Residence in Urban and Rural Areas over the Life Course and Depression among Ghanaian and South African Older Adults

Abstract

This study assessed associations between depression and urban/rural residence from a life-course perspective within African settings. Data on Ghanaian and South African adults aged 50 years and older were taken from wave 1 of the World Health Organization Study on Global Ageing and Adult Health (SAGE). Neither urbanicity of childhood nor adulthood residence was associated with later-life depression in either country. Significant differences were also not observed for residence changes over the life course, but there were trends in the data suggestive of higher depression prevalence in Ghanaian recent rural-urban migrants and lower prevalence among South African recent urban-rural migrants.

Introduction

Aging and urbanization are demographic and social changes occurring rapidly worldwide (Beard et al., 2016; National Institute on Aging and World Health Organization, 2011; United Nations Population Fund, 2016). These processes have consequences for rates of depression, a principal cause of disease burden internationally (Ferrari et al., 2013; World Health Organization, 2020). Namely, older adults often face many factors that put them at risk for the disorder and, particularly among the oldest age groups, are affected by depression at higher rates than younger individuals (Alexopoulos, 2005; Cole & Dendukuri, 2003; Gatz, Kasl-Godley, & Karel, 1996). Additionally, living in urban environments has also been linked to an increased risk of depression (Peen, Schoevers, Beekman, & Dekker, 2010), and many studies have shown links between depressive outcomes and greater urbanicity as measured by the population size or density of cities as well as other features both in cross-sectional and longitudinal data (Arvind et al., 2019; Bui, Vu, & Tran, 2018; Generaal et al., 2019; Kovess-Masféty, Alonso, De Graaf, & Demyttenaere, 2005; Peen et al., 2010; Pun, Manjourides, & Suh, 2019; Sundquist, Frank, & Sundquist, 2004). Potential explanations have suggested physical and social factors that may function as stressors and are generally more common in urban areas, such as air and noise pollution, traffic, decreased access to green space, crime, and socioeconomic disadvantage (Gruebner et al., 2017).

Although Africa is urbanizing and its aging population size is increasing at rapid rates (United Nations Department of Economic and Social Affairs Population Division, 2014; Velkoff & Kowal, 2006; World Health Organization, 2014), research on the urbanicity-depression relationship is very limited in this region, so the role urbanicity plays in depression in African contexts remains unclear. Moreover, the majority of these studies have only used current residence as a measure of urbanicity and have therefore not taken into account the role of urbanicity at other stages of life and across the life course.

The life-course perspective suggests that there may be critical or vulnerable periods in human development for the etiology of disease, and risk to a given exposure can accumulate over time (Ben-Shlomo & Kuh, 2002). Early life exposures have already been implicated in the development of some mental illness (Angelini, Howdon, & Mierau, 2019; Kessler et al., 2010; van Os, Kenis, & Rutten, 2010). Furthermore, the experience of migration itself can also impact health and the risk of mental illnesses (Bhugra & Jones, 2001). Scholars have recognized the role aging is playing in the growth of non-communicable diseases in poorer nations; and as exposures emerge, change, or intensify in these settings, Tollman et al. (2016) argue that it becomes increasingly important to adopt a life-course perspective to understand, predict, and address health challenges in populations of all ages, including older adults and not just from early life to adulthood. Indeed, health effects of the neighborhood environment have been shown to intensify with age, which may be a reflection of longer exposure both over extended periods of time as well as in the short-term because older adults are more likely to be home-bound or spend more time within their communities (Cagney & Cornwell, 2010). Thus, similar processes may apply to the larger urban context overall, and an examination of the effects of urbanicity on depression that not only addresses understudied geographic locations but that also encompasses longer periods of the life course is important to better understand this relationship.

The following study therefore aims to assess whether residence in urban compared to rural areas across the life course and at specific life stages affects the likelihood of depression in later life within the context of two middle-income African countries, Ghana and South Africa. These two nations serve as suitable settings for investigating this topic not only due to the availability of data, but also because they offer points of comparison and contrast. Although both Ghana and South Africa are rapidly growing economies in sub-Saharan Africa, they also differ in many respects, including geographic location, climate, historical backgrounds, culture, political and social conditions, and demographic makeup. Moreover, their experience of urbanization is not the same.

