Very few studies have investigated mental health in sub-Saharan Africa (SSA). Using data from Malawi, this article provides a first picture of the demography of depression and anxiety (DA) among mature adults (aged 45 or older) in a low-income country with high HIV prevalence. DA are more frequent among women than men, and individuals affected by one are often affected by the other. DA are associated with adverse outcomes, such as poorer nutrition intake and reduced work efforts. DA also increase substantially with age, and mature adults can expect to spend a substantial fraction of their remaining lifetime—for instance, 52 % for a 55-year-old woman—affected by DA. The positive age gradients of DA are not due to cohort effects, and they are in sharp contrast to the age pattern of mental health that has been shown in high-income contexts, where older individuals often experience lower levels of DA. Although socioeconomic and risk- or uncertainty-related stressors are strongly associated with DA, they do not explain the positive age gradients and gender gap in DA. Stressors related to physical health, however, do. Hence, our analyses suggest that the general decline of physical health with age is the key driver of the rise of DA with age in this low-income SSA context.
Depression and anxiety (DA) are two important dimensions of mental health that have a significant and growing contribution to the global burden of disease (Collins et al. 2013; Murray et al. 2012; Prince et al. 2007). DA are an integral part of global population health (Susser and Patel 2014). In resource-poor contexts, DA have been widely recognized as having important implications for individual productivity, individual- and family-level well-being, and overall economic development (Canavan et al. 2013; Lancet2011; Sorsdahl et al. 2011). The World Bank and the World Health Organization (WHO) have started an initiative to move mental health to the forefront of the international development agenda (Carey 2016; Chisholm et al. 2016). Despite their growing relevance, however, DA continue to be poorly documented and inadequately understood in low-income countries and sub-Saharan African (SSA) countries affected by HIV/AIDS (Lund 2014; WHO 2013). To help fill this research gap, we address essential but underresearched questions about the demography of DA among mature adults in rural Malawi. For example, in this low-income country with high HIV prevalence, why does the prevalence of DA increase markedly with age? Why do women have higher levels of DA than men across all ages? To what extent does exposure to a living environment characterized by frequent social or economic shocks explain variation in DA within this population?
Gaps in Understanding the Demography of Depression and Anxiety in SSA Low-Income Countries
Our focus in this article is on the demography of DA among mature adults, defined here as adults aged 45 or older, who represent an increasingly important subset of societies in SSA. The population of mature adults will grow more rapidly in many SSA countries than any younger 10-year age group in the next decades (own calculations based on UN Population Division 2012). In the next 50 years, 80 % of the additional person-years lived among adults as a result of increasing life expectancies in SSA low-income countries will occur among individuals aged 45 or older (Fig. 1, panel a). These recent and projected future improvements in the life expectancy of adults in SSA low-income countries with high HIV prevalence have reversed the previous trend of declines in adult survival caused by the HIV/AIDS epidemic during the 1990s and early 2000s (Bor et al. 2013; Floyd et al. 2012; Payne and Kohler 2017) (Fig. 1, panel b). Despite these gains, the overall well-being of mature adults is often low (McKinnon et al. 2013; Zimmer and Das 2014). A recent Malawi Longitudinal Study of Families and Health (MLSFH) study, for example, showed that 45-year-old women in Malawi can expect to spend 58 % of their remaining 28 years of life with physical limitations affecting their day-to-day activities and work efforts, and 45-year-old men can expect to live 41 % of their remaining 25.4 years subject to such limitations (Payne et al. 2013). Mature adults are also frequently affected by DA (McKinnon et al. 2013; Thapa et al. 2014). Evidence from high-income settings suggests that experiencing physical limitations on daily activities is associated with DA (Loeb and Jonas 2015; Sareen et al. 2006) and that effective treatment of these mental health conditions may improve physical functioning scores (Niles et al. 2013). However, fundamental research on the linkages between mental and physical health in low-income contexts is lacking, and the high physical requirements of daily life may lead to an amplification of the association between physical and mental health in SSA. A better understanding of the health of mature adults, including how it is affected by DA and other noncommunicable diseases, is therefore critical (Chisholm et al. 2016; Daar et al. 2014). Indeed, mature adults constitute an economically important subset of society: they have almost universal labor force participation in Malawi (98 % at ages 50–64, and 90 % at age 65+) (Malawi National Statistical Office 2010), make important contributions to intergenerational transfers to both children and elderly parents (Kohler et al. 2012), and have pivotal caretaking roles in families affected by HIV/AIDS (Reijer 2013).
