Cognitive lifestyle jointly predicts longitudinal cognitive decline and mortality risk
Cognitive lifestyle measures such as education, occupation, and social engagement are commonly associated with late-life cognitive ability although their associations with cognitive decline tend to be mixed. However, longitudinal analyses of cognition rarely account for death and dropout, measurement error of the cognitive phenotype, and differing trajectories for different population sub-groups. This paper applies a joint latent class mixed model (and a multi-state model in a sensitivity analysis) that accounts for these issues to a large (n = 3,653), population-based cohort, Paquid, to model the relationship between cognitive lifestyle and cognitive decline. Cognition was assessed over a 20-year period using the Mini-Mental State Examination. Three cognitive lifestyle variables were assessed: education, mid-life occupation, and late-life social engagement. The analysis identified four latent sub-populations with class-specific longitudinal cognitive decline and mortality risk. Irrespective of the cognitive trajectory, increased social engagement was associated with a decreased mortality risk. High education was associated with the most favourable cognitive trajectory, and after adjusting for cognitive decline, with an increased mortality risk. Mid-life occupational complexity was also associated with more favourable trajectories but not with mortality risk. To realistically examine the link between cognitive lifestyle and cognitive decline, complex statistical models are required. This paper applies and compares in a sensitivity analysis two such models, and shows education to be linked to a compression of cognitive morbidity irrespective of cognitive trajectory. Furthermore, a potentially modifiable variable, late-life social engagement is associated with a decreased mortality risk in all of the population sub-groups.
KeywordsCognitive lifestyle Mortality Cognitive decline Longitudinal modelling
There is increasing evidence to relate measures of cognitive lifestyle, such as education, occupation, and social engagement to late-life cognitive ability and incident dementia in later life [1, 2, 3, 4]. Understanding the relationship between these potentially modifiable variables and longitudinal population-based cognitive trajectories is vital as even a modest delay to the onset of severe cognitive impairment and dementia could have a tremendous public health impact . Studies have shown that cognitive lifestyle variables in tandem, but not individually, are linked to a decrease risk of incident dementia  via multiple possible pathways . Basic measures of education, occupation, and social engagement have also been linked to a compression of cognitive morbidity in longitudinal analyses .
However, there are still some mixed findings, particularly for models that examine education and cognitive change [8, 9, 10, 11, 12], which may stem from methodological issues. For example, it is rare for longitudinal analyses of cognitive decline in the older population to account for death and dropout, measurement error of the cognitive phenotype, ceiling and floor effects in the cognitive test , and the possibility of cognitive recovery, especially from the mild cognitive impairment state . Furthermore, modelling a single trajectory for a large population will not account for heterogeneous sub-groups with different rates of both initial ability and decline . There may also be differing effects of covariates on these different sub-populations. Other factors to consider include cohort effects when a population age-range spans several years , and practice effects when the same cognitive tests are administered over multiple waves .
The application of models that account for these issues will enable a more realistic assessment of the disease process and better identify the true effects of cognitive lifestyle on cognitive decline. Two statistical approaches that can help do this are joint latent class models  and multi-state models . Previously, a longitudinal multi-state analysis of 13,004 subjects from the Cognitive Function and Ageing Study (CFAS) found cognitive lifestyle factors to be linked to a compression of cognitive morbidity . However, this analysis did not account for population heterogeneity or missing cognitive data. It also required the definition of cognitive states, which both reduces power to detect subtle cognitive changes and limits its application to outcomes that can be translated into clinically meaningful states. In contrast, a joint latent class mixed model uses all of the information from the quantitative cognitive score(s) and can model trajectories and co-occurring events, such as death, in multiple sub-populations. However, it is unable to explicitly model cognitive recovery.
The aim of this paper was to assess (1) the relationship between cognitive lifestyle and late-life cognitive decline accounting for death, and (2) the relationship between cognitive lifestyle and death adjusting for cognitive trajectories in the French cohort, Paquid using a joint latent class mixed model. As a sensitivity analysis and for methodological comparison, a multi-state model was also examined.
Materials and methods
Data stem from the Paquid cohort, which is a longitudinal study of ageing with up to 20 years of follow-up . Recruitment of 3,777 participants occurred across 75 civil parishes of the Gironde and Dordogne regions of south west France. Subjects were aged 65 years and above, residing at home at the study baseline in 1988. There have been up to nine subsequent waves of data collected on each individual at 1, 3, 5, 8, 10, 13, 15, 17, and 20 years after the baseline assessment. At the first follow-up, only subjects from Gironde were interviewed.
At each wave of Paquid, a multi-test cognitive battery was administered to the participants by a trained neuropsychologist. For this analysis we only consider scores from one of the tests, the Mini-Mental State Examination (MMSE) . The MMSE is a test of global cognitive function that is brief to administer and has a scoring range of 0–30. The MMSE scores were treated as continuous data for the joint latent class mixed model analysis. For the multi-state model of cognitive change, MMSE categories were specified as follows: no impairment was defined as a score between 27 and 30, slight impairment for scores between 23 and 26, and moderate/severe impairment for scores below 23 . An MMSE score between 23 and 26 has been shown to be as effective at predicting conversion to dementia as other more detailed clinical measures of mild cognitive impairment .
