Variations in self-reported mental health in low-income urban areas during COVID-19: Insights from South Africa

doi:10.21203/rs.3.rs-4271486/v1

Download PDF

Research Article

Variations in self-reported mental health in low-income urban areas during COVID-19: Insights from South Africa

https://doi.org/10.21203/rs.3.rs-4271486/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Many studies have reported poor mental health coinciding with lockdowns and experiences of the COVID-19 pandemic. The harshest lockdowns tended to happen in cities, and yet differing urban forms are likely to have mediated their effects. South Africa, characterised by varying urban forms, had high levels of poor mental health pre-pandemic, and yet experienced one of the toughest lockdowns. We explored the differences in the prevalence of, and the factors associated with, self-reported changes in mental health, in socio-economically complex and heterogenous urban areas in South Africa, to understand how urban form may have mediated the effects of the pandemic. We specifically looked at Atteridgeville, an old formal township; Melusi, a newer, and less dense informal settlement; and Hillbrow, a long-standing and deteriorating high-rise high-density formal area in central Johannesburg. We used cross-sectional data from 1332 respondents who participated in GRT-INSPIRED’s face-to-face interviews from March to October 2021. Weighted logistic regression models were used to determine the relationships between socioeconomic and demographic variables with self-reported mental health during the pandemic. We found that being from Melusi, the informal location, as compared to being from Hillbrow, the inner-city location, was associated with increased odds of worsening mental health during the pandemic. It was also only in Melusi where receiving a government grant had a protective effect against worsening mental health. Interestingly, we also found that the factors that are associated with worsening mental health are not the inverse of the factors that improve mental health. The differences in the effects of explanatory variables by location suggest that the effects of COVID-19 on mental health may have been mediated through urban form as well as socioeconomic factors such as social networks. This study provided both the factors that can be targeted to prevent poor mental health and to sustain good mental health.

Self-reported mental health

socioeconomic determinants

COVID-19 containment measures

urban form

low-income countries

The outbreak of COVID-19 in December 2019¹ brought significant changes in human interactions that profoundly affected wellbeing. To reduce the spread of, and morbidity and mortality from COVID-19, many countries implemented various restrictions such as lockdown, social distancing, and closure of “non-essential services”. The degree of stringency of these measures varied, depending on the COVID-19 reproduction number, and the capacity of health facilities to care for existing and new cases². Whilst COVID-19 itself is a physical condition, a considerable amount of research found high levels of poor mental health coinciding with lockdowns and experiences of the pandemic^3,4.

Beyond behavioural and genetic factors, the environment to which people are exposed contributes to their mental health^5,6. Urban characteristics provide both protective and risk factors for mental health⁷. For example, planned urban areas, where people have access to health care, access to and use of green spaces, and access to safe drinking water, sanitation, and hygiene, are associated with good mental health⁸. On the other hand, exposure to urban socioeconomic disadvantages such as informality, poor housing, overcrowding, pollution, poor diet, high crime, low social cohesion, poverty, inequality, and daily challenges to meet basic needs are associated with poor mental health^9,10.

Most low to middle-income countries (LMICs), which include South Africa, are characterised by high unemployment, poverty, inequality, hunger, and a high burden of disease, among other societal ills, which are all linked to mental health¹¹. Despite LMICs housing the largest share of the global population and being characterised by an array of risk factors for poor mental health, research on mental health in LMICs during the COVID-19 pandemic is less compared to high-income countries¹². In addition to exacerbating these socioeconomic ills and the burden of disease, COVID-19 and its related containment measures may have magnified disparities in mental health in LMICs.

The COVID-19 containment measures confined people to their dwellings and restricted important human interactions and the ability to work – which are all risk factors for poor mental health, especially for vulnerable people in already disadvantaged urban areas. The fear of COVID-19 infection, and morbidity and mortality from COVID-19, as well as readjustment of COVID-19 restrictions, posed a higher risk of developing negative emotional states¹³ and added to the stress burdens that people bear daily. This COVID-19-induced stress, increased vulnerability to mental illnesses¹⁴ or exacerbated pre-existing mental conditions¹⁵. This study adds to the literature on COVID-19 and mental health by exploring the differences in the prevalence of, and examining factors associated with, self-reported changes in mental health, in a socio-economically complex and heterogenous society, Gauteng Province, South Africa.

South Africa is an upper middle-income country with documented non-decreasing historical and contemporary inequalities¹⁶, quadruple burden of diseases¹⁷, high level of persistent unemployment and poverty¹⁶, and high levels of violent crimes¹⁸, which are all risk factors for poor mental health⁶. Against this background, South Africa implemented some of the world’s harshest COVID-19 containment measures¹⁹, such as a ban on tobacco and alcohol sales²⁰, a ban on social gatherings, restrictions on mobility, curfews, social distancing, and closures of “non-essential sectors” ²¹, coupled with complex restrictions of what goods could be sold as they were deemed “essential items” and which were not²², all of which may well have impacted mental health.

