This study identified trajectories in both physical and cognitive functioning among Dutch older adults aged 75 and older. Using the innovative methodology of Group-Based Trajectory Modelling, modelling trajectories jointly with mortality, we were able to estimate more precise group sizes. We identified five trajectories in functional limitations and four trajectories in cognitive decline. A considerable proportion of the Dutch 75 + experienced high levels of functioning over the course of three years. For physical functioning, 15% of the sample experienced continuous high levels of physical functioning, and 31% of the sample experienced high moderate physical functioning. As for cognitive functioning, 30% of the sample experienced high cognitive functioning and 43% experienced moderately high cognitive functioning. But, adverse trajectories were present as well. For the physical functioning trajectories, 26% of the participants experienced moderate decline and 6% experienced steep decline followed by slight recovery. For cognitive functioning 16% experienced rapid cognitive decline. The most adverse trajectories showed continuous low physical functioning with at least 2 severe ADL-limitations (22%), and continuous low cognitive functioning (11%) with probable dementia. These trajectories had high mortality levels (~ 14%). The declining and low functioning trajectories are the trajectories where the requirement for care is probably highest.
Despite our study using shorter time intervals, examining older participants, and incorporating mortality risk, the trajectories seem to reflect patterns that were also identified in previous studies on trajectories of functioning in old age. Among populations of the same age group similar trajectories were identified (10, 13). Whereas our relatively old study sample resulted in a low trajectory for cognition, that is not identified among younger study samples (7, 8), but is also identified among older study samples (39, 40). It can be concluded that our study corroborates that there is considerable diversity in health trajectories among the 75-plus.
Taking mortality into account resulted in bigger group sizes for the more adverse trajectories, while it led to smaller group sizes for the more favourable trajectories, which is in line with what could be expected based on the studies conducted by Haviland et al. (23) and Zimmer et al. (41), who used the same methodology. However, although we expected that modelling mortality would result in bigger group sizes for the trajectories that showed decline, this was only the case for cognitive decline, but not modelling mortality only resulted in a very slight underestimation of the declining functional limitation trajectories. This is in line with previous studies reporting very low mortality probabilities for people with increasing functional limitations (18).
The second aim of our study was to explore how the trajectories varied for several background variables (gender, age, level of education and partner status) and types of diseases. What is clear from these results is that the persons following the three most favourable trajectories (with either high or high moderate levels of physical functioning, or high levels of cognitive functioning) had rather favourable characteristics. They were younger, middle or high educated, lived independently, and did not have diabetes or heart- and lung disease.
Yet, there was no common denominator between the people following the three declining trajectories. Older age, having cancer, or rheumatic disease increased the probability of the steady declining physical functioning trajectory, which is understandable as these chronic diseases in more advanced stages limit mobility. Older age also increased the chance of cognitive decline, as did having a low education, or diabetes. The link with diabetes can be explained by the adverse effects of hyperglycaemia, inflammatory cytokines, and neuropathic processes (42). Having suffered a stroke reduced the chance of steep decline in physical functioning, but increased the chance of continuous functional limitations. This shows the severe debilitating effects of CVA, since it reduces the level of functioning in such a severe way that the chance of following a trajectory that starts with high functioning is rather low (18). The finding that CVA is only (negatively) associated with high cognitive functioning is in line with previous findings that CVA does not necessarily lead to dementia, but reduces cognitive functioning, thereby resulting in mild cognitive impairment for most (43, 44). People experiencing mild cognitive impairment are likely to be found in all other three cognition trajectories (45), which probably leads to the absence of other associations.
While rheumatic diseases increased the chance of gradually declining physical functioning, a finding also reported by for example Botes et al. (46), they also increased the probability of high levels of cognitive functioning. This positive relation between rheumatic diseases and cognition has been widely studied, and despite the growing body of evidence suggesting that aspirin does not have a protective effect on cognition (47, 48), studies do indicate that non-steroidal anti-inflammatory drugs (NSAIDs) decrease the risk of cognitive decline (49).
As expected based on previous studies, living in an institution, having diabetes and/or CVA were significantly associated with the two trajectories of poor functioning: with severe functional limitations, and with severe cognitive problems (45, 46, 50). In addition, people who suffered from heart- and lung disease were more likely to have continuous cognitive dysfunction. Also, being older or low educated increased the probability of having a continuously high number of functional limitations, but not of low cognitive functioning. This might be due to the inclusion of ‘being institutionalized’ as an explanatory variable. This variable might have led to a stronger attenuation of the associations of age and education for severe cognitive decline than for severe functional limitations, since severe cognitive decline leads to institutionalization more often than severe functional limitations do. All in all, these trajectories seem to contain persons that experienced the deleterious effects of chronic diseases, and about half of them had to be taken into residential care due to the resulting limitations.