A nascent body of research has started to investigate the shift in SSA’s disease burden toward disabling chronic conditions and noncommunicable diseases (Dalal et al. 2011; Ebrahim et al. 2013). The Lancet’s Movement for Global Mental Health, for example, accentuated the multiple interrelations between mental and other health disorders, and the implications of their frequent co-occurrence (Patel et al. 2008). DA, for example, are associated with communicable and noncommunicable diseases (including HIV/AIDS) and often result in poor health management; higher co-morbidity; poorer disease prognosis, including more HIV risk behaviors and lower antiretroviral therapy (ART) treatment adherence; social exclusion; and increased poverty (Lund 2010; Mayston et al. 2012; Prince et al. 2007). Scholars and NGOs have argued that health research and health policies in SSA low-income countries devote far too few resources to DA and poor mental health (Chisholm et al. 2016; Jack et al. 2014). Leading development economists have also claimed that a poverty trap exists in low-income countries because poverty begets poor mental health and therefore low productivity (Banejee and Duflo 2007) and because there is “no health without mental health” (Prince et al. 2007). Malawi’s leading national newspaper recently wrote about the poor mental health of the nation as a “mental bomb that we are comfortably sitting on, without realizing its consequences in the near future” (Nation2012), and others have argued for the strengthening of mental health in the National Health Sector Strategic Plan (Bandawe 2010). Yet, research on DA that could inform mental health–related changes in health policy or health infrastructure in SSA low-income countries is scant; the scarce research that exists is often based on clinical/convenience samples rather than population-based studies and often uses measures that have unknown sensitivity and specificity in relation to a clinical diagnosis (Lund 2014). Although the WHO STEPwise approach to noncommunicable disease risk factor surveillance (STEPS) and Study on global AGEing and adult health (SAGE) surveys measure DA and other noncommunicable diseases in SSA (He et al. 2012; Msyamboza et al. 2011), these data lack information on older individuals and/or are available only for higher-income SSA countries, the findings of which do not necessarily generalize to the low-income country context studied here.
Data and Measures
Malawi is ranked 174 of 187 in terms of the human development index (United Nations Development Programme (UNDP) 2014), with approximately 15 % of its population considered ultra-poor (UNDP 2010). Life expectancy at birth was age 51 for men and age 55 for women in 2010, and healthy life expectancy at birth is estimated to be 44 years for males and 46 years for females (Salomon et al. 2012). In rural areas, where our study population is based, the majority of individuals engage in home production of crops, complemented by some market activities. HIV/AIDS is widespread, as are worries about HIV and the household-level experience of AIDS-related morbidity and mortality. However, the vast majority of the population—more than 85 % of adults aged 15–49, and an even higher fraction among adults aged 50 and older—are HIV-negative (Freeman and Anglewicz 2012).
The MLSFH (Kohler et al. 2015) is one of very few long-standing publicly available longitudinal cohort studies in an SSA low-income country context, with eight data collection rounds during 1998–2013 for up to 4,000 individuals. The MLSFH is collected in three districts: Balaka in the south, Mchinji in the center, and Rumphi in the north. The MLSFH cohorts were selected to represent the rural population, where the majority of Malawians (85 %) live. The MLSFH Cohort Profile provides detailed information on sampling procedures, survey methods, and survey instruments (Kohler et al. 2015); pertinent information is also provided in Online Resource 1 (section S1).
Our analyses focus on the MLSFH mature adults sample, consisting of respondents aged 45 and older who participated in the 2012 (N = 1, 266) and 2013 (N = 1, 257) MLSFH surveys (see Online Resource 1, sections S1.2–S1.4 for additional details on the MLSFH mature adult sample). Specifically, the 2012 and 2013 MLSFH mature adult surveys collected extensive assessments of different dimensions of mental health and allow us to assess both the presence and the severity of depression and anxiety disorders.
The SF12 mental health score, included in MLSFH since 2006, is a widely used measure of overall mental health that has been validated in many different contexts (Macran et al. 2003; Ware et al. 1998). Higher SF12 scores reflect better mental health. The SF12 score, however, does not allow an assessment of the presence and/or severity of clinically defined mental disorders, such as depression or anxiety.
To overcome this limitation, the MLSFH collected additional measures of mental health in 2012 and 2013: the depression and anxiety modules of the Patient Health Questionnaire (PHQ). The PHQ refers to the self-administered version of the PRIME-MD diagnostic instrument for making criteria-based diagnoses of common mental disorders encountered in primary care (Kroenke and Spitzer 2010; PHQ 2011).
The depression module (PHQ9) includes nine questions regarding whether a respondent has been bothered during the last two weeks by aspects such as the following: (1) little interest or pleasure in doing things; (2) feeling down, depressed, or hopeless…; (6) feeling bad about yourself—or that you are a failure or have let yourself or your family down…; (9) thoughts that you would be better off dead or of hurting yourself in some way.
The anxiety module (GAD7) includes seven questions regarding whether a respondent has been bothered during the last four weeks by aspects such as the following: (1) feeling nervous, anxious, or on edge; (2) not being able to stop or control worrying…; (6) becoming easily annoyed or irritable; (7) feeling afraid as if something awful might happen. Response categories for all questions in the PHQ9 and GAD7 modules range from 0 (not at all) to 3 (nearly every day).
An overall depression score (PHQ9 score) was computed as the total score calculated from the PHQ9 instrument. Based on this PHQ9 score, official guidelines classify the clinical significance of depression as follows (with proposed treatment actions in parentheses): 0–4 = none/minimal depression (no treatment); 5–9 = mild depression (watchful waiting, with repeat PHQ9 at follow-up); 10–14 = moderate depression (treatment plan, considering counseling, follow-up and/or pharmacotherapy); 15–19 = moderately severe depression (active treatment with pharmacotherapy and/or psychotherapy); and 20–27 = severe depression (immediate initiation of pharmacotherapy and, if severe impairment or poor response to therapy, expedited referral to a mental health specialist for psychotherapy and/or collaborative management) (Kroenke and Spitzer 2010).
