1. Introduction
Ageing is associated with deterioration in cognitive function. This decline is in turn associated with difficulties in performing tasks of everyday living, leading ultimately to disability and dependence. Perseverance of cognitive abilities, through advancing age, is of great importance for individuals to sustain a good quality of life, as well as for societies and public health systems, especially in a setting of population ageing which is observed currently in most developed countries due to the increasing life expectancy. Given the variation that is observed in age related cognitive decline, a better understanding of the predictors and the determinants of cognitive performance and cognitive deterioration that is associated with ageing holds great importance, as it could set the basis for targeted interventions that could reduce the burden of dependence among older adults.
Cognitive performance is conceptualized as a set of different domains organized in a hierarchical order, with the bottom referring to basic sensory and perceptual abilities and the top referring to more complex functional abilities [
1]. These different domains are not independent from one another as lower order functions enable the individual to receive, understand and remember information which will in turn be utilized through higher order functions of analysis and eventually synthesis. However, when examining different domains of cognition separately, there is variation in how these domains change with advancing age [
2]. Generally, abilities such as executive function, memory and processing speed are reported to decline from midlife, while general knowledge remains more stable through ageing [
3,
4].
Processing speed has been described as an inherent ability similar to the clock speed of a computer in the sense that it dictates how long it will take for any cognitive task to be completed [
5]. Consequently, processing speed is regarded a fundamental part of the cognitive system and it has been suggested that it is the reduction of this speed that mainly contributes to the impairment of cognitive functioning that is associated with age [
6]. In fact, processing speed tends to be the strongest predictor of overall cognitive performance, loading highest in single factor solutions of cognitive ability [
1]. However, little is known regarding the trajectories of processing speed performance through aging and their determinants, with a recent study by Bott et al. [
7] associating a more stable course with genetic factors, lower inflammation and lifestyle characteristics such as physical activity.
The aim of this study is to identify latent groups of individuals with similar trajectories in processing speed as measured in the English Longitudinal Study of Aging (ELSA) dataset as well as to evaluate potential time-stable and time-varying determinants of each trajectory. Given the large sample and high quality of ELSA along with the significance of processing speed as a potential predictor of cognitive function in general, this analysis might provide valuable insight in the course of this specific cognitive domain through aging as well as age-related cognitive impairment in general.
4. Discussion
The present study identified four latent groups of processing speed scores trajectories in our sample. All latent groups presented a slight declining trend over the years. The trajectory with the steepest decreasing trend was the “High” one. This might be mainly attributed to the fact that it is the trajectory that consists of fewer individuals (4.3% of the total sample), so a more rapid decline in some participants will have a greater impact in the overall performance of the group. Previous studies investigating latent classes of cognitive performance assessed with various neuropsychological tests have also identified four or three trajectories [
14,
15,
16] with similar shapes. The slightly less decreasing and more stable trend observed in the trajectories in the present study might be explained by the shorter duration of observation of our study, as cognitive decline is a slow process that might require a longer period to be reflected or the different course of processing speed compared to other cognitive domains. However, what is common in all these studies, including ours, is that the trajectory lines do not mix and higher performance at baseline predicts higher performance also at the end of study, depicting that cognitive performance with advanced age is at some degree dependent on the mental capacity that an individual has at their prime.
Level of education was found to play an important role in the probability of belonging to a favorable trajectory. In particular, medium level of education was a predictor of greater likelihood to belong to all three favorable trajectories, compared to the low level that was set as reference, with a p-value that indicates statistical significance (<0.001 in all three groups). What is interesting is that the OR, that reflects the magnitude of this association, presented an increasing trend from the lowest to the highest trajectory (1.88 in the Low/Stable, 3.48 Middle/Stable, 4.04 in the High). The same trend was observed in the effect of the High level of education, differing but at an even greater scale. The OR of the High level of education versus the low that was set as reference was 2.267, 8.87, 12.39 in the Low/Stable, Middle/Stable and High group respectively. Conclusively, based on these findings, it can be hypothesized that education is a strong predictor of predictor of better cognitive performance and the magnitude of this association increases not only with the increasing level of education but also with the higher cognitive performance. These findings are in line with a recent study be Ferraro et al. [
17] where higher number of years of education was associated with better physical, functional and cognitive performance. Education is consistently reported a protective factor for cognitive impairment and dementia [
18,
19]. The findings of the current study confirm this beneficial effect and highlight the importance of education as a potential modifiable factor that can reduce the burden of cognitive impairment across the ageing population.
