The development and preliminary psychometric properties of a new instrument called the Brief Test of. Attention (BTA) are described. In terms of the conceptual ...
0920-1637/96/1 00 1-80$12.00 @ Swets & Zeitlinger
The Clinical Neuropsychologist 1996, Vol. 10, No.1, pp. 80-89
Development and Psychometric Properties of the Brief Test of Attention * David Schretlenl, Julie Hoida Bobholz2, and Jason Brandt3 IJohns Hopkins University School of Medicine, 2Chicago Medical School, and 3Johns Hopkins University School of Medicine
ABSTRACT The development and preliminary psychometric properties of a new instrument called the Brief Test of Attention (BTA) are described. In terms of the conceptual model proposed by Cooley and Morris (1990), the BTA is best described as a measure of auditory divided attention. The test consists of two parallel forms that require less than 5 minutes each to administer and score. The results of testing 926 patients and normal control subjects support the BTA's reliability, equivalence of forms, and construct validity. Coefficients alpha for the entire test range from .82 to .91, while between-form correlations range from .69 to .81. Neither practice nor interference effects were found to influence performance from the first to second form administered. Correlation and principal components analyses showed that the BTA correlates more strongly with widely accepted measures of attention than with other cognitive tasks, and more strongly with complex than simple attention tasks.
Impairments of attention characterize most neuropsychiatric disorders. Attentional deficits are central to attention-deficit hyperactivity disorder (Barkley, Grodzinsky, & DuPaul, 1992) and delirium (Mesulam, 1985), but also have been associated with schizophrenia (Braff, 1993), depression (Hartlage, Alloy, Vasquez, & Dykman, j 1993), traumatic brain injury (Stuss & Gow, 1992), learning disabilities (Fleisher, Soodak, & Jelin, 1984), cortical and subcortical dementia syndromes (Cummings & Benson, 1992), epilepsy (Mirsky, 1989), and alcohol abuse (OscarBerman & Bonner, 1985), to name just a few. I The majority of widely used clinical tests of attention were developed without respect to theories of attention. Elegant cognitive and neuroanatomical models of attention, such as those proposed by Triesman (1961), Atkinson and Shiffrin (1968), Gibson (1969), Posner and ISnyder (1975), Wickens (1984), and Mesulam
.
*
(1985), captured scientific attention long after most attention tests enjoyed a secure place in clinical neuropsychology. Few experts would question the clinical utility of the Digit Span and Digit Symbol (Wechsler, 1939), Trail Making (Reitan, 1958), Stroop Color-Word (Stroop, 1935), or continuous performance tests. However, aside from Mirsky's (1989) factor analytic approach to defining components of attention, clinical tests rarely are described in terms of these conceptual models. One possible explanation for this dissociation between clinical practice and experimental neuropsychology is that accepted clinical measures mix different components of attention, and confound attention with other, often complex cognitive processes. Cooley and Morris (1990) described a framework for conceptualizing the ta~k demands of various clinically accepted measures of attention. They argue that both sustained and
The Brief Test of Attention@ may be obtained by contacting the first author. The authors wish to thank two
anonymous reviewers for their helpful comments on an earlier draft of this manuscript. Address correspondence to: David Schretlen, Ph.D., Johns Hopkins Hospital, 600 N. Wolfe St., Meyer 218, Baltimore, MD 21287-7218, tel: (410) 955-3268, USA. Accepted for publication: April 26, 1995.
THE BRIEF TEST OF ATTENTION
divided attention tasks can be conceptualized as special cases of a basic selective attention process. Every selective attention task involves two components: target identification (attend) and distractor inhibition (inhibit). According to this model, sustained attention tasks are conceptualized as the extension of these two component processes over time. Divided attention is understood as requiring the performance of two simultaneous selective attention tasks. Cooley and Morris conceptualize factors that influence the "inhibit" and "attend" components in terms of four levels of processing. At the most basic level, tonic arousal regulates general information processing (Mesulam, 1985). Modalitylinked sensory registration processes, as required by the detection of light flashes or buzzes, define the second level of processing. Modality-specific perceptual processes, such as matching designs or letters, define the third level. The fourth level involves conceptual processing. Relevant tasks typically involve the use of multiple cognitive processing systems, crossmodal comparisons, and/or the allocation of processes between systems. Finally, Cooley and Morris argue that each of the above levels of processing may be represented within five (verbal, spatial, memory, motor, and executive) functional neuropsychological systems. In this article, we use the model proposed by Cooley and Morris (1990) to describe a new measure of auditory divided attention called the Brief Test of Attention (Schretlen, 1989). We examine its psychometric properties, and compare its processing requirements to those of other tests of attention.
BRIEF TEST OF ATTENTION Development of the Brief Test of Attention (BTA) began in 1989. The primary aim was to devise a brief, relatively simple and easily administered test of auditory divided attention that would be sensitive to subtle attentional impairments. A secondary goal was to reduce confounding task demands such as psychomotor speed or conceptual reasoning.