Similarly, an overall anxiety score (GAD7 score) was computed as the total score calculated from the GAD7 instrument. The official guidelines specify scores of 5, 10, and 15 as cut points for mild, moderate, and severe anxiety, respectively (Kroenke and Spitzer 2010). A score higher than 10 is recommended for further evaluation when GAD7 is used as a screening instrument for anxiety disorders.
Although evidence suggests some universality in the experience of depression and anxiety in SSA when measured with scales that have primarily been validated in high-income settings, some variation also exists in the expression and salience of symptoms across contexts that may reduce the sensitivity of PHQ9, GAD7, and related instruments to assess DA in an SSA low-income country context (Kessler and Bromet 2013; Sweetland et al. 2014). To reduce these concerns about the validity of established cut points for classifying DA, we assume for most of our analyses that mental health does not involve a qualitative discontinuity between depressed and nondepressed status (or no/mild/severe anxiety disorder) that occurs at specific PHQ9 or GAD7 official classification levels. Instead, we use a linear specification of the PHQ9 depression and GAD7 anxiety scores. In both cases, higher scores indicate that an individual experiences higher number of depressive or anxiety symptoms. By using the full variation in the PHQ9 and GAD7 scores, we are able to better capture the range of depressive and anxiety symptoms without relying on a classification scheme that has predominately been validated for Western and more affluent contexts and that may or may not be fully applicable to the low-income countries context studied here (Jacob and Patel 2014; Sweetland et al. 2014).
To facilitate longitudinal analyses of DA based on the MLSFH waves 2006–2013, of which only the last two waves include PHQ9 and GAD7 scores, we also define a combined depression/anxiety index (DAX) derived from two questions that are part of the SF12: (1) “How much time of the time during the past 4 weeks have you felt calm and peaceful?,” and (2) “How much of the time during the past 4 weeks have you felt downhearted and depressed?” Both questions are specifically related to DA and are available in the MLSFH since 2006. The response categories range from 1 = all the time to 5 = none of the time. The DAX is then computed as follows: DAX = 0 (no depression/anxiety) when Q1 ≤ 2 and Q2 ≥ 4; DAX = 2 (moderate/severe depression/anxiety if Q1 ≥ 4 and Q2 ≤ 2; and DAX = 1 (mild depression/anxiety) otherwise. The DAX is related to the SF12 mental health score, with a correlation of approximately –.8 in our data (Table 2), but it has the advantage for our analyses that it is more explicitly focused on DA (see Online Resource 1, Table S18).
MLSFH Mature Adults: Summary Statistics
Table 1 (columns 1–3) reports summary statistics of the analytical sample (N2012 = 1,246 respondents aged 45 and older). Approximately 40 % of the sample is aged 45–54, with the remainder approximately evenly split between ages 55–64 and 65 and older. Schooling attainment is generally low: 36 % have no formal schooling, and very few respondents have attended secondary school. Approximately 27 % of the study population is Muslim. Almost all the men are currently married, in contrast to only 63 % of women (two-thirds of the not currently married women are widowed, with the remainder being divorced or separated). Only approximately 5 % of the study population is HIV-positive, reflecting possibly high levels of mortality among HIV-positive individuals for these cohorts in the past (Table S2 reports HIV prevalence by age).
Patterns of Depression and Anxiety Among Mature Adults
The top panel of Table 2 reports summary statistics for PHQ9 depression score, GAD7 anxiety score, and SF12 score for overall mental health in 2012. Subjective well-being is reported for comparison, as is our depression/anxiety index (DAX) derived from two questions of the SF12. Columns 1–3 report the mean scores and standard deviations, and columns 4–6 report the correlations among the four scores in 2012 (similar patterns prevail in 2013, see Table S3). In addition, column 7 reports the correlation of each measure between the 2012 and 2013 waves. Columns 4–7 in Table 1 report a multivariate regression of each of the three mental health scores and subjective well-being on basic respondent’s characteristics.
All three measures indicate marked gender difference, with women having worse mental health and subjective well-being than men (Table 2), and a significant decline of mental health occurs with age (Table 1, columns 4–5): the PHQ9 depression and the GAD7 anxiety scores are elevated among 65- to 74-year-olds by approximately 33 % and 46 % of a standard deviation respectively, and by approximately 70 % and 80 % of a standard deviation among those aged 75 years or older. Overall mental health declines with age as does well-being (Table 1, columns 6–7). Higher levels of schooling are associated with better mental health outcomes, as is being currently married (Table 1). Being HIV-positive is not significantly associated with mental health—possibly related to the low HIV prevalence (5 %)—nor is being Muslim.
