Validation of the automated neuropsychological assessment metrics for assessing cognitive impairment in systemic lupus erythematosus. (January 2022)

• Record Type: Journal Article
• Authors: Yuen, Kimberley; Beaton, Dorcas; Bingham, Kathleen; Katz, Patricia; Su, Jiandong; Diaz Martinez, Juan Pablo; Tartaglia, Maria Carmela; Ruttan, Lesley; Wither, Joan E.; Kakvan, Mahta; Anderson, Nicole; Bonilla, Dennisse; Choi, May Y.; Fritzler, Marvin J.; Green, Robin; Touma, Zahi
• Is Part Of: Lupus. Volume 31:Number 1(2022)
• Page Start: 45
• Page End: 54
• Publication Date: 2022-01
• Subjects: systemic lupus erythematosus -- cognitive impairment -- CART analysis -- ANAM -- lupus -- screening; Systemic lupus erythematosus -- Periodicals; 616.772005
• DOI: 10.1177/09612033211062530
• Languages: English
• ISSNs: 0961-2033

Abstract

Objective: We previously demonstrated the utility of the Automated Neuropsychological Assessment Metrics (ANAM) for screening cognitive impairment (CI) in patients with systemic lupus erythematosus (SLE) and developed composite indices for interpreting ANAM results. Our objectives here were to provide further support for the ANAM's concurrent criterion validity against the American College of Rheumatology neuropsychological battery (ACR-NB), identify the most discriminatory subtests and scores of the ANAM for predicting CI, and provide a new approach to interpret ANAM results using Classification and Regression Tree (CART) analysis. Methods: 300 adult SLE patients completed an adapted ACR-NB and ANAM on the same day. As per objectives, six models were built using combinations of ANAM subtests and scores and submitted to CART analysis. Area under the curve (AUC) was calculated to evaluate the ANAM's criterion validity compared to the adapted ACR-NB; the most discriminatory ANAM subtests and scores in each model were selected, and performance of models with the highest AUCs were compared to our previous composite indices; decision trees were generated for models with the highest AUCs. Results: Two models had excellent AUCs of 86 and 89%. Eight most discriminatory ANAM subtests and scores were identified. Both models demonstrated higher AUCs against our previous composite indices. An adapted decision tree was created to simplify the interpretation of ANAM results. Conclusion: We provide further validity evidence for the ANAM as a valid CI screening tool in SLE. The decision tree improves interpretation of ANAM results, enhancing clinical utility.