Midlife financial stress and cognitive and physical impairments in older age: The role of potentially modifying factors

Abstract

Background:

Financial stress is an important source of chronic stress and has been associated with cognitive and physical impairments. The goal of this study was to investigate whether financial stress is associated with cognitive and physical impairment and their combination, the role of potential modifiable factors and potential sex differences.

Methods:

The Cardiovascular Risk Factors, Aging, and Dementia population-based cohort study from Finland was used (n = 1497) (baseline data collected 1972–1987, mean age 50 years). Two late-life re-examinations (mean total follow-up 21 years). Midlife financial stress was measured using two questions on financial situation. Cognitive functioning was based on six cognitive domains. Physical impairment was self-reported, including activities of daily living and mobility. Potential mediation factors investigated were smoking, alcohol, physical activity, cohabitant status, non-manual work, and sleep disturbances. Sex differences were investigated. We used path analyses with full information maximum likelihood estimation.

Results:

Midlife financial stress was associated with worse cognitive functioning, physical impairment and their combination. Smoking and sleep disturbances mediated the associations between financial stress, physical impairment, and combined impairments. For men: Among smokers financial stress was associated with worse cognitive functioning; alcohol interacted with financial stress on combined impairments; cohabitation and non-manual work mediated associations to worse cognitive functioning. Among women, sleep disturbances moderated the association to worse cognitive functioning.

Conclusions:

Midlife financial stress is associated with late-life impairments, and lifestyle/sociodemographic factors may modify these associations. Sex differences were observed. Interventions promoting healthier lifestyle and psychosocial factors may buffer against the deleterious role of financial stress.

1. Background

The global advancements in medicine and health promotion init’[iatives have contributed to longer life expectancy and better quality of life among older adults. Nevertheless, despite these advancements, cognitive and physical impairments are the leading causes of disability, dependency and mortality among older adults (Goldman et al., 2014; “World Health Organization. Risk reduction of cognitive decline and dementia: WHO guidelines. (2019).”). These conditions are the source of immense healthcare and social burdens. To reduce the risk and prevent cognitive and physical impairments, a better understanding is needed regarding risk and protective factors across the lifespan.

While short-term or acute stress can be adaptive for responding to challenges, chronic psychological stress is an important determinant of adverse physical and mental health outcomes among older adults (Franks et al., 2021; McEwen, 2017). Various sources of stress across the lifespan have been associated with cognitive and physical impairments, including work-related stress (Andel et al., 2011; Nilsen et al., 2014, 2017; Pandey et al., 2020; Sindi et al., 2017b), early life stress/adversity (Alastalo et al., 2013; Grainger et al., 2020; Hedges and Woon, 2011), and widowhood (Clouston et al., 2014; Singham et al., 2021; Wu-Chung et al., 2022). Midlife financial stress - an important source of chronic stress (Park N et al., 2017) - has also been recognized for its associations with poor global self-reported health and quality of life among older adults (R. Huang et al., 2020).

Various studies have shown that financial stress is associated with cognition. Income volatility (changes in income levels) among younger adults (age 24–35 years) was associated with worse cognitive function in midlife (Grasset et al., 2019). Higher levels of financial stress earlier in life were associated with worse performance on executive function and episodic memory later in life (Chen et al., 2022). Financial stress is commonly experienced among individuals with low socioeconomic position, and a recent meta-analysis showed that low socioeconomic position was associated with an increased risk for cognitive impairment (Wang et al., 2022). Middle-aged adults who reported an improvement in their financial situation in midlife had a lower risk for dementia, when compared with individuals who reported no change in financial situation, or individuals who reported worse financial situation (Sindi et al., 2020). A meta-analysis reported that financial scarcity in adult life (including midlife) was more strongly associated with cognition than financial scarcity during childhood, highlighting the importance of midlife adversity (de Almeida et al., 2024). Some studies have shown that midlife income was not associated with late-life dementia risk nor dementia mortality risk (Anttila et al., 2002; Strand et al., 2015). This may highlight the importance of subjective financial stress, including an individual’s appraisal and perception of the stressor, in addition to income levels (Glei et al., 2018).

Financial stress has also been associated with poor physical health and physical impairments, which refer to difficulties or restrictions in physical abilities or activities. For example economic hardship (indicating low household income, challenges to pay expenses, and lacking cash reserves) was associated with poor self-reported health and musculoskeletal disorders (Ahnquist et al., 2012). One longitudinal study showed that higher levels of midlife family financial stress among couples was associated with poorer physical health after 15 years (Wickrama and O’Neal, 2021). Financial stress also appears to have a cumulative influence on late-life physical health. The number of financial stress periods in young, middle-age and early older adulthood (up to age 65 years) are associated with more self-reported functional limitations, illnesses, symptoms, and poorer self-reported health, with more detrimental associations among those with continuous episodes (Kahn and Pearlin, 2006). Importantly, this study showed that such chronic periods of financial stress represent an important risk factor, regardless of current financial situation among older adults (Kahn and Pearlin, 2006). Similarly, some previous studies have shown that subjective financial stress in midlife predicts self-reported health, despite adjusting for objective measures including income and education (Arber et al., 2014; Singh-Manoux et al., 2005), further supporting the important role of subjective financial stress. A recent review showed that financial stress is highly prevalent in midlife, and is associated with health and health behaviours, which highlights the importance of midlife financial stress (Gomez-Bernal et al., 2019).

Due to sex differences in education and occupational attainment, sex differences are also observed in income and financial stress, and women have been found to report more economic hardships (Ahnquist et al., 2012). Sex differences have been reported in the response to stress (McEwen, 2017) and coping strategies. For example, it was shown that although women report more stress, they also score higher on emotional expression, and report more reliable and higher quality of social support networks (Kneavel, 2021; Rubio et al., 2016). In the aforementioned study on financial stress among couples, the associations were more pronounced among men (Wickrama and O’Neal, 2021). However, other studies showed similar associations between economic hardships and health outcomes in both sexes. With regards to the role of financial situation, socioeconomic position in childhood was associated with worse late-life cognitive impairment among women (Wolfova et al., 2021).