The continuous readjustments of COVID-19 containment measures resulted in significant job losses especially for those individuals from poor households, the less-skilled and low-wage workers, workers in non-essential services, and workers in the informal sector²³, leading to substantial losses of livelihoods²⁴, which may, in turn, have effects on mental health. The implementation and adjustment of COVID-19 containment measures were guided by the successive waves of the evolution of the SARS-CoV-2 virus, the number of cases and rate of transmission, and the capacity of health facilities, among other factors²¹. Security forces were deployed to enforce these restrictions, with some of the security sector operations reported to have been violent and to have claimed lives in high-density urban residential areas²⁵. However, even though pre-existing socioeconomic ills such as deprivation, poor public infrastructure, inadequate housing, and overcrowding in urban areas may have increased vulnerability to COVID-19, in some instances they also impeded the enforcement of containment measures²⁶. This means that some people went on with their lives largely uninterrupted as compared to other communities where the containment measures were fully enforced. As a result, the impact of the pandemic on mental health differs between communities.

Gauteng, South Africa’s most populous province²⁷, mirrors South Africa’s national socio-economic ills such as accelerated urbanisation not matched by public infrastructure and service delivery, a high unemployment rate, and high crime rates²⁸. Besides prevalent inequalities, recent statistics show that Gauteng appears in the top three of the nine provinces in terms of different poverty measures in South Africa²⁹. Despite it being the smallest province by land size³⁰, Gauteng houses more than 26% of South Africa’s population^27,30 making it South Africa’s most densely populated province. These features made Gauteng a hotspot for COVID-19 – with the first case of the highly transmissible Omicron variant identified in the province on the 25th of November 2021³¹ and the highest number of COVID-19 cases and deaths in South Africa consistently recorded in the province³². As a result, the mental health effects of the pandemic are likely to be high in Gauteng.

Our study sites, Atteridgeville, Hillbrow, and Melusi, which all fall in the Gauteng Province, present stark socio-economic and infrastructural differences. Two of the three study sites are in the City of Tshwane. Atteridgeville is an old formal township that also includes a large informal area³³. Because this settlement is well-established, people have squeezed into every available space which resulted in a very high density of dwellings, with many backyard dwellings, very few stand-alone shacks, and very little space between dwellings as shacks are joined³³. In comparison, Melusi is a relatively new, fast-growing, and less dense informal settlement with stand-alone shacks³³. Available spaces and informality (such as unplanned road networks) in Melusi likely made it difficult for law enforcement to enforce COVID-19 restrictions (unless they moved on foot, which they chose not to). This means that freedom of movement was relatively higher in Melusi than in Atteridgeville and Hillbrow, which have formal road networks.

In the City of Johannesburg, Hillbrow is a long-standing high-density formal area in the city centre, which was once a ‘white’ zoned area but is now characterised as a deteriorating, high violence, and high crime area, to the extent that some buildings have been ‘captured’ by criminal syndicates^33,34. The predominant feature of this location is multiple-storey buildings, which accommodate multiple households. Some are well-managed buildings, with appropriately managed density. Others are being rack-rented, with multiple households per apartment, and limited, if any, services (like water and sanitation). With a population density above 68 000 people per square kilometre, Hillbrow has the highest local population density in Africa^34,35.

The socio-economic and infrastructural differences across the three sites may be associated with the disparities in the burden of COVID-19 and related measures. Previous research in the three sites by Delport et al.³⁶ showed that higher COVID-19 infection rates were recorded in the crowded formal settlements, Hillbrow and Atteridgeville, than in the less crowded informal settlement, Melusi. As a result, urban form was likely to mediate the effect of COVID-19 on mental health.

Ethics approval and consent to participate

Ethics approval was obtained from the University of the Witwatersrand, the University of Johannesburg, and the University of Pretoria research ethics committees (respective references: 200907; 01-20-2020; 625/2020). Informed consent was obtained from all individual participants included in the study.

Data

We used cross-sectional data from 1332 respondents who participated in GRT-INSPIRED’s face-to-face interviews from March – October 2021.

Figure 1 below shows the dates for data collection and how they relate to COVID-19 variants and subsequent restriction stringency per alert level. The five alert levels were a risk-adjusted approach to the gradual easing of the lockdown (see Appendix A, Table 5), which was guided by the number of cases and rate of transmission, the capacity of health facilities, and the economic and social impact of continued restrictions, among other factors²¹. Fieldwork in the three sites commenced after the Delta wave of COVID-19 had receded and restrictions on movement were lifted. The fieldwork in Hillbrow was completed just before Omicron led to a renewal of lockdown restrictions.

The survey collected information on perceived changes in anxiety, depression, and isolation, and socio-demographic variables of adults aged 18 and above. Respondents were randomly selected from randomly selected households (i.e. a group of people who live in the same dwelling and eat together 3 to 4 times per week) residing within a randomly selected dwelling unit (i.e. physical building). To ensure the sample represented the three study sites, dwelling units were sorted according to their geographic coordinates, followed by a systematic sampling of visiting points within each stratum. A sample of 360 visiting points was selected from each stratum, with two sets of 90 oversample points to be used when the selected unit was empty or refused. To reduce fieldwork bias, the oversample units were randomly selected.

