
	Rating (Summary)		(?) Evidence for sufficient structural validity not met	(?)			(+) 5 out of 5
2 Daisy Figure	GRADE			Very Low			Very Low
	Rating (Summary)			(?) Inter-rater r = 0.98			(?)
Baking Tray Task	GRADE			Very Low	Very Low		High
	Rating (Summary)			(+) Test-retest(1 h): ICC = 0.87	(?) LoA = −5 to +5 (−0.21 bias)^b		(+) 13 out of 15 ^a
Apples Test	GRADE		Low		Low		Moderate
	Rating (Summary)		(?) Evidence for sufficient structural validity not met		(?) Ego 88%, Allo 94% ^b		(+) 13 out of 17 ^a
Hearts	GRADE		Low	Low			Moderate
	Rating (Summary)		(?) Evidence for sufficient structural validity not met	(+) Test-retest(3d): ICC = 0.733			(+) 5 out of 8 ^a
Ota’s Task	GRADE				Very Low		Moderate
	Rating (Summary)				(?) Test-retest 64–75%		(+/−) 6 out of 11
		Structural validity	Internal Consistency	Reliability	Measurement Error	Criterion Validity	Construct/ Known Groups Validity (No. of hypotheses confirmed)
VR-DiSTRO	GRADE			Very Low			Very Low
	Rating (Summary)			(?)			(+) 7 out of 9
Indented Paragraph Reading Test	GRADE			Very Low			Low
	Rating (Summary)			(?) Inter-rater: r = 0.99			(+) 7 out of 12
Fluff Test	GRADE			Very Low			Low
	Rating (Summary)			(?) Test-retest (2w) r = .79			(+) 4 out of 6 ^a
Video Occulography	GRADE			Very Low	Very Low		Moderate
with FVE	Rating (Summary)			(+) Test re-test (24 h): ICC = 0.96	(?) Test re-test: LoA =  + 3 to – 3 SEM =0.49.		(+) Mean gaze position 18 out of 21 ^a
Mon Amour	GRADE			Very Low	Very Low		Very Low
	Rating (Summary)			Test – retest (2–5d) (+) Ego: ICC = 0.79–0.95 (−) Allo: ICC = −0.13	(?) MDC: = 5% total omissions, = 13% omissions asymmetry		(+) 6 out of 6
NIHSS neglect item	GRADE			Moderate			Moderate
	Rating (Summary)			(−) Inter-rater wK = 0.61			(+/−) 1 out of 4
DSS	GRADE			Very Low			Very Low
	Rating (Summary)			(?) Test-retest r = 0.67–0.77			(+) Tactile: 1 out of 1 (−) Auditory: 0 out of 1
Halifax Visual Scanning Test	GRADE						Very Low
	Rating (Summary)						(+/−) 7 out of 12
Landmark Task	GRADE						Very Low
	Rating (Summary)						(+/−) 3 out of 5

Note: MIC = Minimally Important Change; LI = Lateralised Index; LoA = 95% Limits of Agreement; wK = Weighted Kappa. Summary of Findings table describes the overall psychometric properties of each spatial neglect tool categorized as per the COSMIN (COnsensus-based Standards for the selection of health Measurement Instruments) standards for good measurement properties, which evaluates them as sufficient (+) = evidence meets the minimum criteria for good measurement properties; indeterminate (?) = not all information reported/ statistical method used not appropriate; insufficient (−) = evidence does not meet the criteria for good measurement properties; or inconsistent (+/−). The modified GRADE (Grading of Recommendations, Assessment, Development and Evaluation) criteria has been reported as per the COSMIN guidelines as either High, Moderate, Low or Very Low. The pooled results in this table have been drawn from the data charted in Online Resource 4. The gaps in the table are due to no data being available.

^a Unconfirmed hypotheses come from doubtful and/or inadequate quality studies.

^b SEM (Standard Error of Measurement) was not determined.

^c One study showed no correlation with bisection of 5 cm lines. There is some evidence that bisection of short lines is not a valid Ax of neglect.

^d Inconsistency unexplained – Quality graded down for inconsistent results

The assessment tools with at least sufficient evidence of both validity and reliability are presented in Table 3 according to the neglect subtype(s) for each tool as per the scoping review categorisations (Williams et al., 2021). There were 23 included tools with some evidence of reliability. However, 20 of these tools (87%) had only “low” or “very low” quality evidence. For construct validity, 22 tools had some evidence of sufficient construct validity, however only nine of these tools had high quality evidence: (1) Behavioural Inattention Test – behavioural and conventional subtests; (2) BIT Star Cancellation Test; (3) BIT Line Bisection; (4) Schenkenberg Line Bisection (5) Bells Test (6) Baking Tray Task (7) Catherine Bergego Scale (CBS); (8) CBS via Kessler Foundation Neglect Assessment Process (KF-NAP); (9) Mobility Assessment Course.

