Socioeconomic status as a risk factor for motoric cognitive risk syndrome in a community‐dwelling population: A longitudinal observational study

Motoric cognitive risk (MCR) is a syndrome characterised by measured slow gait speed and self‐reported cognitive complaints. MCR is a high‐risk state for adverse health outcomes in older adults, particularly cognitive impairment and dementia. Previous studies have identified risk factors for MCR, but the effect of socioeconomic status has, to date, been insufficiently examined. This study explored the association between MCR and socioeconomic status, as determined by occupational social class and years of education.


INTRODUC TI ON
Dementia is a major global public health concern. Much research now focuses on identifying the early predementia stages when intervention may be most effective [1]. Slow gait speed and subjective cognitive complaints are among the earliest reported findings in the preclinical stage of dementia, occurring approximately 10 years before dementia diagnosis [2]. Motoric cognitive risk (MCR) is a predementia syndrome defined as objective (measured) slow walking speed and subjective (self-reported) cognitive complaint in the absence of significant functional impairment and dementia [3].
First defined by Verghese in 2013 [4], MCR demonstrates prognostic value as a high-risk state for developing dementia [5][6][7][8]. MCR is a quick, inexpensive, and easy-to-measure clinical construct that can reliably identify individuals at high risk for dementia, but its mechanisms are not yet fully understood. Effective dementia treatments remain elusive. Identifying high-risk individuals would allow for addressing modifiable risk factors and organising future care.
It would also assist research trials with cohort recruitment and ultimately contribute to a reduction in the prevalence of dementia.
Even a small decrease in dementia prevalence, or delaying the age of onset, would significantly impact the huge associated public health costs [9,10].
Having a low socioeconomic status is a powerful predictor of ill health [11]. Early-and mid-life low socioeconomic status has been associated with an increased risk of dementia in later life (adjusted hazard ratio [aHR] 1.45, 95% confidence interval [CI] 1.15-1.83), and this difference persists to the oldest-old ages [11,12].
The 2020 Lancet Commission on dementia highlighted many risk factors that are, at their core, primarily of social origin [13]. Our recent study of MCR showed that socioeconomic status, as defined by occupational social class, was lower for individuals with MCR [14].
Researchers have operationalised socioeconomic status in various ways, most commonly as education level, social class, income, or occupation [15]. A Swedish study examining the relative importance of these different socioeconomic indicators found that each measure is associated with late-life health, with only minor differences in the effect sizes [15]. However, a 12-year follow-up study in England found that lower wealth in late life, but not education, was associated with increased risk for dementia and that no substantive differences were identified concerning area-based neighbourhood deprivation [16]. Hence, we include both education and occupation class in the present study. By examining the role of socioeconomic status in MCR we can better understand the mechanisms of MCR.
If low socioeconomic status contributes to MCR, individuals of low socioeconomic status could be supported to reduce their dementia risk. This study explores the longitudinal association between socioeconomic status, as determined by highest education and occupational social class during working age, with MCR later in life (aged 76+ years).

Setting and sample size
This longitudinal prospective study used baseline socioeconomic status data and 6-year follow-up data (wave 3) from the Lothian Birth Cohort 1936 (LBC1936). The LBC1936 has been profiled in detail elsewhere [17,18]. In summary, between 2004 and 2007, 1091 Scottish adults born in 1936 were recruited for baseline interviews, cognitive tests, questionnaires, blood tests, and physical measures (mean age of 69.5 years [SD = 0.8]) [17,18]. There is an almost equal sex split; all participants are white. These participants have been reassessed approximately every 3 years since. Wave 6 is ongoing at the time of writing, and a seventh wave is planned. To minimise loss to follow-up between waves, the LBC1936 researchers re-contact those unable to attend a wave due to a temporary illness and see them at a later, more appropriate time [19]. Figure 1 illustrates sample selection and reasons for dropout and exclusion.
We used data from wave 3, when participants had a mean age of 76.3 years (n = 696), for our MCR and covariates data, as this was the first and largest wave to measure all the criteria necessary for deriving MCR. We excluded participants with dementia and those missing data in any MCR criteria.

Outcome measure-MCR
We used the original MCR definition of subjective cognitive complaints and objective slow gait in older individuals without significant functional disability or dementia [4]. We defined slow gait using the typical approach for MCR of walking speed one standard deviation (SD) below age-and sex-matched means in this cohort. LBC researchers timed participants with a stopwatch walking 6 metres along a corridor. A subjective cognitive complaint was recorded if the participant replied affirmatively to the question: "Do you currently have any problems with your memory?". Loss of independence was determined by scoring over 1.5 SD above the mean on the Townsend Disability Scale (a higher score indicates greater disability) [20]. Dementia was assessed by self-report or the Mini-Mental State Examination (MMSE), with a score < 24/30 indicating possible dementia [21].