Measurement of cognitive lifestyle
Three baseline measurements of cognitive lifestyle were included in this analysis: early-life education, mid-life occupation, and late-life social engagement. Education was split into three groups: no education or a non-validated primary school degree; a validated primary degree up to a non-validated secondary degree; and a validated secondary degree or higher. Mid-life occupation was dichotomised into intellectual (craftsmen and shopkeepers, policemen, nurses and white collar workers, and professional workers) versus non-intellectual type work (housewives, farm workers, farm managers, domestic service employees, and blue collar workers). Four binary response questions were combined to create a scale for late-life social engagement. They focussed on: membership of a group or association; visits from family and friends; membership of a golden age club; and membership of another club. A three category response was created after summing the scores from all four variables: low social engagement (sum score 0); medium social engagement (sum score 1); and high social engagement (sum score 2, 3, or 4). Although a more detailed list of social engagement questions was available for analysis, those chosen are analogous to the questions used in CFAS where such a measure was linked to cognitive decline .
Description of the Paquid cohort
Total (n = 3,653)
Age (mean, s.d.)
Sex (n, %)
Education (n, %)
Mid-life occupation (n, %)
Late-life social engagement (n, %)
MMSE group (n, %)
No impairment (27–30)
Slight impairment (23–26)
Moderate/Severe impairment (0–22)
Joint latent class mixed model output for the association between cognitive lifestyle and cognitive decline and mortality
Immediate decliners (1)
Slow decliners (2)
Low baseline cognition (3)
High baseline cognition (4)
Number assigned to each class
Class membership probability—OR (95 % CI)
Sex (female vs. male)
Education (medium vs. low)
Education (high vs. low)
Occupation (intellectual vs. non-intellectual)
Social engagement (medium vs. low)
Social engagement (high vs. low)
Risk of death—HR (95 % CI)
Sex (female vs. male)
Education (medium|high vs. low)
Occupation (non-intellectual vs. intellectual)
Social engagement (medium|high vs. low)
Estimates for the survival sub-model are also shown in Table 2 and illustrated in Fig. 3b. The high baseline cognition group had the highest survival probabilities whilst the immediate and slow decliners had the greatest mortality risk: 11.9 (95 % CI 6.4, 22.3) and 7.0 (95 % CI 3.5, 14.0) times greater than the high baseline cognition group. The class-specific hazard ratios (HRs) for survival showed an increased mortality risk for those with a higher level of education, although this was only statistically significant in the high baseline cognition group: HR 1.4 (95 % CI 1.1, 1.8). After adjustment, there was no association between occupation and survival but being socially engaged was associated with a decreased mortality risk in all except the immediate decliners (HR range 0.7–0.8). Simplification of the model to allow class-common effects of education and social engagement on survival resulted in an increased mortality risk for those with medium and high education—HRs 1.3 (95 % CI 1.1, 1.4) and 1.2 (95 % CI 1.0, 1.4), respectively—and a decreased mortality risk for those with medium or high social engagement—HRs 0.8 (95 % CI 0.7, 0.9) and 0.7 (95 % CI 0.6, 0.8), respectively.
Hazard Ratios and 95 % CIs for education, mid-life occupation and late-life social engagement on late-life cognitive change
Education (med vs. low)
Education (high vs. low)
Occupation (int vs. non-int)
Social engagement (med vs. low)
Social engagement (high vs. low)
State 1—State 2
State 2—State 1
State 2—State 3
Two additional sensitivity analyses examined potential cohort effects and lagged effects of social engagement. The former included baseline age as a covariate on each state transition. This had no effect on the associations for the cognitive lifestyle covariates (results not shown). Similarly, a model that considered cognitive decline and mortality 5 years after the baseline assessment found similar protective effects of social engagement on transitions to death (results not shown).
This study applied two statistical models to a large, population-based cohort with up to 20 years of follow-up. They found late-life social engagement to be strongly associated with a decreased mortality risk irrespective of cognitive decline profile or cognitive state (no, slight, or moderate/severe impairment). By contrast, high education was associated with an increased mortality risk irrespective of cognitive trajectory. It was also associated with the most favourable cognitive trajectory in the mixed model, and with a reduced risk of cognitive decline in the multi-state model. In both models, mid-life occupation tended to be associated with more favourable cognitive change but not with mortality.
These findings are compatible with a cognitive reserve hypothesis . We found a compression of cognitive morbidity in participants who had high levels of education, a non-manual occupation in mid-life, and an active social engagement status in late-life. This could be due to neurocompensation i.e., the provision of coping strategies that alleviate the impact of underlying damage until this becomes overwhelming, resulting in an accelerated terminal decline.
