Abstract
Background:
Wealth and income are potential modifiable risk factors for dementia, but whether wealth status, which is composed of a combination of debt and poverty, and assessed by wealth and income, is associated with cognitive impairment among elderly adults remains unknown.
Objective:
To examine the associations of different combinations of debt and poverty with the incidence of dementia and cognitive impairment without dementia (CIND) and to evaluate the mediating role of depression in these relationships.
Methods:
We included 15,565 participants aged 51 years or older from the Health and Retirement Study (1992–2012) who were free of CIND and dementia at baseline. Dementia and CIND were assessed using either the modified Telephone Interview for Cognitive Status (mTICS) or a proxy assessment. Cox models with time-dependent covariates and mediation analysis were used.
Results:
During a median of 14.4 years of follow-up, 4,484 participants experienced CIND and 1,774 were diagnosed with dementia. Both debt and poverty were independently associated with increased dementia and CIND risks, and the risks were augmented when both debt and poverty were present together (the hazard ratios [95% confidence intervals] were 1.35 [1.08–1.70] and 1.96 [1.48–2.60] for CIND and dementia, respectively). The associations between different wealth statuses and cognition were partially (mediation ratio range: 11.8–29.7%) mediated by depression.
Conclusion:
Debt and poverty were associated with an increased risk of dementia and CIND, and these associations were partially mediated by depression. Alleviating poverty and debt may be effective for improving mental health and therefore curbing the risk of cognitive impairment and dementia.
INTRODUCTION
Dementia and cognitive impairment are common contributors to severe disability, increased demands for medical and personal care, and premature death in older people, resulting in a major societal burden [1, 2]. Worldwide, approximately 50 million people had dementia in 2017, and these numbers are expected to increase to 152 million by 2050 [3]. Since effective treatments for dementia and cognitive impairment are lacking, prevention has been highlighted as a pivotal component to reduce its burden, and research on the etiology of these conditions and modifiable risk factors is warranted [4].
Previous studies have shown that both wealth and income are modifiable risk factors for dementia, as economic factors have an impact on digital literacy, which is in turn associated with the risk of dementia [5–7]. There is a difference between income and wealth; for instance, retired older adults may have little income but may have accumulated substantial wealth through a lifetime of work, while those who bear substantial debt may have little wealth despite having a high level of income. However, research on the association between wealth and income combined is limited. In addition, although poverty, defined as a low income, has been suggested to be associated with an increased risk of poorer cognitive function [8, 9], the relationship between wealth status considering poverty combined with debt, which is estimated by negative wealth, and the development of dementia and cognitive impairment has not been investigated.
Despite growing evidence linking wealth to cognition [5, 6], the exact mechanism is not well understood. In developed countries, it has been established that wealth is associated with worse mental health outcomes, including depression and other psychiatric disorders, and these mental conditions were linked to the development of dementia and cognitive impairment [10, 11], which suggests the possibility of a pathway from wealth status to cognition via mental health [12, 13]. To our knowledge, very few studies have examined whether depression plays a mediating role in the association of wealth status with dementia and cognitive impairment.
To provide evidence that policy makers can use to implement preventive interventions to improve cognitive performance, the current study comprehensively investigated whether wealth status, which is composed of a combination of poverty and debt, is a potential risk factor for the development of dementia and cognitive impairment and whether depression symptoms mediates these associations among older adults in the U.S.
METHODS
Study participants
Data were drawn from 1992–2012 Health and Retirement Study (HRS), which is a nationally representative longitudinal survey of individuals aged 51 years or older in the U.S., with response rates ranging from 68.8% to 92.3% [14]. The initial interviews took place in 1992 and HRS participants were reinterviewed every 2 years until 2014 on a range of sociodemographic, economic, health, and cognitive measures. Detailed information about the HRS is available elsewhere [15]. Since the measures of cognition used in HRS waves 3–12 were adapted from the modified Telephone Interview for Cognitive Status (mTICS), wave 3 was set as the baseline. In the analysis, we excluded participants who attended only one visit (n = 125). We further excluded participants who had dementia or CIND at baseline (n = 1,491). The final analytic sample included 15,655 participants who were free of dementia or CIND at the baseline survey (Supplementary Figure 1). The HRS was approved by the Health Sciences and Behavioral Sciences institutional review board at the University of Michigan, and written informed consent was provided by all participants.