The relatively low levels of the PHQ9 and GAD7 scores in Table 2 are noteworthy given that the scales range from 0 to 27 (for the PHQ9) and 0 to 21 (for the GAD7), and informal observations during fieldwork indicated relatively widespread problems with poor mental health. The mean SF12 mental health score of 53 is also not substantially different from the mean levels that this score yields in many higher-income contexts (e.g., the SF12 score is calibrated to have a mean of 50 and a standard deviation of 10 in the U.S. population), and only approximately 15 % of mature adults reported being very or somewhat unsatisfied with life. Based on the official PHQ9 and GAD7 classification guidelines, approximately 25 % of respondents (29 % among women and 19 % among men) exhibit mild or higher levels of depression, and 22 % (25 % among women, 28 % among men) express mild or higher levels of anxiety. Moderately severe and severe depression, and moderate to severe anxiety are fairly rare, affecting only approximately 2 % of mature adult respondents (Table 2, columns 1–3).
These low reported levels of DA may be related to the fact that individuals in a context such as Malawi are often not very sensitized toward issues related to poor mental health (see also Kessler and Bromet 2013) and thus tend to underreport anxiety, depression, and poor mental health. Alternatively, they may tend to understate feelings of DA because they are relatively common in this context, and individuals use their immediate social environment as their reference group. The influence of such reference group has been well documented with respect to subjective well-being and subjective health (Carrieri 2012), where subjective measures often accurately reflect within-population variation in well-being and health but often do not substantially vary across populations with very different levels of objective health or well-being.
Nevertheless, even at the fairly modest levels, the presence of DA is importantly associated with lower subjective well-being, lower food (protein) consumption, less sexual activity, lower earnings and savings, and reduced work efforts in this study population (Models 1 and 2 in Table 3). The effects are sizable, with mild depression associated with 11 % decline in the number of days on which individuals consume chicken, fish, or meat; a 15 % reduction in annual earnings; and a 15 % reduction in the number of hours devoted to work on one’s own farm or in domestic work. Reductions are larger—often by approximately 50 % to 100 %—for moderate and more severe levels of depression. In addition, for several outcomes in Table 3, DA have independent effects in multivariate analyses (Model 3).
Within each wave, the correlation coefficients (absolute values) for the PHQ9, GAD7, and SF12 scores range between .55 and .68 (columns 4–5 in Table 2), indicating that depression, anxiety, and overall mental health are affected by common stressors, such as social or economic shocks (e.g., poor crop yields and morbidity/mortality of adult household members), health concerns (including worries about HIV/AIDS), and poor physical health. The correlation of the mental health measures—depression, anxiety, and overall mental health (SF12)—with subjective well-being is approximately .30, indicating that subjective well-being does not adequately capture these dimensions of mental health. Our DAX is correlated at .50–.59 with the more detailed PHQ9 and GAD7 measures of depression and anxiety, indicating that this simple index already captures significant variation in DA.
The correlation of the mental health scores between 2012 and 2013 is, however, relatively weak (.27–.23; Table 2, column 7), indicating that many of these influences on depression, anxiety, and overall mental health are relatively short-term, often dissipating during the course of one year. These low correlations between the DA scores in 2012 and 2013 are consistent with the expectation that mild and moderate DA are more likely to be short-term and fluctuate over time, which is consistent with our upcoming multistate life table analyses, whereas severe depression is generally more persistent (Kessler and Bromet 2013).
Age Patterns of Mental Health
Figure 2 depicts the age pattern of DA, SF12, and overall mental health scores (based on pooled 2012–2013 regressions, controlling for schooling, region, and MLSFH wave). DA scores increase markedly for individuals aged 55 and older, with women exhibiting higher levels of DA across all mature adult ages. The increases with age are substantial: the average depression score for a 70-year-old woman is 1.9 points (52 % of a standard deviation) above that of a 50-year-old women, and the average anxiety score is 1.5 points (55 % of a standard deviation) higher (for men, the respective differences are 29 % and 35 % of a standard deviation). Overall mental health, as reflected by the SF12 score, declines accordingly, and the diminishing mental health among older adults in rural Malawi is part of a broader pattern of declining subjective well-being with age.
The coefficients estimated via linear regressions of our mental health measures on age and selected controls in Table 4 (column 1) show that, on average, the depression and anxiety scores increase by .093 and .077, respectively, per year of age for mature adults in the 2012–2013 MLSFH data. In addition, overall mental health (SF12) and subjective well-being decline by .21 and .02 per year of age. Except for the anxiety score, no significant differences are evident in these linearized age gradients between males and females.
The analyses in Fig. 2 and Table 4 (column 1) are cross-sectional, making it conceivable that the reported age gradient results from cohort differences in mental health (Blanchflower and Oswald 2008). Given that the MLSFH includes longitudinal information on the depression/anxiety index (DAX), the SF12 mental health score, and subjective well-being for the period 2006–2013 (five waves, available for these three outcomes only), we are able to estimate hierarchical age-period-cohort (APC) growth curve models (Yang and Land 2013: chap. 9) that estimate longitudinal changes in mental health with age while allowing for cohort differences (for similar analyses, see Yang 2007; Yang and Lee 2010). Because the MLSFH provides only five data points across a seven-year time period (2006–2013), it is realistic to identify only a linear age gradient within each cohort along with a linear trend across cohorts in both the intercept and the age gradient. Essentially, our APC growth curve model estimates yit = β0i + β1iAgeit + εit, where yit is one of our mental health outcomes, and the intercept β0i = γ00 + γ01Cohorti + γ02Femalei + γ03Schoolingi + γ04Regioni + ν0i allows for level differences across cohorts and controls for systematic differences by schooling, region, and gender. The growth rate β1i = γ10 + γ11Cohorti + ν1i allows the age gradient in the mental health outcome yit to differ by cohorts. Five birth cohorts are specified as 1939 or earlier, 1940–1946, 1947–1953, 1954–1960, and 1961–1967. A detailed discussion of our APC model is provided in Online Resource 1 (section S2).