Stress is associated with numerous modifiable lifestyle factors, such as smoking, physical activity, alcohol, and sleep (Fransson et al., 2012; Nyberg et al., 2013; Sparrenberger et al., 2009). While stress may promote unhealthy lifestyle behaviors due to a perceived lack of time or resources (e.g. physical inactivity), individuals may also adopt unhealthy coping mechanisms (e.g. alcohol consumption) in response to stress. Numerous social factors such as social resources may also alter one’s sensitivity to the long-term effects of stress. Social resources, including social network and support, may promote resilience against stress (Ellwardt et al., 2020). Living alone (non-cohabiting) and poor social relationships increase the risk for cognitive decline whereas cohabiting and better social relationships had a protective role (Hakansson et al., 2009; Piolatto et al., 2022). Studies have also shown that blue-collar/manual work (which is often characterised by passive jobs) is associated with a higher risk for cognitive impairment (L. Y. Huang et al., 2020; Tomioka et al., 2020) and adverse physical outcomes including musculoskeletal conditions (Yang et al., 2023). In the context of aging, it is unclear how individual differences in lifestyle behaviors and social resources influence inter-individual differences in sensitivity to stress, and the effects of stress on cognitive and physical function.

The co-existence of cognitive and physical impairments is common among older adults. Previous evidence has demonstrated that cognitive function declines are associated with physical function declines, with potential bidirectional associations between fine motor and processing speed (Finkel et al., 2016). Yet, it is unclear whether self-reported midlife financial stress is longitudinally associated with the co-existence between cognitive and physical impairment among older adults, and whether sex differences are present in these associations. The goal of this study was to investigate whether financial stress is associated with late-life cognitive and physical impairments, and their combination, and whether lifestyle factors (such as smoking, alcohol consumption, or physical activity) or psycho-sociodemographic factors (such as cohabitant status, manual work, or sleep disturbances) modify these associations. We additionally aimed to examine potential sex differences in these associations.

2. Methods

2.1. Sample

Data were derived from The Cardiovascular Risk Factors, Aging and Dementia (CAIDE) study. Midlife (baseline) examinations were conducted within the North Karelia Project and the FINMONICA (later FINRISK) population-based studies (Finland). Participants performed baseline assessments during one of the following years: 1972, 1977, 1982, or 1987 (Borodulin et al., 2018; Puska, 2010). At baseline, participation rates ranged from 82% to 90%. The follow-up study (CAIDE) included older adults aged 65–79 by the end of 1997. For study flow chart see Fig. 1. The distribution of baseline assessments and follow-ups are presented in Table 1.

Fig. 1.

Flowchart representing the study population and examinations in the CAIDE Study.

Table 1.

Distribution of baseline assessments and follow-ups.

Baseline year	Follow-up
	First	Second	Total
1972	494	294	788
1977	528	292	820
1982	221	137	358
1987	158	106	264
Total	1401	829	2230

Of 3559 eligible individuals, 2684 (75.4%) remained in this region (with 875 (24.6%) having died before the end of 1997). A random sample of 2000 survivors were invited to participate in the first reexamination in 1998. A total of 1449 individuals (72.5%) participated and 1409 underwent cognitive assessments. The mean follow-up time was 21 years (SD = 4.9; range 11–26 years).

Between 2005 and 2008, participants returned for the second re-examination. In 2005, 1426 (of the original sample of 2000) were alive and living in the same region. A total of 909 (63.7%) of those invited accepted to participate, and completed cognitive assessments were available for 852. The mean follow-up time to the second follow-up was 29 years (SD = 5.1; range 18–37 years). In total, 1511 participated in at least one re-examination, and 733 participated in both re-examinations. After excluding those without information about financial stress, 1,497, out of 1,511, remained in the analyses (see Table 2). Of those 733 where included in both re-examinations. For the mean ages at the different re-examinations, see Table 2. The CAIDE study adheres to the Declaration of Helsinki, and local ethics committees approved the study. Informed consent was obtained from all participants.

Table 2.

Sociodemographic and lifestyle characteristics of the study sample.

Measured at baseline	Missing in low/in between/high financial stress		Financial stress (range 0–4)
		Range	Low (n = 393)	In between (n = 888)	p-value1	High (n = 216)	p-value2
N in 1st follow-up (t1)			369	833		199
N in 2nd follow-up (t2)			223	486		120
Women (%)	0/0/0		66	61	0.095	60	0.142
Age	0/0/0	39–64	51	50	<0.001	49	<0.001
Time to 1st follow-up	0/0/0	11–26	20	21	0.007	22	<0.001
To 2nd follow-up	0/0/0	18–37	28	29	0.057	30	0.005
Education, years	8/13/3	0–23	8.9	8.4	0.014	8.7	0.466
Smoker (%)	0/0/0		39	44	0.074	47	0.041
Alcohol	132/241/41	0–30.4	1.0	0.9	0.466	1.4	0.189
Physical activity	12/22/5	0–365	101	87	0.026	101	1.000
Cohabiting (%)	0/1/0		81	81	0.756	72	0.017
Non-manual work (%)	12/16/13		53	47	0.043	41	0.009
Sleep disturbances	7/73/14	0–2	0.48	0.45	0.324	0.60	0.015

Financial stress: Low, range 0–1; In between, range 1.5–2.5; high, range 3–4. P-values are based on χ² tests for % and multinomial logistic regressions for numeric variables.

^{1 –}difference between low and in between;

^{2 –}difference low/high.

Variables: Alcohol – number of occasions of alcohol consumption per months; physical activity – number of days performing physical activity per year; Sleep disturbances is a mean index, based on insomnia and nightmares, coded 0 = not at all, 1 = sometimes, and 2 = often.

2.2. Financial stress

Financial situation in midlife (baseline) was measured using the following two questions with 5-point likert scales. Each question was initially analyzed separately:

1. Satisfaction with financial situation at midlife: The question asked: ‘How satisfied are you with your financial situation?’. Response options were: 1 = Very satisfied, 2 = Satisfied, 3 = Somewhat satisfied, 4 = Dissatisfied, 5 = Very dissatisfied.