Measurement of assets and income
At each HRS interview, participants were asked questions assessing the net value of their financial assets and debts. The total net assets of each participant were calculated by subtracting the sum of all imputed debts from the sum of all imputed assets. Participants were defined as being in debt if the value of total net assets was negative. Poverty status was assessed using the poverty thresholds which are updated annually by the U.S. Census Bureau and was used to define and quantify poverty in the U.S. [16]. Participants were categorized as living in poverty if the household income, which was assessed by the sum of all earnings in a household in the last year, was below the weighted average poverty threshold (Supplementary Table 1). The total net assets and per capita household income (household income divided by household size) were divided into quartiles among participants without debt and without poverty, respectively. Participants were also divided into 4 groups based on combinations of wealth status: no debt and nonpoverty, in debt but nonpoverty, no debt but in poverty, and in debt and in poverty.
Cognitive assessment
The measurement of cognition in the HRS differs for self-respondents (95.61%) and proxy respondents (4.39%). For self-respondents, cognitive performance was assessed based on the mTICS, and a global cognition score was obtained by summing the following three items: immediate and delayed recall of a list of 10 words randomly assigned to each participant (0–20 points), backward counting (0–2 points), and the serial 7 s test (participants are asked to count down from 100 by 7 for a total of five trials, 0–5 points) [17, 18]. The final summary score ranges from 0 to 27, and lower scores indicate greater cognitive impairment. According to score cut-offs developed by Langa and Weir, we categorized respondents who scored 0 to 6 points as having dementia, those who scored 7 to 11 as having CIND, and those who scored 12 to 27 as having normal cognition [19]. For proxy respondents, an 11-point scale proxy assessments was performed based on previous studies [20]; those with scores of 6–11 were classified as having dementia, and those with scores of 3–5 were classified as having CIND. The mTICS had a weighted accuracy of 69% in the correct classification of participants as normal, with CIND, or with dementia [19], and the Langa and Weir cut-offs including self-reports and proxy-reports were validated with a sensitivity of 75% (95% CI, 47–60%), a specificity of 83% (95% CI, 80–83%), and an accuracy of 81% (95% CI, 78–83%) [2, 21].
Mediator assessment
We examined baseline depression symptoms as a potential mediator between different combinations of debt and poverty and cognition. Depression symptoms in the HRS were measured separately via the eight-item Center for Epidemiologic Studies Depression Scale (CESD-8), which asks respondents if they had experienced eight specific depressive symptoms over the past week. Depression was defined when the respondent had a CESD-8 score greater than 3 [22]. The sensitivity and specificity of the CESD-8 cutoff were 71% and 79%, respectively [22].
Covariates
We obtained a range of baseline characteristics of the participants to include as covariates in the models. Sociodemographic information included age, gender, race (white/Caucasian, black/African American, or other), marital status (married or not married), and educational level completed (less than high school, high school or above). Lifestyle factors included smoking status (nonsmoker, past smoker or current smoker) and alcohol consumption (nondrinker, light drinker or heavy drinker) based on average weekly alcohol assumption [23] and regular exercise (yes or no). Health conditions were collected by self-report of doctor-diagnosed medical comorbidities, including hypertension, diabetes, heart diseases, stroke, lung diseases, and cancer.