The relevant coefficients determining the age gradient in this gender-pooled APC growth curve model by age within each cohort and the change in the age gradient across cohorts are reported in column 2 of Table 4 (the full model is reported in Online Resource 1, Table S11). Column 3 in Table 4 also reports the results of an individual fixed-effect model of each longitudinal mental health indicator on age.
The conclusion obtained from the APC growth curve model and fixed-effect analyses in Table 4 is that the declines in mental health with age illustrated in our cross-sectional analyses (Fig. 2 and column 1 in Table 4) also occur within cohorts as individuals age, and the correspondence between the cohort age pattern and cross-sectional age pattern is relatively close.
Although we can conduct the longitudinal APC and fixed-effect analyses only for the bottom three indicators in Table 4 (DAX, SF12 mental health score, and subjective well-being), the moderately strong correlation of these measures with both the depression and anxiety scores (Table 2) suggests that findings would be similar also for the PHQ9 depression and GAD7 anxiety scores. The results in columns 2–3 in Table 4 hence provide evidence that the observed declines of mental health with age are not the result of cohort differences; instead, both the longitudinal analyses and the cross-sectional analyses consistently suggest that older individuals have worse mental health—including higher levels of DA and lower levels of well-being—than middle-aged individuals.
This marked decline of mental health is in sharp contrast to the age pattern of psychological well-being that has been documented in the United States and other high-income countries, where it is generally U-shaped, showing increasing well-being after an age of 50 years. A large-scale study in the United States, for example, showed that stress and anger steeply declined beginning in the early 20s, worry was elevated through middle age and then declined, and sadness was essentially flat (Stone et al. 2010). Depression has been shown to have a similar inverted U-shape, declining markedly after middle age (López-Ulloa et al. 2013). We find the opposite in our data: DA increase with age among mature adults in Malawi, along with a decline in overall mental health and subjective well-being. In summary, two important findings emerge from the analyses of mental health among mature adults in Malawi. First, across all measures—depression, anxiety, overall mental health, and subjective well-being—a consistent gender gap prevails, with women having worse mental health (but not necessarily a stronger age gradient) than men (a pattern that is also often observed in other populations; Kessler and Bromet 2013). Second, mature adults have an age pattern of mental health that is opposite to the one documented in high-income contexts: mental health among mature adults declines with age, with a particularly marked decline after age 55. This pattern does not conform to the inverted U-shape of well-being and U-shape of mental health that has been documented across the life course in the United States and other developed countries (Stone et al. 2010).
Mental Health, Age, and Socioeconomic Stressors
Mature adults in rural Malawi are regularly exposed to social and economic shocks (stressors) that are likely to affect mental health, including mortality of household members, disease and health shocks, volatile crop yields and incomes, and others. Beyond the direct mental health effects of these shocks, stress due to the perception of uncertainty and exposure to substantial risks can result in DA (Carleton et al. 2012; Lund et al. 2010). We investigate whether the gender gap and age gradients are due to differential exposure to such stressors, either by sex or age, or whether other factors provide the underlying cause of mental health differences by age and sex. Here, as well as in the subsequent sections, our main analyses will focus on DA, as measured in the 2012–2013 MLSFH Mature Adult Surveys through the PHQ9 and GAD7 scores. Corresponding analyses of overall mental health and subjective well-being are included in the Online Resource 1 (Tables S12–S14).
Measures of socioeconomic shocks in the MLSFH are based on a set of questions about whether a respondent’s household was affected in the last year by (1) death/serious illness of an adult member; (2) poor crop yields, loss of crops due to disease or pests, or loss of livestock; (3) loss of income source; (4) breakup of the household; or (5) damage of house due to fire, flood, or other unexpected event. Panel A in Table 5 shows the mean number of shocks experienced by respondents, conveying clearly that they are common: more than 30 % of mature adults experienced a death or serious illness of an adult household member, more than 50 % were affected by poor crop yields or crop failures, and more than 20 % lost an income source within the one to two years prior to the survey. On average, respondents experienced approximately 1.2 of the shocks listed in columns 1–6 of Table 5. Although experience of such socioeconomic shocks is common, no strong age patterns are evident because these socioeconomic shocks are essentially household-level or community-level events, hence affecting individuals mostly irrespective of their age.
Panel B in Table 5 reports the regression coefficient for each of these socioeconomic stressors obtained from linear regressions of the PHQ9 depression score on the respective stressor, age, female, and selected controls (region, schooling, and MLSFH wave). Panel C reports the change in the age gradient and the female-male difference in depression when each socioeconomic stressor is included in the estimation.
