Recognition subtests for the Repeatable Battery for the Assessment of Neuropsychological Status: Preliminary data in cognitively intact older adults, amnestic Mild Cognitive Impairment, and Alzheimer’s disease

Abstract

The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) has three delayed recall subtests (list, story, figure), but only one delayed recognition subtest (list). Since comparisons between delayed recall and recognition can be useful in clinical neuropsychology, the current study sought to develop and preliminarily examine two proposed new subtests for Form A of the RBANS, Story Recognition and Figure Recognition. A sample of older adults who were cognitively intact (n=48) or classified with amnestic Mild Cognitive Impairment (MCI, n=29) or mild Alzheimer’s disease (AD, n=24) were administered the RBANS and the two new recognition subtests. In the primary analyses, cognitively intact participants performed significantly better than the two memory-impaired groups on all twelve scores (one recall and three recognition [total, hits, false positive errors] for the list, story, and figure). For amnestic MCI and AD participants, they showed statistically comparable scores on 7 of the 12 variables, where those with MCI performed better than those with AD on the other five scores. Across the three groups, effect sizes were large (e.g., Cohen’s d=1.0 – 2.9). In secondary analyses, all of the List Recall and Recognition scores significantly correlated with one another, and this pattern was observed for all of the Story Recall and Recognition scores and most of the Figure Recall and Recognition scores. Although preliminary, these new recognition scores appear to provide useful information and may improve the sensitivity of the RBANS in identifying cortical/subcortical profiles in clinical and research settings.

Introduction

Within clinical neuropsychology, one of the more widely considered memory profiles is that of “cortical” vs. “subcortical” dementia (Bonelli & Cummings, 2008), in which the “cortical” profile shows deficits in both recall and recognition and the “subcortical” profile shows deficits in recall but relatively preserved recognition. Although this distinction of cortical/subcortical profiles is not definitive within neuropsychology (Arango-Lasprilla et al., 2006; Salmon & Filoteo, 2007), many neuropsychological disorders present with varying levels of difficulty on recognition memory testing, including Alzheimer’s disease (AD) (Ally, 2012; Salmon, 2012), Huntington’s disease (Montoya et al., 2006), Parkinson’s disease (Garcia-Ptacek & Kramberger, 2016; Whittington, Podd, & Kan, 2000), Dementia with Lewy Bodies (Kemp et al., 2017; Salmon et al., 2015), vascular disease and dementia (Libon, Price, Davis Garrett, & Giovannetti, 2004), cancer-induced cognitive impairment (Lindner et al., 2014), cerebral anoxia (Caine & Watson, 2000), epilepsy (Guedj et al., 2010), HIV-related cognitive impairments (Schiller, Foley, Burns, Sellers, & Golden, 2009), and schizophrenia (Bowie et al., 2004; Putnam & Harvey, 1999). Although not universal, patients with cortical memory profiles often poorly discriminate between hits and distractors, leading to an increased number of false positive errors, whereas those with subcortical memory profiles are more adept at discriminating hits from distractors, leading to fewer false positive errors (Salmon & Filoteo, 2007). As such, recognition memory seems to be an important neuropsychological factor when making differential diagnoses.

The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) is a popular brief battery of cognitive tests that evaluates a number of cognitive domains, including immediate and delayed memory (Randolph, 1998). On the Delayed Memory Index of the RBANS, there are three free recall trials: one for a list of ten previously-learned words, one for a previously-learned brief story, and one for a previously-copied complex figure. On this Index, there is also an auditory list recognition trial, in which the examinee needs to identify the ten words from the original list embedded among ten distractor words that were not on the list. Across a range of clinical diseases and disorders, patient sample present with varying levels of deficits on the RBANS List Recognition subtest, including AD (Duff et al., 2008; McDermott & DeFilippis, 2010; Randolph, Tierney, Mohr, & Chase, 1998), Huntington’s disease (Duff, Beglinger, Theriault, Allison, & Paulsen, 2010; Randolph et al., 1998), vascular dementia (McDermott & DeFilippis, 2010), Parkinson’s disease (Badenes et al., 2018), and multiple sclerosis (Beatty, 2004). However, the RBANS does not contain recognition trials for the story or the figure, and, to our knowledge, there are no published papers on such recognition trials for the RBANS.