Variables

Our outcome variables (better mental health and worse mental health) were generated from the ordered responses (3-point Likert scale) to questions on anxiety, depression, and isolation. Specifically, respondents were asked if they thought their level of anxiety, feelings of depression, and feelings of isolation from family or friends had gotten better, or worse, or stayed the same as compared to 2019 (pre-pandemic). The response took the value 1 if the respondent perceived a mental health domain to have improved; took the value 2 if a mental health domain remained the same; and took the value 3 if a mental health domain got worse. We then restricted our sample to respondents who responded to all three mental health domain questions and generated two binary mental health variables. The first variable, better mental health, took the value 1 if a respondent reported better in all three mental health domains, and 0 otherwise. The second variable, worse mental health, took the value 1 if a respondent reported worse in all three mental health domains, and 0 otherwise.

The main explanatory variable was residential area, so we included the three study sites in our regression analyses, using Melusi, the stand-alone informal location, as the reference site. We also estimated separate regression results for each of the three study sites. In addition, we controlled for several socio-economic and demographic variables. These were: education level completed, employment status, recipient of social grant, number of diagnosed medical conditions, violent partner, changes in community violence, victim of crime, age group, sex, and household size. We also controlled for the number of assets that a respondent reported, such as television, radio, mobile phone, landline telephone, computer/laptop, access to internet, functioning toilet, access to electricity, and access to clean water.

Statistical analyses

A logistic regression model was used to determine factors associated with mental health domain changes during the pandemic. A logistic regression model regresses a binary dependent variable on relevant explanatory variables to estimate the probability of our dependent variable to be 1. In our case, that is the probability that an individual screened positive for worsening mental health or screened positive for improving mental health. We report the adjusted odds ratios.

In our descriptive analyses, we present percentages for all variables. To represent the target population, all our analyses were weighted using weights that were calculated using the population-estimated average number of persons aged 18 years and above per household. Calibration weights³⁷ were used to adjust design weights to the latest sampling frame of GeoTerra Image, which is based on the 2022 Statistics South Africa mid-year district population estimates and GTI's new development database.

Descriptive analyses

Demographic and socio-economic characteristics

Table 1 shows that, overall, the survey population is almost balanced in terms of sex and predominantly aged 25–64 years. With 98% of the sample being black African, our sample is almost homogeneous in terms of race. A third (33%) of our sample lives alone. In terms of education levels, only 1% of the sample reported having no formal education and the largest share (43.3%) reported having matriculated from high school. About 70% of the population is unemployed, with approximately 38% receiving a government grant, and 31% reported having at least one medical condition. Concerning socioeconomic status, approximately 1 in every 5 respondents in our sample has only between zero and two assets. About 38% of the sample receive a government grant. The share of people who reported being victims of crime was 10%. Almost half (45%) of our sample reported that their partners were violent.

There are, however, differences in the distribution of our variables by site. Melusi recorded the smallest share (2.8%) of people with education above matric, the largest share (58.9%) of people actively looking for jobs but not employed, and more than 50% of households are one-person. The smallest share (5%) of people who were victims of crime and the largest share (52.3%) of people who had violent partners were recorded in Melusi. Melusi had more people with less than two assets and no people with more than six assets. As an informal settlement, Melusi is one location where people live when they first move to the city from rural or peri-urban areas. Mostly on their own, with few assets, and without support from nearby family or friends, residents are actively looking for jobs.

Our analysis shows that Hillbrow had the largest share (67%) of people with matric and more education, the largest proportion (53.4%) of people aged 18 and above who were employed, and the smallest share (13.5%) of people receiving government grants. The largest share (18%) of people who reported being victims of crime and the smallest share (28.4%) of people with violent partners were recorded in Hillbrow. Hillbrow had no people with fewer than three assets, and more people (46%) at the higher end of the socioeconomic status measure (7–9 assets). In relation to comorbidities, Hillbrow had the largest share (18.3%) of people reporting at least two medical conditions. (This may reflect differences in health-seeking behaviour among the sites. For example, Hillbrow as an inner-city site has better access to health care services, and these may result in more people seeking healthcare and subsequently getting diagnosed with their medical conditions.)

Atteridgeville recorded the smallest share (0.1%) of people with fewer than three assets, and more people (5.1% of the sample) aged 65 and above. In relation to employment status, the smallest share (22%) of people were employed, and the largest share (33.4%) of people who were unemployed but not actively looking for jobs were recorded in Atteridgeville. Given the age, and the employment status distribution, Atteridgeville understandably had the largest share (48.2%) of people who received government grants. Also, Atteridgeville is a well-established township where people are likely to have their official documentation, and so can access grants.