2. Change in financial situation at midlife: The question asked: ‘Is your financial situation now better or worse than before?’. Response options were: 1 = Much better, 2 = A little bit better, 3 = Same as before, 4 = A little bit worse, 5 = Much worse.

The final measure was constructed as an index based on the mean of the two variables and subtracting one to obtain a centered variable.

2.3. Outcome variables

The outcome variables in this study are based on physical functioning, referred to as physical impairment, the inverse of cognitive functioning, referred to as worse cognitive functioning, and the co-occurrence of both physical impairment and worse cognitive functioning, referred to as combined impairments.

2.4. Physical impairment

Physical function at follow-up was based on Activities of Daily Living (ADL) and mobility. ADL is based on two questions: Washing without help; Dressing without help. Response options to each of these questions were: 1 = not able to, 2 = able to, but with difficulties, 3 = able to without any difficulties.

Mobility was based on four questions: Climbing stairs without help; Walk about half a kilometer without help; Run about a hundred meters; Run over half a kilometer. Response options to each of these questions were: 1 = not able to, 2 = able to, but with difficulties, 3 = able to without any difficulties.

Mobility and ADL were computed into two separate mean indices. From these two indices a new combined mean index was created. The three indices on physical function were reversed so that higher scores indicated more physical function limitations (thereafter referred to as physical impairments), transformed with zero-skewness log-transformation and transformed to have zero as the lowest value and a standard deviation of 1.0. Preliminary analyses showed stronger associations between financial stress and the combined measure of impairments with physical functioning which was used in the analyses.

2.5. Worse cognitive functioning

Cognitive function was based on five measures on cognitive functioning: episodic memory, pegboard, stroop32, verbal fluency, and the letter-digit substitution test, as described previously (Sindi et al., 2017) and below.

At the first and second re-examinations, a comprehensive battery of neuropsychological tests was administered to measure several cognitive domains. For this study, we used the following measures that were available from both re-examinations:

(1) episodic memory measured with an immediate word recall test (10-word list); (2) executive functioning measured through the Stroop test (time difference between the task of naming the color of the ink used to write the name of a different color, and task of naming colors of dots); (3) verbal fluency measured through category fluency test (number of correct animal names generated in 60 s); (4) psychomotor speed measured by the letter digit substitution test; (5) manual dexterity measured through the bimanual Purdue Pegboard test, commonly used for fine motor skills, coordination and manual dexterity measures (Tiffin and Asher, 1948).

Each of the tests was transformed with zero-skewness log-transformation and z-transformed to have zero as the mean and a standard deviation of 1.0. The variables on cognitive functioning were reversed so that higher scores reflected worse cognition. A mean index was created from these variables, again transformed with zero-skewness log-transformation and transformed to have zero as the lowest value and a standard deviation of 1.0. in the index, as in the included variables on cognitive functioning, a higher score reflects worse cognition, hereafter referred to as worse cognitive functioning. The lowest value, which is zero, will hereafter be referred to as the best cognitive functioning.

2.6. Combined impairments

Finally, an outcome variable showing the combined presence of physical impairment and worse cognitive functioning was created. A multiplicative variable was created by multiplying the indexes on physical impairments with the index on worse cognitive function (both having zero as lowest value). This multiplicative term was transformed with zero-skewness log-transformation and transformed to have zero as the lowest value and a standard deviation of 1.0. An interaction term in regular regressions is a multiplicative term added on top of the main effects. The multiplicative term shows the additional association that the combination of two variables has on the outcome. The multiplicative term in our analyses corresponds to a multiplicative term in a regression, but for the dependent variables. The association to the multiplicative term shows the additional association the independent variables have to a combination of the outcomes.

2.7. Control variables

We adjusted for age, sex and education (model 1). Research has consistently shown that lower education is associated with cognitive decline (Clouston et al., 2020) and that education is associated with late-life cognitive function through its influence on cognitive skills throughout adulthood (from early adulthood to older age) (Lovden et al., 2020). Age is among the most important risk factors for cognitive decline (Dahan et al., 2020). Men and women show different patterns and trajectories of cognitive performance and decline (Ferretti et al., 2023; Levine et al., 2021). Age, sex and education are also associated with poorer physical function (Garber et al., 2010). Models 2 and 3 adjusted for modifying lifestyle and sociodemographic factors, as described below.

2.8. Modifying lifestyle and sociodemographic factors

This study aimed to assess the role of potential modifying variables in the associations between financial stress and physical impairment, worse cognitive functioning and the combination of physical impairment and worse cognitive functioning (referred to as ‘combined impairments’). All potentially modifying factors are analyzed both for mediation and moderation. The potential modifying factors were: smoking, alcohol consumption, physical activity, cohabiting/not, being manual/non-manual worker, and sleep disturbances. Our data include three commonly used measures of socioeconomic position: income, education, and manual/non-manual workers. Income is not suitable to use as a possibly modifying factor since it is related to financial stress. We use manual/non-manual worker as our measure of socioeconomic position. All modifying variables are from baseline. These variables were coded as follows:

Smoking: presently smoking (at baseline) vs. not smoking; alcohol is given linear representation, coded as number of occasions of alcohol consumption per months based on a question with the response categories, never (coded 0), once a year or less (coded 0.08), 3–4 times a year (coded 0.29), once every second month (coded 0.5), once/month (coded 1), 2 times/month (coded 2), once a week (coded 4.33), a few times a week (coded 10.83), daily (coded 30.44). The numerical values were assigned in order to give the variable linear representation; physical activity is given linear representation, coded as number of days performing physical activity per year based on the response alternatives not at all (coded 0), a few times a year (coded 7), 2–3 times per month (coded 30), once a week (coded 52), 2–3 times a week (coded 130), and daily (coded 365); manual/non-manual work was measured through asking participants to select their longest-held occupation among the following categories: office/service, farming/forestry, mining/industrial/construction work, housewives, or other. Office/service was considered non-manual, while farming/forestry, mining/industrial/construction work, housewives, and others were categorized as manual work, and the final variable was dichotomous (as described previously (Sindi et al., 2017a)). Cohabitant status refers to whether the participant lives alone or is cohabiting (as described previously (Hakansson et al., 2009)). Sleep disturbances is a mean index, based on insomnia and nightmares, coded 0 = not at all, 1 = sometimes, 2 = often, given linear representation (for further descriptions of the measures, see (Sindi et al., 2018a)).