The multi-state modelling results for education and occupation closely resemble those reported previously in a UK-based cohort . However, there was no overlap in the findings for social engagement. Possible reasons for this include sampling differences: at baseline CFAS included subjects living in both the community and in institutions and had an 81 % response rate, Paquid included only community-dwelling subjects and had a 68 % response rate. The use of a single measure of social engagement at baseline is a limitation in both studies.
Comparison of the statistical models
While the interpretation of the results for the multi-state model and the joint latent class mixed model are similar, relative strengths and weaknesses of the approaches can still be discussed. The main benefits of the joint latent class mixed model are three-fold. Firstly is its ability to use all available information from quantitative response variables, which avoids the need to generate cut-points and to define states. Moreover, extension of the latent class model can allow the combination of information from multiple tests to create a general cognitive factor that evolves over time . Secondly the model accounts for population heterogeneity by identifying latent classes and display underlying trajectories of cognitive decline. However, individuals are not allowed by the model to change latent class. Thirdly, the joint latent class model is particularly robust to missing data. It explicitly deals with death, and missingness due to dementia dropout is handled via the latent cognitive trajectories. An advantage of using either method for longitudinal analyses of MMSE scores is that they overcome curvilinearity via cognitive state definitions or a nonlinear transformation of the MMSE scores. Both methods are also able to model death. Despite commonly being used in longitudinal models of cognitive ageing [8, 11, 12], a potential limitation is our use of the MMSE. It is typically used as a screening tool for dementia prediction, which provides rationale for its use in the multi-state model. Furthermore, while it is insensitive to change at the ceiling and floor of the distribution , our mixed modelling approach was able to account for this.
Limitations of the joint latent class mixed model analysis include a less intuitive interpretation of the results compared to the multi-state model, which yields a single set of HRs for each covariate. A unique feature of the multi-state model was the explicit modelling of cognitive recovery from slight impairment back to no impairment. An alternative approach to the joint latent class model could be a two-stage approach where a latent class mixed model would be estimated first using only repeated measures of MMSE to identify the subpopulations and then a survival model would be estimated in each posterior class. However, the joint model is a better option whenever possible because it accounts for truncation by death when estimating the trajectories and because variances of the regression parameters in the survival model are better estimated by taking into account the uncertainty of the estimates of the latent class mixed models.
Both multi-state models and joint latent class mixed models have flexible properties that enable a realistic modelling of longitudinal cognitive decline although subtle differences exist in terms of interpreting output. For future analyses the most appropriate model will depend on the hypothesis to be tested, the cognitive test under study and the study design. If the objective is to study cognitive recovery or life expectancies, and if clinically meaningful cognitive states can be defined then the multi-state model is ideal. On the other hand, a joint latent class mixed model, which uses all of the information provided by quantitative scores, is ideal when change in the dependent variable is likely to be small, when cognitive states are difficult to define, or when multiple cognitive outcomes are being assessed simultaneously.
There are two main epidemiological findings from this analysis. Firstly, social engagement was associated with a decreased mortality risk in Paquid. While causality cannot be implied from this analysis, a model that considered cognitive decline and mortality 5 years after the assessment of social engagement found similar effects. Hence, even if it is only a marker of general health and well-being, social engagement may be a useful predictor of mortality risk. Secondly, there is evidence to link education to a decreased risk of cognitive decline but an increased mortality risk from severe cognitive impairment. These findings were also observed in CFAS , with the latter finding also replicated in the US Health and Retirement Study . The former finding is in contrast with some but not all previous analyses where education has been linked to initial cognitive level but not the rate of decline [12, 29, 30]. However, there are major methodological problems related to the analysis of cognitive decline of the older population, such as competing risks for mortality, missing data, unequal sensitivity to change (with floor and ceiling effects), cognitive recovery, and measurement error. The application of methods that account for these issues is vital in order to provide a more valid and accurate assessment of cognitive ageing.
Methodological developments for this work were partly funded by the ANR-funded project MOBIDYQ; Grant No.: 2010 PRSP 006 01. PAQUID study was funded by NOVARTIS pharma, SCOR insurance, Agrica, Conseil régional d’Aquitaine, Conseil Général de la Gironde and Conseil Général de la Dordogne. Fiona Matthews is funded under MRC programme grant UC_US_A030_0031. Riccardo Marioni is an Alzheimer’s Research UK Fellow (ART-RF2010-2) and a member of the University of Edinburgh Centre for Cognitive Ageing and Cognitive Epidemiology, part of the cross council Lifelong Health and Wellbeing Initiative (G0700704/84698), for which funding from the Biotechnology and Biological Sciences Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, and Medical Research Council is gratefully acknowledged.
- 24.Proust-Lima, C, Séne, M, Taylor, JM, Jacqmin-Gadda, H. Joint latent class models for longitudinal and time-to-event data: A review. Stat Methods Med Res. 2014;23(1):74–90Google Scholar
- 25.R Development Core Team: R: A Language and Environment for Statistical Computing R Foundation for Statistical Computing. Vienna, Austria; 2010 (ISBN) 3-900051-07-0 http://www.R-project.org.
- 26.Jackson CH. Multi-state models for panel data: the msm package for r. J Stat Softw. 2011;38:8.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.