Statistical analysis
Wealth status and cognitive impairment and dementia
The study cohort was described using mean (standard deviation: SD) for continuous variables and frequencies (%). Since the proportional hazards as-sumption was violated in Schoenfeld’s residuals tests, we used Cox proportional hazards regression models with time-dependent covariates to assess the association between cognitive status and risk of CIND and dementia, and the associations were analyzed at time intervals between baseline and the end of follow-up. Since only HRS waves 3–12 used cognitive measures adapted from the mTICS, for participants who were enrolled in waves 1-2, wave 3 was set as the baseline and the date of entry was set as the baseline for participants who were enrolled in wave 3 or after. Participants were followed from baseline until the time of CIND or dementia diagnosis or censoring. Those who remained free of CIND or dementia were censored at the time of death, loss to follow-up, or at the end of the follow-up period (April 15, 2015), whichever occurred first. Incidence rates of dementia or CIND (per 100 person-years) according to different wealth statuses were calculated by dividing the number of cases in each wealth status category by the number of person-years at risk of participants within that category. The following three Cox proportional hazards models with time-dependent covariates were used to assess these associations: model 1 was adjusted for age at entry, sex, and race; model 2 added marital status, education level, physical activity, smoking status, and alcohol assumption; and model 3 was additionally adjusted for hypertension, diabetes, heart diseases, stroke, lung diseases, and cancer. Missing values accounted for < 1.10% of all the covariates. Participants with missing values for any of the categorical covariates were assigned to a separate “unknown” category. Missing BMI (1.06%) was imputed using bootstrapped multiple imputation. All covariates were treated as time-dependent covariates.
Mediation analysis
To test the mediating effect of depression on the association between different wealth statuses and cognition, we further conducted mediation analysis using Lavaan, an R Package for Structural Equation Modeling (SEM) [24]. The direct model included the cognition scores as the outcome (Y) and the independent variable (X = debt, poverty, or a combination of the two) as the predictor. The standardized regression coefficient of wealth status was the total effect c. The mediation models had the same outcome and predictor, with depression (M) added as mediator. The mediation effect of depression symptoms was the product of a and b, which represent the indirect pathway between X and Y. The mediation ratio was the indirect effect ab divided by the total effect c. Bootstrapped confidence interval (CIs) of the direct effect and indirect effect was computed to test the significance of the mediation effect. It can be concluded that effects are significant when zero is not in the 95% CIs. The schematic mediation model is shown in Supplementary Figure 2. Only the cognitive scores of the self-respondents (95.61%) were used in mediation analysis.
Subgroup and interaction analysis
Interactions between wealth status and the following baseline characteristics of the participants with respect to CIND and dementia were further explored: age (< 65 and ≥65 years), gender (male and female), education level (high school and above or not), smoking history (yes/no), depression (yes/no), diabetes (yes/no), and stroke (yes/no). Analyses stratified by the aforementioned characteristics were then presented using the Cox regression models with time-dependent covariates to derive the stratum-specific HRs.
All analyses were conducted using Stata (version 15; StataCorp, College Station, Texas) and RStudio Software, and p < 0.05 was considered statistically significant.
RESULTS
Baseline characteristics
A total of 15,655 participants in the resulting analytical sample were followed up for a median of 14.39 (interquartile range [IQR], 5.9–17.8) years. Of the included HRS participants, 4,484 (28.64%, 2.6 cases per 100 person-years) incident CIND cases and 1,774 (11.33%, 9.3 cases per 1000 person-years) incident dementia cases occurred (Table 1). The study population was divided into four groups based on different combinations of debt and poverty status. Specifically, 13,068 (83.47%), 1,624 (10.37%), 681 (4.35%), and 282 (1.80%) participants were classified into the “no debt and nonpoverty”, “no debt but in poverty”, “in debt but nonpoverty” and “in debt and poverty” groups, respectively. The mean cognitive scores across groups were 17.27, 15.05, 16.41, and 14.54, respectively. Comparisons of baseline characteristics between included and excluded participants are shown in Supplementary Tables 2 and 3.
Baseline characteristics of the study population according to wealth status combinations, HRS, 1992–2012
HRS, the Health and Retirement Study; CIND, cognitive impairment without dementia; BMI, body mass index. aIn debt and poverty versus no debt and nonpoverty. bThe per capita household income of each participant was calculated as the household income divided by the number of family members.