Socioeconomic status
Our risk factor of interest was socioeconomic status as determined by years of education and the occupational social class of the participants during their working years. This approach for classifying socioeconomic status in LBC1936 has been used many times previously in the literature [22][23][24][25]. Occupational social class is based upon principal occupation, coded in line with the 1980 census, and was categorised as professional, managerial, skilled non-manual, skilled manual, or semiskilled/unskilled [17,18]. As such, it categorises individuals into a social class grouping rather than giving specific information about the occupation role. Typical for the time, married women were assigned a social class based on the highest occupation of the household, be that their own or their husband's. Although we could not determine which women were coded based on their husband's occupation and which were based on their own occupation, it remains typical practice to assign a socioeconomic status to a household based on the highest occupational social group, as opposed to separating a married couple into different social groups.
For our study, we collapsed the occupational classes of professional, managerial, and skilled non-manual into a 'non-manual' category and the skilled manual and semiskilled/unskilled classes into a 'manual' category. This preserved the distinction between higher and lower socioeconomic status while improving the data distribution and providing a relevant comparator for a measure like MCR, which contains a motor/manual component such as gait speed.
Table S1 presents the descriptive characteristics of the non-manual and manual occupation groups for comparison. The non-manual and manual occupation groups differed significantly in several key characteristics, including age, sex, years of education, smoking status, depression symptoms, and grip strength (Table S1). Body mass index (BMI) is also significantly different between manual (29.1 ± 4.6 kg/ m 2 ) and non-manual (27.4 ± 4.4 kg/m 2 ) occupational social classes, but dietary pattern was not available to explore this further. Our final model adjusts for all these potential confounders. Age 11 intelligence also differs significantly between non-manual and manual groups; but as expected, age 11 intelligence and occupational social class were collinear, so it was not included in our model.

Other risk factors
Based primarily on reported risk factors for MCR [8,9,26], we included the following risk factors in our analysis: age, sex, self-

Statistical methods
We summarised the participants' characteristics using means and SD or frequencies and percentages, as appropriate. We compared non-MCR and MCR groups using χ 2 tests with a continuity correction for categorical variables. We performed an F-test (ANOVA) by default for continuous explanatory variables. We used a Kruskal-Wallis test when variables were considered non-parametric, except when expected counts were less than five, where Fisher's exact test was more appropriate.
We estimated the odds ratio (OR) association with 95% CIs of socioeconomic status, as determined by occupational social class and years of education, with MCR using logistic regression models. We performed a sensitivity analysis with the sexes separated. We assessed for possible bias due to missing data by comparing the distribution of missingness in the final model variables amongst MCR and non-MCR groups. The main analysis was based on participants with no missing data. We performed variance inflation factor analysis to ensure no multicollinearity between variables in the final regression model. We considered p values <0.05 to be statistically significant. All statistical analyses were conducted in R version 4.0.2, using the 'finalfit' package version 1.0.4 [27,28].
Code is openly shared on Github [29]. LBC1936 data are available on request from the LBC1936 team (https://www.ed.ac.uk/lothi an-birth -cohor ts/data-acces s-colla boration) [30]. The reporting of this study conforms to the STROBE statement [31].  [19]. Figure 1 illustrates the participant flow in this study. MCR prevalence at wave 3 was 5.6% (95% CI 4.0-7.6; n = 39/696). Table 2 presents the results of the final logistic regression analysis model. The final and most adjusted model (Model 4) demonstrates that working in a manual rather than a non-manual occupation earlier in life is associated with a greater than three-fold increased risk of having MCR later in life (adjusted odds ratio [aOR] 3.55, 95% CI 1.46-8.74; p = 0.005). This is illustrated in Figure 2 with full details given in Table 2. improves this to a 75% chance (C-statistic 0.75).

RE SULTS
None of the main explanatory variables are collinear as illustrated in Figure S1. Variance factor analysis suggests no evidence of multicollinearity among any of the variables in our final model (Table S3).

Sensitivity analysis
We performed a sensitivity analysis with the sexes separated to explore their effect on the true association between socioeconomic status and MCR. We used the same analysis approach and factors as our main analysis, except we excluded the sex variable. In the male cohort (Table S4), the effect of manual occupational social class on future MCR risk was higher than in the overall cohort when unadjusted (OR 3.79, 95% CI 1.48-10.45; p = 0.007) but the same when adjusted (aOR 3.55, 95% CI 1.19-11.73; p = 0.028). In the female cohort, only three female participants had both MCR and a manual occupation, which was too small for further analysis.
As so few participants had MCR data missing (n = 4), we have not included here our sensitivity analysis on those with missing data for MCR derivation. However, for completeness, this information is provided in Table S5.

DISCUSS ION
Our main finding is that lower socioeconomic status as defined by non-manual versus manual occupation (and not years of education) is associated with a more than three-fold increased risk of having MCR later in life (aOR 3.55, 95% CI 1.46-8.74; p = 0.005). The strong association between manual occupation and increased risk of MCR remained despite adjusting for an extensive array of potential confounds. This is the first study to focus on socioeconomic status as a risk factor for MCR, which adds to its value and importance but makes it difficult to place it in context. However, a recent study examining manual and non-manual occupational social class found that manual occupation is associated with dementia-related death [12]. As MCR is highly prognostic of dementia in later life, our main finding has good face validity.