Ghana is a lower middle-income country in West Africa with a slight majority of its overall population (56%) living in urban areas and an estimated annual average urbanization rate of 1.2% (United Nations Department of Economic and Social Affairs, 2018). Urbanization is primarily a result of internal migration from rural areas and natural population growth in urban areas (Songsore, 2009), and much of its urban growth is geographically restricted to a few major urban locations in the country (Yankson & Bertrand, 2012). In contrast, South Africa is an upper middle-income country with a higher proportion of the population urban (66%) but a lower annual urbanization rate (0.8%) (United Nations Department of Economic and Social Affairs, 2018). Most of its urbanization is a result of migration, both internal as well as external (Turok, 2012). In terms of aging, 2016 estimates of life expectancy at birth in both countries stood at around 63.5 years (World Health Organization, 2018).

There are also qualitative differences in the urban environments of the two countries. While both have Dutch and British colonial history, the establishment of South Africa as a settler colony in contrast to indirect rule in Ghana has led to a more racially diverse population in South Africa with a high degree of racial residential segregation persisting as a result of apartheid policies (Hall et al., 2018). Informal settlements as well as stark inequality between the rich and poor also characterize both countries’ urban areas, although to a much greater extent in South Africa (United Nations Human Settlement Programme, 2013). However, as a poorer nation, urban areas in Ghana may suffer more from a lack of infrastructure (United Nations Human Settlement Programme, 2013).

Despite the varying country characteristics and situations, given the connection identified between urbanicity early in life and depression (Lundberg, Cantor-Graae, Rukundo, Ashaba, & Ostergren, 2009; Vassos, Agerbo, Mors, & Pedersen, 2016), we hypothesize that both childhood urban residence and adulthood urban residence are associated with depression. Further, using the concept of cumulative exposure as a guiding principle, we also hypothesize that individuals who have resided in urban areas during multiple time periods in their lives are more likely to experience depression than those who have not.

Methods

Data

This study uses data from wave 1 of the World Health Organization Study on Global Ageing and Adult Health (SAGE), a longitudinal study of adults focusing on the age group 50 years and older conducted in the six middle-income countries of China, Ghana, India, Mexico, Russia, and South Africa (Kowal et al., 2012). Wave 1 SAGE data for the two African countries represented in the study, Ghana and South Africa, were collected from 2007 to 2008 through a household survey that employed a stratified, multistage cluster design (Biritwum, 2013; Phaswana-Mafuya & Peltzer, 2013). Data were collected using paper survey instruments (Biritwum, 2013; Phaswana-Mafuya & Peltzer, 2013). The details of the survey are described elsewhere (Kowal et al., 2012). The overall samples for the study included 4,304 Ghanaian and 3,278 South African adults aged 50 years and older with information on former residence.

Measures

Exposure Variables:

For part one of the analysis comparing the role of childhood residence and adulthood residence in later-life depression, respondents were classified as urban or rural residents during these time periods if they reported living in either urban or rural locations in response to survey questions on where they spent most of their childhood (before age 10) and most of adulthood (18+ years).¹ Urban and rural categories were based on official classifications in each country, with areas containing populations of 5,000 or more considered urban in Ghana and urban areas in South Africa based on legal designations that consider factors such as settlement type and land use (Ghana Statistical Service, 2012; Statistics South Africa, 2003).

For part 2 of the analysis, we used a similar approach to Kim et al. (2004) and Yiengprugsawan et al. (2011) to create categories of lifetime residence using a sequence of childhood, adulthood, and current time periods with responses classified as urban or rural. This was based on the survey questions on where individuals spent most of their childhood and most of adulthood, along with the location of their current household residence, and resulted in 8 possible life-course patterns: 1) rural-rural-rural; 2) rural-rural-urban; 3) rural-urban-urban; 4) rural-urban-rural; 5) urban-rural-rural; 6) urban-urban-rural; 7) urban-rural-urban and 8) urban-urban-urban. Individuals with missing or unknown responses on any of the three components were excluded from analysis; thus, the analysis included 4,199 Ghanaian and 3,174 South African adults at least 50 years of age with complete residence information.

Outcome Variable:

Depression over the past 12 months was defined using the International Classification of Diseases, 10th revision (ICD-10) (World Health Organization, 1993) criteria for a depressive episode based on respondents’ reported experience of depressive symptoms for the last 12 months. In addition, respondents who self-reported treatment for depression in that time period were also considered to have depression.