Mental health and other changes during COVID variables

Across all three sites, around 60% of respondents reported no changes in perceptions of isolation, anxiety, and depression, whilst 20–24% reported worsening, and around 18% reported improvements (17.5–18.4%), in each of the three domains. The mental health of respondents in Melusi was more likely to remain constant (65–68%), compared to the other sites (45–63%). In Atteridgeville people were more likely to report worsening anxiety and worsening depression, while isolation was more of a problem in Hillbrow (25% compared to 18% in Melusi and 20% in Atteridgeville).

When combining the three mental health domains into ‘better’ or ‘worse’, across all sites, 11% reported improvements in all three domains, and 13% reported a worsening in all domains. In comparison to Atteridgeville (11%) and Melusi (8%), people in Hillbrow (14%) were more likely to report improvements, and less likely to report things getting worse, in all three mental health domains.

In the disaggregated analyses by sex (data not presented here), we found that women were more likely to report both improvements - and worsening in isolation, anxiety, and depression. This means that women dominated the extreme ends of the three mental health domains scale. Men were more likely to report that their isolation, anxiety, and depression levels remained the same as pre-COVID-19 levels.

Across all three sites, around 60% of respondents reported no change in community violence. Respondents in Hillbrow were more likely to have perceived community crime as worsening during COVID-19 (43% compared to 22% in Atteridgeville and 5% in Melusi), matching the higher numbers of people reporting being a victim of crime in Hillbrow.

The proportion of people reporting worries during the pandemic

As shown in Table 2 below, 55–63% of respondents were very or extremely worried about employment, crime, children’s education, meeting basic expenses, and catching COVID-19, whilst 28–34% were not worried or were only a little worried. Overall, being alone was the least of the worries recorded in our sample, with the worry about catching COVID-19 the highest. Compared to Atteridgeville and Hillbrow, people in Melusi were less likely to be very, or extremely, worried about other things except for not having a job. As previously highlighted by the demographic and socio-economic characteristics of our sample, people in Melusi are focused on getting jobs, such that isolation is less of a concern.

Regression results

Regression results by site

In this analysis (presented in Table 3), we restricted our sample to those who responded to all three mental health questions. In Models 1 to 4, we present factors associated with improving mental health during the pandemic. We found that being female, being older than 16–24 years, being employed, being from a household of three, not having other diagnosed medical conditions, perceiving less community violence, and not being a victim of crime were associated with increased odds of improved mental health during the pandemic in our study population (Model 1). Education, socioeconomic status, having a non-violent partner, and residence (site) had no significant effects on improving mental health.

In Melusi (Model 2), being aged between 55 and 64, not having other diagnosed medical conditions, and perceiving less community violence were associated with increased odds of improved mental health during the pandemic. Being aged 55 to 64 and perceiving less community violence were the only two factors significantly associated with increased odds of improved mental health during Hillbrow (Model 4). In Atteridgeville (Model 3), being female, being older than 16–24 years, being employed, being from households of two and three, not having other diagnosed medical conditions, perceiving less community violence, not being a victim of crime, and having a non-violent partner were associated with increased odds of improved mental health during the pandemic.

Models 5 to 8 present factors associated with worsening mental health. We found that being older than 24 years, not receiving any government grant, being from a household of two, having more assets, having more diagnosed medical conditions, perceiving less community violence, having a violent partner, and being from Melusi (as compared to being from Hillbrow) were associated with increased odds of worsening mental health (Model 5). We did not find significant sex, education, employment status, and crime gradients in the odds of worsening mental health.

In regressions by sites, we found that in Melusi receiving a grant had a protective effect, but being female, being older, having a violent partner, and living with a larger number of people increased a person’s vulnerability, as did having a greater number of assets. In Atteridgeville, having a greater number of assets, and living in a large family had a protective effect. However, being older, and being unemployed increased the odds of worsening mental health. In Hillbrow, having a violent partner, unexpectedly, was associated with lower odds of worsening mental health.

Interestingly, results presented in Table 3 also suggest that the factors that make the three self-reported mental health domains seem worse are not the inverse of the factors that make things better. For example, whilst being older was associated with improved mental health in the full study sample (Model 1), it was also associated with worsening mental health (Model 5). Being unemployed and being a victim of crime made respondents less likely to say their mental health got better, but it also did not mean that they would say that their mental got worse. Understandably, grants are associated with people being less likely to say things got worse, but they are not associated with things getting better. Whilst there was a sex profile in the odds of improving mental health in the study population, sex was non-significant in explaining worsening mental health in the study population. We, found, however, that being female increased the odds of both improving mental health and deteriorating mental health in Melusi.

Regression results by sex

Because women’s responses were clustered mainly at the extreme ends of the mental health spectrum, we also wanted (in Table 4) to find if mental health for men and women was explained by the same factors. Regarding reporting improved mental health, among women (Model 1), being older, having a non-violent partner, and living in a larger household increased the odds of reporting better mental health, but having primary education only, being unemployed, having diagnosed medical conditions, and reporting increases in community violence decreased the odds. Men who reported improved mental health were more likely to be aged above 65, less likely to receive a grant, more likely to report less community violence, and more likely to have a greater number of assets (Model 2).