Most of the included studies only used the conventional subscale (a composite of 6 subtests) of the Behavioural Inattention Test (BIT) (Wilson et al., 1987). The conventional subscale alone demonstrated sufficient validity and reliability (see Table 3), however there was only “very low” evidence for reliability. The whole BIT battery, and the behavioural subscale (of 9 subtests) in isolation have not demonstrated sufficient reliability from the included studies.

Regarding the utility of the assessment tools (whether the tool specifies the neglect severity and subtypes of neglect identified): Nine tools did not state the subtype of neglect – simply stating the tool identified spatial neglect. 24 of the 33 tools stated the subtype of neglect, however, these 24 tools did not indicate all relevant subtypes, according to our research team. For example, many cancellation tools state they identify visual neglect (not mentioning peripersonal neglect or that visual neglect cannot be disentangled from motor neglect with these tools). Therefore, a column was added in Table 2 on the neglect subtypes for each tool, taken from a previously published scoping review that systematically identified the neglect subtypes of all spatial neglect tools (Williams et al., 2021).

Fourteen studies reported on some aspect of feasibility (completion time, resources and training required), with six studies reporting the average completion time and six studies reporting the percentage of the sample that was able to complete the tool. Therefore, some of the timings in Table 2 are approximations, which were sourced from a combination of the clinical experience of the authors and the stroke engine website (strokengine.ca). One study reported carers’ comments of the acceptability of the tool by the stroke survivors, and one study asked the stroke survivors themselves about their satisfaction and ability to understand the instructions.

No evidence was found on the measurement invariance or the cross-cultural validity of the included tools. Additionally, no tool showed evidence of sufficient internal consistency, with 90% (9/10) of tools not demonstrating sufficient structural validity (a prerequisite for the evaluation of internal consistency) (See Table 3). A summary of the characteristics of each psychometric property extracted from each study, the summarized results and the risk of bias are presented in the Online Resource 4.

Discussion

Accurate and reliable identification of spatial neglect can be the essential gateway to effective intervention. Therefore, this systematic review critically appraised the evidence for the psychometric properties of the 33 tools most often used for identifying spatial neglect in clinical practice and research. We reviewed 164 studies aimed to identify the assessment tools with the best psychometric properties, categorized per subtype of neglect. However, not enough evidence was available to strongly recommend one assessment tool over another, as the general state of the evidence was poor, or absent for some subtypes such as motor, tactile, auditory or representational neglect. There is no one tool that has sufficient high-quality evidence of both validity and reliability. There are many gaps in the summary tables where no evidence was available, for example 21 tools have no evidence of measurement error reported. Even when the evidence exists, the majority (66%, n = 63) of the psychometric properties were rated as either low or very low quality (Refer to Table 3). This paper describes the best available evidence, then identifies the gaps in the evidence and makes recommendations for improvements through future research.

Tools with the best available evidence

Table 4 specifies the eleven assessment tools with the best available evidence, therefore it is recommended that clinicians and researchers refer to this table to guide their decisions on which tools to use. However, the quality of the evidence should be taken into account, as labels of “low” or “very low” quality evidence suggests that the true psychometric property may be substantially different to what is reported (Mokkink, Prinsen, et al., 2018). Other considerations such as the completion time, being able to access the tool and cost of each tool have been integrated into Table 4 as these may further guide clinicians and researchers on which tools to use.

Table 4. Tools with the best available evidence of both validity and reliability according to the COSMIN criteria for good measurement properties, categorized by neglect subtype.