The key finding obtained from these analyses in Table 5 is that although several of the socioeconomic shock variables are strongly and significantly associated with depression, controlling for socioeconomic shocks in the estimation neither changes the age gradient of depression nor reduces the estimated female-male difference. Similar findings prevail for anxiety (panels D–E in Table 5), overall mental health (SF12), and subjective well-being (Online Resource 1 Table S12). The experience of shocks, such as mortality or morbidity of household members, poor crop yields, or income loss, importantly contributes to DA among mature adults in Malawi, as one would expect based on the existing literature (Baird et al. 2011; Lund et al. 2010). However, a differential exposure to such socioeconomic stressors explains neither the gender gap in mental health nor the rise of depression or anxiety with age among mature adults.
Mental Health, Age, and Risk/Uncertainty–Related Stressors
Table 6 continues the approach of the previous section, investigating whether stressors related to perceived risks and uncertainty provide an explanation for the increase of DA with age as well as the gender gap in mental health among mature adults. Perceived risks and uncertainties are well-known factors contributing to depression and anxiety (Carleton et al. 2012; Lund et al. 2010), and the MLSFH contains several measures of perceived risk that are particularly pertinent to the Malawian context: worries about HIV/AIDS; perceived local level of AIDS mortality (how many people the respondent knows who have died from AIDS); and several measures of perceived risk based on subjective probabilities, including the subjective probability of own HIV infection, HIV infection of spouse or partner, risk of dying within one year of the survey, and risk that a local person of the respondent’s age and sex dies within one year (general mortality).
Panel A in Table 6 shows that mature adults in Malawi perceive substantial risks and uncertainty. For example, mature adults are substantially worried about AIDS, know between 3.3 and 4.1 persons who (they suspect) have died of AIDS in the last year, estimate their own risk of HIV infection as 12 % to 21 % and that of their spouses/partners as 14 % to 24 %, and perceive a one-year mortality risk of 28 % to 38 %—the latter being a substantial overestimation of the actual mortality risk (Delavande and Kohler 2009, 2016). Moreover, these stressors exhibit modest age patterns (panel A): worries about HIV/AIDS, and the subjective probability of the respondent or his/her spouse being infected with HIV decline with age, whereas the perceived local AIDS mortality and the subjective probability of dying (self) increase with age. There is no age pattern in the perception of general mortality.
All these stressors pertaining to the respondent or his/her spouse (i.e., columns 1–5) are strongly and significantly associated with DA in the expected direction: the higher the perceived risk or uncertainty, the more depressed or anxious the respondent is (Table 6, panels B and D). However, differential exposure to these risk-/uncertainty-related stressors by age does not explain the increase in DA by age. Controlling for these stressors in analyses of the age gradient of depression or anxiety does not substantially reduce the age gradient in any of the models (Table 6, panels C and E); in several cases, the age gradient becomes even steeper after controlling for risk-/uncertainty-related stressors. Similarly, sex differences in risk-/uncertainty-related stressors do not explain the gender gap in depression or anxiety. Hence, our analyses suggest that risk-/uncertainty-related stressors are important factors contributing to DA (and more generally, poor mental health and subjective well-being), but they do not explain why older individuals tend to have worse mental health (including higher levels of DA) than middle-aged individuals, nor do they explain why women have worse mental health than men.
Mental Health, Age, and Physical Health–Related Stressors
Table 7 investigates whether stressors related to the respondents’ physical health contribute to poor mental health among mature adults and whether these stressors possibly explain the female-male difference in DA and the increases in DA with age.
Physical health–related stressors available in the MLSFH include the following: the respondent accomplished less in the past four weeks due to his/her physical health, the respondent experienced work limitations due to physical health, or pain interfered with his/her work. In addition, the MLSFH mature adult data include measures of grip strength, body mass index (BMI), and blood pressure. All these stressors indicate that older individuals experience pain or are limited in their work due to poor health more often than younger mature adults. Grip strength and BMI decline substantially with age, while blood pressure increases. Only 4 % of MLSFH mature adults have a BMI ≥ 30 in 2012 (6.5 % of women and 0.8 % of men), most (67 %) have normal BMI (62 % of women and 74 % of men), and 17.7 % are underweight (BMI < 18.5; 18 % of women and 17 % of men).
The physical health–related stressors in columns 1–4 of Table 7 are strongly and significantly associated with both depression and anxiety, as expected (Kessler and Bromet 2013), indicating that physical and mental health are closely related in this context. Particularly noteworthy is the association with measured grip strength, which has been suggested as a strong predictor of functional limitations, limitations in activities of daily living (ADLs), morbidity, and mortality (Nybo et al. 2001; Rantanen et al. 2000).