Regarding reporting worse health, women who reported worsening mental health were older, less likely to receive a grant, but more likely to have more assets, more likely to have a violent partner, more likely to be from a bigger household, more likely to have a diagnosed medical condition (Model 3). Men who reported worsening mental health were likely to be 25–64 years old, live in a household of two, and have a greater number of assets (Model 4). Concerning socioeconomic status, a high number of assets increased the odds of worse mental health for both men and women. Except for a modest effect on women’s improved mental health, we did not, in general, find significant associations between education levels and mental health.

Even though COVID-19 is a physical condition, several studies have reported high levels of poor mental health coinciding with lockdowns and experiences of the pandemic. However, most of these studies only focused on the negative effects of the pandemic on mental health. These studies, for example, Kola et al.¹², Nochaiwong et al.³⁸, and Nguse and Wassenaar³⁹ explained how COVID-19 and its associated containment measures increased mental health problems. In this study, we used cross-sectional data from three heterogenous urban locations in South Africa to explore the differences in prevalence of, and the factors explaining self-reported changes in mental health during the pandemic. We used logistic regression models to determine the associations between self-reported changes in mental health and various demographic and socio-economic variables.

Despite all expectations of a strong relationship between the reported elements of a pandemic and a decline in mental health, our analysis also indicated improvements in mental health. Specifically, we found that being older than 24 years, not having other diagnosed medical conditions, and perceiving less community violence were associated with increased odds of improved mental health during the pandemic. Concerning age, Westerhof and Keyes⁴⁰ (in the Netherlands) and Thomas et al.⁴¹ (in San Diego, USA) also found a linear relationship between age and improving mental health. This result may be explained by older people being more secure financially and having survived other challenges which made them more resilient⁴¹; and perhaps through benefit-finding during adversities^42–44, which are all associated with better mental health. Thus, although the pandemic brought about numerous hardships and mental health challenges, people may have also developed some positive experiences that may have helped them to cope. Benefit-finding promotes appreciation of simple things, resilience and adaptation, reflection and reprioritisation, personal growth, digital innovation and connectivity, and community solidarity^42–44, which may improve mental health during the pandemic.

Interestingly, our results also suggest that the factors that are associated with improved mental health are not the inverse of the factors that worsened mental health. For example, and understandably so, we found that social grants were associated with reduced odds of worsening mental health, but they were not associated with mental health getting better. In this vein, Keyes⁴⁵ noted that poor mental health and good health were not extreme ends on a single continuum, but different and yet correlated dimensions that make mental health a complete state. This was also confirmed globally by Iaseiello et al.⁴⁶ and in Africa by Khumalo et al.⁴⁷. In the UK, Kinderman et al.⁴⁸ found that good mental health and poor mental health were differentially affected by different individual and social factors. To this end, a complete state of mental health may only be attained by addressing both the factors that reduce mental health problems and the factors that improve mental health.

The disaggregated analysis by site suggested differential effects of explanatory variables on self-reported changes in mental health. We found that education was not associated with mental health, while socioeconomic status can either be a risk factor in Melusi, or a protective factor in Atteridgeville. It was also only in Melusi where receiving a government grant had a protective effect against worsening mental health. We also found that being from Melusi, the informal location, as compared to being from Hillbrow, the inner-city location, was associated with increased odds of worsening mental health during the pandemic. These differences in the effects of explanatory variables on mental health can be explained by the differences in the characteristics of the study sites. Differences in urban form translate to differences in risk and protective factors of mental health^26,49. The effects of COVID-19 on mental health can be mediated through urban form and socio-economic factors such as social networks, among other city characteristics^26,49. For example, people in Melusi have smaller social networks from which they could draw support during COVID-19, as they were all more likely to be poorer, living on their own, and more likely to be unemployed and looking for work. On the other hand, Hillbrow is dense, with bigger households, and more people likely to be employed from whom an individual various forms of support such as social, financial, and food support during the pandemic.

We also found sex differences in the factors explaining mental health. For example, whilst the number of diagnosed medical conditions and being a victim of crime were not significantly associated with the odds of men’s improved mental health, they were negatively associated with the odds of improving mental health for women. Because of the stigma around mental health, men are less likely to report internalising disorders such as anxiety and depression, but more likely to present externalising disorders like substance abuse and violence^50–52. More so, if we consider men as providers, those men who qualify for grants are vulnerable to economic pressure and the pressure of not being able to provide so they are less likely to have good mental health. On the other hand, receiving grants can protect vulnerable women against the stresses of not being able to meet basic needs.

We found that higher socioeconomic status is associated with worse mental health for both men and women in the three urban areas. Despite us using a different measure of socioeconomic status (total number of assets), this result is consistent with Oyenubi and Kollamparambil⁵³ who also found a disproportionately higher mental health burden among the affluent (measured using per capita household income) in South Africa using a large nationally representative sample. A possible explanation is that the lowest-ranked in terms of socioeconomic status are likely to have experienced more hardships than their counterparts, which might have made them more resilient – a process known as “steeling”^54,55, or stress innoculation⁵⁶.