Covariates:

Independent variables considered for adjustment in the analysis included sex (male/female); age (50–59, 60–69, 70–79, and 80+); marital status (never married, currently married/cohabiting, separated/divorced, widowed); education (no formal education, some primary, primary completed, secondary completed, post-secondary completed); ethnicity (Akan, Ewe, Ga-Adangbe, Gur, and other for Ghana and African/Black, White, Coloured, Indian/Asian, and other for South Africa); residential mobility (ever moved); employment status (currently working, not currently working, never worked); household size (number of people); bereavement (household deaths in the past 24 months); permanent income quintiles based on country-specific household goods and assets; functional disability score based on the WHO Disability Assessment Schedule-II (Sousa et al., 2010) which includes 12 items evaluating difficulty carrying out common tasks ranging from walking and bathing to sustaining friendships; and cognitive performance calculated as composite z-scores based on five cognitive tests (immediate and delayed verbal recall, forward and backward digit span, and verbal fluency).

Statistical Analysis

Data were analyzed separately for each country. Characteristics of the samples were calculated as proportions for categorical variables and means for continuous variables based on the sample data, and corresponding national estimates were also produced using individual-level survey weights. These weights were based on the individual probability of selection and were post-stratified by region, locality, sex, and age in Ghana and by province, sex, and age in South Africa and are not normalized (Biritwum, 2013; Phaswana-Mafuya & Peltzer, 2013). Bivariate logistic regression models between the outcome and each independent variable were run to determine their unadjusted association. Multivariable logistic regression models accounting for the clustered and stratified survey design were employed to examine the association between depression and urbanicity of residence in early childhood and adulthood as well as across the life course, adjusting for sex, age, and variables significantly associated with the depression outcome in each country. Residential mobility was also included as a covariate to isolate the effects of urbanicity from those of migration. To separate the effect of childhood and adulthood residence from current residence in the first part of the analysis, additional models were run controlling for current residence.

Weights were not applied in the main regression analyses. However, weighted analyses were conducted for comparison, and these produced similar results for Ghana and did not change the conclusions of the analysis for the variables of interest in the South Africa models although it appeared to have a greater impact on the estimates. Data were analyzed using Stata 13.

Results

Descriptive statistics on both the Ghana and South Africa samples are displayed in Table 1. In Ghana, roughly half of the older adult population was male and about half had received no formal education. The majority were currently married and currently working. Most had not experienced a recent death in the household, and about one third of the older adult population had relocated at some point in time. The average household size was between 5 and 6 members, and respondents had mild functional disability on average. In South Africa, about 40% of the population was male, around a quarter were not formally educated, half were currently married or cohabiting, and close to a third were currently working. Approximately one third of the population had moved in their lifetime, and essentially none of the population had a household death in the past 24 months. Households had an average size of 4 people, and respondents were on average mildly functionally disabled.

Table 1:

Sample Characteristics by Country

	Ghana	South Africa
	% (w%)a	% (w%)a
Male	52.2 (52.4)	39.9 (40.2)
Age
50–59	39.2 (39.8)	43.8 (49.9)
60–69	28.0 (27.5)	32.3 (30.6)
70–79	22.9 (23.1)	17.5 (14.1)
80+	9.9 (9.6)	6.4 (5.4)
Education Level
None	55.1 (53.9)	25.8 (23.6)
Some Primary	10.1 (10.4)	24.9 (25.2)
Primary Completed	10.8 (10.9)	24.0 (22.7)
Secondary Completed	20.5 (21.1)	20.3 (22.8)
Post-Secondary Completed	3.5 (3.6)	5.0 (5.8)
Marital Status
Never Married	1.2 (1.3)	14.0 (15.2)
Married/Cohabiting	56.8 (59.3)	50.8 (52.7)
Separated/Divorced	14.2 (12.9)	6.4 (6.3)
Widowed	27.9 (26.5)	28.8 (25.8)
Employment Status
Currently Working	71.8 (71.4)	27.6 (31.1)
Never Worked	1.5 (1.6)	13.0 (13.6)
Not Working	26.8 (27.0)	59.4 (55.4)
Recent Household Death	1.3 (1.3)	0.2 (0.2)
Ever Moved	34.1 (33.9)	29.9 (33.4)
	mean (meanw)a	mean (meanw)a
Total Household Members	5.5 (5.6)	4.1 (4.0)
Functional Disability Scoreb	21.7 (21.6)	20.0 (20.7)

^aw% and mean_w are weighted to account for survey design and provide national population estimates

^brange is 12 (none) to 60 (extreme)

Part 1: Urbanicity of Residence at Different Life Stages

38.9% and 41.4% of Ghanaian older adults lived in urban areas in childhood and adulthood, respectively, and overall depression prevalence was 7.5% based on national population estimates. In South Africa, 60.8% and 62.1%, respectively, were childhood and adulthood urban residents. Depression prevalence was 4.1%. Among the sample with urban childhood residence, the rate of depression was 7.2% in Ghana and 4.7% in South Africa, while the rate among those with rural childhood residence was 7.9% and 4.0% in Ghana and South Africa, respectively. The rates of depression in the samples based on type of adulthood residence were 7.1% for urban and 8.1% for rural in Ghana and 4.8% for urban and 3.8% for rural in South Africa (Figure 1).

