This study was conducted from May to December 2017. Participants were selected using convenience sampling. Those who visited the hospital for a health examination were asked to participate and were interviewed for eligibility according to the inclusion and exclusion criteria. Then, they were asked to complete the KCNT. To control confounding variables such as age, sex, and education, every participant repeated the battery of the KCNT using four different test devices (later referred to as Systems) in a randomly assigned order. This study was approved by the institutional review board of Yeungnam University (IRB File No. YU 2017-04-001-001). Seventy-four people volunteered and none were ineligible to participate. However, data from two participants were inappropriate for analysis and were excluded. Therefore, 72 participants were included in this study.

People who were aged over 19 and under 65 years were selected if they did not meet the exclusion criteria. They were excluded if they had any of the following [16–20]: a past medical history of or present serious condition that could affect neurobehavioral performance such as head trauma or neurological disease; potential occupational exposure to neurotoxins revealed during an interview; and physical disabilities that could influence the neurobehavioral test, such as hearing impairment, color vision deficiency, or severe lower back pain.

General characteristics were collected during the interview, including age, sex, and years of formal education. Participants' typing speed was also tested to objectively evaluate computer familiarity. Typing speed was defined as the number of Korean characters typed in a minute.

Four different testing systems were used in this study (Fig. 1): a desktop computer equipped with a conventional 106-key keyboard (System 1), a desktop computer equipped with a simplified keyboard (System 2), a tablet PC with a simplified 17-key on-screen keyboard (System 3), and a tablet PC equipped with a simplified keyboard (System 4). In System 3, the tablet PC display was a capacitive screen digitizer and an on-screen keyboard was used as the input device. The tablet PC used in this study had a display with a diagonal length of 10 in., whereas the monitor connected to the desktop computer had a display with a diagonal length of 24 in..

The test battery comprised five subtests selected by the authors [9, 14, 15]: Simple Reaction Time (SRT) test, Choice Reaction Time (CRT) test, Digit Addition (DA) test, Symbol-Digit Substitution (SDS) test, and Finger Tapping Speed (FTS) test. Since every participant performed the KCNT multiple times, there was a risk of biases due to mental fatigue and learning effect [21, 22]. To minimize these biases, participants performed the test in a completely counter-balanced, randomly and evenly assigned order. That is, there were 24 possible combinations of the four systems, and every participant was assigned a random sequence in which to serially perform the KCNT.

All subtests, except for the FTS test, had three common parameters: correct response rate (Rate_CR), mean reaction time of correct responses (RT_mean), and standard deviation of the reaction time (SD_RT), where reaction time is expressed in milliseconds. In contrast, there were only two parameters for the FTS test: average number of taps during 10-s trials using the dominant hand and the non-dominant hand, respectively (FTS_D and FTS_ND).

Statistical analyses were performed using IBM SPSS Statistics Version 22. General characteristics were described using frequencies, percentages, means, and standard deviations. To assess performance, parameters generated from each subtest were analyzed. Repeated measures analysis of variance (ANOVA) and the Friedman test were used to compare performance for the four different systems. For the ANOVA and Friedman test, Bonferroni test and Wilcoxon signed-rank test were applied as post-hoc procedures, respectively. Analysis of covariance (ANCOVA) was used to compare performance between groups with different computer familiarity. For ANCOVA, Bonferroni test was applied as a post-hoc procedure. A p-value below 0.05 was considered statistically significant.

General characteristics of all 72 participants are listed and summarized in Table 1. Participants were on average 40.3 ± 12.8 years old and 50% were male. Seventy-one participants (98.6%) completed high school education or higher; one participant completed only middle school education. Their typing speed was 258.2 ± 164.5 characters per minute on average. Forty participants (55.6%) could type more than 200 characters per minute, and were classified as Group III (very familiar with computers). Twenty-three participants (31.9%) had a typing speed of lower than 200 characters per minute and were classified as Group II (relatively familiar with computers). Nine participants (12.5%) with a typing speed of near zero were classified as Group I (no competency using computers).