Therefore, we sought to develop and preliminarily examine the performance of two new subtests for the RBANS, Story Recognition and Figure Recognition, in a sample of older adults who were cognitively intact or classified as having amnestic Mild Cognitive Impairment (MCI) or mild AD. In this cross-sectional sample, it was hypothesized that individuals with MCI would perform more poorly than their cognitively intact peers but better than individuals with AD on these clinically-relevant recognition scores. If supported in this sample and validated in others, then such an addition to the RBANS may improve its sensitivity in identifying cortical/subcortical profiles in clinical and research settings.

Methods

Participants.

One hundred one older adults (≥65 years) participated in the current study. These individuals were recruited from senior centers, independent living facilities, and cognitive disorders clinics to prospectively study cognitive change and biomarkers in older adults. Based on criteria from the Alzheimer’s Disease Neuroimaging Initiative – 2 (ADNI-2) study (Petersen & Weiner, 2015), participants were classified as cognitively intact, amnestic MCI, or mild AD. ADNI criteria depend on performance on three measures (Mini Mental Status Examination [MMSE], Wechsler Memory Scale – Revised [WMS-R] Logical Memory II Paragraph A only, Clinical Dementia Rating Scale [CDR]). For the MMSE, scores of 24 – 30 would classify someone as intact or MCI, whereas scores of 20 – 26 would classify someone as mild AD. For WMS-R Logical Memory II Paragraph A, raw scores at or above 9, 5, or 3 would classify someone as intact if his/her educational attainment was ≥16 years, 8 – 15 years, or <8 years, respectively. Raw scores below those cutoffs would classify someone as MCI or mild AD. For the CDR, an overall rating of 0 would classify someone as intact, 0.5 would classify someone as MCI, and 0.5 or 1 would classify someone as mild AD. Even though there is some overlap within criteria, individuals need to meet all three criteria to be classified as intact, MCI, or mild AD. Additionally, since ADNI criteria were solely used to determine the groups, scores from the RBANS (neither original nor newly developed) were not used for this purpose.

Exclusion criteria for this study (which also required brain imaging) included:

• History of major stroke, head injury with loss of consciousness of >30 minutes, or other neurological/systemic illness that may affect cognition;

• Current or past major psychiatric illness (e.g., schizophrenia, bipolar affective disorder);

• Raw scores on the 15-item Geriatric Depression Scale ≥6’

• History of substance abuse;

• Current use of antipsychotics or anticonvulsant medications;

• Known allergic or hypersensitivity reactions to previously administered radiopharmaceuticals;

• Requiring monitored sedation or anesthesia for PET or MRI scanning;

• Claustrophobia to a degree that the individual cannot undergo PET or MRI imaging;

• History of metal injury which precludes the individual from undergoing MRI imaging; and

• Inadequate vision, hearing, and manual dexterity to participate in the cognitive assessments.

Demographic information for the three groups is presented in Table 1.

Table 1.

Demographic information and recall and recognition scores.