Figure 1:

Proportion of the Sample with Depression by Country and Location of Residence

In the Ghana sample, sex, age, education, household wealth, employment status, ethnicity, residential mobility, bereavement, functional disability, and cognitive function were individually associated with the depression outcome and included in the multivariable model. The adjusted odds ratio (OR) for depression comparing urban to rural residents was 0.80 (95% CI: 0.53–1.20) for both primary childhood and adulthood residence. After adjusting for current residence, the adjusted ORs reduced to 0.75 (95% CI: 0.43–1.32) for childhood residence and 0.66 (95% CI: 0.35–1.23) for adulthood residence (Table 2).

Table 2:

Odds Ratios (95% CI) for Depression as a Function of Childhood and Adulthood Residence

	Ghana			South Africa
	Unadjusted	Model 1	Model 2	Unadjusted	Model 1	Model 2
Childhood Residence
Rural	Ref	--	--	--	--	--
Urban	0.91 (0.61–1.35)	0.80 (0.53–1.20)	0.75 (0.43–1.32)	1.18 (0.83–1.69)	1.00 (0.66–1.53)	0.98 (0.56–1.72)
Adulthood Residence
Rural	Ref	--	--	--	--	--
Urban	0.87 (0.59–1.29)	0.80 (0.53–1.20)	0.66 (0.35–1.23)	1.28 (0.87–1.88)	1.01 (0.65–1.56)	0.99 (0.57–1.70)

Note. Model 1 adjusts for significant covariates and model 2 adjusts for covariates and also controls for current residence

In the South Africa sample, age, marital status, education, employment status, ethnicity, household wealth, and functional disability were significantly associated with depression in bivariate analyses and included in the multivariable model along with sex and residential mobility, which were included a priori. The adjusted OR for depression comparing urban to rural residents was 1.00 (95% CI: 0.66–1.53) for primary childhood residence and 1.01 (95% CI: 0.65–1.56) for primary adulthood residence. After controlling for current residence, the adjusted ORs reduced to 0.98 (95% CI: 0.56–1.72) and 0.99 (95% CI: 0.57–1.70) for childhood and adulthood residence, respectively (Table 2).

Part 2: Urbanicity of Residence across the Life Course

Based on depression rates within each lifetime category of childhood-adulthood-current residence, the highest percentages of depressed individuals in the Ghana sample were among more recent rural-urban migrants (rural-rural-urban, 8.28%), followed by lifetime rural (rural-rural-rural, 8.22%) and then lifetime urban residents (urban-urban-urban, 7.58%) (Table 3).

Table 3:

Distribution of Residence Patterns and Rates of Depression by Life-Course Residence

	Ghana		South Africa
	Distribution	Depressed	Distribution	Depressed
	% (n)	% (n)	% (n)	% (n)
Rural-Rural-Rural	53.20 (2234)	8.22(183)	29.58 (939)	3.81 (34)
Rural-Rural-Urban	4.05 (170)	8.28 (14)	4.47 (142)	3.55 (5)
Rural-Urban-Urban	2.67(112)	3.60 (4)	1.76 (56)	7.55 (4)
Rural-Urban-Rural	1.43 (60)	3.33 (2)	0.38 (12)	8.33 (1)
Urban-Rural-Rural	1.43 (60)	5.00 (3)	0.38 (12)	8.33 (1)
Urban-Urban-Rural	3.00(126)	5.56 (7)	2.74 (87)	1.16(1)
Urban-Rural-Urban	0.21 (9)	0.00 (0)	0.19(6)	0.00 (0)
Urban-Urban-Urban	34.01 (1428)	7.58 (108)	60.49 (1920)	4.87 (91)
Total	100 (4199)	7.67(321)	100(3174)	4.46 (137)

Note. Life-course categories represent childhood-adulthood-current residence. Denominators for the percent depressed may differ from the total sample size in each category due to missing data. Percentages reflect values based on the samples and corresponding sample sizes for each category rather than population estimates

