Development of embedded performance validity indicators in the NIH Toolbox Cognitive Battery
| Autor: | William J. Panenka, Noah D. Silverberg, Christopher A. Abeare, Laszlo A. Erdodi, Grant L. Iverson, Isabelle Messa, Douglas P. Terry |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Adult
Male Adolescent Psychometrics NIH Toolbox Test validity Neuropsychological Tests Machine learning computer.software_genre Sensitivity and Specificity Young Adult Cognition medicine Humans Cutoff Prospective Studies Neuropsychological assessment Brain Concussion Aged medicine.diagnostic_test business.industry Univariate Reproducibility of Results Middle Aged Psychiatry and Mental health Clinical Psychology Cross-Sectional Studies Symptom validity test Multivariate Analysis Female Artificial intelligence business Psychology Neurocognitive computer |
| Zdroj: | Psychological Assessment. 33:90-96 |
| ISSN: | 1939-134X 1040-3590 |
| DOI: | 10.1037/pas0000958 |
| Popis: | To assess noncredible performance on the NIH Toolbox Cognitive Battery (NIHTB-CB), we developed embedded validity indicators (EVIs). Data were collected from 98 adults (54.1% female) as part of a prospective multicenter cross-sectional study at 4 mild traumatic brain injury (mTBI) specialty clinics. Traditional EVIs and novel item-based EVIs were developed for the NIHTB-CB using the Medical Symptom Validity Test (MSVT) as criterion. The signal detection profile of individual EVIs varied greatly. Multivariate models had superior classification accuracy. Failing ≥4 traditional EVIs at the liberal cutoff or ≥3 at the conservative cutoff produced a good combination of sensitivity (.57 to .61) and specificity (.92 to .94) to MSVT. Combining the traditional and item-based EVIs improved sensitivity (.65 to .70) at comparable specificity (.91 to .95). In conclusion, newly developed EVIs within the NIHTB-CB effectively discriminated between patients who passed versus failed the MSVT. Aggregating EVIs within the same category into validity composites improved signal detection over univariate cutoffs. Item-based EVIs improved classification accuracy over that of traditional EVIs. However, the marginal gains hardly justify the burden of extra calculations. The newly introduced EVIs require cross-validation before wide-spread research or clinical application. (PsycInfo Database Record (c) 2021 APA, all rights reserved). |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|