Variable	Intact (n=48)	MCI (n=29)	AD (n=24)	p	d
Age (years)	72.3 (4.8) [65 – 91]	73.5 (4.6) [65 – 84]	77.8 (5.8) [66 – 89]	0.002 1,2<3	0.7
Education (years)	16.7 (2.2) [12 – 20]	15.3 (2.7) [12 – 20]	16.3 (2.3) [12 – 20]	0.04 1,3>2	0.5
Sex (% female)	60%	59%	54%	0.88	0.1
Race (% white)	100%	94%	100%	0.32	0.2
List Recall (max = 10)	6.2 (2.9) 4/7/9	1.1 (1.7) 0/1/1	0.8 (2.0) 0/0/1	<0.001 1>2,3	2.0
List Recognition Total (max = 20)	19.7 (0.6) 19/20/20	17.0 (2.2) 15/17/19	14.7 (2.4) 13.25/15/16	<0.001 1>2>3	2.9
List Recognition Hits (max = 10)	9.7 (0.5) 9.25/10/10	8.8 (1.5) 8/9/10	7.3 (1.9) 6.25/8/8.75	<0.001 1>2>3	1.8
List Recognition FPE (max = 10)	0.1 (0.4) 0/0/0	1.7 (2.0) 0/1/3	2.6 (1.7) 1.25/3/4	<0.001 1>2,3	1.7
Story Recall (max = 12)	10.3 (1.7) 9.25/11/12	3.8 (3.4) 1/3/6	1.4 (1.1) 1/1/2	<0.001 1>2>3	2.5
Story Recognition Total (max = 20)	18.9 (1.4) 18/19/20	14.7 (2.7) 12.5/15/16	12.6 (1.9) 11/12/13.75	<0.001 1>2>3	2.3
Story Recognition Hits (max = 10)	9.4 (0.8) 9/10/10	8.2 (1.2) 7/8/9	7.3 (1.5) 6/8/8	<0.001 1>2,3	1.5
Story Recognition FPE (max = 10)	0.6 (0.9) 0/0/1	3.7 (2.2) 2/4/5	4.7 (1.5) 4/5/5.75	<0.001 1>2,3	2.2
Figure Recall (max = 20)	14.6 (3.5) 12.25/15/17	5.0 (5.3) 0/4/7.5	1.4 (1.8) 0/0.5/3	<0.001 1>2>3	2.4
Figure Recognition Total (max = 20)	17.3 (1.7) 16/17/19	14.3 (2.6) 12/14/16	13.2 (1.7) 12/13/14	<0.001 1>2,3	1.7
Figure Recognition Hits (max = 10)	8.9 (1.0) 8/9/10	7.3 (1.9) 6.5/8/9	6.9 (2.5) 6/7/9	<0.001 1>2,3	1.0
Figure Recognition FPE (max = 10)	1.8 (1.4) 1/1/2	3.1 (1.7) 2/3/4	3.8 (1.9) 2/4/5.75	<0.001 1>2,3	1.0

Note. AD = Alzheimer’s disease, d = effect size (Cohen’s d) between the groups, FPE = false positive errors, max = maximum possible score, MCI = Mild Cognitive Impairment, p = overall p-value from Kruskal-Wallis test. Post-hoc differences: 1 = Intact, 2 = Mild Cognitive Impairment, 3 = Alzheimer’s disease. All cognitive scores are raw scores, with higher scores indicating better performance, except for FPE (false positive errors), in which lower scores indicate better performance. In the second, third, and fourth columns, means and standard deviations (in parentheses) and ranges for age and education [in brackets] are listed in the top row and percentiles of scores within each group are in the bottom row (25^th/50^th [median]/75^th percentiles).

Procedures.

All individuals provided informed consent or assent prior to participation, and all procedures were approved by the local institutional review board. During a baseline visit, all participants completed Form A of the RBANS as part of a larger cognitive battery.

The RBANS (Randolph, 1998) is a brief battery of 12 neuropsychological subtests that assess the domains of immediate memory, visuospatial perception and construction, language, attention, delayed memory, and global cognition. The Delayed Memory Index contains a recognition trial for the list of ten words learned during the Immediate Memory Index, in which the examinee needs to identify the words from the list (hits) among distractor words (false positive errors). However, the RBANS does not contain recognition trials for the story or figure.

In developing a story recognition subtest, we attempted to keep it similar in format to the RBANS List Recognition subtest, in which the examinee identified story elements from RBANS Story Recall among distractor elements. However, some adjacent story elements seemed too difficult to separate (e.g., “two” and “hotels” are two individual elements in Story Recall, but they were combined as “two hotels” in the recognition subtest). By combining such adjacent story elements, it resulted in 10 story recognition elements (vs. the 12 elements in Story Recall). Ten distractor elements were also generated that could have been true for the story or that were close to being true in the story (e.g., Cincinnati instead of “Cleveland,” March instead of “May”). This resulted in an equal number of story recognition elements (hits) and distractors (false positive errors). The items were randomized using a random number generator.

In developing a figure recognition subtest, we also attempted to keep it similar in format to the RBANS List Recognition subtest. The 10 individual Figure Recall drawing/placement elements were reproduced on individual 3” × 5” cards. Ten distractor elements were also generated that were either similar to the drawing elements (e.g., an oval instead of the “circle,” arrow with a closed point instead of an “arrow” with an open point) or similar to the placement elements (e.g., three small circles in a triangle pointing up instead of “three small circles” pointing down, large horizontal cross instead of “diagonal cross”). This resulted in an equal number of figure recognition drawing/placement elements (hits) and distractors (false positive errors). Both drawing/placement elements and distractors were removed from the overall context of the figure (e.g., a “curving line,” but not connected to the larger figure). The items were randomized using a random number generator, and they were combined into a flip booklet. Readers can contact first author for copies of the Story and Figure Recognition items, as well as instructions for administering these new subtests.