Performance of all participants was evaluated by comparing test parameters among Systems 1, 2, 3, and 4 (Table 2). The mean reaction time of the SRT test showed a significant difference between systems (p < 0.001). Post-hoc analysis showed that the mean reaction time for Systems 3 and 4 was greater than that of Systems 1 and 2. The mean reaction time for the CRT test also showed differences between systems (p < 0.001). In contrast to the results from the SRT test, post-hoc analysis showed that the mean reaction time for the CRT test using System 3 was significantly greater than that of System 4. The mean reaction time for the CRT test using Systems 3 and 4 was significantly greater than that of Systems 1 and 2. Although the mean reaction time for the SDS test was similar between systems (p = 0.961), the correct response rate significantly differed (p < 0.001); there was no difference between System 1 and System 2, but the rate was lower for System 3. Performance on the DA test did not show significant differences by system type for all parameters (RT_mean, p = 0.364; SD_RT, p = 0.664; Rate_CR, p = 0.751). Similarly, for the FTS test, results using dominant hand did not differ between systems (p = 0.350), but the results using non-dominant hand showed a decreased performance in System 3 compared to Systems 1, 2 and 4 (p < 0.001).

If performance differed by computer familiarity group, further analysis would be required to compare test systems stratifying by computer familiarity. Therefore, test results from Groups I, II, and III were compared for different systems (Table 3), even though this study did not primarily aim to assess the effect of computer familiarity on the performance of computerized neurobehavioral tests. Since the general characteristics differed by each group, the results had to be adjusted by age and the length of formal education.

The mean reaction time for the SRT test, correct response rate for the DA test, and mean reaction time for the SDS test showed significant differences among the computer familiarity groups. Post-hoc analyses revealed that the performance of Group II for the SRT test was consistently lower than that of Group I throughout the test systems. Group I had the lowest correct response rate for the DA test when tested with Systems 1, 2, and 3. For the SDS test, Group I showed lower performance in terms of reaction time than Group II when tested with System 2 and 3.

Since performance differed by computer familiarity group (Table 3), the performance of each group classified by computer familiarity was evaluated by comparing test parameters among Systems 1, 2, 3, and 4 (Table 4).

The mean reaction time for the SRT test showed a significant difference between systems for all three groups, and the post-hoc analyses demonstrated that the mean reaction time for the SRT test was greater for Systems 3 and 4 than for Systems 1 and 2 for all three groups, which was consistent with the results reported in Table 2. However, there were significant differences between Systems 1 and 2 for Group I and between Systems 3 and 4 for Group II. The mean reaction time for the CRT test also showed a significant difference between systems for all groups, and the post-hoc analyses showed the similar results to those reported in Table 2. In all groups, the mean reaction times for the CRT test using Systems 1 and 2 did not show significant differences and were greater than that for System 3. Performance using System 4, however, was superior to that using System 3 for Group II and Group III. The mean reaction time for the SDS test showed no statistically significant difference between systems for all three groups. Finally, performance on the DA and FTS tests did not differ by type of system, for all three groups.

Performance on the DA test did not differ significantly by test system. As for the FTS test, the performance using dominant hand did not demonstrate significant differences among test systems, but the test performed using non-dominant hand showed a significantly decreased performance in System 3 compared to Systems 1, 2 and 4. The DA test and the FTS_D did not seem to be influenced by the type of computers and input devices. On the other hand, for the SRT and CRT tests, and to some extent the SDS test, performance decreased on the tablet PC versus the desktop computer.

We postulated that participants who were unfamiliar with computers might show inferior performance because they would find it more difficult to adapt to the newly introduced tablet PC system. However, the performance of each computer familiarity group on the test systems showed similar patterns. For the SRT and CRT tests, all three computer familiarity groups uniformly showed decreased performance when tested with Systems 3 and 4. Moreover, reaction time for the SDS test also showed homogenous results among these groups. Therefore, computer unfamiliarity did not appear to lead to the decreased performance on the KCNT when using the tablet PC.

Accordingly, use of a tablet PC for the KCNT to evaluate performance on the SRT and CRT tests is not recommended; however, tablet PCs with an on-screen keyboard may be used to administer the DA test, and only limitedly the SDS and FTS tests.