Associations for sex were not present. Although previous studies stratified by sex a priori (7, 41), stratifying by sex would have reduced our statistical power substantially due to our small sample. However, analysis stratified by sex showed comparable trajectories for men and women (see figures S2 and S3 in additional file 2). The absence of sex differences might be explained by the finding that these differences are most pronounced in the level of functional impairment, while rates of change are similar for men and women (51). It could be possible that due to our shorter measurement intervals the rate of change has had a bigger impact in defining the trajectories. On the other hand, the absence of sex differences is not entirely anomalous; for functional limitations Bolano et al. (52) and Holstein et al. (53) do not report any statistically significant sex differences, and Comijs et al. (8) do not always identify sex differences for trajectories in cognition. Moreover, our analyses included mortality, various diseases, age and level of education, which are all factors that differ by sex, which may have decreased the effect of sex itself to non-significance.
Low education being associated with low levels of physical functioning is a finding also reported by Boyd et al. (54) and Kingston et al. (13). The finding that education is negatively associated with moderate or declining levels of cognitive functioning, and positively associated with high cognitive functioning, corroborates the link between education and cognition. Furthermore, it is partly in line with the MMSE being less sensitive for cognitive decline among higher educated people (55), but also in line with education having a protective effect on cognitive decline (56), and people having more cognitive capacities having pursued more education.
Associations for partner status were absent. This might have been caused by not differentiating between coresiding and noncoresiding partners. Second, it is possible that the protective effect of having a partner diminishes with age, since this usually results in the partner requiring more care as well. Lastly, studying a population that could be either institutionalized or community-dwelling might have resulted in absent associations for partner status.
Strengths and Limitations
The main strength of this study was the use of the 75PLUS LASA-data, containing a representative sample of the Dutch oldest old: the study has a high response and cooperation rate, and enabled for studying both community dwelling and institutionalized people. Accounting for attrition by jointly modelling mortality is a strength as well, enabling us to estimate the group sizes correctly. Third, defining ADL as a scale forms a strength in opposition to previous studies that compressed the range of the severity of ADL-limitations by dichotomizing ADL. Because the overall degree of limitations decides the need for care, it is precisely this degree that is of vital importance for policymakers, and by measuring ADL as a scale we were better at capturing the existence and the range of need for care that follow from functional limitations.
The first limitation of the study was not being able to conduct a multi-trajectory model to study the interconnectedness between cognitive decline and ADL-limitations that is implied by previous studies (57, 58). We instead decided to report the estimates of the trajectories separately, since jointly modelling mortality in a multi-trajectory model was not possible, and accounting for decease is necessary in a very old population. Second, although the use of proxy data allowed us to also include severely cognitively impaired respondents, this resulted in two different measurements for cognition (the sMMSE and the IQCODE) (31). Although different ways of harmonizing did not affect the trajectories much, the absence of guidelines on how to harmonize the sMMSE and IQCODE leaves some uncertainty on whether the eventual scores are an accurate reflection of cognitive functioning among our participants. Although we did not have a considerable amount of missing items for ADL or sMMSE, we are mindful of the slight overestimation of both cognitive and ADL-levels in which the imputation of these items might have resulted. On the flip side, not including these participants in the analysis would have likely resulted in an overestimation of favourable trajectories as well. Third, although our sample size was sufficient for performing analyses, analysing a bigger sample size would have allowed for performing all of the analysis stratified by sex, and would have increased statistical power.
Implications
This study has implications for policymakers in health and long term care. Despite this study showing that a considerably large group experiences little to no functional limitations and/ or cognitive decline, this study also identifies groups that, based on their low or declining levels of physical and cognitive functioning, have a high or increasing care need. The trajectories corresponding to the highest requirements of care are the two stable low trajectories (11–22%), and part of those groups are already living in residential care. This simultaneously shows how half of these people apparently have a high requirement of care, but do still live in independent housing, probably with a large demand on care from informal and formal caregivers. The declining trajectories (6%, 16%, 26%) are of most interest due to the increasing care need over time. This increase makes this group vital for policies aimed at future care planning, since they require more adjustments in care provision than the stable trajectories do. Our study does not provide one indicator to target all these groups, but shows old age, low education, diabetes, and CVA as the best indicators for targeting risk groups. Previous studies and policymakers should aim at finding indicators to identify the people that experience declines in functioning.