Archives of Clinical Neuropsychology 30 (2015) 99–104
Abbreviating the Finger Tapping Test Lee Ashendorf1,2,*, Julie E. Horwitz3, Brandon E. Gavett4 1
Edith Nourse Rogers Memorial Veterans Hospital, Psychology 116B, Bedford, MA, USA 2 Boston University School of Medicine, Boston, MA, USA 3 Memorial Hospital, University of Colorado Health, Colorado Springs, CO, USA 4 Department of Psychology, University of Colorado, Colorado Springs, CO, USA
*Corresponding author at: Psychology 116B, 200 Springs Road, Bedford, MA 01730, USA. Tel: +1-781-687-3042. E-mail address:
[email protected] (L. Ashendorf). Accepted 9 December 2014
Abstract The Finger Tapping Test (FTT) has a strong empirical base but its procedures are inconsistent and the test can be time-consuming. To simplify and abbreviate administration procedures, several potential abbreviated procedures were evaluated using a sample of 71 individuals presenting to a VA Hospital for neuropsychological evaluation. A short version using the mean score of Trials 3– 5 for each hand was found to be a strong predictor of full-test performance. The abbreviated version also had stronger reliability than the full version, and it accurately predicts impairment and performance validity classification based on the full version. This abbreviated version appears to be more efficient and sufficiently accurate to be considered for use in lieu of the traditional and potentially longer version of the FTT. Keywords: Neuropsychology; Finger Tapping Test; Psychometrics; Short Form
The Finger Tapping Test (FTT; Reitan & Wolfson, 1993), also called the Finger Oscillation Test, has a long history in the field of neuropsychology, both as part of the Halstead-Reitan Neuropsychological Battery (HRNB) and as a standalone test. It is among the more sensitive tests to brain damage, and as a result is also included in the HRNB screening batteries for adults and for children (Reitan & Wolfson, 2006). It is commonly used in evaluations for traumatic brain injury (TBI) and other neurological disorders (Ashendorf, Vanderslice-Barr, & McCaffrey, 2009; Leckliter & Matarazzo, 1989; Prigatano, Johnson, & Gale, 2004). Alternative computerized versions of the FTT have also been developed and are included in some clinically and research-oriented computerized test batteries (e.g., Hubel, Reed, Yund, Herron, & Woods, 2013). The traditional FTT is said to take 10 min to administer (Strauss, Sherman, & Spreen, 2006). However, if all 10 trials are given with each hand, and all breaks between trials are allotted as instructed (30 s between trials, 1 – 2 min break after every 3 trials), the test would have to take at least 15 min. Availability of a shorter version that measures the same outcome as the full version would make it more appealing for use as a standalone measure and would help to reduce total testing time. Another reason to abbreviate the test is the high degree of inter-trial scatter for those who receive all 10 trials, which calls into question the reliability of performance and the meaningfulness of the mean FTT score in such cases. Gill, Reddon, Stefanyk, and Hans (1986) appear to be the only investigators to have explored consistency of performance across trials. These authors showed, graphically, that performance among women tends to be fairly consistent, but men tend to exhibit a small but steady improvement over the first five trials and then remain relatively steady for the remaining five trials. The current investigation seeks to create an abbreviated administration method for the FTT that closely approximates the full version. It is hypothesized that a three-trial version will perform sufficiently for clinical use. However, a score based on the first three trials (as used, e.g., by Arnold et al., 2005) is not expected to be reflective of total output, especially in light of the above observations of Gill and colleagues (1986). Therefore, a score based on later trials near the middle of the test instead is expected to be a more accurate predictor of full-test score. Published by Oxford University Press 2015. This work is written by (a) US Government employee(s) and is in the public domain in the US. doi:10.1093/arclin/acu091 Advance Access publication on 5 January 2015
100
L. Ashendorf et al. / Archives of Clinical Neuropsychology 30 (2015); 99–104