81
The Brief Test of Attention consists of two parallel forms that are presented via audio cassette. Subjects are administered both forms; each requires 4 minutes to administer and score. On Form N, a voice reads 10 lists ofletters and numbers (e.g., "M-6-3-R-2") that increase in length from 4 to 18 items. The subject's task is to disregard the letters and count how many numbers are read aloud. Each list is followed by 5s of silence, during which the subject reports how many numbers were recited. The same 10 lists are presented as Form L, for which the subject's task is to disregard the numbers and count how many letters are read aloud. Unlike digit span tests, the subject is not asked to recall which numbers (or letters) are presented. The number of correctly monitored lists is summed across forms; thus, total BTA scores can range from 0 to 20. The BTA was tape recorded in a sound-attenuated studio. The voice is that of a female radio broadcast professional who read test stimuli at the rate of one letter or number per second. Each trial is followed by 5s of silence. After all 10 lists were recorded, the resulting production was re-mastered using digital audio recording to filter out noise between trials. The first 10 trials are introduced on tape as, "Brief Test of Attention, Part A." However, the same 10 trials are duplicated on the tape and subsequently introduced as, "Brief Test of Attention, Part B." In this way, either form N or L can be presented first. The effect of order of presentation is discussed below. Prior to beginning the tape, directions are read to the subject by the examiner. The examiner also reads two short sample lists. If the respondent fails all three trials of both sample lists, the test is discontinued; otherwise, the entire form is administered. The tape is never stopped during the administration of a given form. However, in some circumstances, the use of only one form will yield reliably interpretable results.
82
DAVID SCHRETLEN
STUDY 1: RELIABILITY AND DEMOGRAPHIC CORRELATIONS
The purpose of this study was to document basic psychometric properties of the Brief Test of Attention in normal and clinical samples.
ET AL.
were administered the BTA as part of their examinations. The BTA was not administered in a systematic order relative to other tests. Although demographic information was obtained for most subjects, the available psychometric data varied greatly, depending on research protocols and patient referral questions.
RESULTS
METHOD Subjects Normal sample This sample consisted of 275 adults and 74 children who served as normal control subjects in one of five separate studies. The children included all second (n = 24), fifth (n = 25), and eighth (n = 25) grade students of a local elementary school. Although the elementary (n
= 74)
and college
(n
= 62)
students, as well as 45
adult normal control subjects did not undergo any particular screening procedures, the remaining (n = 213) adult subjects were screened for dementia, severe psychiatric disorders, and current substance abuse. Clinical sample This sample consisted of 577 patients drawn from studies and clinics conducted at the Johns Hopkins University and Hospital. Included were patients with schizophrenic (n = 55), affective (n = 105), sexual (n = 24), substance abuse (n = 16), eating (n = 14), mental retardation (n = 16), and other (n = 40) psychiatric disorders, as well as Huntington's disease (n = 56), dementia (n = 32), traumatic brain injury (n = 44) and adrenoleukodystrophy(n = 51). The primary diagnosis was either unavailable or not recorded for 124 patients. Demographic characteristics of both normal and patient samples are shown in Table 1. Procedure After giving voluntary informed consent, control subjects were administered the BTA along with whatever other measures were included in the protocol in which they served. Patients referred for clinical evaluations
Internal Consistency Based on the 349 normal adults and children, internal consistency analyses yielded a coefficient alpha of .82 for the BTA (.71 for Form L and .66 for Form N). When these data, were pooled with those of 480 patients for whom item scores were recorded, and internal consistency analyses were repeated using this combined sample (n = 829), the coefficient alpha increased to .91 for the BTA (.83 for Form Land .84 for Form N).
Form Equivalence Based on the 349 normal adults and children, the Pearson correlation (r) between forms Nand L was .69. However, the between-forms correlation increased to .81 for the combined (n
= 926)
normal and clinical samples. Despite the modest correlation between forms, paired-sample t-tests revealed that normal adults produced virtually identical scores on Forms Land N (8.5, SD = 1.6 vs. 8.5, SD = 1.5; t(274)= 0.67; p = .50). The patients produced lower and more variable scores overall, but they also showed virtually identical performance on Forms Land N (5.6, SD = 2.9 vs. 5.5, SD = 2.9; t(596)= 1.02; p = .31), as did the normal
children
(t(73)
= -0.62;
P
=
.54).
Table 1. Demographic Characteristics of Normal and Clinical Samples. Demographic Variable Normal Samples Adults Children Clinical Sample 'Race: W = White; B
n
Age M (SD)
Sex M/F
Race W/B/O'
Education M (SDi
275 74 577
47.1 (19.3) 10.1 (2.5) 40.4 (15.7)
112/163 38/36 327/250
205/30103 70/410 336/201/73
14.4 (2.7) 4.0 (2.5) 12.2 (3.2)4
= Black; a
= Other. 2Education expressed in years completed. 3Race was not recorded for
40 normalcontrolsubjectsand33 patients.4Educationnotrecordedfor 37 patients.
83
THE BRIEF TEST OF ATTENTION
The effect of order of administration
was ex-
amined by subtracting each subject's score on whichever form was administered first from that obtained on the form administered second. A practice effect should result in positive difference scores; an interference effect should lead to negative difference scores. However, if intraindividual variability across forms is random, then the difference scores should be normally distributed with a central tendency that approaches zero. In fact, the mean difference score of the 765 patients and normal subjects for whom order was recorded did approach"zero (M = .15, SD = 1.8 points). The distribution of difference scores was mildly leptokurtic (.30), indicating that more subjects than expected produced difference scores of zero. The distribution showed negligible skewness (-.07), suggesting that exposure to the first form given neither facilitated nor impeded performance on the second form. In fact, over 97% of normal subjects and 93% of patients produced scores that differed by 7/l 0 on the first form given thereby demonstrates normal performance. However, the present analyses indicate that, regardless of age, no patient who earned> 7/1 0 points on the first
100 75 55 40 23 15 9 3
form given produced an abnormal total BTA score.
DISCUSSION The results of this study indicate that the Brief Test of Attention has adequate internal consistency. Correlations between Forms Nand L ranged from .69 to .81. Between-form difference scores were normally distributed with a nearzero mean, and were