Our study is not without limitations. We used self-reported changes in isolation, anxiety, and depressive symptoms in our analysis. Whilst these self-reported mental health domains can give a good understanding of the COVID-19-associated changes in mental health in South Africa, we cannot rule out biased results emanating from unobserved heterogeneity. Reporting internalising disorders also entails issues such as stigma in mental health and health-seeking behaviour. This may have resulted in underreporting or overreporting of mental health problems by people of different socioeconomic classes. However, analyses of survey data that measure self-reported health can rarely be spared such criticism. The strengths of the present study emanate from the breadth of the analysis undertaken.

In addition to demographic and socioeconomic risk factors, differences in living environment explain the disparities in mental health by residence. This study provided both factors that can be targeted to prevent poor mental health and to sustain good mental health. Our study, therefore, reiterates the urgent need for investments in public infrastructure and services that may have protective effects against poor mental health. To achieve national mental health targets, policymakers should address both the factors that reduce mental health problems and the factors that improve mental health. The policies should also be sex sensitive.

Contributors

All authors conceived and planned the paper. TG cleaned and analysed the data and wrote the article, with JG, IM, DE, and RD reviewing drafts. All authors reviewed the final draft of this article.

Funding information

Partial financial support was received from:

Department of Science and Innovation / Medical Research Council South African Population Research Infrastructure Network (SAPRIN)
Gauteng Research Triangle Initiative for the Study of Population, Infrastructure and Regional Economic Development (GRT-INSPIRED). DE and RD appointed within this funding body were involved in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Wits RHI, University of the Witwatersrand
University of the Witwatersrand
University of Johannesburg
University of Pretoria
TG and JG’s contributions were supported by the South African National Research Foundation through the SARChI Programme for Health Systems and Policy Research at the Centre for Health Policy at the School of Public Health, University of the Witwatersrand.

National Research Foundation Grant (RD; N01808)

Competing interests

We declare that we have no personal or financial relationships that may have inappropriately influenced us in writing this article.

Ethics approval and consent to participate

Data availability

Data analysed during this study is available from the corresponding author upon reasonable request.

Acknowledgements

The contribution of fieldworkers from the University of the Witwatersrand, the University of Johannesburg, and the University of Pretoria is gratefully acknowledged. We thank Dr Ariane Neethling for computing the weights used in the analysis. GeoTerra Image (PTY) Ltd are thanked for availing their data free of charge.