Data Analyses.

Prior to the primary analyses, the three groups were compared on age, education, and gender via ANOVA and chi-square tests. An initial examination of the total score, hits, and false positive errors for the List Recognition, Story Recognition, and Figure Recognition indicated that some were not normally distributed (e.g., skewness well above or below 0, kurtosis tended to be well above 3), especially for the intact participants. Therefore, nonparametric statistics were utilized. The primary analyses examined group differences (intact vs. MCI vs. AD) on relevant recall and recognition scores with the Kruskal-Wallis test. In the event of an overall effect, post-hoc tests (Mann-Whitney U) were examined to compare all pairs of groups. Secondary analyses examined: 1) internal reliability of the total scores for List Recognition, Story Recognition, and Figure Recognition with Cronbach’s alpha and split-half reliability, 2) the relationship between the respective recall and recognition scores for the entire sample with Spearman’s rho correlations, and 3) the relationship between all recall and total recognition scores for the entire sample with Spearman’s rho correlations. Given the number of statistical comparisons across the three studies, an alpha level of 0.01 was used for the primary and secondary analyses. Effect sizes (Cohen’s or converted to Cohen’s d) were calculated for all analyses (Lenhard & Lenhard, 2016).

Results

As can be seen in Table 1, there were statistically significant group differences on age and education, with the AD participants being 4 – 5 years older than the other two groups, and the MCI group having about one year of education less than the other two groups. The three groups were comparable in the number of males and females.

In the primary analyses of between-group differences on the various recall and recognition scores, cognitively intact participants performed significantly better than the two memory-impaired groups on all scores. For example, as seen in Table 1, they had the highest scores on List Recall, List Recognition total, and List Recognition hits, but made the fewest List Recognition false positive errors. This same pattern was observed across all Story Recall and Recognition and Figure Recall and Recognition scores. For the amnestic MCI and AD participants, they showed statistically comparable scores on 7 of the 12 variables (see Table 1). On the other five variables, the participants with MCI performed better than those with mild AD. Across the three groups, effect sizes on all variables were large (e.g., Cohen’s d = 1.0 – 2.9).

In the secondary analyses, internal reliability of the total scores for List Recognition, Story Recognition, and Figure Recognition were examined with Cronbach’s alpha and split-half reliability. Data for all participants were used in these analyses. For List Recognition, which was examined for comparison, its Cronbach’s alpha was 0.78, and its split-half coefficient was 0.84. For Story Recognition, its Cronbach’s alpha was 0.78, and its split-half coefficient was 0.76. For Figure Recognition, its Cronbach’s alpha was 0.58, and its split-half coefficient was 0.61.

Additionally, when data was pooled across all participants, all of the List Recall and Recognition scores significantly correlated (all ps<0.001, see Table 2). This pattern was also observed for all of the Story Recall and Recognition scores (all ps<0.001, see Table 3). These same relationships were seen for nearly all Figure Recall and Recognition scores (most ps<0.001, see Table 4).

Table 2.

Correlations between scores on RBANS List.

	List Recall	List Recognition Total	List Recognition Hits
List Recognition Total	0.71	-	-
List Recognition Hits	0.57	0.78	-
List Recognition FP	−0.60	−0.82	−0.39

Note. FP = false positive errors, RBANS = Repeatable Battery for the Assessment of Neuropsychological Status. All Spearman’s rho correlations are statistically significant at p<0.001. n= 101.

Table 3.

Correlations between scores on RBANS Story.

	Story Recall	Story Recognition Total	Story Recognition Hits
Story Recognition Total	0.85	-	-
Story Recognition Hits	0.66	0.78	-
Story Recognition FP	−0.82	−0.92	−0.57

Note. FP = false positive errors, RBANS = Repeatable Battery for the Assessment of Neuropsychological Status. All Spearman’s rho correlations are statistically significant at p<0.001. n= 101.

Table 4.

Correlations between scores on RBANS Figure.

