Discriminant power of socio-demographic characteristics and mood in distinguishing cognitive performance clusters in older individuals: a cross-sectional analysis

ABSTRACT

Objectives: Identification of predictors of cognitive trajectories has been a matter of concern on aging research. For this reason, it is of relevance to infer cognitive profiles based on rapid screening variables in order to determine which individuals will be more predisposed to cognitive decline.

Method: In this work, a linear discriminant analysis (LDA) was conducted with socio-demographic variables and mood status as predictors of cognitive profiles, computed in a previous sample, based on different cognitive dimensions. Data were randomly split in two samples. Both samples were representative of the Portuguese population in terms of gender, age and education. The LDA was performed with one sample (n = 506, mean age 65.7 ± 8.98 years) and tested in the second sample (n = 548, mean age 68.5 ± 9.3 years).

Results: With these variables, we were able to achieve an overall hit rate of 65.9%, which corresponds to a significant increment in comparison to classification by chance.

Conclusion: Although not ideal, this model may serve as a relevant tool to identify cognitive profiles based on a rapid screening when few variables are available.

KEYWORDS:

• Linear discriminant analysis
• mood
• neurocognitive function
• aging

Acknowledgements

The work was funded by the European Commission (FP7): ‘SwitchBox’ (Contract HEALTH-F2-2010-259772) and co-financed by the Portuguese North Regional Operational Program (ON.2 – O Novo Norte) under the National Strategic Reference Framework (QREN), through the European Regional Development Fund (FEDER). N.C. Santos is supported by a SwitchBox post-doctoral fellowship, and P.C. Moreira and T.C. Castanho by doctoral fellowships from Fundação para a Ciência e Tecnologia (FCT, Portugal). We are thankful to all study participants. The authors would like to acknowledge Drs Pedro Cunha and Jorge Cotter, and all colleagues who assisted with participant recruitment and evaluation. The authors declare no conflicts of interest.

Disclosure Statement

The authors declare no conflicts of interest. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Notes

1. The latent classes were derived from a Bayesian latent class analysis (LCA). LCA is a model-based approach that produces a probability-based classification. It relies on the use of starting values to ensure that the best solution is achieved. In contrast with other clustering approaches, LCA does not rely on traditional modeling assumptions, such as linear relationship or normal distribution of the variables.

Additional information

Funding

European Commission (FP7): ‘SwitchBox’ [contract number HEALTH-F2-2010-259772]; Portuguese North Regional Operational Program (ON.2 – O Novo Norte) under the National Strategic Reference Framework (QREN); European Regional Development Fund (FEDER).