Tool name	Main Neglect Subtype(s)	Secondary Subtype(s)	Time to complete	Cost and access to tool	Quality of reliability evidence	Quality of validity evidence
BIT-conventional subtest	Peripersonal Egocentric	Visual/motor/Representational/Allocentric	Approx 20–30 min	Pearsonassessments.com $	Very low	High quality
BIT Star Cancellation	Peripersonal Egocentric	Visual/motor	<5 min	Pearsonassessments.com ($ via BIT test kit)	Moderate	High quality
Bells Test			<5 min	strokengine.ca (Gauthier et al., 1989)	Very low	High quality
BIT Line Bisection Task			<5 min	Pearsonassessments.com ($, via BIT test kit)	Moderate	High quality
Baking Tray Task			<5 min	Tham and Tegner (1996)	Very low	High quality
Schenkenberg Line Bisection Task			<5 min	Schenkenberg et al. (1980)	Low quality	High quality
Broken Hearts (Oxford Cognitive Screen)	Egocentric Allocentric Peripersonal	Visual/motor	<5 min	http://www.ocs-test.org (via OCS test kit)	Low quality	Moderate
Video-oculography with Free visual exploration	Visual Peripersonal Egocentric		<10 min	Kaufmann et al. (2020)	Very low	Moderate
Catherine Bergego Scale (CBS)	Egocentric Personal Extrapersonal	Visual/motor/tactile/ Representational/ Auditory/ Peripersonal/	30–40 min	Azouvi (1996)	Low quality	High quality
CBS Via KF-NAP			30–40 min	https://www.kflearn.or g/courses/KF-NAP	Very low	High quality
Mobility assessment course	Egocentric	Visual/motor Peripersonal/ Extrapersonal	2–9 min	Ten Brink et al. (2018)	Low quality	High quality
No tools*	Motor
No tools*	Tactile
No tools*	Auditory
No tools*	Representational

Note: Main subtype(s) = Tool differentiates behaviours for the neglect subtypes listed; Secondary subtype(s) = Deficient performance may be attributable to the subtype, but the scoring does not allow for it to be differentiated from other subtypes; No tools* = No tools have demonstrated sufficient reliability and validity (according to the COSMIN guidelines); $ = tools are available at a cost.

Gaps in the evidence

Table 4 highlights the lack of tools with sufficient validity and reliability for motor, tactile, auditory and representational neglect. One of the main limitations with most tools is the inability to differentiate motor from visual neglect, as is the case with all pen and paper cancellation tasks (Husain et al., 2000; Làdavas, 1994; Plummer et al., 2003). These tasks are unable to differentiate visual and motor neglect as they require viewing visual information across the page along with a motor response with the hand to cross out or draw items on the contralesional side (Husain et al., 2000). Previous studies have been unable to consistently classify visual and motor neglect across timepoints (Hamilton et al., 2008; Harvey & Olk, 2004). It is important to distinguish visual from motor neglect as they may respond differently to interventions and may have different recovery profiles (Barrett & Burkholder, 2006). The use of accelerometers worn on the wrist has shown promise for objectively measuring motor neglect by recording differences in spontaneous use of both limbs despite normal strength (Toba et al., 2021), however the reliability, validity and utility of this approach in clinical practice has not yet been explored.

It is surprising that none of the included tools had high-quality evidence of reliability. Reliability is essential to determine if a change in neglect scores reflects real change of neglect signs or other factors, such as measurement error. The true reliability of the included tools may be substantially different to what is reported due to potential sources of bias such as missing information or not stating if the assessors were blinded (50%, n = 25), not including information on the sampling methods or inclusion and exclusion criteria (40%, n = 20), or not stating the time between measurements (4%, n = 2). A possible explanation for the missing information is that studies included data on psychometrics of a tool without it being the focus of the study.

The whole BIT battery, and the behavioural subscale (of 9 subtests) in isolation has not demonstrated sufficient reliability from the included studies. There was very little evidence for the whole BIT battery as the majority of studies used only the conventional scale and very few studies also used the behavioural subscale of the BIT. This is possibly due to some elements of the behavioural scale being outdated such as the use of an old dial telephone.

Reliability of neglect tools in this population can be difficult to evaluate due to the varying presentations of neglect over time (Anderson et al., 2000; Woods et al., 2017). For example, Hamilton et al.’s (2008) study investigating the subtype-defined patterns of performance over time, found inconsistent patterns of performance, with 61% of participants demonstrating a change of neglect subtypes, that was not consistent with spatial neglect improvement. However, the reliability and validity of the tools used in Hamilton et al.’s (2008) study were not evaluated; therefore, it cannot be determined if this was due to a real change in neglect signs, or other factors such as measurement error.