Associations of wealth with CIND and dementia risk
At baseline, 963 (6.15%) participants were in debt, and 1,906 (12.18%) were in poverty (Table 2). In the multivariable adjusted analyses, the hazard ratios (HRs) and 95% confidence interval (CI) associated with debt were 1.29 (95% CI 1.13–1.47) for CIND and 1.29 (1.05–1.57) for dementia, and the HRs (95% CI) associated with poverty were 1.17 (95% CI 1.06–1.28) for CIND and 1.86 (1.66–2.09) for dementia. Among participants without debt/poverty, individuals with more total assets/higher income quartiles were less likely to experience CIND or dementia than those in the lowest quartiles, associations that remained, though weakened, in the fully adjusted model. Furthermore, an inverse relationship between the value of total assets and income and the risk of dementia was observed: the HR (95% CI) comparing the extreme assets and income quartiles was 0.42 (95% CI: 0.35–0.50; p trend < 0.001) and 0.55 (95% CI: 0.45–0.67; p trend < 0.001), respectively. A similar pattern was displayed for the relationship between the value of total assets and income and the risk of CIND.
Hazard ratios and 95% CIs for the associations between wealth status and the risk of CIND and dementia
Model 1: adjusted for age, sex and race. Model 2: adjusted for model 1 plus marital status, education level, physical activity, smoking status, and alcohol consumption. Model 3: adjusted for model 2 plus hypertension, diabetes, cancer, lung disease, heart disease, and stroke. aThe linear trend was tested by treating the mean values for each quartile of total assets or income as a continuous variable. CIND, cognitive impairment without dementia; CI, confidence interval.
We also examined the associations of different combinations of wealth status and risk of CIND and dementia (Table 3). Participants living in debt and poverty had the highest risk of CIND (fully adjusted HR [95% CI] comparing those with no debt and nonpoverty: 1.35, 95% CI 1.08–1.70), with the highest incidence rate of 4.4 per hundred years (95% CI: 3.5–5.4). A similar association was observed between wealth status and the risk of dementia, with participants living in debt and poverty having the highest risk (HR: 1.96, 95% CI 1.48–2.60) and incidence rate (2.7 per hundred person-years, 95% CI: 2.1–3.6).
HRs (95% CI) for CIND and dementia stratified by different combination of wealth status
Model 1: adjusted for age, sex and race. Model 2: adjusted for model 1 plus marital status, education level, physical activity, smoking status and alcohol consumption. Model 3: adjusted for model 2 plus hypertension, diabetes, cancer, lung disease, heart disease and stroke. CIND, cognitive impairment without dementia; HR, hazard ratio; CI, confidence interval.
Mediation effect: Depression as the mediator
Figure 1 presents the estimated SEM models for exposure (different wealth status) and cognition, and standardized coefficients (95% CI) based on bootstrapping are displayed. The results suggest that depression had a significant mediating effect on the association between three different wealth statuses and cognition. Specifically, depression mediated 29.70% of the debt-cognition association (mediation effect = –0.49, 95% CI: –0.58, –0.39), 11.83% of the poverty-cognition association (mediation effect = –0.32, 95% CI: –0.39, –0.24), and 25.34% of the debt and poverty-cognition association (mediation effect = –0.67, 95% CI: –0.82, –0.52).
Mediation analysis models showing the direct and indirect (mediation) effects of wealth status on cognition, with depression as mediator. A) The direct and indirect effects of debt on cognition; B) the direct and indirect effects of poverty on cognition; C) the direct and indirect effects of debt and poverty on cognition. Total effect c = c’ + ab, association between the independent variables (X = in debt/in poverty/in debt and poverty) and the outcome variable (Y = cognition); a, association between X and the potential mediator (M = depression); b, association between M and Y (cognition), adjusted for X; c’, direct association between X and Y after controlling for M; mediation effect = ab; mediation ratio = ab/c×100%. *p < 0.05, **p < 0.001.