Methods Participants The current study included 71 referrals to a Veterans Affairs Medical Center neuropsychology clinic. The mean age was 36.9 years (SD ¼ 12.0) and mean education level was 13.2 (SD ¼ 2.0). There were 67 men (6% women). The sample’s racial composition was 82% white, 11% Latino, 6% black, and 1% Asian. Referral questions varied, but most were referred for evaluation of diagnosed TBI (79%). No cases were excluded due to performance validity test failure (12 cases [17% of the sample] scored below standard cutoffs on the Test of Memory Malingering; Tombaugh, 1996), as invalidity is not expected to affect the predictive validity of the abbreviated version of the FTT, and a good brief version should be able to accurately estimate FTT scores for invalid cases. Additionally, the revised version was to be evaluated on its ability to predict an existing FTT validity index (Larrabee, 2003). Therefore, all 71 individuals were included in the analyses. Measures The FTT uses a wooden board with a small lever attached to an analog counter, as in Reitan and Wolfson (1993). Each trial lasts for 10 s. The standard discontinue rule is after 5 consecutive trials fall within 5 taps of each other, or at 10 trials if that criterion is not reached. The resulting score is the mean of the “5 within 5” trials, or else the mean of all 10 trials. Administration procedures deviated slightly from standard HRNB instructions; after 3 consecutive dominant-hand trials, 3 nondominant-hand trials were completed, and then this went back and forth until the test was complete; in essence, the allotted 1 – 2 min breaks after every three trials were filled by switching to the other hand. A minor deviation from standard procedures in this way is not unprecedented (e.g., Prigatano & Borgaro, 2003), and it is not considered likely to have interfered with study outcome. The abbreviated versions evaluated for this study were the average and maximum trial scores based on all consecutive 3-, 4-, and 5-trial combinations from among the first five trials of the FTT. Only the first 5 trials were considered because the purpose was to create a form that would allow for administration that is briefer than the existing 5- to 10-trial version. Data Analysis To calculate proportion of total variance contributed by each potential abbreviated-form score, a series of linear regressions was performed using total FTT score for each hand as the dependent variable and abbreviated version scores as the independent variables. Intraclass correlation coefficients (ICCs) were calculated for determination of reliability across trials. Classification accuracy statistics were used to compare performance on the revised score with performance of the full FTT; sensitivity (hits/total positive cases), and specificity (true negatives/total negative cases) values were calculated for accuracy at identifying impairment and performance validity. Results Average FTT scores in this sample were 48.5 (SD ¼ 11.9) for the dominant hand and 44.2 (SD ¼ 11.0) for the nondominant hand. Regression scores and deviation scores for each abbreviated version are provided in Table 1. As it displayed the most consistently strong correlations with the full FTT score, it was determined that the Trials 3 – 5 mean score held the most promise, so this was investigated further. To determine its clinical accuracy, the number of cases that fell in each 0.33 z-score increment from the full FTT score (using existing regression-based normative data; Mitrushina, Boone, Razani, & D’Elia, 2005) was calculated (see Table 2). Three cases’ dominant-hand performances were extreme outliers, with estimated scores more than a full SD below true scores. To further assess these outliers, SDs were calculated for each participant’s performance across trials. Those three outliers had 3 of the 5 highest intraindividual SDs, and on closer inspection were specifically characterized by first, a very slow “warmup” on the first few trials; second, abnormally low performance on Trials 3 and 4 (well below the previous and subsequent trials); and third, considerable improvement beyond Trial 3. As the discrepancy among Trials 3 – 5 clearly drove the deviation scores for these three cases, it was observed that the latter two of these represented the only cases in the sample for whom Trial 5 performance was .15 taps faster than Trial 3 performance. The abbreviated form therefore may not be appropriate to use when a discrepancy of this magnitude occurs. Nondominant-hand consistency was greater, as intraindividual SDs were lower and no extreme outliers were observed. The variable ICCs were calculated in order to determine the reliability of the FTT across trials. The ICC for those who required all 10 trials was not very high (ICC ¼ 0.663). In contrast, the reliability for Trials 3 – 5 was high (ICC ¼ 0.898), and was still better
L. Ashendorf et al. / Archives of Clinical Neuropsychology 30 (2015); 99–104
101
Table 1. Variance in total score accounted for and margin of error by abbreviated versions of the finger tapping test (FTT) Dominant hand
Nondominant hand
Trials
R
2
SEE
R2
SEE
1– 3 average 1– 4 average 1– 5 average 2– 4 average 2– 5 average 3– 5 average 1– 3 max 1– 4 max 1– 5 max 2– 4 max 2– 5 max 3– 5 max
.868 .906 .936 .922 .953 .959 .865 .934 .818 .948 .820 .965
4.35 3.68 3.02 3.35 2.59 2.43 4.41 3.08 2.56 2.74 5.35 2.23
.919 .944 .959 .955 .968 .966 .891 .932 .931 .937 .936 .944
3.16 2.64 2.26 2.35 1.98 2.04 3.67 2.90 2.93 2.79 2.82 2.64
Note: SEE ¼ standard error of the estimate; best form is in bold font.
