Related Experiment Video
Updated: Oct 30, 2025
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Is computerized screening for processing speed impairment sufficient for identifying MS-related cognitive impairment
Rachel Galioto1, Gabrielle Macaron2, John W Lace3
1Mellen Center for Multiple Sclerosis Treatment and Research, Neurological Institute, Cleveland Clinic, OH, United States; Department of Neurology, Neurological Institute, Cleveland Clinic, OH, United States.
Background:
Annual screening for processing speed impairment (PSI) is recommended for patients with multiple sclerosis (pwMS). However, cognitive deficits in pwMS are heterogeneous, and whether PSI screening identifies patients with impairment in other cognitive domains is unclear. The objective of this study was to examine sensitivity and specificity of the self-administered, computerized Processing Speed Test (PST) in identifying cognitive impairment defined by a comprehensive neuropsychological battery (NPT).
Methods:
Ninety-one pwMS completed PST and NPT, with raw scores demographically adjusted. Cognitive impairment on NPT was defined as performance <5th percentile in at least one domain. Receiver operating characteristic (ROC) analyses were performed to determine the ability of the PST to discriminate between cognitively normal (CN) and cognitively impaired with PSI (CI-PSI) and cognitively impaired with normal processing speed (CI-PSN) groups.
Results:
Cognitive impairment was observed in 23.1% of pwMS on PST and in 42.9% on NPT. PST demonstrated excellent ability to discriminate between CN (57.1%) and CI-PSI (20.9%) groups (Area Under the Curve [AUC] = 0.86, p < 0.001). In contrast, PST was unable to discriminate CN and CI-PSN (22.0%) groups (AUC = 0.42, p = 0.32).
Conclusion:
The PST demonstrates excellent ability to detect PSI in pwMS but is unable to identify cognitively impaired pwMS without PSI, highlighting the importance of developing additional screening measures.