Additional analysis
We performed stratified analyses for associations of the wealth status combinations with incident CIND and dementia according to potential risk factors using the fully adjusted model (Fig. 2). Women and participants with a younger age (< 65 y) had higher adjusted HRs for the risk of CIND and dementia than men (p for interaction were 0.000 and 0.003, respectively). However, the associations between wealth status and CIND and dementia were not significantly modified by other factors. Since gender and age tended to be effect modifiers, we performed additional analyses regarding the sex-specific or age-specific effects of wealth status on CIND and dementia risk (Supplementary Tables 4 and 5). Compared to men, women tended to exhibit a stronger effect on the risk of dementia but a weakened association with CIND. Additionally, the association of wealth status with the risk of CIND and dementia among men became null, while that of women remained significant. In addition, participants with younger age had a stronger magnitude effect than those of older individuals, which became null after multivariable adjustment.
Subgroup analysis of the associations between the wealth status combinations and the risk of cognitive impairment without dementia and dementia (in debt and poverty versus no debt and nonpoverty). Models were adjusted for age, sex, race, marital status, education level, physical activity, smoking status, alcohol consumption, hypertension, diabetes, cancer, lung disease, heart disease and stroke. CIND, cognitive impairment without dementia; HR, hazard ratio; CI, confidence interval.
DISCUSSION
In a nationally representative prospective cohort of U.S. adults aged 51 years or older who were free of cognitive impairment at baseline, we found that lower levels of wealth and income were independently associated with an increased risk of CIND and dementia. We also revealed a novel phenomenon that living in both debt and poverty was significantly associated with a 35% higher risk of CIND and a 96% higher risk of dementia, and these associations were partially mediated by depression. Furthermore, the adverse association of living in both debt and poverty with CIND and dementia appeared to be somewhat stronger in women than in men.
Our results confirmed other findings from several previous studies, which indicated an adverse effect of living in worse financial situations on dementia risk. In the National Health and Aging Trends Study, lower income and greater financial strain indicated an increased risk of incident dementia among older American adults [25]. In another community-based longitudinal study of older Mexican Americans, Hazzouri et al. demonstrated that life-course poverty was related to late-life dementia and CIND [26]. Moreover, our findings are in accordance with the result from a 12-year follow-up in the English Longitudinal Study of Aging, which suggested a positive association between lower wealth in late life and dementia incidence [5]. Our current study, with 2.39 additional years of follow-up, has provided further evidence supporting the association between low wealth and poverty and a higher risk of dementia/CIND. In addition, we found a novel positive association between living in debt defined as negative net wealth, not just poverty, and dementia/CIND, and the strength of the association was augmented when both factors were present together.
Lacking evidence has investigated the mechanisms underlying the association between wealth status and cognition. Our study moves beyond previous studies by examining potential pathways (i.e., depression symptoms) linking wealth status (i.e., debt, poverty, and a combination of both) and cognition. We found that depression symptoms explained part (25%) of the higher cognitive decline risk among those living in both debt and poverty, which suggested that undesirable wealth status promotes the development of cognitive impairment by affecting mental health. Other potential mediators that might play a role in the pathway from wealth status and cognition included less healthy lifestyles (e.g., habitual alcohol consumption), poorer general health, lifestyle-related diseases (e.g., diabetes) [27] and psychiatric symptoms [28]. Further investigations are warranted to explore other medications that underling the associations between specific economic dilemmas on cognitive impairment.
Poverty often parallels debt as a risk factor or a consequence, and people with long-term exposure to poverty tend to have overall debt [29]. People living both in poverty and debt face substantial uncertainty and income volatility, and sustained long-term exposure to the stress arising from managing this volatility may threaten mental health [30]. Evidence has suggested that long-term exposure to poverty and the subsequent debt can result in depressive symptoms due to exposure to more stressful life events, such as unemployment and poor ability to afford living expenses, which may contribute to a higher risk of cognitive impairment and dementia [10, 32]. One important potential explanation is that depression caused by long-term economic stress increases the risk of dementia through hyperactivity of the hypothalamic-pituitary-adrenal axis, and prolonged exposure to glucocorticoids in patients with depression leads to hippocampal atrophy and the development of dementia [33]. Another possible explanation for this association is that the biological underpinnings of depression and neurodegenerative diseases overlap considerably at a molecular level, including insufficient defenses against antioxidants and neurogenesis, increased apoptosis, and immune system dysregulation [34]. Furthermore, the risk factors common to both depression and dementia could explain the observed associations between these conditions, such as memory impairment, sleep disturbances, and impaired social functioning, which share common pathophysiological pathways including neurodegeneration, inflammation, vascular risk factors, and hypothalamic-pituitary-adrenal axis dysregulation [35–37].