World Health Organization. Coronavirus disease (COVID-19) pandemic [Internet]. 2023 [cited 2023 Aug 21]. Available from: https://www.who.int/europe/emergencies/situations/covid-19
World Health Organization. Considerations for implementing and adjusting public health and social measures in the context of COVID-19 [Internet]. COVID-19: Critical preparedness, readiness and response. 2023 [cited 2023 Aug 17]. Available from: https://www.who.int/publications/i/item/who-2019-ncov-adjusting-ph-measures-2023.1
Sun Y, Wu Y, Fan S, Dal Santo T, Li L, Jiang X, et al. Comparison of mental health symptoms before and during the covid-19 pandemic: Evidence from a systematic review and meta-analysis of 134 cohorts. BMJ. 2023;380:e074224.
Word Health Organization. COVID-19 pandemic triggers 25% increase in prevalence of anxiety and depression worldwide [Internet]. 2022 [cited 2023 Jul 25]. Available from: https://www.who.int/news/item/02-03-2022-covid-19-pandemic-triggers-25-increase-in-prevalence-of-anxiety-and-depression-worldwide
U.S. Department of Health and Human Services. Mental health: A Report of the Surgeon General [Internet]. Rockville, MD: U.S. Department of Health and Human Services, Substance Abuse and Mental Health Services Administration, Center for Mental Health Services, National Institutes of Health, National Institute of Mental Health; 1999 [cited 2023 Aug 17]. Available from: https://profiles.nlm.nih.gov/spotlight/nn/catalog/nlm:nlmuid-101584932X120-doc
World Health Organization, Calouste Gulbenkian Foundation. Social determinants of mental health [Internet]. Geneva: World Health Organization; 2014 [cited 2023 Aug 17]. Available from: https://www.who.int/publications/i/item/9789241506809
Caracci G. General concepts of the relationship between urban areas and mental health. Curr Opin Psychiatry. 2008;21:385–90.
Yang W, Yang R, Li X. A canonical correlation analysis study on the association between neighborhood green space and residents’ mental health. Journal of Urban Health. 2023;100:696–710.
Halbreich U. Impact of urbanization on mental health and well being. Curr Opin Psychiatry. 2023;36(3):200–5.
Mrazek PJ, Haggerty RJ, editors. Reducing risks for mental disorders: Frontiers for preventive intervention research. Washington, D.C: National Academy Press; 1994. 178 p.
Kumar M, Kumar P. Impact of pandemic on mental health in lower- and middle-income countries (LMICs). Global Mental Health. 2020;7.
Kola L, Kohrt BA, Hanlon C, Naslund JA, Sikander S, Balaji M, et al. COVID-19 mental health impact and responses in low-income and middle-income countries: Reimagining global mental health. Lancet Psychiatry. 2021;8(6):535–50.
Bhattacharjee A, Ghosh T. COVID-19 pandemic and stress: Coping with the new normal. J Prev Health Promot. 2022;3(1):30–52.
Lahav Y. Psychological distress related to COVID-19 – The contribution of continuous traumatic stress. J Affect Disord. 2020;277.
Rohde C, Jefsen OH, Norremark B, Danielsen AA, Ostergaard SD. Psychiatric symptoms related to the COVID-19 pandemic. Acta Neuropsychiatr. 2020;32(5):274–6.
World Bank. Overcoming poverty and inequality in South Africa: An assessment of drivers, constraints and opportunities. Washington, D.C: World Bank; 2018.
National Department of Health. Strategic Plan 2020/21–2024/25. Pretoria: National Department of Health; 2020.
Statistics South Africa. Victims of Crime Survey, 2018/19. Statistics South Africa. Pretoria: Statistics South Africa; 2019.
Greyling T, Rossouw S, Adhikari T. The good, the bad and the ugly of lockdowns during Covid-19. PLoS One. 2021;16(1):e0245546.
Moultrie TA, Dorrington RE, Laubscher R, Groenewald P, Parry CDH, Matzopoulos R, et al. Unnatural deaths, alcohol bans and curfews: Evidence from a quasi-natural experiment during COVID-19. South African Medical Journal. 2021;111(9):834–7.
South African Government. COVID-19/Novel Coronavirus [Internet]. 2023 [cited 2022 Jul 1]. Available from: https://www.gov.za/Coronavirus
National Department of Health. Essential services during the lockdown period [Internet]. 2020 [cited 2023 Oct 10]. Available from: https://sacoronavirus.co.za/2020/03/20/essential-services-during-the-lockdown-period/
Köhler T, Bhorat H, Hill R, Stanwix B. Lockdown stringency and employment formality: Evidence from the COVID-19 pandemic in South Africa. J Labour Mark Res. 2023;57(1).
Schotte S, Zizzamia R. The livelihood impacts of COVID-19 in urban South Africa: A view from below. Soc Indic Res. 2023;165(1):1–30.
Trippe K. Pandemic policing: South Africa’s most vulnerable face a sharp increase in police-related brutality [Internet]. 2020 [cited 2023 Jul 20]. Available from: https://www.atlanticcouncil.org/blogs/africasource/pandemic-policing-south-africas-most-vulnerable-face-a-sharp-increase-in-police-related-brutality/
Okem AE, Makanishe TB, Myeni SL, Roberts DC, Zungu S. A scoping review of COVID-19 in the context of informal settlements. Urbanisation. 2022;7(2):147–62.
Statistics South Africa. Mid-year population estimates 2022 [Internet]. Pretoria; 2022 Jul [cited 2023 Aug 7]. Available from: https://www.statssa.gov.za/?page_id=1854&PPN=P0302&SCH=73305
Gauteng Provincial Government. Gauteng Province Socio-economic Review and Outlook (SERO) 2020. Marshalltown: Gauteng Provincial Government; 2020.
Statistics South Africa. Subjective poverty in South Africa: Findings from General Household Survey, 2019 [Internet]. Statistics South Africa Published by Statistics South Africa. Pretoria; 2022. Available from: www.statssa.gov.za