	Figure Recall	Figure Recognition Total	Figure Recognition Hits
Figure Recognition Total	0.83	-	-
Figure Recognition Hits	0.57	0.75	-
Figure Recognition FP	−0.59	−0.70	−0.19* p=0.06

Note. FP = false positive errors, RBANS = Repeatable Battery for the Assessment of Neuropsychological Status. All Spearman’s rho correlations are statistically significant at p<0.001, except where marked with an asterisk and a p-value is provided. n= 101.

Finally, when data was pooled across all participants, all recall and total recognition scores were significantly correlated (all ps<0.001, see Table 5). Within recall scores (i.e., List Recall, Story Recall, Figure Recall, correlations tended to be slightly higher than within recognition scores (i.e., List Recognition, Story Recognition, Figure Recognition): 0.78 and 0.72, respectively. However, these latter two correlations were not significantly different (z=0.96, p=0.34).

Table 5.

Correlations between scores on all recall and recognition scores.

	List Recall	List Recognition	Story Recall	Story Recognition	Figure Recall
List Recognition	0.71
Story Recall	0.76	0.80
Story Recognition	0.73	0.78	0.85
Figure Recall	0.72	0.79	0.86	0.79
Figure Recognition	0.60	0.70	0.77	0.69	0.83

Note. All Spearman’s rho correlations are statistically significant at p<0.001. n= 101.

Discussion

Given the utility of recognition memory scores in many conditions evaluated in clinical neuropsychology, especially in identifying cortical and subcortical profiles (Duff, Schoenberg, Mold, Scott, & Adams, 2007; Randolph et al., 1998), the current study sought to develop and preliminarily examine the performance of two new recognition subtests for Form A of the RBANS (Story Recognition and Figure Recognition) in a sample of older adults who were cognitively intact or classified as having amnestic MCI or mild AD. Consistent with hypotheses in this cross-sectional sample, cognitively intact older adults performed better than the two memory-impaired groups on all of the recognition scores (total, hits, false positive errors) for the list, story, and figure of the RBANS. This pattern was also observed on the delayed recall trials of these same three subtests. For the amnestic MCI and AD participants, they showed statistically comparable scores on 6 of the 9 recognition scores (and 1 of the 3 recall scores). On the other three recognition scores (List Recognition total, List Recognition hits, Story Recognition total), the participants with MCI performed better than those with mild AD. On the other two recall scores (Story Recall, Figure Recall), those with MCI also scored higher than their AD peers. When looking at differences between all three groups, effect sizes on all variables (recall and recognition) were large (e.g., Cohen’s d = 1.0 – 2.9).

When looking at the unique types of scores (recall, recognition total, hits, false positive errors) collapsed across the three pieces of information being remembered (list, story, figure), the largest effects were seen on the recall and recognition total scores (Cohen’s d = 2.4 and 2.3, respectively), with relatively smaller effects for hits and false positive errors (Cohen’s d = 1.5 and 1.0, respectively). It is possible that hits and false positive errors were relatively less powerful in separating these three groups because such scores typically are limited in range. For example, for List Recognition hits, the scores in intact subjects ranged from 8 – 10, with 75% of them having a perfect score of 10. Similarly, for List Recognition false positive errors in the intact group, scores ranged from 0 – 2, with over 90% having a perfect score of 0. Such limited range of scores likely contributed to its relatively poorer effect.

Despite the promising preliminary validity results on the two new recognition subtests, the reliability data was less encouraging. The new Story Recognition subtest had internal reliability in the 0.76 – 0.78 range, which is generally considered adequate (Strauss et al., 2006). However, for Figure Recognition, internal reliability was notably lower (0.58 – 0.61), and raises the question about if it should be used in clinical decision-making. For comparison in the same sample, List Recognition’s reliability tended to be in the adequate to high range (0.78 – 0.84). Interestingly, in the RBANS manual (Randolph, 2012), test-retest reliabilities are lower for Figure Recall than Story Recall in 70 – 79-year-olds (0.55 and 0.72, respectively). Given this information, it might not be surprising that recognition subtests based on these recall subtests are generally lower. Although test-retest reliability is needed, users of these new recognition subtests should be aware of their lower internal reliability.