Line cancellation (Albert, 1973; Wilson et al., 1987) was the only tool with inconsistent evidence of construct validity. There was conflicting evidence that this tool was correlated with other neglect measures, which could mean that it does not consistently measure the construct it purports to measure (Mokkink, Prinsen, et al., 2018). The inconsistency could not be explained by further inspection of the data, as there were no patterns with the type of tools that were correlated or not correlated. A potential reason may be that this is the only cancellation task that does not require selective attention due to the absence of distractor items. This is consistent with previous research that task-related factors such as the number of distractor stimuli and the type of response required can adversely impact spatial exploration (Sarri et al., 2009). This means that line cancellation is relatively insensitive to mild or moderate forms of spatial neglect in comparison to other cancellation tasks such as the star cancellation test (Halligan et al., 1990) or the Bells Test (Gauthier et al., 1989). Therefore, line cancellation may not be correlated with other neglect tools if the sample includes a significant proportion of mild or moderate cases of spatial neglect.

It is important for screening tools to have high sensitivity so that potential cases of spatial neglect are identified for further diagnostic assessment (Grech et al., 2017). It was not possible to directly compare the sensitivity of the included tools in this review as there is no gold standard or reference standard to compare against. Therefore many studies either used a clinical judgement for the criterion scores for example (Grech et al., 2017; Rengachary et al., 2009), or compared to healthy control participants for example (Nurmi et al., 2010), or compared the sensitivity to the other tools in the study for example (Azouvi et al., 2002; Moore et al., 2018). Sensitivity analyses are constrained by the quality of the criterion scores; therefore, only validated tools should be used. Consensus on a gold standard is urgently needed, which may not be a single test, but a composite of cancellation and drawing tests and more naturalistic observation of daily tasks, that tap into different presentation/subtypes of neglect. This is in line with surveys of current practice, which have identified cancellation and drawing tests as the most popular cognitive tests to identify spatial neglect in clinical practice, along with unstructured functional observation (Checketts et al., 2021; Evald et al., 2021; Menon-Nair et al., 2007).

Utility and feasibility

We have summarized the limited information on the utility and feasibility of all the tools in Table 2. The time required to complete each assessment tool was only mentioned in a few studies with no information on the feasibility or interpretability of completing or scoring the tools for clinicians. Although not a focus of this study, it was noted that there was no information in the included studies on whether the tools are acceptable and meaningful for stroke survivors themselves. Further research is needed to gain the perspectives of clinicians and stroke survivors on what factors are important for them.

Strengths and limitations of this review

This review guides clinicians and researchers on the psychometrics of commonly used spatial neglect tools, which has not been summarized since the last review 19 years ago (Menon & Korner-Bitensky, 2004). In 2004, only 62 standardized and non-standardized spatial neglect tools existed, compared to almost 300 tools which exist today. This study is the first to report on the quality of the psychometric evidence for spatial neglect tools, as standards for evaluating the evidence such as COSMIN (Mokkink et al., 2020) have only recently been developed. It was outside the scope of this review to search any grey literature (such as test manuals) or contact authors to obtain data not reported due to potential reporting or publication biases. Only peer reviewed studies were included in this review however, this was not considered a limitation due to the potential bias of reporting in test manuals from the lack of peer review.

Why do we not have stronger psychometrics?

It is concerning that we do not have stronger psychometrics, considering the high proportion of stroke survivors with spatial neglect, and the plethora of neglect research that has been completed across multiple discipline areas. This may be due to many reasons, such as: the complex and varying nature of the condition itself, leading to current assessment tools not meeting the needs of researchers and clinicians, new tools being developed each year, and use of unvalidated batteries. Additionally, the COSMIN standards have only recently been developed, and many studies include statistical approaches that are not recognized as “acceptable”, such as Pearson correlations for evidence of reliability. There was also huge variability in the sample sizes, with almost a third of the included studies with very small samples (50 studies had under 30 participants), and another 23 studies were excluded due to having fewer than 5 participants. The small sample sizes are possibly the result of research being completed by individual centres. Moving forward it will be important to foster collaboration, enabling larger-scale studies and the development of a consensus on the most appropriate tools for screening and assessing spatial neglect.

Implications for research

High quality research studies on the validity and reliability of spatial neglect assessment tools guided by the COSMIN methodology is needed. Direct comparisons of the validity and reliability of assessment tools is required so that the best tools for each neglect subtype can be identified. The tools with the best evidence to date are summarized in Table 4. Notably, there are no tools with sufficient evidence for identifying tactile, auditory, motor or representational neglect. Therefore, it is essential for further research to focus on confirming the best tools for these subtypes of neglect (see Table 4).