Persons living with debt and poverty may also be at higher risk of developing dementia due to a preoccupation with pressing budgetary concerns [8]. These preoccupations can be present and distracting, and because the human cognitive system has limited capacity, they leave fewer cognitive resources available to guide choice and action [38, 39]. Another possible mechanism is that these financial conditions might influence cognitive load by changing people’s affective state.
Another interesting result was found in the subgroup analyses. Among women, a higher impact of debt and poverty on risk of incident cognitive impairment and dementia was observed, which was in line with the results from several cohort studies revealing that women in low-income areas have a higher incidence of dementia/cognitive impairment than men do [40, 41]. It was speculated that women are much more susceptible to anxiety than their peers when coping with the effects of living with financial hardships; this anxiety is related to inflammation and immune activation, leading to an increased anxiety-associated risk of cognitive decline [42, 43]. Women tend to stay at home more and to perform more of the domestic duties than men and are more affected by the stress and strain of living in poverty and debt. Other mechanisms underlying the results may be that women are more likely to experience inflammatory dysregulation, which is a risk factor for dementia [44], and have increased neurobiological vulnerability postmenopause [45]. In addition, younger participants (< 65 years) were found to have a higher risk of dementia and CIND than older elderly participants (≥65 years). This finding can be explained by the fact that younger adults are more likely than older adults to experience chronic stress and financial worry [46], such as the risk of unemployment, stress of supporting a family, which may contribute to more stress which increases the risk of dementia and cognitive impairment. It can also be speculated that among older adults, younger age is the significant risk factor for depression, which may contribute to a higher risk of cognitive impairment and dementia [47].
The strengths of our study include a prospective study design and lengthy follow-up period in a relatively large nationally representative sample of older adults, providing an adequate setting for detecting the long-term association between wealth status and the risk of dementia or cognitive impairment. Nevertheless, our study also has some limitations. First, our measure of dementia/cognitive impairment was based on cognitive tests and proxy reports rather than clinical diagnosis. Although the algorithms and cut-offs used in our study have been validated in previous studies [2, 21], there is still potential for differential misclassification that could underestimate the onset of dementia/CIND, which might underestimate the association between wealth status and dementia/cognitive impairment. Of course, the questionnaire is used for cognition test in large sample community population is more feasible. Second, although adjustments were made for a wide array of covariates, including major risk factors for dementia/cognitive impairment in the Cox regression models, data for a potential confounder (antidepressant use) were not available, which may have resulted in some residual confounding. However, evidence has shown that antidepressant use is not significantly associated with a change in cognitive function over a 6-year period, which makes our results more convincing [48]. Third, we excluded participants who had fewer than 2 follow-up visits or had dementia/CIND at baseline in our study, but their characteristics were different from the included participants: the excluded participants were younger, poorer, have a lower education level and less likely to regular exercise, and included a higher percentage of current smokers, diabetic, and had more depressive symptoms (Supplementary Table 2). The association between wealth status and cognition might have been diluted. Therefore, it might have led to selection bias, which may affect the internal validity of estimates and limit generalization to other populations. Finally, the observational design of our study prevented us from directly drawing a causal inference.
CONCLUSION
In conclusion, our study indicates that living in debt and poverty was associated with an increased risk of dementia and CIND, and these associations were partially mediated by depression. Public health policy should aim to improve mental health by alleviating poverty/debt and thereby reducing dementia risk, especially in undeveloped areas.
Footnotes
ACKNOWLEDGMENTS
This work was supported by the Open Project of the Guangdong Provincial Key Laboratory of Tropical Disease Research (G818330002). The funders played no role in study design or implementation; data collection, management, analysis, and interpretation; manuscript preparation, review, or approval; or the decision to submit the manuscript for publication.