The only difference between Systems 1 and 2 was the input method, that is, the keyboard. The results of this study showed that overall performance in four out of five subtests was slightly higher using a simplified keyboard than using a conventional one, but these differences in performance were not shown to be statistically significant for all subtests (Table 2).

Considering these results, there was no significant difference between the conventional full-key keyboard and the simplified keyboard in this study. Nevertheless, previous studies demonstrated that using a relatively complex conventional 106-key keyboard may lower examinees' performance compared to using a simplified keyboard, and therefore, use of a simplified keyboard was recommended [9, 14].

Systems 2 and 4 were a desktop computer and a tablet PC, respectively, both equipped with a simplified keyboard. The difference between these two systems was the size of the display with other conditions kept equivalent. The performance results between Systems 2 and 4 suggested that the size of the display did not influence the tests related to higher cognitive functions (i.e., DA and SDS tests) but did influence the tests related to simple and basic cognitive functions (i.e., SRT and CRT tests). Similarly, previous studies also reported that diminished stimuli dimension led to a latency in reaction time [23]. Moreover, size, contrast, and luminance of visual stimuli have been shown to be major determinants of detection threshold affecting neurobehavioral performance on computerized tests [24].

On the other hand, Kim et al. used a simplified keyboard and found a laptop and desktop computer only showed marginal differences in performance, which were not statistically significant [15]. However, we believe that only minor differences were found because there was not much difference in the size of the display: the monitor had a display with a diagonal length of 17 in. and that of the laptop computer was 15 in. Kim's study implies that, if the size of the display is similar, the platform of the KCNT system, whether a desktop or laptop computer, will not affect performance significantly.

Despite the aforementioned efforts to explain the results, it is not possible to claim with certainty that the size of stimuli was the only difference between System 2 and System 4 influencing the participants' performance, because we have not compared a desktop computer against a tablet PC with similar screen sizes. However, it is certain that performance significantly differed when using the desktop computer versus tablet PC.

Systems 3 and 4 were based on a tablet PC with the same display size but different input devices. An on-screen keyboard was implemented in System 3 and a simplified keyboard was used in System 4. To our surprise, Systems 3 and 4 did not show any performance differences for almost all parameters except for the CRT and FTS tests, similar to how Systems 1 and 2 showed similar performances (Table 3). Although the difference in mean reaction time for the CRT test between Systems 3 and 4 was only 58.5 ± 81.2 ms, it was indeed statistically significant at p < 0.05. FTS_ND also showed a difference (3.2 ± 5.9 taps, p < 0.001), whereas FTS_D did not.

It seems that change of input method does not greatly influence the results of tests involving higher-order cognitive functions that require longer reaction times, such as the DA and SDS tests. Likewise, basic tasks such as the SRT and FTS_D tests barely require examinees to scan the keyboard because tapping a spacebar or a control key is all that is needed to complete the tests. Hence, no differences were observed between an on-screen keyboard and a simplified keyboard.

The CRT test, on the other hand, demands examinees to perceive stimuli on the display, scan arrow keys on the keyboard, and give correct responses as quickly as possible. Our interpretation of the results is that the simplified keyboard with tactile feedback was superior to the on-screen keyboard in such a test. While physical keyboards offer visual-auditory-tactile feedback, on-screen keyboards only provide visual-auditory feedback. The results of this study implied that the contribution of tactile feedback to the test performance was more substantial on non-dominant hand than dominant hand for the FTS test and on the CRT test than the SRT test. Numerous previous studies reported that tactile feedback improves performance of various tasks [25–28]. The fact that a conventional 106-key keyboard and a 17-key simplified keyboard provide the same type of feedback also explains why there was no significant difference for the CRT between Systems 1 and 2.

The relatively small number of participants in the computer familiarity groups is a limitation of this study. There were only 9 participants in Group I, whereas Groups II and III had 23 and 40 participants, respectively. This was because most participants were somewhat familiar with the use of computers. With the current high level of computer literacy in the population, obtaining a large number of participants unfamiliar with computers would require a much larger number of overall participants.