2.9. Statistical analyses

Analyses were performed using Stata 17.0 (Stata Corp, College Station, TX, USA). The significance level for all analyses was set at p < 0.05.

The analyses were conducted with path analyses with full information maximum likelihood estimation (Allison, 2012; Enders, 2001). Data were organized in long format. Results are presented as standardized β-coefficient and p-values. For an example of path analyses model, see Fig. 2, where all modifiers are tested both as potential mediators and moderators.

Fig. 2.

Example of path analyses model.

For all participants, baseline was from midlife (age 39–64 years). Some participants had follow-up data from one of the follow-ups, some participants had data from both follow-ups. As a result, some participants were included twice in the analyses, which could lead to erroneously low standard errors. To adjust for this, we used cluster correlated standard errors. The results presented were the linear associations between financial stress as independent variable and impairments as dependent variables.

We simultaneously analyzed physical impairment, worse cognitive functioning, and a multiplicative term between physical impairment and worse cognitive functioning (physical impairment * worse cognitive functioning). The coefficient for physical impairment shows the estimated association between financial stress and physical impairment for those with the best cognitive functioning, the coefficient for worse cognitive functioning shows the estimated association between financial stress and worse cognitive functioning for those with no physical impairment. The coefficient for the association between financial stress and the multiplicative term physical impairment * worse cognitive functioning shows the association to the co-occurrence of physical impairments and worse cognitive functioning (combined impairments).

The analyses were performed in different models. Model 1: Controlling for sex, age at baseline, follow-up time, years of education, and the covariance between worse cognitive functioning, physical impairment, and cognitive * physical impairment (combined impairments); Model 2: Additionally controlling for one possible modifying factor in each model; smoking, alcohol consumption, physical activity, cohabitation, non-manual work, and sleep disturbances. Total, direct, and indirect effect (showing mediation) for the association between financial stress and impairment are presented. Model 3: additionally controlling for interactions (showing moderation) between the modifying factors and financial stress. Only interaction and direct effects are presented. Only interactions with p < 0.10 are presented.

We also analyzed the interaction between financial stress and sex, and in further sensitivity analyses, we also assessed the interaction between financial stress and age. No interaction between financial stress and age had p < 0.202, and these results are not presented. There were interactions between sex and the outcomes that had p < 0.10, and these results including interactions are presented. Models 1, 2, and 3, were as above but also including interactions between financial stress and sex. In model 1 the β-coefficient and p-values are presented for the interaction between sex and financial stress. For model 2 and model 3 the reference category for sex was alternated and the results are presented separately for women and men.

3. Results

3.1. Sample characteristics

The age range of the sample at baseline was 39–64 years. Table 2 presents the sociodemographic and lifestyle characteristics of the study sample. When comparing the three financial stress groups (Low, medium, high), there were significant differences in mean baseline age (low: 51 years, middle: 50 years, high: 49 years), time to first follow-up and to second follow-up (the longest follow-up durations were observed for the group with high financial stress, while the shortest follow-up durations were observed for the group with low financial stress), proportion of smokers, which was highest in the group with high financial stress and lowest in the group with low financial stress, the proportion of participants who were cohabiting (which was lowest for the group with high financial stress), the proportion of those who were physically active (which was higher in the group with lower financial stress compared to the group with medium financial stress), the proportion of participants who have non-manual work (which is highest in the group with low financial stress, and highest in the group with high financial stress), and sleep disturbances which were highest among the group with high financial stress. No significant differences were observed for the proportion of women and for alcohol consumption across the groups.

3.2. Associations between financial stress and impairments

The associations between financial stress and impairments are presented as standardized β-coefficients in Table 3. While adjusting for multiple confounders, the results from model 1 show that the association between financial stress and physical impairment (β = 0.14) was stronger than the association to worse cognitive functioning (β = 0.09) and between financial stress and combined impairments (β = 0.16).

Table 3.

Associations between financial stress as independent variable with physical impairments, worse cognitive functioning, and combined impairments as outcomes.

The main independent variable is financial stress	Physical impairment		Worse cognitive functioning		Physical * cognitive1
Model	β2	p-value	β2	p-value	β2	p-value
Model 1 (no modifying factors)	0.14	<0.001	0.09	0.001	0.16	<0.001
Model 2, including smoking
Direct effect from financial stress (regardless of smoking)	0.14	<0.001	0.09	0.002	0.15	<0.001
Indirect effect from financial stress, mediation through smoking	0.01	0.013	0.00	0.419	0.01	0.017
Total effect (direct + indirect)	0.15	<0.001	0.09	0.001	0.16	<0.001
Model 3: interaction (showing moderation) between financial stress and smoking
Interaction (different associations depending on smoking)			−0.11	0.093
Smokers			0.15	0.001
Non-smokers			0.04	0.314
Model 2, including alcohol
Direct effect	0.14	<0.001	0.09	0.004	0.15	<0.001
Indirect effect, mediation	0.00	0.372	0.00	0.862	0.00	0.414
Total effect	0.14	<0.001	0.09	0.004	0.16	<0.001
Model 2, physical activity
Direct effect	0.14	<0.001	0.09	0.004	0.16	<0.001
Indirect effect, mediation	0.00	0.354	0.00	0.620	0.00	0.332
Total effect	0.14	<0.001	0.09	0.004	0.15	<0.001
Model 2, cohabiting
Direct effect	0.14	<0.001	0.10	0.004	0.15	<0.001
Indirect effect, mediation	0.00	0.354	0.00	0.620	0.00	0.332
Total effect	0.14	<0.001	0.09	0.004	0.16	<0.001
Model 2, non-manual worker
Direct effect	0.14	<0.001	0.10	0.005	0.15	<0.001
Indirect effect, mediation	0.00	0.113	0.01	0.057	0.01	0.084
Total effect	0.15	<0.001	0.10	0.002	0.16	<0.001
Model 2, sleep disturbances
Direct effect	0.13	<0.001	0.09	0.006	0.14	<0.001
Indirect effect	0.01	0.030	0.00	0.125	0.01	0.034
Total effect	0.14	<0.001	0.09	0.004	0.15	<0.001

Linear associations between financial stress as independent variable and impairments as dependent variables, estimated with path analyses with maximum likelihood estimation using cluster correlated standard errors.

