To the best of our knowledge, this is the first study to examine the test-retest reliability and practice effect of the TVPS-4 in individuals with schizophrenia. Our findings demonstrated excellent test-retest reliability and trivial practice effect for the overall scale. The TVPS-4 subscales revealed moderate to good test-retest reliability and trivial to small practice effects. These findings are critical for interpreting visual perception using the TVPS-4 in individuals with schizophrenia.
The overall scale of the TVPS-4 displayed excellent test-retest reliability (ICC > 0.90), which is suitable for use in clinical settings [19]. The subscales exhibited sufficient test-retest reliability, except for the visual memory subscale (ICC < 0.60). The relatively lower ICC value for the visual memory subscale was similar to that of the TVPS-3 in individuals with stroke (ICC = 0.53) [14]. A possible reason for the moderate test-retest reliability of the visual memory subscale is the time constraint during testing. In the visual memory subscale, participants are required to memorize a target within 5 seconds and then identify the target within 20 seconds. Individuals with schizophrenia show deficits in visual perception and require more time for visual processing [22, 23]. Further studies are warranted to evaluate whether not timing or giving more time for the visual memory subscale could enhance the test-retest reliability in individuals with schizophrenia. Four subscales of the TVPS-4 (form constancy, sequential memory, visual figure-ground, and visual closure) showed much more reliable ICC values (ICC = 0.70–0.75) compared to those of the TVPS-3 (ICC = 0.55–0.77) [24]. The overall scale and the other three subscales displayed similar ICC values. As a whole, the newest edition demonstrates better test-retest reliability.
In this study, the MDC values at the 95%, 90%, and 80% confidence levels in the overall scale and subscales of the TVPS-4 were provided. These values can assist clinicians and researchers when interpreting score changes (improvement or deterioration) in visual perception more reasonably. For example, a score change for an individual with schizophrenia exceeding the MDC95 (16.3) of the overall scale can be explained as a real change with 95% certainty. Substantial random measurement errors in two subscales (i.e., visual discrimination and form constancy, MDC% > 30%) were observed and the other subscales had MDC% of 26.1%-29.7%, which was close to 30%. The reason behind the substantial random measurement errors may be due to the long administration time which could result in fatigue and lead to a tendency of guessing answers. In this study, participants took 30–40 minutes to complete the seven subscales of the TVPS-4. The MDC% value was estimated at the 95% confidence level, which is more robust. People with schizophrenia may not achieve a real change with 95% certainty in the subscales. The MDC values at the other confidence levels (i.e., 90% and 80%) were estimated in this study. Future users could consider which confidence level they would like to assert to explain the score change for an individual with schizophrenia over two assessments.
Small effect sizes were observed, and the paired t-tests revealed significant differences between the two assessments in three subscales: visual discrimination, visual memory, and visual figure-ground. These findings suggest the presence of small practice effects in these three subscales. The results indicate that participants in this study may have easily memorized the answers or developed strategies to identify them. Two possible methods to reduce the practice effects are as follows: (1) repeating similar items more times to reach a plateau phase in the practice effect [25]; and (2) developing a computerized adaptive test, which can decrease the need to administer the same item in repeated assessments. Moreover, a computerized adaptive test can shorten the administration time which may improve the test-retest reliability of the TVPS-4. Future studies could consider developing a computerized adaptive test of visual perception to assess people with schizophrenia in a more precise manner.
This study had two limitations. First, we used a convenience sample recruited from a psychiatric center in northern Taiwan, which may restrict the generalizability of our results. Second, we did not examine the validity (e.g., construct validity with larger sample size), limiting the explanation of the visual perceptual constructs of the TVPS-4. Further validation is needed to confirm the construct validity of the TVPS-4 in people with schizophrenia.