A-071 Latent Classes of Cognitively Normal Individuals have Unique Relationships between Demographic and Neuropsychological Variables
| Autor: | Rossetti H, Carlew A, Schaffert J, Lacritz L, Goette W |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Mini–Mental State Examination
medicine.diagnostic_test Homogeneity (statistics) Trail Making Test Neuropsychology General Medicine Speech fluency Verbal learning Psychiatry and Mental health Clinical Psychology Neuropsychology and Physiological Psychology Boston Naming Test Linear regression medicine Psychology Clinical psychology |
| Zdroj: | Archives of Clinical Neuropsychology. 35:862-863 |
| ISSN: | 1873-5843 |
| DOI: | 10.1093/arclin/acaa068.071 |
| Popis: | Objective Determine whether clinically normal (CN) individuals represent a single homogeneous group prior to normative adjustment. Method Data from 1,055 CN participants (Mage = 68.0, SD = 8.68; Meducation = 14.9, SD = 2.90; white = 92.7%) from the Texas Alzheimer’s Research and Care Consortium were used. Participants had no recorded neurological, cognitive, or psychiatric diagnoses. Raw scores from the AMNART, Animal Fluency, Boston Naming Test (BNT), CERAD verbal learning test, CLOX1 and CLOX2, MMSE, and Trail Making Test (TMTA and B) were examined with finite mixtures of general linear regression models using age, education, race, and gender as predictors. Each test was modeled with up to 10 latent classes with the Bayesian Information Criterion used to select best fit. Results Animal Fluency, CLOX2, and TMT A errors were best fit by 1 underlying group. The remaining tests required 2 (CERAD, CLOX1, MMSE, and TMT-B errors), 3 (BNT and TMT-A), and 5 (AMNART and TMT-B) latent classes. Generally, latent classes for tests differed in coefficients for race, gender, and intercepts, though results differed from test-to-test (Supporting Figure). Conclusions Latent classes of CN individuals were identified for which the predictive power of certain demographic variables differed depending on the latent class. Further research is needed to identify who may belong to distinct latent classes so the appropriate regression-based norms are used. Different latent class coefficients for race and gender suggest heterogeneity within these variables that can be addressed to produce more accurate models. Findings suggest that regression-based norms could be improved by identifying latent classes and finding ways of predicting who belongs to which latent class. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|