Model 1: Controlling for sex, age at baseline, follow-up time, years of education, and the covariance between worse cognitive functioning, physical impairment, and cognitive*physical impairment (combined impairments); Model 2: Additionally controlling for one possible modifying factor in each model. Total/direct/indirect effect refers to the association between financial stress and impairment/worse functioning; Model 3: additionally controlling for interaction between modifying factor and financial stress. Only interaction and direct effects are presented. Only interactions with p < 0.10 are presented.

¹⁾multiplicative term physical impairment * worse cognitive functioning, standardized to have the standard deviation of 1.0. The term reflects the combination of physical impairment and worse cognitive functioning.

²⁾standardized β-coefficient.

Results from model 2 show that the only smoking and sleep disturbances had significant (indirect) effects, i.e., they mediated the effect from financial stress on impairments. Smoking mediated a small but significant proportion of the association between financial stress and physical impairment (β = 0.01; p = 0.013) and between financial stress and the combined impairments (β = 0.01; p = 0.017). The association between financial stress and worse cognitive functioning was not only mediated by smoking. Smoking also moderated (an interaction with p = 0.093) the association (β = 0.15), i.e., the effect from financial stress was reinforced among smokers. Sleep disturbances also mediated a small portion of the association between financial stress and physical impairment (β = 0.01; p = 0.030) and with the combined impairments (β = 0.01; p = 0.034). For alcohol consumption, physical activity, cohabiting, and non-manual work, results from respective models 2 did not show significant mediation (indirect effects) or moderation (interaction effects). We also investigated potential interactions with age, and there was no interaction between age at baseline and financial stress with p < 0.20.

3.3. Sex and age differences

We first investigated potential interactions with age, and there was no interaction between age at baseline and financial stress with p < 0.20.

There were interactions (with p < 0.10) between financial stress and sex for worse cognitive functioning without the presence of physical impairment (β = 0.12; p = 0.067), and for the combined impairments (β = 0.12; p = 0.066) (Table 4, Model 1). Considering these findings, separate results for the sexes are presented for all three outcomes. The association between financial stress and all three outcomes was moderately but not significantly stronger among men than women. Among women significant direct effect from financial stress was only found for physical impairments and the combined impairments, but not worse cognitive functioning (Model 1).

Table 4.

Associations between financial stress as independent variable with physical and cognitive functioning as outcomes, interactions with sex.

The main independent variable is financial stress	Physical impairments		Worse cognitive functioning		Physical * cognitive1
Model	β2	p-value	β2	p-value	β2	p-value
Model 1, interaction between financial stress and sex3	0.10	0.146	0.12	0.067	0.12	0.066
Women	0.11	0.003	0.05	0.199	0.11	0.003
Men	0.21	<0.001	0.17	0.001	0.23	<0.001
Model 2, smoking Women, direct effect	0.10	0.007	0.05	0.215	0.11	0.006
Indirect effect, mediation	0.00	0.371	0.00	0.886	0.00	0.388
Total effects (direct + indirect)	0.10	0.005	0.05	0.214	0.12	0.002
Men, direct effect	0.19	<0.001	0.17	0.001	0.22	<0.001
Indirect effect, mediation	0.02	0.078	0.01	0.295	0.02	0.069
Total effect	0.21	<0.001	0.17	0.001	0.24	<0.001
Model 3, men, interaction (showing moderation) between financial stress and smoking
Interaction			−0.19	0.099
Smokers			0.20	<0.001
Non-smokers			0.01	0.893
Model 2, alcohol
Women, direct effect	0.10	0.005	0.05	0.215	0.11	0.005
Indirect effect, mediation	0.01	0.298	0.00	0.952	0.00	0.601
Total effect	0.10	0.006	0.05	0.215	0.11	0.005
Men, direct effect	0.20	<0.001	0.17	0.001	0.23	<0.001
Indirect effect, mediation	0.00	0.644	0.00	0.805	0.00	0.265
Total effect	0.20	<0.001	0.17	0.001	0.23	<0.001
Model 3, men, interaction (showing moderation) between financial stress and alcohol
Interaction	−0.03	0.086			−0.04	0.040
Non-drinkers4	0.27	<0.001			0.31	<0.001
Weekly drinkers4	0.13	0.057			0.15	0.022
Model 2, physical activity
Women, direct effect	0.11	0.004	0.05	0.184	0.11	0.004
Indirect effect, mediation	0.00	0.276	0.00	0.239	0.00	0.242
Total effect	0.10	0.006	0.05	0.221	0.11	0.005
Men, direct effect	0.20	<0.001	0.17	0.001	0.23	<0.001
Indirect effect, mediation	0.00	0.861	0.00	0.816	0.00	0.851
Total effect	0.20	<0.001	0.17	0.001	0.23	<0.001
Model 2, cohabiting
Women, direct effect	0.10	0.005	0.05	0.211	0.11	0.005
Indirect effect, mediation	0.00	0.641	0.00	0.826	0.00	0.683
Total effect	0.21	<0.001	0.18	0.001	0.24	<0.001
Men, Direct effect	0.19	0.001	0.14	0.005	0.21	<0.001
Indirect effect, mediation	0.02	0.201	0.03	0.017	0.03	0.087
Total effect	0.10	0.005	0.05	0.213	0.11	0.005
Model 2, non-manual work
Women, direct effect	0.11	0.005	0.05	0.211	0.11	0.005
Indirect effect, mediation	0.00	0.806	0.00	0.805	0.00	0.806
Total effect	0.10	0.005	0.05	0.201	0.11	0.004
Men, direct effect	0.20	<0.001	0.16	0.002	0.23	<0.001
Indirect effect, mediation	0.00	0.809	0.03	0.022	0.01	0.399
Total effect	0.21	<0.001	0.18	<0.001	0.24	<0.001
Model 2, sleep disturbances among women
Women, direct effect	0.09	0.016	0.05	0.258	0.09	0.012
Indirect effect, mediation	0.01	0.123	0.00	0.264	0.01	0.133
Total effect	0.10	0.006	0.05	0.217	0.11	0.006
Model 3, women, interaction (showing moderation) between financial stress and sleep disturbances
Interaction			0.12	0.093
No sleep disturbances			−0.09	0.490
Sleep disturbances			0.15	0.054
Model 2, sleep disturbances among men
Direct effect	0.19	0.001	0.16	0.001	0.22	<0.001
Indirect effect, mediation	0.02	0.101	0.01	0.298	0.01	0.118
Total effect	0.20	<0.001	0.17	0.001	0.23	<0.001