Table 2. Cumulative frequency values for Trials 3 –5 versus total FTT z-score discrepancies SD discrepancy
Dominant hand
Nondominant hand
.3.00 .2.67 .2.33 .2.00 .1.67 .1.33 .1.00 .0.67 .0.33 .0.05
0 0 1 (1.4%) 1 (1.4%) 2 (2.8%) 3 (4.2%) 3 (4.2%) 9 (12.7%) 23 (32.4%) 64 (78.9%)
0 0 0 0 0 1 (1.4%) 2 (2.8%) 8 (11.3%) 18 (25.4%) 58 (66.2%)
Note: Discrepancy scores are based on the formula z-scoreshort 2 z-scorefull, using z-scores derived from the normative equations provided in Mitrushina and colleagues (2005).
than the reliability of the full test even if one only examines the participants who required all 10 trials (ICC for Trials 3 –5 ¼ 0.752). Bland – Altman plots are also provided in Fig. 1 (dominant hand) and Fig. 2 (nondominant hand) to demonstrate the agreement between the full and T3– 5 scores. As can be seen in Figs. 1 and 2, the majority of the estimates produced by the short form are within the 95% confidence interval (i.e., +1.96 SD). The Trials 3 – 5 average score’s accuracy at predicting impairment status based on traditional cutoffs (as in Bornstein, 1986) was calculated. Of the 71 cases in this sample, using the dominant hand, 34 (48%) fell below the cutoff for impairment. The revised score misclassified 3 of these people as intact, yielding a sensitivity of 91%. Of the remaining 37 individuals whose full FTT scores were not impaired, 3 were misclassified by the revised score as impaired, yielding a specificity of 92%. Using the nondominant hand, 29 (41%) fell below the cutoff for impairment. The revised score misclassified one of these as intact, resulting in a sensitivity of 97%. Of the remaining 42 individuals, none were classified as impaired by the revised score, yielding a specificity of 100%. Using a cutoff of 21.5 SD to denote “impairment,” the revised test performed similarly. These data are summarized in Table 3. The Trials 3 – 5 average score’s accuracy at predicting performance validity status based on Larrabee (2003) was calculated. Of the 71 cases in this sample, 8 (11%) fell below the cutoff and were considered invalid. The 35 average score accurately predicted classification for all but one of the participants, who was misclassified as failing the index. These data are also summarized in Table 3. Discussion These findings describe a method of predicting total FTT score using a sometimes-briefer administration format. The results demonstrate the effectiveness of a score that is based on the mean number of taps on Trials 3 through 5. This score would permit administration of only Trials 1 – 5, regardless of consistency (negating the “5 within 5” rule). While the abbreviated form fared very
102
L. Ashendorf et al. / Archives of Clinical Neuropsychology 30 (2015); 99–104
Fig. 1. Bland–Altman plot showing total Finger Tapping Test (FTT) dominant-hand score (x-axis) and its difference from Trials 3 –5 dominant-hand mean score (y-axis). The middle, top, and bottom horizontal lines represent the mean difference and +1.96 SD (95% confidence interval), respectively.
Fig. 2. Bland– Altman plot showing total FTT nondominant-hand score (x-axis) and its difference from Trials 3– 5 nondominant-hand mean score (y-axis). The middle, top, and bottom horizontal lines represent the mean difference and +1.96 SD (95% confidence interval), respectively.
well overall, for situations in which Trials 3 – 5 differ by .15 taps, administration of the traditional (full) form of the test will likely be necessary. Those few individuals whose full-form scores were not accurately predicted had very inconsistent output across 10 trials, and so even their mean total scores were not likely accurate and reliable reflections of their true physical ability.