Moreover, contrary to our earlier analyses of socioeconomic and risk-/uncertainty-related stressors, the inclusion of physical health–related stressors (columns 1–4) in estimates of the age gradient of DA substantially reduces the estimated coefficient for age. The results in panels C and E of Table 7 thus suggest that declines in physical health (as indicated by grip strength) as well as the implications of declining physical health for daily activities and work (as indicated by the self-reported limitations in columns 1–3) are very important contributors to the rise in DA with age among mature adults. For example, the age gradient in both depression and anxiety is reduced by approximately one-third if the analyses control for grip strength. This reduction is remarkable as grip strength is generally seen as an indicator of functional limitations. The age gradient is reduced even more, by approximately two-thirds, if the self-reported work limitations in columns 1–3 are included. In addition to explaining a substantial fraction of the decline of mental health with age, controlling for the physical health–related stressors in columns 1–4 also substantially reduces the female-male difference in DA. In contrast, BMI does not seem to be associated with either depression or anxiety among mature adults, and higher blood pressure seems to be associated with lower levels of anxiety (for depression, the point estimate is also negative, but not statistically significant). Neither help explain the age gradient or female-male difference in DA.
These physical health–related stressors differ importantly from the socioeconomic and risk-/uncertainty-related stressors analyzed in the previous sections. The latter were strongly associated with DA but did not help explain the age gradient and female-male difference in DA. Stressors related to physical health, in contrast, are strongly associated with DA, and do help explain the age gradient and female-male difference in DA. The age gradient is reduced by up to two-thirds, as is the female-male difference, after analyses control for these stressors. Moreover, the differential physical health of females and males, highlighted in panel A of Table 7 and Payne et al. (2013), also seems to contribute importantly to the fact that women experience higher levels of DA than men across all mature adult ages.
Our analyses cannot identify the causal direction, but it seems likely that in a context such as Malawi, the relatively poor physical health of mature adults and the decline of physical health with age are related to the long-term exposure to disease and poor nutrition that mature adults in this context have experienced. If this is the case, our analyses suggest that the general decline of physical strength and health with age (as indicated by grip strength) and the interference of poor physical health with day-to-day activities (as indicated by the self-reported limitations in columns 1–3) importantly contribute to the rise of DA with age among mature adults. This pattern is similar to that observed in the Tsimane forager-farmers that share with our Malawian mature adult population a decline of mental health with age (Stieglitz et al. 2015). DA arising from poor physical health in contexts such as Malawi may be importantly related to individuals’ perception of being an adult and functioning member of their family and community (Freeman 2012).
Mental Health Transition Probabilities
In our final section, we shift from the analyses of respondents’ mental health status in 2012–2013 to dynamic multistate life table analyses of transitions between different DA levels during 2006–2012. The specific aim is to derive health expectancies (HEs)—that is, estimates of the remaining person-years a mature adult spends with depression or anxiety, as well as age- and gender-specific transition probabilities between different DA levels. Online Resource 1 (section S4) provides a detailed description of the estimation of this multistate life table model.
The key insight of these multistate life table analyses of our DAX measure is that mature adults in rural Malawi can expect to live a significant fraction of their remaining life expectancy subject to some anxiety and depression (Fig. 3). The model estimates that, on average, a 45-year-old woman will live almost 55 % of her remaining life with some anxieties or depressive symptoms; this figure is approximately 40 % for a 45-year-old man. Approximately one-half of this time is subject to moderate to severe DA—levels of DA that likely have substantial effects on individuals’ well-being and social/economic lives. Figure 3 also highlights the clear and progressive increase in the amount of remaining life spent with DA. By age 75, women are expected to live approximately 73 % of their remaining life with some DA; this figure is 62 % for men. Approximately one-half of this time is subject to moderate to severe levels of DA. The analyses of the underlying transition rates between DAX states (Figs. S2–S4 in Online Resource 1) further reveal that transitions between DAX states are fairly common and are strongly age-patterned: the probabilities of recovering from mild or moderate/severe DA (DAX = 1 or 2) decline strongly with age, whereas the probabilities of transitioning to a worse mental health state increase with age. Moreover, the fairly high rates of transition between DAX states during the mature adult life course mean that time spent with DA does not occur solely at the end of life; instead, these DAX states are dynamic, and especially at younger mature adult ages, individuals will often recover from DA.
Summary and Discussion
Our analyses focus on the demography of mental health among mature adults (aged 45 or older) in rural Malawi. Using innovative data on DA collected in 2012–2013 as part of the MLSFH (Kohler et al. 2015), our analyses show that despite substantial socioeconomic hardship and often poor physical health, few mature adults have severe or moderately severe levels of DA. Still, mild to moderate levels of DA are common among mature adults. Even at this modest severity, DA are systematically associated with adverse outcomes, such as poorer nutritional intake, less sexual intercourse, and substantially reduced work efforts and earnings.
DA are more frequent among women than men, and individuals are often affected by both. DA also increase substantially with age for both genders, along with substantial declines in overall mental health and subjective well-being. Mature adults can also expect to spend a substantial fraction of their remaining lifetime—for instance, 52 % for a 55-year-old woman—affected by DA. The positive age gradients of DA observed in the cross-sectional (pooled 2012–2013) MLSFH data are not primarily due to cohort effects given that marked increases in DA with age are also found in fixed-effects and hierarchical APC growth curve analyses using up to five waves of MLSFH data from 2006–2013 and measure longitudinal changes in DA with age allowing for cohort differences. This positive age gradient of DA in our study population is in sharp contrast to the age pattern of mental health that has been shown in high-income contexts where older individuals often reflect lower levels of DA and better subjective well-being than middle-aged individuals (Blanchflower and Oswald 2008; López-Ulloa et al. 2013; Steptoe et al. 2015; Stone et al. 2010).