Linear associations between financial stress as independent variable and impairment as dependent variables, estimated with path analyses with maximum likelihood estimation using cluster correlated standard errors. Separate results for women and men are from similar models, just changing reference category for sex.

Model 1: Controlling for age at baseline, follow-up time, years of education, and the covariance between worse cognitive functioning, physical impairment, and cognitive*physical impairments (combined impairments); Model 2: Additionally controlling for one possible modifying factor in each model. Total/direct/indirect effect refers to the association between financial stress and impairment; Model 3: additionally controlling for interaction between modifying factor and financial stress. Only interaction and direct effects are presented. Only interactions with p < 0.10 are presented.

¹⁾multiplicative term physical impairment * worse cognitive functioning, standardized to have the standard deviation of 1.0. The term reflects the combination of physical impairment and worse cognitive functioning.

²⁾standardized β-coefficient.

³⁾ref = women.

⁴⁾Alcohol was given linear representation, centered at non-drinkers and weekly drinkers respectively. Sleep disturbances was given linear representation, centered at no sleep disturbances and fairly severe sleep disturbances (1.5 on the scale ranging 0–2) respectively – the value 1.5 indicates severe disturbances on one of the items and moderate on the other.

Significant indirect effects (mediation between financial stress and the outcomes) from possible modifying factors were found for cohabiting among men in relation to worse cognitive functioning (without the presence of physical impairment) (β = 0.03; p = 0.017) and manual vs non-manual workers, also in relation to worse cognitive functioning (β = 0.03; p = 0.022).

The interaction found between financial stress and smoking in relation to worse cognitive functioning (presented in Table 3) was only significant among men (Table 4). Among male smokers there was a significant association between financial stress and worse cognitive functioning (β = 0.20; p < 0.001), not among male non-smokers and among women. Among both women and men there were significant associations between financial stress and physical impairments and combined impairments. Among women there were no significant association between financial stress and worse cognitive functioning.

Sleep disturbances as a mediating factor in the relation between financial stress and physical impairments or the combined impairments (Table 3), was no longer significant when stratifying the analyses by sex. However, an interaction with p = 0.093 indicates that the association between financial stress worse cognitive functioning only was present among women with sleep disturbances (Table 4).

Interactions between financial stress and possible modifying factors with p < 0.10 were also found for alcohol in relation to physical impairment (β = −0.03; p = 0.086) and to the combined impairments among men (β = −0.04; p = 0.040). Direct associations between financial stress and impairment were found among women only in relation to physical impairments (β = 0.10) and to the combined impairments (β = 0.11). Among men, significant direct effects were found in relation to worse cognitive functioning (β = 0.17; p = 0.001), to physical impairments among non-drinkers (β = 0.27; p=<0.001, and a trend among weekly drinkers β = 0.13; p = 0.057), and to the combined impairments among both non-drinkers and weekly drinkers – but the association was significantly stronger (P = 0.040) among non-drinkers (β = 0.31 for non-drinkers and β = 0.15 for weekly drinkers).

4. Discussion

The aim of this study was to assess the associations between midlife financial stress and late-life cognitive and physical impairments (separately and combined), and the role of modifiable lifestyle and sociodemographic factors, and potential sex differences. Our results are the first to show that midlife financial stress was associated with combined cognitive and physical impairments, as well as worse cognitive functioning and physical impairment separately. Among the potentially modifiable factors that we investigated, smoking and sleep disturbances had a significant mediating effect in the associations between financial stress, physical impairment, and the combined impairments. Smoking also moderated the association between financial stress and worse cognitive functioning.

Interaction analyses showed that financial stress was only significantly associated to worse cognitive functioning among male smokers. Among men alcohol consumption interacted with financial stress on physical and combined impairments; among non-drinkers, financial stress was associated with physical impairment and combined impairments, and to a lesser extent among weekly drinkers. Among men, the associations between financial stress and worse cognitive functioning were mediated by cohabitation and socioeconomic position (manual vs. non-manual work). Among women with more sleep disturbances, there was a trend that financial stress was related to worse cognitive functioning.