L. Ashendorf et al. / Archives of Clinical Neuropsychology 30 (2015); 99–104
103
Table 3. Classification accuracy for prediction of traditional impairment and validity scores using the abbreviated (Trials 3– 5 mean) FTT Dominant hand Impairment criterion a
Traditional z-score cutoffb Validityc
Nondominant hand
Sensitivity
Specificity
Sensitivity
Specificity
0.91 1.00 Sensitivity 1.00
0.92 0.92
0.97 0.95 Specificity 0.98
1.00 0.98
a
Impairment cutoffs based on traditional criteria as in Bornstein (1986). “Impaired” n ¼ 34 (dominant hand) and n ¼ 29 (nondominant hand). Using a fixed cutoff of .1.5 SDs below the mean (based on meta-analytic equations from Mitrushina et al., 2005). “Impaired” n ¼ 18 (dominant hand) and n ¼ 21 (nondominant hand). c Validity cutoffs are combined dominant + nondominant hand performance based on Larrabee (2003). b
In comparison with other common motor tests such as the Grooved Pegboard Test, the FTT is a purer measure of motor speed, without interference from visuospatial and executive functions that affect performance on the former task (Ashendorf et al., 2009). As the GPT might ordinarily be a more appealing measure to use due to its ease and brevity of administration, this briefer version of the FTT might encourage its more frequent use as an independent measure of motor speed. In predicting “impairment” on the full test based on traditional cutoff scores, classification accuracy statistics (sensitivity and specificity) for the revised version were strong. In addition, based on the current data, the revised score predicted full score faithfully enough to accurately predict whether the full score fell above or below the Larrabee (2003) PVT cutoff in 99% of the sample. Given the finding that in most cases, the revised score deviated ,2/3 standard deviation from the full-test score, this appears to be a clinically viable procedure. The primary objective of this investigation was to reduce administration time. This study assumed that examiners would use 30-s breaks between trials and alternate hands after every three trials (rather than taking 1– 2 min breaks and completing the preferred hand before proceeding to the nonpreferred hand). This procedure takes ,5 min to complete the five trials, which compares favorably to the potentially 15-min-long 10-trial procedure. Limitations of this study include a fairly narrow clinical sample, mostly consisting of veterans with histories of mild TBI. Future cross-validation efforts might benefit from inclusion of different populations with greater motor involvement, such as Parkinson’s disease or multiple sclerosis. In addition, only four participants were women, which is potentially relevant given the aforementioned sex differences in learning curves over trials (Gill et al., 1986). However, it is noteworthy that none of the three outliers noted above were women, and given that performance among women has actually been found to be more consistent than that among men, there should in theory be less concern about the validity of this abbreviated measure when administered to women. Overall, this study demonstrates a reliable and simple method for making the FTT briefer and more clinician-friendly. Conflict of Interest None declared. References Arnold, G., Boone, K. B., Lu, P., Dean, A., Wen, J., & Nitch, S., et al. (2005). Sensitivity and specificity of finger tapping test scores for the detection of suspect effort. The Clinical Neuropsychologist, 19, 105– 120. Ashendorf, L., Vanderslice-Barr, J. L., & McCaffrey, R. J. (2009). Motor tests and cognition in healthy older adults. Applied Neuropsychology, 16, 171–176. Bornstein, R. A. (1986). Classification rates obtained with “standard” cut-off scores on selected neuropsychological measures. Journal of Clinical and Experimental Neuropsychology, 8, 413–420. Gill, D. M., Reddon, J. R., Stefanyk, W. O., & Hans, H. S. (1986). Finger tapping: Effects of trials and sessions. Perceptual and Motor Skills, 62, 675–678. Hubel, K. A., Reed, B., Yund, E. W., Herron, T. J., & Woods, D. L. (2013). Computerized measures of finger tapping: Effects of hand dominance, age, and sex. Perceptual and Motor Skills, 116, 929– 952. Larrabee, G. J. (2003). Detection of malingering using atypical performance patterns on standard neuropsychological tests. The Clinical Neuropsychologist, 17, 410–425. Leckliter, I. N., & Matarazzo, J. D. (1989). The influence of age, education, IQ, gender, and alcohol abuse on Halstead-Reitan neuropsychological test battery performance. Journal of Clinical Psychology, 45, 484–512. Mitrushina, M., Boone, K. B., Razani, J., & D’Elia, L. F. (2005). Handbook of normative data for neuropsychological assessment (2nd ed.). New York: Oxford University Press. Prigatano, G. P., & Borgaro, S. R. (2003). Qualitative features of finger movement during the Halstead Finger Oscillation Test following traumatic brain injury. Journal of the International Neuropsychological Society, 9, 128– 133.
104
L. Ashendorf et al. / Archives of Clinical Neuropsychology 30 (2015); 99–104
Prigatano, G. P., Johnson, S. C., & Gale, S. D. (2004). Neuroimaging correlates of the Halstead Finger Tapping Test several years post-traumatic brain injury. Brain Injury, 18, 661 –669. Reitan, R. M., & Wolfson, D. (1993). The Halstead –Reitan neuropsychological test battery: Theory and clinical interpretation (2nd ed.). Tucson, AZ: Neuropsychology Press. Reitan, R. M., & Wolfson, D. (2006, October). Issues in neuropsychology: Where have we succeeded? Where have we failed? What are the clinical and forensic implications? Presented at the 26th annual conference of the National Academy of Neuropsychology, San Antonio, TX. Strauss, E., Sherman, E. M. S., & Spreen, O. (2006). A compendium of neuropsychological tests (3rd ed.). New York: Oxford University Press. Tombaugh, T. N. (1996). Test of memory malingering. North Tonawonda, NY: Multi-Health Systems.