Correlations of scores within a subtest (e.g., list, story, or figure) were largely statistically significant, moderately large, and heading in the expected direction. For example, as seen in Table 2, the List Recall score was positively correlated with the List Recognition total and List Recognition hits scores and negatively correlated with the List Recognition false positive errors. Similar patterns were observed within the Story Recall/Recognition and Figure Recall/Recognition correlation matrices (see Tables 3 and 4, respectively). Although the correlations within each subtest were similar, they were slightly higher for the Story Recall/Recognition matrix (i.e., average of all six correlations within the matrix; Story rho = 0.77, List rho = 0.65, Figure rho = 0.61). The only non-significant correlation was within the Figure Recall/Recognition matrix (i.e., rho = −0.19 [p = 0.06] between Figure Recognition hits and Figure Recognition false positive errors). Such findings are consistent with studies in multiple patient groups that have identified that story recall can dissociate from list recall and figure recall (Badenes et al., 2018; Bahar-Fuchs et al., 2013; Beatty, 2004; Ober, Shenaut, & Taylor, 2019; Zahodne et al., 2011).

Correlations between subtests (e.g., recall or recognition) were all statistically significant, moderate to large, and heading in the expected direction. For example, List Recall and Recognition scores were positively correlated with Story and Figure Recall and Recognition scores (see Table 5). The relationships within recall scores were slightly higher than those within recognition scores, but there were not statistically different. These relatively strong correlations (0.60 – 0.86), however, raise some concern about the incremental utility of these new recognition tasks. Although the current study was not designed to see what these new scores might add to clinical decision making, future studies might examine this with some external criterion. For example, do recall or recognition scores tell us more about an older patient’s ability to manage his/her medications or finances? Similarly, do recognition scores provide unique information above and beyond recall scores?

Although the current sample of intact participants is relatively small (n=48), these data can be used as preliminary information for the new recognition scores until larger and more diverse samples can be collected. In Table 1, means and standard deviations are presented for each of the new recognition scores; however, given the non-normal distributions of many of these scores, means and standard deviations might not be the best descriptive information to use. Also presented in Table 1 are the 25^th, 50^th (i.e., median), and 75^th percentiles for each score for each group. For example, the Story Recognition total score has a maximum value of 20. In the intact group, the 25^th percentile was a score of 18, the 50^th percentile was a score of 19, and the 75^th percentile was a score of 20. Therefore, an obtained Story Recognition total score of 17 would below the 25^th percentile of the current intact group. Similarly, a single false positive error on this same recognition trial would be at the 75^th percentile of the current sample (with higher scores indicating worse performance on this trial). Again, these data should be viewed as preliminary until larger and more diverse samples are collected.

Despite the potential contribution of these new recognition scores, there are some limitations that deserve to be mentioned. First, the format of the two new recognition subtests modeled that of the existing List Recognition subtest, with yes/no responses. Although this provides consistency between all three recognition subtests, it may have missed an opportunity to examine recollection and familiarity (Schoemaker, Gauthier, & Pruessner, 2014). Recollection accesses memories by invoking contextual aspects of the encoding event, it tends to be slower and more effortful, and it engages the hippocampal regions (Leube et al., 2008). Conversely, familiarity appears to rely less on the context of the encoding event, it tends to be faster and more automatic, and it primarily engages perirhinal brain regions (Wolk, Dickerson, & Alzheimer’s Disease Neuroimaging, 2011). In a meta-analysis on these different aspects of recognition memory, Koen and Yonelinas (2014) found that patients with AD showed significant deficits in both recollection and familiarity compared to healthy peers, whereas patients with amnestic MCI only showed clear decrements in only recollection. Future recognition tests might be structured to assess both recollection and familiarity (e.g., yes/no responses, as well as rating the confidence of those responses). Second, as noted earlier, there is a need for normative data on these recognition scores, including larger and more diverse normative samples. Third, additional clinical samples are needed to better validate the prospective utility of these recognition scores in making diagnoses, especially those with classic “subcortical” profiles. Fourth, there is a need for short- and long-term longitudinal data to establish the reliability of these scores, as well as to see how they change across time. Finally, these recognition scores only apply to Form A of the RBANS, and future studies might develop similar recognition subtests for the other forms of the RBANS. Regardless of these limitations, if the current recognition scores are supported and validated in future samples, then such an addition to the RBANS may improve its sensitivity in identifying cortical/subcortical profiles in clinical and research settings.