Intermediate depression rates were seen among the two groups with urban childhood but current rural residence (5.00–5.56%). Lowest rates were seen among the groups with rural childhood but urban adulthood (3.33–3.60%). In the adjusted analyses, the highest odds ratio for depression compared to lifetime rural residents was among the rural-rural-urban group (OR=1.76, 95% CI: 0.75–4.16). The lowest odds of depression compared to lifetime rural residents was among the rural-urban-rural group (OR=0.62, 95% CI: 0.13–2.88) followed by the rural-urban-urban group (OR=0.75, 95% CI: 0.28–2.03). The odds of depression in lifetime urban residents compared to lifetime rural residents was OR=0.80, 95% CI: 0.50–1.26 while urban-rural migrants had ORs close to null (Table 4). These estimates did not reach statistical significance.

Table 4:

Adjusted Odds of Depression in Ghana Based on Life-Course Residence

Covariates	OR (95% CI)
Life-Course Residencea
Rural-Rural-Rural	Ref
Rural-Rural-Urban	1.76 (0.75–4.16)
Rural-Urban-Urban	0.75 (0.28–2.03)
Rural-Urban-Rural	0.62 (0.13–2.88)
Urban-Rural-Rural	1.01 (0.27–3.77)
Urban-Urban-Rural	1.09 (0.51–2.32)
Urban-Urban-Urban	0.80 (0.50–1.26)
Sex
Male	Ref
Female	1.43 (1.09–1.90)*
Age
50–59	Ref
60–69	1.27 (0.87–1.87)
70–79	1.29 (0.82–2.03)
80+	1.52 (0.86–2.68)
Education Level
None	Ref
Some Primary	1.25 (0.82–1.91)
Primary Completed	0.85 (0.51–1.40)
Secondary Completed	1.65 (1.06–2.58)*
Post-Secondary Completed	1.28 (0.47–3.48)
Ethnicity
Akan	Ref
Ewe	0.51 (0.26–0.98)
Ga-Adangbe	0.62 (0.36–1.07)
Gur/Northern	1.18 (0.69–2.02)
Other	0.77 (0.46–1.29)
Ever Moved	0.52 (0.33–0.83)**
Employment Status
Currently Working	Ref
Never Worked	0.19 (0.03–1.47)
Not Working	1.62 (1.19–2.20)**
Permanent Income Quintile
1st	Ref
2nd	1.58 (1.08–2.30)*
3rd	1.02 (0.66–1.58)
4th	1.08 (0.69–1.70)
5th	0.82 (0.51–1.30)
Recent Household Death	2.30 (0.91–5.79)
Functional Disability	1.00 (0.98–1.02)
Cognitive Performance	0.93 (0.89–0.96)***

Note. Analyses account for the clustered and stratified design.

^aLife-course categories represent childhood-adulthood-current residence. The urban-rural-urban category dropped out of the analysis due to an absence of depression cases.

^*p<0.05,

^**p<0.01,

^***p<0.001

In South Africa, the highest rates of depression were seen among the rural-urban-rural and urban-rural-rural groups (8.33%), followed by rural-urban-urban residents (7.55%) (Table 3). Lifetime urban residents had the next highest depression rates (4.87%), followed by lifetime rural residents (3.81%). The lowest rates of depression were seen among the recent urban-rural migrants (1.16%). After adjusting for covariates, the urban-rural-rural category had the greatest likelihood of depression (OR=1.62, 95% CI: 0.15–17.50). Based on adjusted results, the lowest likelihood was observed among the urban-urban-rural group (OR=0.24, 95% CI: 0.04–1.66) followed by the rural-rural-urban group (OR=0.55, 95% CI: 0.19–1.60). Lifetime urban residents had essentially similar odds of depression compared to lifetime rural residents. Estimates did not reach statistical significance. Complete adjusted results are displayed in Table 5.

Table 5:

Adjusted Odds of Depression in South Africa Based on Life-Course Residence

Covariates	OR (95% CI)
Life-Course Residencea
Rural-Rural-Rural	Ref
Rural-Rural-Urban	0.55 (0.19–1.60)
Rural-Urban-Urban	1.04 (0.29–3.75)
Urban-Rural-Rural	1.62 (0.15–17.50)
Urban-Urban-Rural	0.24 (0.04–1.66)
Urban-Urban-Urban	0.97 (0.61–1.54)
Sex
Male	Ref
Female	0.97 (0.65–1.44)
Age
50–59	Ref
60–69	0.38 (0.24–0.61)***
70–79	0.24 (0.13–0.46)***
80+	0.07(0.02–0.30)***
Marital Status
Never Married	Ref
Married/Cohabiting	1.32 (0.71–2.46)
Separated/Divorced	2.43 (1.06–5.59)*
Widowed	2.02 (1.01–4.06)*
Education Level
None	Ref
Some Primary	1.25 (0.72–2.16)
Primary Completed	1.18 (0.67–2.09)
Secondary Completed	1.01 (0.52–1.98)
Post-Secondary Completed	0.79 (0.23–2.69)
Ethnicity a
African/Black	Ref
White	1.47 (0.62–3.44)
Coloured	1.17 (0.75–1.82)
Indian/Asian	1.38 (0.74–2.59)
Ever Moved	1.30 (0.83–2.01)
Employment Status
Currently Working	Ref
Never Worked	1.97 (0.93–4.18)
Not Working	2.81 (1.57–5.02)**
Permanent Income Quintile
1st	Ref
2nd	1.01 (0.51–2.00)
3rd	1.72 (0.98–3.03)
4th	1.71 (0.88–3.32)
5th	2.19 (1.11–4.31)*
Functional Disability	1.05 (1.02–1.07)***

Note. Analyses account for the clustered and stratified design.

^aLife-course categories represent childhood-adulthood-current residence. The rural-urban-rural and urban-rural-urban residence categories and other ethnicity category dropped out of the multivariable analysis due to an absence of depression cases.

^*p<0.05,

^**p<0.01,

^***p<0.001

In both countries, the urban-rural-urban group was very small and had no cases of depression and was therefore not retained in the analyses.

Discussion

Urbanicity of Residence at Different Life Stages

Contrary to our hypothesis, as well as other findings, neither urban residence in childhood nor adulthood was significantly associated with depression in Ghanaian older adults, though depression rates were slightly higher in Ghanaian rural residents. This is similar to the adjusted effect of current residence reported in Adjaye-Gbewonyo et al. (2019), which demonstrated a comparable OR of 0.85 (95% CI: 0.55–1.31). Results of this analysis suggest a slightly stronger—though still non-significant—magnitude of effect for both childhood and adulthood residence than for current residence, which may imply a lag period in the observation of effects. When adjusting for current residence, the estimates in the present analysis became more protective for urban compared to rural residents, particularly for primary adulthood residence, although the direction of effect for current residence changed. Though all effect estimates remained non-significant, the stronger magnitude of effect for adulthood residence may suggest that it could be more relevant for depression in later life than childhood residence.

In South African older adults, depression rates were in fact slightly higher in urban residents in both childhood and particularly adulthood but did not reach significance; however, adjusted results attenuated the effects such that the overall model results indicated essentially no effect of childhood or adulthood residence on depression, with ORs around one.

In a study of young adults aged 18–30 years in Uganda, urban birthplace was significantly associated with recent experience of depressive symptoms in the past week, as well as with symptoms of anxiety and psychosis and lifetime experience of delusional ideation (Lundberg et al., 2009). Likewise, another study showed that the incidence rate for depressive disorder was significantly higher among Danish individuals born in urban as compared to rural environments (Vassos et al., 2016). Although the SAGE survey does not provide information on place of birth, the results of this analysis for childhood residence do not support an association between urban location in early life and later depression as found in these studies. A study of urbanicity and depressive symptoms in British children at age 12 also did not find a significant association (Newbury et al., 2016), but adulthood outcomes were not assessed.

Contextual differences likely play a role in the failure to observe expected effects of urbanicity of residence in childhood or adulthood on depression in this study, and other studies that did not support the hypothesized association have likewise suggested context-specific explanations. For instance, Cheng et al. (1989) noted that Taiwanese rural residents had greater health events and more chronic stressors, which could explain their higher rates of depression compared to urban residents. Out-migration and neglect of rural areas has also been suggested for a lack of observed urban-rural differences in Finland (Jokela, Lehtimäki, & Keltikangas - Järvinen, 2007), and Stickley et al. (2015) also note economic disadvantage and social problems as risk factors in rural former Soviet areas. Indeed, a recent meta-analysis focusing specifically on depression in older adults did not find significant urban-rural differences in lower-income countries, although in high-income countries odds of depression were significantly greater in older adults residing in urban compared to rural areas; and differences in characteristics across settings are also offered as an explanation (Purtle et al., 2019). Though no studies in African countries were included in that analysis, the results of the present study are consistent with these findings for lower-income countries.

