Association of Multidimensional Poverty With Dementia in Adults Aged 50 Years or Older in South Africa

Key Points Question Is multidimensional poverty associated with dementia in older adults in low- and middle-income countries? Findings In this cross-sectional study of 227 adults aged 50 years or older living in Soweto, Johannesburg, South Africa, 20% of adults with dementia were poor compared with 10% of those without dementia, with a difference of 118% on the Multidimensional Poverty Index. The risk of dementia was 2.31 times higher for adults who were multidimensionally poor. Meaning These results suggest that public policies addressing factors associated with poverty may be helpful to delay the onset and reduce the prevalence of dementia among older adults.


Introduction
Dementia is a global health challenge that affects high-income countries but has profound repercussions for low-and middle-income countries (LMICs). 1 The World Health Organization estimates that adults with all-cause dementia (estimates: 5%-7% for those Ն60 years) 2 will more than triple to reach 152 million by 2050, with at least 60% of those individuals living in LMICs. 3 This increase is associated with the overall growth of the population and the aging demographic. 4 Dementia-related costs will increase during the next 3 decades, and LMICs will be responsible for shouldering this economic burden. 5 Families will have to absorb these costs because formal geriatric care is in its infancy in LMICs. 6 Older adults are both the recipients and the providers of informal care, in which the reciprocity of care between generations has been shown to improve well-being. 7,8 This advantage disappears with dementia. Older adults with dementia become unable to provide care for other family members, a major problem in societies in which few formal systems of care exist. Similarly, adults with dementia need more financial, social and emotional, and physical support from other family members. This increased need directly affects the caregiver's ability to work and attend school, especially because many children are tasked with providing care. 9 The 2020 Lancet Commission on dementia prevention, intervention, and care identified 12 potentially modifiable risk factors for dementia, including poor education, hypertension, depression, low social contact, smoking, air pollution, and traumatic brain injury. 10 It is well documented that poor social determinants of health are directly associated with disease. [11][12][13][14][15][16] However, to our knowledge, limited research exists investigating how social determinants of health are associated with dementia. Context-specific studies using multidimensional poverty measures are required to better understand this association. 13,[17][18][19] A multidimensional approach to poverty encompassing various components of well-being is measured in terms of individuals' functioning and capabilities instead of resources or utility. This approach offers a more precise account of risk factors that eventually can trigger multiple conditions, including dementia, at a later stage in life and offers insight into how to improve care and policy. 20,21 For instance, multidimensional poverty has been shown to be associated with chronic health conditions. 22 Furthermore, various dimensions of deprivation, including chronic infectious diseases, poor health status, malnutrition, prenatal stress, 23 poor mental health, low access to health care, low level of literacy, 24,25 and lack of occupation, 26,27 are associated with income-based poverty and have been shown to be associated with aging and dementia. In LMICs, identifying specific dimensions of deprivation may help target interventions and public policies to mitigate an immediate risk of cognitive impairment, 28 support older adults and their caregivers, 29 and slow cognitive decline.
We searched PubMed and Scopus for scientific publications about poverty and dementia in LMICs since 2000. We used the following combination of search terms: dementia, Alzheimer, Alzheimer disease*, cognitive disorders AND poor OR poverty, and social determinants of health. The literature on dementia in LMICs is scarce, and little is known about the quantitative association between dementia and multidimensional poverty in LMICs. One study investigated risk prediction models of dementia applied to LMICs. 30 One study showed that individuals with Alzheimer disease have a lower level and smaller range of capabilities. 31 Since the end of apartheid nearly 3 decades ago, overall poverty levels have decreased in South Africa, the second-largest economy in sub-Saharan Africa. However, more than half the population continues to live with an income under the poverty line, and since 1994, almost 2.5 million South Africans, many of whom are Black or of mixed race, became poor. 32 Moreover, older age is widely recognized as the main nonmodifiable risk factor for dementia in South Africa and LMICs. 33 Estimates from studies conducted in South Africa indicate that the prevalence of dementia ranges from 3.8% to 11.0% for adults aged 65 years or older. [33][34][35] In this study, we examined multidimensional poverty and dementia in older adults in Soweto, South Africa. Our objective was to examine the association between multidimensional poverty and dementia and the role of various dimensions of deprivation in overall poverty. Using multidimensional poverty measures, we hypothesized that social determinants of health are positively associated with dementia in older adults.

Methods
Ethical approval for this cross-sectional study was obtained from the Faculty of Humanities Research Ethics Committee at the University of Johannesburg in South Africa and the Human Research Protection Office of Washington University in St Louis, Missouri. Written informed consent was obtained from all participants except for those who lacked the capacity to provide it; for these participants, verbal assent was obtained using a simplified text, and their caregiver also provided written consent. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for standard reporting in cross-sectional studies.