Marital status also presented a consistently significant association with our outcome of interest. The categories “Married” and “Divorced or Separated” presented a significantly higher probability of belonging to all three favorable trajectories (except for “High” group with a p value of 0.075 and 0.055 respectively) compared to the “Never Married” category that was set as reference. On the contrary, the category “Widowed” was associated with lower probability of belonging to any favorable group compared to the reference category, which was only significant in the “High” group. Previous studies have also reported the association between widowhood and greater risk for cognitive decline [
20] as well as a greater risk for cognitive impairment in single adults compared to those cohabiting with a partner [
21]. The findings can be explained by viewing marital status’s effect as part of the overall effect that social life and social isolation have on the cognitive function of older adults [
22]. The exact mechanisms behind these effects are not clear, but since this a real-world study any consistent association is worth highlighting as it might set the basis for future research.
The level of self-reported physical activity also presented an interesting association with the trajectories of processing speed in our cohort. “Mild” level of physical activity presented a consistently significant association with greater probability of belonging to any of the favorable trajectories compared to the “no physical activity” that was set as reference. On the contrary, the association between the categories of “Moderate” and “Vigorous” physical activity was just as consistent but inverse, interpreted as less probability of belonging to any of the three favorable groups for both these categories. Physical activity is mainly regarded as a protective factor for the cognitive function of older adults [
23]. However, there is no clear consensus regarding its characteristics (type, frequency, duration, intensity) that seem to provide the most beneficial results. In fact, it has been suggested that high intensity exercise might even worsen cognitive performance [
24]. From that view, the findings of our study are partly in line with published literature, however based on these alone no further implications can be made.
Finally, level of household wealth also presented a consistent association, as higher level of wealth was a significant predictor of belonging to all three favorable trajectories. This does not come as a surprise as higher socioeconomic status has been associated with better cognitive function in older adults [
25] and this is confirmed in the present study.
Regarding time-varying covariates, incidence of cardiovascular events presented a statistically significant negative correlation with processing speed scores across all groups except for the “Low/Decline” where it presented a marginal p-value of 0.094. There is significant evidence in published literature linking cardiovascular fitness and cognitive performance [
26]. Aside from vascular dementia where there is the obvious association of cardiovascular health with brain perfusion and cognitive performance, it has been shown that managing cardiovascular risk factors might decrease the risk of other forms of dementia such as Alzheimer’s disease [
27]. Although the incidence of diabetes mellitus follows the same principles with the rest of cardiovascular risk factors regarding its association with cognitive function, it seems to hold an exquisite importance [
28]. This was reflected in the current study too, where incidence of diabetes mellitus was significantly associated with lower processing speed scores across all trajectories, except for the “High”, where the same association was close to statistical significance with a p-value of 0.062. These findings are in line with those of the study of Marseglia et al. [
29] where in a sample of 793 adults over 50 years of age, those with diabetes presented a steeper decline in perceptual speed. Conclusively, based on our findings in addition to what is already known, management of cardiovascular risk factors and especially of diabetes mellitus is an essential target for interventions in order to decrease the burden of cognitive decline in older adults.
Presence of depressive symptoms showed a negative correlation with processing speed scores across all trajectories, which was statistically significant in the two lower groups. Cognitive impairment is mentioned as core feature of depression and not an epiphenomenon [
30]. Moreover, it is quite common depressive symptoms (such as memory problems) to be mistakenly attributed to cognitive impairment. Although one cannot establish a clear relationship of cause and effect between cognitive decline and depression, our study confirms the association between these two conditions. The lack of a consistently statistically significant relationship in more trajectories may be partly due to the fact that presence of depressive symptoms was self-reported and not accurately measured with specific neuropsychological scales. Nevertheless, when thinking of targeted interventions, depression should be considered as a potential modifiable factor. Finally, self-reported sleep disturbances did not present a consistent relationship with processing speed scores.
The results of this study should be viewed in the light of its limitations. Firstly, the participants were all living in England, weakening the generalizability of the findings in populations from other countries or continents. Moreover, information on certain variables, such as depressive symptoms, was self-reported and greater detail regarding the duration of these symptoms was not available. Additionally, all participants were included in the analysis. This might result in a degree of heterogeneity, as some of the participants might have already presented mild cognitive impairment or dementia at the beginning of the study, differing significantly in terms of cognitive decline over the years from the rest of the participants. However, despite its limitations this study was based on real world data from a big sample of high-quality longitudinal study and the analyses were made utilizing unbiased techniques. Moreover, this is the first analysis to our knowledge investigating the trajectories of processing speed using the GBTM. Taking these into account, our findings provide useful insight in the course of the processing speed through aging, which in addition to similar analyses of different cognitive domains [
9], strengthen our knowledge about the epidemiology of age-related cognitive decline in general.