Urbanicity of Residence across the Life Course

Results for the Ghana sample likewise do not support a significant difference in depression in older adults based on different residence and migration patterns over the life course, and patterns do not confirm the cumulative exposure hypothesis. By contrast, findings from a longitudinal study indicate that certain demographics of rural-urban migrants may be more susceptible to depressive symptoms and that this is particularly the case for those coming from rural to urban settings more recently, suggesting that the timing of the move matters (Lu, 2010a). Individuals who made the rural-urban transition earlier may be more acclimated to their new environments and better able to reap the benefits of their new setting, which may have prompted their move, such as employment opportunities; and these situations may therefore make them less prone to depression.

Though results did not reach statistical significance, the direction of effects for the Ghana sample in this analysis appears to align with these findings as more recent urban migrants (rural-rural-urban) had the highest crude depression prevalence and had an adjusted odds ratio above one while rural-urban migrants who moved earlier (i.e., from childhood to adulthood) had the lowest prevalence and adjusted odds ratios below one compared to lifetime rural residents. Moving in general during one’s lifetime seems to have a beneficial effect in terms of depression in Ghanaian older adults, as the significant OR for the residential mobility variable demonstrates; and apart from the more recent rural-urban migrants, Ghanaian individuals who experienced both urban and rural settings likewise showed tendencies towards a lower likelihood of later-life depression than lifetime residents of either setting. This was particularly true of return rural migrants, which goes against some literature on rural-urban migration indicating that unhealthier or unhappier individuals may selectively return to their original locations (Nauman, VanLandingham, & Anglewicz, 2016). Given that estimates were not statistically significant, which may be attributable to small sample sizes across lifetime exposure groups, additional analyses would be needed to verify these tendencies.

Similarly, no significant differences in the odds of depression were found among South African older adults based on residency in rural and urban areas across the life course. Although clear, consistent patterns were not observed, based on the rate of depression in each category, highest rates were observed among those whose location of residence changed from childhood to adulthood, regardless of initial or current residence. However, these rates may be affected by small numbers in these categories. Nonetheless, some research does point to a significant negative impact of relocation during childhood on the development of depressive disorder and other psychiatric conditions in teenage years to middle age (Mok, Webb, Appleby, & Pedersen, 2016). Similarly, childhood residential instability was significantly associated with lifetime development of depression in another study, but the effect was modified by age of onset and only significant among those with onset of depression by age 14 (Gilman, Kawachi, Fitzmaurice, & Buka, 2003). However, individuals in this onset age group were also most likely to have recurring episodes, which may indicate that effects of childhood residential instability on depression may extend into adulthood. As no information on depression onset is included in the SAGE survey, it is not possible to determine whether effects of life-course urbanicity on current depressive outcomes relate to age at onset in the present study.

Lowest rates of depression in South Africa were observed among those who moved more recently in their adulthood, especially the urban-urban-rural group followed by the rural-rural-urban group. Based on adjusted results, more recent migrants seemed to have better outcomes in terms of depression in the South African context, though estimates again did not reach statistical significance.

Several studies have examined the effects of rural-urban migration on depression and other psychological outcomes (Albers, Kinra, Radha Krishna, Ben-Shlomo, & Kuper, 2016; Lu, 2010a, 2010b; Mou, Griffiths, Fong, & Dawes, 2013), with the majority demonstrating that rural-urban migrants suffer more from depression than individuals who do not make this transition. Study authors point to migration as a stressful and disruptive life event in explaining these findings. However, these studies focus on migration as the main factor and do not address life-course exposure to urbanicity. Thus, little distinction is made based on the timing of migration. In one of the few studies that purport to take a life-course approach, Kim et al. (2004) also found higher rates of depression among recent rural-urban migrants in Korean older adults, which agrees with the literature and the present results in the Ghana sample. Although the lowest depression rates in their study were observed among lifetime rural residents as we hypothesized for this study, other observed patterns differ from those hypothesized as well as from the other findings of the current analysis. However, among Thai adults, Yiengprugsawan et al. (2011) found a dose-response relationship with the rural-rural group showing the lowest psychological distress and diagnosed depression, followed by the rural-urban group, and highest among the urban-urban group. This is in line with our original hypothesis suggesting longer duration of residence in urban areas may be detrimental. However, neither of the life-course studies assessed urban-rural migrants because of small numbers and a primary interest in rural-urban migration patterns. A cumulative impact of length of urban residence in early life has also been reported in the case of schizophrenia, which is in agreement with life-course theory (Pedersen & Mortensen, 2001).