Study Design and Sample
Participants were first selected from a 2-stage, random cluster cross-sectional study conducted in

Multidimensional Poverty Measures
To measure multidimensional poverty, we included 7 dimensions that are central to well-being: education, health, economic activity, living standards, social participation, fair treatment, and psychological well-being (eTable 1 in the Supplement). Each dimension contained indicators identified in the literature as crucial to human development.
Education is associated with one's ability to gain employment and earn an income, which may have affected participants in this study who were exposed to the Bantu education system, Unemployment is an indicator of deprivation 46 and economic hardship 46 and a major challenge in South Africa, with one-third of the population unemployed. 47 The cutoff was used if an adult was unemployed, looking for a job, or not looking for a job because the participant was discouraged or could not afford the cost of seeking work or the wages offered were too low.
Household living standards were composed of 3 indicators (water pipe, electricity, and flush toilet), for which deprivation within the compound was the cutoff, which are important indicators of quality of life in South Africa. 48,49 Involvement in community groups has been shown to enhance life satisfaction. 50 Social participation, resulting in community interconnectedness, is a central feature of ubuntu (a philosophy that emphasizes the importance of community) in South Africa, and study participants who were not involved in any group were considered deprived. 51,52 Discrimination and stigma were measured using the validated 22-item Unfair Treatment subscale of the Discrimination and Stigma Scale. 53 Content and face validity tests were conducted. Moderate discrimination was the cutoff. 53,54 Finally, measures of depression and self-esteem represented deprivation of psychological wellbeing and are often associated with poverty, particularly among older adults. 55-58 Depression was measured using the 10-item Center for Epidemiologic Studies Depression Scale Revised (CESD-R-10), 59 already validated in South Africa. 60 A score of 10 or higher was the established cutoff.
Self-esteem was measured using the 10-item Rosenberg Self-Esteem Scale, validated in South Africa, 61,62 with a score below 15 as the established cutoff. Studies have shown that aging is associated with a decrease in self-esteem 63 and with multidimensional poverty. 64 There is no known association between dementia and low levels of self-esteem. However, dementia may lead to a sense of insecurity, affecting a person's ability to make decisions and to maintain personal relationships, employment, health, and finances. 58 As a result, inclusion of self-esteem in the multidimensional poverty measure is important.

Statistical Analysis
Statistical analysis was conducted from August 1 to 30, 2021. We used an unmatched, multidimensional poverty measure to identify differences in levels of poverty comparing those with and those without dementia. Dimensions of deprivation were independently assessed, and the method focused on dimensional shortfalls. 65 We aggregated dimensions of multidimensional poverty measures that consisted of 2 cutoffs: 1 for each dimension and 1 associated with crosscutting dimensions. Older adults were considered deprived if they fell below the cutoff on a given dimension. The second poverty cutoff determined the number of dimensions in which an older adult had to be deprived to be deemed multidimensionally poor.
One-way analyses examined differences in the poverty level between adults with no dementia and those with a score above the threshold for either the AD8 or the RUDAS, or for both AD8 and RUDAS, and comparing gender, age group, and marital status. We adjusted for post hoc pairwise comparisons. 66 We also performed a correlation analysis to assess the overlap of dimensions of deprivation.
We We calculated unadjusted and adjusted logistic regression models to identify the association between dementia and multidimensional poverty. The binary outcome compared adults with and adults without dementia as determined by either the AD8 or the RUDAS. Multidimensional poverty exposure was defined as being deprived in 4 dimensions, in which each adult is deprived corresponding to the highest gap between adults with and adults without dementia and a prevalence of poverty of 17% among all adults. This was based on estimates of 25.4% of urban South African individuals living below the national lower bound of the poverty line. 67 We investigated whether multidimensional poverty was associated with dementia in older adults. Models adjusted for gender (man or woman), age (continuous), marital status (living alone or living with a partner), and household size (continuous). All P values were from 2-tailed tests, and results were deemed statistically significant at P < .05. Missing values (n = 10) were treated as being missing completely at random. Data were analyzed using Stata, version 16.0 (StataCorp). There was a higher prevalence of deprivation among adults with dementia (particularly when considering those identified by both the AD8 and the RUDAS) than among adults without dementia. We estimated Spearman rank correlation coefficients between each pair of indicators of deprivation (eTable 3 in the Supplement). We found no evidence of a strong correlation between indicators, illustrating the absence of association between them except for type of housing, access to clean water, and sanitation. We kept those indicators as part of the living standards dimension of poverty but gave them each one-third the weight of other indicators in the equal nested weights