These results show that financial stress in midlife is associated with cognitive and physical impairments in later life, both separately and combined. These results support previous findings showing that financial stress, income volatility and lower socioeconomic position are associated with worse cognitive performance or higher dementia risk (Chen et al., 2022; Grasset et al., 2019; Wang et al., 2022). The current results also support studies showing that financial stress is associated with poor physical function, including musculoskeletal disorders, poor physical health, and functional limitations (Ahnquist et al., 2012; Kahn and Pearlin, 2006; Wickrama and O’Neal, 2021). The current study adds to the literature by demonstrating that financial stress is also associated with combined cognitive and physical impairments, i.e., co-existence of both conditions. The results are relevant for the literature on cognitive frailty (which consists of physical frailty and cognitive impairment measured by the Clinical Dementia Rating scale score of 0.5) (Kelaiditi et al., 2013). This literature showed that cognitive frailty is more prevalent among women (Qiu et al., 2022), and is associated with an increased risk for dementia, disability, hospitalization and mortality (Kojima, 2016; Solfrizzi et al., 2017; St John et al., 2017). Previous results had also shown that cognitive frailty was more likely among those with lower income (Lu et al., 2022), which complements the current results on the role of financial stress.

Among the modifiable factors investigated, smoking showed a significant mediating effect in the associations between financial stress and physical impairment and combined impairments. Among male smokers only (not among women or male non-smokers), there was a significant direct effect between financial stress and worse cognitive functioning. The associations between smoking and financial stress may be bidirectional, as it may be used as a coping mechanisms to handle financial stress and the challenges to meet one’s needs with the available household income, while financial stress may increase the risk for smoking including heavy smoking (Widome et al., 2015). Financial stress may even prevent smokers from being able to quit smoking (Siahpush et al., 2009). Smoking is also a risk factor for cognitive impairment and physical impairment (Amiri and Behnezhad, 2020), including among samples with lower socioeconomic status (Zhang et al., 2016). In the current analyses smoking had an important role in exacerbating the associations between financial stress and late-life impairments, highlighting it as an additional risk factor for participants experiencing financial stress. The sex differences in these associations may reflect smoking quantity, as men in this sample smoke a higher number of cigarettes per day compared to women (Rusanen et al., 2010).

With regards to alcohol consumption, among non-drinker males, financial stress was associated with physical impairments and combined impairments, while among weekly drinkers, similar associations were weaker (a significant association with the combined impairments and a trend for physical impairment). Previous evidence had shown that worsening in financial strain increases the risk for heavy alcohol consumption and smoking, and these associations were more pronounced among men and among participants with lower education levels (Shaw et al., 2011). The authors suggested this to support the tension-reduction hypothesis, as these unhealthy behaviors are perceived to (temporarily) reduce stress and alleviate negative emotions (Shaw et al., 2011; Wolkowicz NR et al., 2022), although they have detrimental long-term health consequences. A specific example was shown from the US 2008–2009 recession, where adults who experienced severe economic loss (related to housing or employment) showed an increase in unhealthy alcohol consumption compared to those who experienced moderate economic losses (Wolkowicz NR et al., 2022).

That male non-drinkers showed stronger associations with physical impairment and the combined impairment may be consistent with the literature regarding the inverted-U or J shaped association between alcohol consumption and certain health outcomes (Rehm et al., 2017). Previous studies showed that non-drinking in mid- or late-life was associated with a higher risk for cognitive impairment, and poorer performance on multiple cognitive domains (Ngandu et al., 2007; Sabia et al., 2018; Zhang et al., 2016), while heavy drinking also increased dementia risk (Sabia et al., 2018). Similar associations were found for non-drinking of alcohol and higher odds of physical limitations (Sabia et al., 2018). The co-existence of several midlife risk factors such as alcohol abuse or alcohol abstinence, smoking and physical inactivity was also related to later disability (Artaud et al., 2016). As described in the literature, low-moderate drinking may be a proxy for other health-promoting factors such as more social interactions, more leisure and cognitively stimulating activities, while those who abstain from drinking may do so due to comorbidities, previous alcohol dependence/abuse, which increase their risk for health outcomes. Interactions may have not been found among women due to their lower alcohol consumption during the measurement periods (Tigerstedt et al., 2020), the type of alcohol consumed (for example women tend to consume less spirits than men) (Makela et al., 2006), or due to having additional coping resources such as tangible or instrumental social support (Aslund et al., 2014), all of which may protect cognitive function in the presence of financial stress and alcohol consumption.

For sleep disturbances, there was a mediation effect for the associations between financial stress and physical impairments and the combined impairments. Among women, there was also an interaction trend where the association between financial stress and worse cognitive functioning was stronger among those with more sleep disturbances, more specifically insomnia. Insomnia tends to be more prevalent among women (Jee et al., 2020), and the pattern of sex differences in the results may have been different if other measures of sleep disorders had been available. Financial stress is associated with sleep disturbances (Hall et al., 2009), and this may be driven by insomnia due to excessive worrying and anxiety about meeting one’s needs. Recent evidence has shown that sleep disturbances may be an important risk factor for dementia, cognitive impairment (Sindi et al., 2018; Sindi et al., 2018b; Xu et al., 2020), and physical impairments (Amiri and B, 2021). Improving sleep habits and managing the anxiety caused by financial stress may buffer against the later risks for cognitive and physical impairments.

Our results showed that among men there was a significant mediation by cohabitation in relation to worse cognitive functioning, which is aligned with previous literature. In our sample, cohabitation included individuals who are living with a partner/relative, regardless of their marital status. Previous results from the US and Sweden showed that unmarried men had a higher risk for dementia, while among women, marital status was not associated with dementia risk (Najar et al., 2021). Another study also showed that non-married participants had a higher risk for dementia, and that divorced men may have a more pronounced risk than divorced women (Sundstrom et al., 2016). Other findings showed that regardless of cohabitation status, unmarried participants had a higher risk for dementia (Liu et al., 2020). It may be that longer cohabitation duration is important to have a protective role. In relation to financial stress, the marital resource conceptual model posits that marriage offers psychological, social and financial benefits, and some of these factors may extend to cohabitation without marriage (Williams K and Carlson, 2010). Cohabitation may also increase cognitive stimulation and contribute to increasing cognitive reserve (Sundstrom et al., 2016), which then postpones the onset of cognitive impairment symptoms regardless of brain pathology. The mediating effect of cohabitation for the associations between financial stress and worse cognitive functioning was more pronounced among men. This may be attributed to the cohabitant’s role in maintaining/promoting social contacts and reducing social isolation, which are more common attributes among women (Cornwell, 2011). As the cumulative role of financial and marital stress were previously shown to be correlated (Wickrama and O’Neal, 2021), it would be interesting for future studies to investigate these interactions in relation to late-life impairments.