Occupational social class
The LBC1936 dataset contains another measure of socioeconomic status, the Scottish Index of Multiple Deprivation (SIMD) [32]. This is a relative measure of deprivation across small areas of Scotland, ranked 1 (most deprived) to 6976 (least deprived). As this is a population-level measurement, individuals living in the same area but with very different jobs and socioeconomic circumstances will have the same SIMD rank. As such, it is less informative on an individual level than occupational social class or years of education, and we did not include it in our model. Income data are unavailable for the participants. However, working-age occupation class is a good proxy measure for individual wealth later in life, often better than other indices such as neighbourhood deprivation or years of education [15,16]. As such, it is likely that the difference in health outcomes between these two groups is partly mediated by economic status.

Other covariates
Despite the narrow age range of LBC1936 participants, we noted a significant association between older age with the presence of MCR. This is in keeping with a recent meta-analysis of factors associated with MCR in 22 studies, which found that a majority reported age as an associated factor for the presence of MCR [26]. There was no significant relationship between MCR status and any other variables tested for, including those traditionally associated with MCR such as obesity, multimorbidity, depression, and anxiety. This is surprising as we chose these variables a priori based primarily on their previous association with MCR. Our study may not have the power to detect a significant relationship, an idea supported by the effect sizes of these other variables, which are generally in the same direction as larger studies [8,9,26]. Further investigation is required using casecontrol studies or much larger cohort studies.
The lack of association between MCR and traditional risk factors for MCR may also be partly due to healthy volunteer bias.
Lothian is a relatively affluent area of Scotland, and the participants in LBC1936 have a higher average number of years of education Participants are also all white. These factors should be considered when generalising these findings to other populations. However, the overall affluence of the LBC1936 sample may underplay the role of socioeconomic status as a risk factor for MCR. Perhaps a more socioeconomically diverse cohort would reveal an even stronger association effect size between low and high socioeconomic status and MCR. This would be an important future study in a more populationrepresentative cohort.
Perceived stress and levels of social support may mediate the association between occupation and MCR. Unfortunately, specific data on these variables were not available. However, the participants in LBC1936 retired long before wave 3 (mean age 76 years) when we first derived MCR. As such, any questionnaires regarding occupational stress would rely on recall from many years before.
Nevertheless, while not stress-specific, we included depression and anxiety symptoms (Hospital Anxiety and Depression Scale) in our final adjusted model even though there were no significant differences in symptoms between the MCR and non-MCR groups. This was an a priori decision based on previously reported associations between these psychological symptoms and MCR in the literature [9,26,33].

Limitations and strengths
Our data had additional limitations to those already discussed.
Growing research identifies head trauma as a risk factor for laterlife subjective and objective cognitive impairment [34]. A potential increased rate of head trauma related to occupation could mediate the relationship between socioeconomic status and MCR, but these data are unavailable in our cohort. Attrition from ill-health or mortality is a common and often unavoidable limitation of longitudinal studies of ageing. The LBC1936 research team attempt TA B L E 2 Final model -logistic regression of motoric cognitive risk (dependent variable) and socioeconomic status at baseline with and without adjustment for potential confounders.

Parameter MCR (n (%)) No MCR (n (%)) OR (univariable) aOR (multivariable)
Occupational to minimise attrition by re-contacting those unable to attend a wave due to a temporary illness and assessing them at a later, more appropriate time where possible. Despite this, attrition in the LBC1936 is approximately 20% between waves, leading to a 37% loss of participants over the 6 years between baseline and wave 3. Although this attrition is substantial, it is within the limit of what is considered acceptable by international quality assessment bodies [35]. Fortunately, our study has an excellent participation rate (99%) of the eligible sample, with only 7 of the 697 available participants excluded, which helps address any selection bias. Despite this, our sample size is still small and a replication study in a larger cohort, or a cohort with a higher prevalence of MCR, would increase confidence in our findings. A replication study is certainly feasible. MCR has been derived in many cohorts globally [7]. Likewise, socioeconomic status has been operationalised in various ways [15], all of which are associated with late-life health, with only minor differences in the effect sizes [15].

Implications
An association between lower socioeconomic status and MCR has several implications. First, the mechanism of this association must  [37]. Finally, public health strategies should target individuals with lower socioeconomic status for earlier dementia detection and intervention. Given that approximately 50 million people worldwide live with dementia, a number projected to triple over the next 30 years [13], even a small reduction in incidence or delaying the age of onset could make a substantial difference to patients, families, and societies globally.

CON CLUS IONS
In conclusion, this prospective study shows an association between lower socioeconomic status, as defined by manual occupation at a younger age, and MCR at an older age, but no effect of educational such as MCR in their social context and early in the life course could be effective strategies for reducing health inequalities in older age.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are available on reasonable request at: https://www.ed.ac.