The marked gender gap and the positive age gradients in DA are not explained by a differential exposure to socioeconomic stressors. Although the experience of a recent income loss, household-level morbidity/mortality, or other socioeconomic shocks is associated with higher levels of DA, individuals across all mature adult ages are exposed to these shocks. A similar pattern prevails for stressors related to risk perceptions and uncertainty, such as worries about HIV and elevated risk-perceptions of mortality. Instead, our analyses suggest that declining mental health with age is importantly related to declines in physical health among mature adults as measured by hand grip strength and limitations in daily activities. Moreover, due to their gender differences and increasing prevalence with age, these factors explain a substantial part of the positive age gradient and gender gap in DA. Most strikingly, the decline in grip strength with age alone explains 32 % to 35 % of the age gradient in DA. Our multistate life table analyses of mental health changes in the MLSFH during 2006–2012 additionally show that mature adults can expect to spend a substantial fraction of their remaining lifetime affected by DA (52 % for a 55-year-old woman, 39 % for a 55-year-old male, and even higher proportions for older individuals). Hence, even if the prevalence of DA at any given time is only moderately high, spells of DA across the mature adult life course accumulate, especially as individuals age; as a result, a substantial fraction of remaining life expectancy for mature adults is lived with DA. Together, our analyses therefore characterize processes of declining mental health with age, with significant gender differences and possibly important productivity and well-being implications, in a rapidly growing but understudied portion of the SSA population.
Our analyses also describe an aging trajectory distinctly different from those observed in higher-income countries: the positive age gradients in DA observed in Malawi are in sharp contrast to the age pattern of mental health shown in high-income contexts where older individuals often report lower levels of DA and better subjective well-being than middle-aged individuals. Although socioeconomic and risk-/uncertainty-related stressors are strongly associated with DA, they do not explain the positive age gradients and gender gap in DA. Stressors related to physical health, however, do. Our analyses cannot identify the causal direction, but it seems likely that in a rural SSA context, the relatively poor physical health of mature adults is importantly related to the long-term exposure to disease and poor nutrition that mature adults have experienced. If this is the case, our analyses suggest that the general decline of physical strength and health with age, and the interference of poor physical health with daily activities, are key drivers of the rise of DA with age among mature adults. Addressing the challenge of poor mental health among mature adults in SSA low-income country contexts may therefore require approaches that address both physical and mental health concerns, including treatment for DA combined with palliative care for pain and physical health problems for which the overburdened health systems in rural SSA low-income countries are currently unlikely to provide effective treatments.
Our results are also important because they help inform the health policies and health sector strategies required for preparing for the growing population of mature adults and elderly individuals in SSA low-income countries. The average annual growth rate of the population age 60 and older in SSA is projected to increase from more than 2 % (already higher than the 60+ growth rate in developing countries) to more than 4 % during the next 45 years—four times the growth rate expected in developed countries. In countries like Malawi, this rapid growth of the mature adult and older population occurs while the overall age structure of the population will remain relatively young. Hence, while rapid population growth continues to be a major social and policy issue in SSA, current demographic and epidemiological trends foreshadow the coming challenge of a growing elderly population in SSA. Because of high levels of morbidity, low levels of economic development, and widespread poverty, individual aging and population aging in SSA will likely be associated with a unique set of demographic and economic concerns. Yet, national and international decision-makers lack an understanding of the magnitude of the aging problem in SSA. Evidence from more developed contexts is often not sufficient for understanding the health issues and healthcare needs associated with a growing aging population in SSA. In Malawi and other SSA countries, health sector strategies are beginning to recognize the importance of mental health disorders that will gain further prominence in coming decades. Our findings provide important insights into the potential gains in well-being and economic productivity that can arise from investments in the mental health of mature adults in SSA low-income countries, and they highlight importance of expanding the identification and treatment of mental health disorders in these contexts.
We gratefully acknowledge the generous support for the Malawi Longitudinal Study of Families and Health (MLSFH) by the Rockefeller Foundation; the National Institute of Child Health and Human Development (NICHD, Grant Nos. R03 HD05 8976, R21 HD050653, R01 HD044228, R01 HD053781); the National Institute on Aging (NIA, Grant Nos. P30 AG12836 and R21 AG053763); the Boettner Center for Pensions and Retirement Security at the University of Pennsylvania; and the NICHD Population Research Infrastructure Program (Grant Nos. R24 HD-044964), all at the University of Pennsylvania. We are also grateful for support through the Swiss Programme for Research on Global Issues for Development (SNF r4d Grant 400640_160374) as well as pilot funding received through the Penn Center for AIDS Research (CFAR), supported by NIAID AI 045008 and the Penn Institute on Aging. Collin Payne also gratefully acknowledges support through a United States National Science Foundation Graduate Research Fellowship (Grant No. DGE-0822).