Multidimensional Poverty
The The adjusted head count ratio was higher for dimensions ranging between 1 and 6 for women with dementia compared with women without dementia and for dimensions ranging between 1 and 8 for men with dementia compared with men without dementia ( Figure 3). No woman without dementia was poor with deprivation in more than 4 dimensions, and no man without dementia was poor with deprivation in more than 5 dimensions, whereas 3.6% of women with dementia (4 of 110)  were poor with deprivation in 5 dimensions and 10.9% of men with dementia (6 of 55) were poor with deprivation in 6 dimensions. Overall, men with dementia were poorer, with deprivation in more dimensions, than women with dementia. The difference in the adjusted head count ratio between men with dementia and men without dementia is significant and reaches 166% for those deprived in 4 dimensions and 505% for those deprived in 5 dimensions.
Deprivations of education, health, and employment separately contributed, respectively, to more than 18%, 19%, and 10% (with 1 exception [8.4%] for those deprived in 3 dimensions) of the adjusted head count ratio for older adults with dementia compared with 16%, 13%, and 8% (with 1 exception [4%] for those deprived in 4 dimensions) for those without dementia for dimensions ranging between 1 and 4 (eTable 5 in the Supplement). Social cohesion contributed highly (>15%) to the adjusted head count ratio for older adults without dementia for dimensions ranging between 1 and 3.
The same 3 dimensions-deprivation of education, health, and employment-were the main contributors to the adjusted poverty head count poverty ratio for older men and women with or without dementia, for dimensions ranging between 1 and 4. Education contributed the most for both men and women with or without dementia, with the highest contribution for women with dementia (26.8% for those deprived in 1 dimension, 23.2% for those deprived in 2 dimensions, and 20.6% for those deprived in 3 dimensions). Employment contributed more to the adjusted head count ratio for both men and women with dementia compared with those without dementia. Poor health contributed more to the adjusted head count ratio for men without dementia, while deprivation of social participation contributed more to the adjusted head count ratio for women without dementia.

Multivariable Regression Analysis
The odds of dementia were 2.31 (95% CI, 1.08-4.95) times higher for adults who were multidimensionally poor (those deprived in 4 dimensions) compared with those who were not multidimensionally poor, even after controlling for gender, marital status, age, and size of household ( Table 2). Being a woman increased the relative odds of dementia by 2.03 (95% CI, 1.00-4.12), while living in a larger household increased the odds by 1.27 (95% CI, 1.05-1.53) for each additional household member. Age and marital status had no significant association with dementia. We tested the interaction between poverty and gender, between poverty and marital status, and between poverty and household size, and the results were not statistically significant.

Sensitivity Analysis
The multidimensional poverty calculation was repeated using the equal nested weight structure between the 7 dimensions, and the results remained unchanged (eTable 6 in the Supplement), including when stratified by gender (eTable 7 in the Supplement). Multidimensional poverty was found to be significantly higher for older adults with dementia compared with older adults without dementia for either AD8 or RUDAS screening for dimensions 1, 2, and 5; and for both AD8 and RUDAS screening for dimensions 1 through 4. The difference existed for other cutoffs but was not statistically significant. Women and men with dementia were poorer than women and men without dementia for any cutoff between 1 and 6 dimensions. Contributions of dimensions to the adjusted head count ratio consistently showed the prominence of education, health, and employment for older adults with or without dementia and for both genders (eTable 8 in the Supplement).

Discussion
This study found that adults with dementia have a higher level of multidimensional poverty in an LMIC context. Considering 7 domains of social and environmental determinants of health (education, health status, employment, living standards, social participation, stigma, and psychological well-being), we found that exposure to multidimensional poverty was associated with dementia.
Men with dementia were poorer and were deprived in a higher number of dimensions than women with dementia. However, being a man was protective against dementia, as previously reported. 69 Living in a large household was also associated with a higher prevalence of dementia. This finding may be due to the fact that an older adult with declining independence or functioning may move back with his or her family, this older adult may receive less support, or fewer resources may be available to this older adult. 31 Despite the fact that dementia prevalence increases with age, we did not find that age significantly strengthened the association with poverty. In addition, deprivation of education, health, and employment were identified as major contributors to multidimensional poverty, which constitutes an important indicator that social and environmental determinants of health are associated with dementia. Access to affordable and quality universal health care has been promoted as essential to reducing the effect of dementia. 70 Moreover, policies promoting a fair economic arrangement and equal access in key sectors of society (eg, education and health care) may prevent health disparities over time. 71 Alleviating multiple poverty indicators has the potential to slow the deterioration of cognitive functioning, which may advance interventions and prevention strategies in LMICs where the highest increase in dementia prevalence is expected but where limited research is taking place.

Limitations
This study has some limitations. First, it was not possible to establish the direction of the association between poverty and dementia-poverty can be a cause as well as a consequence of dementia.
Second, the size of the sample is limited, reducing the statistical power to further investigate the difference in multidimensional poverty by other sociodemographic characteristics of adults, such as age group or marital status. Third, because we could not obtain conventional biomarkers (eg, imaging and cerebrospinal fluid) or a formal assessment of cognitive functioning, we screened participants using the AD8 and the RUDAS. This approach did not establish a formal diagnosis of dementia; thus, we cannot conclusively exclude that the score obtained for both measures may be associated with the level of education or the socioeconomic status of the study participant. Fourth, 22.3% of participants in the sample were missing interviews, which might have introduced bias in our results because we cannot assume they were missing completely at random.

Conclusions
Our study provides evidence for physicians, allied health professionals, and policy makers to consider daily stressors associated with multidimensional poverty and aging. This study offers some valuable insight into LMICs and what public policies (access to quality education, a strong workforce, and quality and free universal health care) could be prioritized that may be associated with dementia prevention and may reduce its effect on families and communities. 70