In the present analysis, the urban-rural-urban group contained no cases of depression and is not discussed. The absence of cases in this category is likely due to its very small sample size in both countries, although it is also possible that there could be other factors related to this migration pattern that may explain this. Interpretation of results is also affected by small numbers and, consequently, wide confidence intervals in other categories, which may have limited power to detect significant differences in depression across the life-course groups in general for this analysis. Furthermore, despite attempts to arrive at a measure of exposure throughout one’s lifetime, SAGE survey questions did not cover the entire lifespan. For instance, the item on childhood residence only focused on the time period up to and including age 9. In addition, because only one response option was included per question, it did not account for the possibility of living in multiple types of locations during childhood or adulthood. As a result, categories in this analysis represent approximations of lifetime residence only.

Life-course residence categories may also represent a wide range of actual time spent in each locality. For example, time since migration for the more recent migrants whose current residence type differed from childhood and adulthood spanned less than a year to over a decade. This heterogeneity in length of time since migration could further obscure the ability to observe significant effects, and future studies that can specifically analyze and distinguish individuals based on time since migration may be able to reveal additional information on the relationship between residence and depression.

It is also important to acknowledge that the binary urban/rural classifications available in the data and often used in studies do not recognize gradations or distinctions between the two ends of the spectrum and therefore could be too broad to allow for enough variation to observe differences between them. Moreover, definitions of urban and rural are usually specific to each country—as was the case in the data used for this study—and thus are not directly comparable across locations. This suggests that the construct does not measure exactly the same exposure in all settings. However, given the wide degrees of urbanization and variety of urban contexts and experiences globally, a standard measure may not necessarily be feasible or appropriate across all situations (Peen et al., 2010; United Nations Department of Economic and Social Affairs Statistics Division, 2017).

Additionally, the use of a standardized depression measure in the SAGE survey does not account for cultural differences in not only how people interpret and respond to questions relating to depression symptoms but also in how depression is experienced and conceptualized. However, studies do support the presence of a similar set of depressive symptoms across societies although the relative contribution of various symptoms may differ (Draguns & Tanaka-Matsumi, 2003). And the use of a standardized measure of depression based on internationally recognized and accepted criteria allows for comparison of results across multiple settings.

Nevertheless, this study provides new insights into the role of urbanicity of residence at different points in time on depression and specifically in African contexts, and it is the only such study the authors were able to identify for the countries in question. Given that low- and middle-income countries, including those on the African continent, and their urban areas differ culturally and structurally from high-income countries, generating locally relevant evidence is imperative in order to better inform programs and policies rather than assuming similarity or transferability of evidence from other locations. This study therefore contributes to increasing the context-specific knowledge base, and the comparative nature of the research can help to illuminate how and why outcomes differ across settings.

The study particularly represents one of a very limited number of studies that incorporates a life-course approach to understanding the urbanicity-depression relationship and more importantly, is novel in that it not only addresses exposure to the different types of settings—urban and/or rural—during the life course overall but specifically considers the timing and order of these exposures and what impact differences in this sequence may have. And unlike other related studies, it includes all possible patterns rather than focusing solely on rural to urban migration. It therefore begins to shed light on variations in the experience of depression based on different migration patterns.

Conclusion

Results of this study do not support a significant difference in later-life depression based on residence in urban compared to rural areas in either childhood or adulthood as well as across the life course. However, further analyses and larger samples will be useful in order to clarify potential trends suggested by these data. Additional quantitative and qualitative studies along with refined measures that better capture urbanicity over the life course are also needed to further explore and confirm these results as well as to elucidate the underlying contexts, migratory choices, and reasons for migration among various subpopulations that may be influencing outcomes. As data from additional waves of the WHO-SAGE and other aging studies become available, they will also provide opportunities for longitudinal research that will lead to better understanding social determinants of health such as urbanicity across the life course.

Although findings may not warrant changes in or differential allocation of mental health services, they may still inform the identification and targeting of interventions for individuals who may be more at risk for depression based on their area of residence and migration patterns. In particular, they may suggest at the very least a need for additional monitoring of the mental health of migrant groups with attention to the timing of setting changes, whether it be a more recent change or in between the childhood to adulthood transition.

Highlights.

• Later-life depression was not significantly associated with life-course urbanicity

• Yet, depression rates differed more by adulthood than childhood residence type

• Nonsignificant results for some lifetime residence groups may be due to low numbers

• Further study is needed to clarify trends in depression rates among recent migrants