This study has categorized the assessment tools according to the subtypes of neglect that they assess. However, assessing all neglect subtypes in clinical practice may not be feasible or necessary. Further research should aim to get consensus on what subtypes of neglect are important to assess consistently, and the best screening and diagnostic methods for doing this across clinical practice settings. Once consensus is achieved the recommended battery of tools should be validated. Many of the included papers used a combination of individual tools, collated specifically for their individual studies, rather than a validated battery. Additionally, the tools being used to screen for spatial neglect, are often also used for the comprehensive assessment of neglect (albeit, often combined with several other tools), then those same tools are also being used as outcome measures (Checketts et al., 2021; Grech et al., 2017; Williams et al., 2021). Ideally the tools used for screening should be distinct from the tools used for diagnosis and then outcome measures. However, we are a long way off from achieving consensus on all of these aspects.

It would be essential for reliability studies of neglect tools to ensure consistency in testing conditions for repeated measures, specifically for (1) time of day, (2) testing environment, including the location of the assessor (3) administration procedures, prompts and instructions, ensuring the tool is assessed at the same point in the session each time, and (4) stroke survivors perceived level of fatigue prior to each test. Additionally, the time between measurements needs to be carefully planned, with consideration for time post injury. Determining how stable neglect is for each of the participants prior to a test-retest reliability study would help in determining the optimal time between measurements. Additionally, it is important for study samples to be representative of the whole stroke population with assessment tools that account for other frequently co-existing stroke sequelae such as aphasia.

Computer-based tools such as eye-tracking, reaction time tasks, and virtual reality tools that are highly sensitive to mild neglect signs have the potential to address some of the limitations with tools used in clinical practice. However, many of these tools are not available or accessible for most clinical practice settings. Further research should focus on the feasibility of integrating low-cost options of virtual reality and computer-based tools into clinical practice.

Implications for clinicians

This review has not found a single tool or battery with high-quality evidence of sufficient psychometric properties. For now, we recommend that clinicians use a battery of tools for identifying neglect rather than relying on one tool. This is consistent with previous research recommending the use of multiple tools for the diagnosis of neglect (Azouvi et al., 2006). The shortlist of 11 tools categorized by subtype outlined in Table 4 have the best psychometrics and should be used as a guide for clinicians. Clinicians can refer to Table 2 for a description of each tool, scoring, completion time, and neglect subtypes identified to further inform their choice. It is important for clinicians to screen for a variety of neglect subtypes with standardized tools using a combination of cognitive tests such as cancellation and drawing tasks, along with more naturalistic observation of daily tasks with tools such as the Catherine Bergego Scale (CBS) (Azouvi, 1996; Chen et al., 2012) and the mobility assessment course (Ten Brink et al., 2018; Verlander et al., 2000). Line cancellation is not recommended for routine use with all stroke survivors as it has not shown to be a valid assessment of neglect.

Conclusion

This systematic review critically appraised the evidence for the psychometric properties of 33 commonly used assessment tools for identifying spatial neglect in clinical practice. The general quality of the evidence was poor and no one tool had high-quality evidence of both validity and reliability. Eleven tools show some promise as they meet the minimum criteria for validity and reliability. However, due to the low-quality of the reliability evidence, the true psychometric property may be considerably different to what is reported. Further high-quality studies using the COSMIN standards as a guide are needed to compare the psychometric properties of multiple neglect assessment tools so that variations can be attributed to the test, rather than other factors related to the study sample. It is essential to use tools with sufficient reliability when determining if interventions are effective, as a change of neglect scores doesn’t necessarily reflect a real improvement of neglect signs if the reliability or measurement error of the tool is unknown.

Competing interests

KH is an author of KF-NAP (one of the tools that was included in this review). KH did not extract, or assess the measurement property/Risk of bias for any studies relating to the KF-NAP or CBS. The other authors have no competing interests to declare that are relevant to the content of this article.

Authors’ contributions

Conceptualization: Lindy Williams, Tobias Loetscher, Jocelyn Kernot, Susan Hillier, Audrey Bowen; Literature search and screening: Lindy Williams, Tobias Loetscher, Jocelyn Kernot, Susan Hillier, Jennifer Jones; Data extraction and analysis: Lindy Williams, Tobias Loetscher, Jocelyn Kernot, Susan Hillier, Jennifer Jones, Kimberly Hreha; Writing – original draft preparation: Lindy Williams; Writing – review and editing: Lindy Williams, Tobias Loetscher, Jocelyn Kernot, Susan Hillier, Audrey Bowen, Kimberly Hreha, Jennifer Jones.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data Availability statement

The data that support the findings of this study are available from the corresponding author upon request.

Additional information

Funding

This work was supported by an Australian Government Research Training Program Scholarship.