Similarly, our results suggested that among men there was a mediation effect by non-manual work - which is an indicator of higher socioeconomic position – in association with worse cognitive functioning. Various studies suggest that non-manual work, which is more cognitively stimulating than manual work, is associated with reduced cognitive impairment (L. Y. Huang et al., 2020; Kim et al., 2020; Tomioka et al., 2020). Evidence also demonstrates that work complexity is associated with reduced cognitive impairment, while passive jobs and those characterized by high strain increase the risk (L. Y. Huang et al., 2020). The current results lend support to recent results showing that among men, passive jobs were associated with worse cognitive functioning and physical impairments, and low job strain was associated with physical impairments (Sindi et al., 2023). Even though we adjusted for education in the current analyses, the associations we observe may be due to the types of occupations that men held, considering their higher education levels and the division of household and care responsibilities. The combination of higher education and occupational complexity enhance cognitive reserve and prevent or postpone cognitive decline (Stern et al., 2020).

In our results we did not find an interaction between age and midlife financial stress. This might indicate that the timing of the stressor (within the midlife stage) is less important for physical and cognitive functioning in old age. It should however be noted that age varied between 39 and 64 years at baseline. Exposure to financial stress in earlier or later ages might have different effects. Indeed, the timing of financial stress has been highlighted in recent literature. For example, in a study that integrated financial strain at different life periods: before age 18 years, 18–35 years, 35–50 years, 50–65 years, and the strongest associations between timing of financial stress and self-reported physical health appears to be for the period 35–50 years (Kahn and Pearlin, 2006), which is also partially captured by the midlife measure of financial stress in the current study. Having additional financial stress between 50 and 60 years was associated with self-reported illness and physical function (Kahn and Pearlin, 2006). Importantly, midlife and early late-life financial stress had a negative association with late-life physical health only if preceded by financial stress before the age of 35 years (Kahn and Pearlin, 2006). The cumulative associations between financial stress across the lifespan and both physical and cognitive impairments are important to consider in future studies with such measures available.

It is noteworthy that stress and lifestyle cannot be fully disentangled from a social determinants of health perspective. According to this approach the emphasis is on societal factors (including inequalities and ideologies), rather than having the responsibility limited to the individual (Short and Mollborn, 2015). In the current study, we focused on self-reported financial stress and the potential modifying role of various health behaviors (e.g. smoking, alcohol consumption, physical inactivity). In addition to investigating such ‘downstream’ (micro level) determinants of health at the individual level, it will be important for future research to examine the role of the next ‘meso’ level, including neighborhoods, workplaces, families and the norms and interpersonal connections in such settings (Short and Mollborn, 2015), as these can either promote or prevent the adoption of healthy behaviors. For example it has been shown that neighborhood deprivation is associated with health and health-related behaviors (Jivraj et al., 2020). These ‘meso’ level factors need to be considered in the context of the macro-level factors such as health care systems.

The mechanisms and pathways through which financial stress are associated with physical and cognitive function warrant further investigation. For example, financial stress is associated with interpersonal stressors (e.g. with spouse, family, and work colleagues) (Sturgeon et al., 2016). This would potentially impact social support and social support networks, which can buffer against high levels of perceived stress, poor psychological wellbeing, and psychosomatic symptoms among individuals experiencing financial stress (Aslund et al., 2014; Park N et al., 2017). Midlife family financial strain can increase the risk for pain-depressive symptoms, which elevates the odds of having subjective memory problems and more physical limitations (Wickrama et al., 2022). It has also been shown that individuals with higher levels of perceived social status had the strongest associations between financial stress and the inflammation marker c-reactive protein (CRP) (Sturgeon et al., 2016). Changes in financial strain have been associated with ambulatory blood pressure and the stress hormone cortisol (Steptoe et al., 2005).

This study has a few limitations worth noting. First, we only had a measure of financial stress in midlife, so we could not examine the associations between change in financial stress and late-life impairments. Second, we did not have a measure of financial social support, which may alter the levels of perceived financial stress. Also, the measure of physical impairment was self-reported, which may have under- or overestimated the symptoms of physical impairment considering the subjective nature of such data. The ADL measure was limited to certain domains (e.g. washing and dressing) rather than including a broader range of items. The strengths of this study include the large representative population-based cohort, long follow-up duration (three decades), the cognitive measure was based on several tests representing multiple cognitive domains, and the possibility to investigate several modifiable factors that may mediate or moderate the associations between financial stress and late-life impairments.

This study has important implications for various stakeholders in society. Middle-aged adults may be aware of the risks associated with work-related stress or other inter-personal conflicts, but an increasing awareness that chronic financial stress may encourage those who experience such stress to enhance buffering factors such as social support, enhanced social networks, physical activity and management of sleep disturbances. Similarly, clinicians can be encouraged to routinely collect information regarding various sources of stress, including financial stress, so that appropriate interventions (e.g. psychological counselling, cognitive behavioural therapy for insomnia, recommendations regarding lifestyle interventions) may be recommended to decrease their risk for cognitive and physical impairments. Finally, policy makers can consider how to further optimise the financial situation for such individuals, through re-assessment of social benefits, income ranges, availability of resources for financial literacy and guidance and current employment policies.

In conclusion, this study shows that midlife financial stress is associated with late-life cognitive and physical impairments, and that various modifiable factors may mediate or modify these associations, including smoking, alcohol consumption, cohabitation, non-manual work and sleep disturbances. We also observed sex differences in the role of the modifying factors, where smoking, alcohol consumption, cohabitation, non-manual work had more pronounced modifying effects (mediation or moderation) among men, while sleep disturbances played a more important role among women. Future research would benefit from integrating longitudinal measures of financial stress and investigating the interactions with other stressors and comorbidities related to physical and mental health.

Data availability

Data will be made available on request.