Government Communications (GCIS). South Africa Yearbook 2021/22 [Internet]. 29th ed. Mokoena M, Honwane M, Mabena L, editors. Pretoria: Government Communications (GCIS); 2022. Available from: www.gcis.gov.za
Madhi SA, Kwatra G, Myers JE, Jassat W, Dhar N, Mukendi CK, et al. Population Immunity and Covid-19 Severity with Omicron Variant in South Africa. New England Journal of Medicine. 2022;386(14):1314–26.
National Department of Health. COVID-19 daily updates and cases. COVID-19 South African Online Portal. 2021.
SAPRIN. Gauteng Research Triangle Initiative for the Study of Population, Infrastructure and Regional Economic Development [Internet]. HDSS Nodes/GRT-INSPIRED. 2023 [cited 2024 Jan 29]. Available from: https://saprin.mrc.ac.za/grtinspired.html
Stadler J, Dugmore C. ‘Honey, Milk and Bile’: A social history of Hillbrow, 1894-2016. BMC Public Health. 2017;17(Suppl 3):444.
Everatt D, Lynge H, Abrahams C. Neighbourhood characteristics and inequality in the City of Johannesburg: A preliminary analysis [Internet]. GCRF Centre for Sustainable, Healthy and Learning Cities and Neighbourhoods (SHLC). 2023. Available from: https://www.centreforsustainablecities.ac.uk/research/city-report-neighbourhood-characteristics-and-inequality-in-the-city-of-johannesburg/
Delport R, Everatt D, Palanee T, Hugo J, Barnard TG. Urban form and health status in Gauteng, South Africa. In: Advances in Science, Technology and Innovation. Springer Nature; 2023. p. 147–50.
Neethling A, Galpin JS. Weighting of household survey data: A comparison of various calibration, integrated and cosmetic estimators. South African Statistical Journal. 2006;40:123–50.
Nochaiwong S, Ruengorn C, Thavorn K, Hutton B, Awiphan R, Phosuya C, et al. Global prevalence of mental health issues among the general population during the coronavirus disease-2019 pandemic: A systematic review and meta-analysis. Sci Rep. 2021;11(1).
Nguse S, Wassenaar D. Mental health and COVID-19 in South Africa. South African Journal of Psychology. 2021;51(2):304–13.
Westerhof GJ, Keyes CLM. Mental illness and mental health: The two continua model across the lifespan. J Adult Dev. 2010;17(2):110–9.
Thomas ML, Kaufmann CN, Palmer BW, Depp CA, Martin AS, Glorioso DK, et al. Paradoxical trend for improvement in mental health with aging: A Community-Based study of 1,546 adults aged 21-100 years. Journal of Clinical Psychiatry. 2016;77(8):e1019–25.
Tennen H, Affleck G. Benefit-finding and benefit-reminding. In: Snyder CR, Lopez SJ, editors. Handbook of positive psychology. New York: Oxford University Press; 2002. p. 584–97.
Cassidy T, Doyle I. Benefit finding and well-being in older adults: The utility of the General Benefit Finding Scale. Int J Psychol Res (Medellin). 2018;11(2):8–18.
Moore JB, Rubin KCR, Heaney CA. Benefit finding and well-being over the course of the COVID-19 pandemic. PLoS One. 2023;18(7):e0288332.
Keyes CLM. Mental illness and/or mental health? Investigating axioms of the complete state model of health. J Consult Clin Psychol. 2005;73(3):539–48.
Iasiello M, van Agteren J, Cochrane EM. Mental Health and/or Mental Illness: A Scoping Review of the Evidence and Implications of the Dual-Continua Model of Mental Health. Evidence Base. 2020;2020(1):1–45.
Khumalo IP, Appiah R, Wilson Fadiji A. Measuring positive mental health and depression in Africa: A variable-based and person-centred analysis of the Dual-Continua Model. Front Psychol. 2022;13.
Kinderman P, Tai S, Pontin E, Schwannauer M, Jarman I, Lisboa P. Causal and mediating factors for anxiety, depression and well-being. The British Journal of Psychiatry [Internet]. 2015;206:456–60. Available from: https://doi.org/10.1192/bjp.bp.114.147553
Jevtic M, Matkovic V, Kusturica MP, Bouland C. Urban places and mental health challenges (lessons learned from Covid-19 crisis). Eur J Public Health. 2022;32(Supplement 3).
McKenzie SK, Oliffe JL, Black A, Collings S. Men’s experiences of mental illness stigma across the lifespan: A scoping review. Am J Mens Health. 2022;16(1).
Rosenfield S, Mouzon D. Gender and mental health. In: Aneshensel CS, Phelan JC, Bierman A, editors. Handbook of the Sociology of Mental Health. Dordrecht: Springer; 2012. p. 277–96.
Hill TD, Needham BL. Rethinking gender and mental health: A critical analysis of three propositions. Soc Sci Med. 2013;92:83–91.
Oyenubi A, Kollamparambil U. COVID-19 and depressive symptoms in South Africa [Internet]. 2020. Available from: https://cramsurvey.org/reports/
Rutter M. Resilience in the face of adversity. British Journal of Psychiatry. 1985;147:598–611.
Carver CS. Resilience and thriving: Issues, models, and linkages. Journal of Social Issues. 1998;54(2):245–66.
McGee SL, Höltge J, Maercker A, Thoma M V. Sense of coherence and stress-related resilience: Investigating the mediating and moderating mechanisms in the development of resilience following stress or adversity. Front Psychiatry. 2018;9:378.

Table 1

Sample characteristics (weighted analysis)
Variables	Melusi	Atteridgeville	Hillbrow	Total
	(N = 298)	(N = 620)	(N = 414)	(N = 1332)
Demographic and socio-economic characteristics
sex
female	51.1%	51.2%	50.7%	51.1%
male	48.9%	48.8%	49.3%	48.9%
age categories
16–24	16.5%	17.1%	16.7%	16.8%
25–34	33.3%	30.0%	36.5%	32.1%
35–44	23.6%	24.5%	26.7%	24.6%
45–54	11.4%	12.7%	11.7%	12.1%
55–64	13.8%	10.6%	5.8%	10.9%
>=65	1.5%	5.1%	2.6%	3.5%
race
African	98.5%	98.3%	99.0%	98.5%
non-African	1.5%	1.7%	1.0%	1.5%
education level
none	2.9%	0.1%	0.4%	1.1%
primary	11.8%	2.9%	6.1%	6.4%
some secondary	46.9%	33.2%	26.6%	36.6%
matric	35.6%	46.4%	49.3%	43.3%
more	2.8%	17.5%	17.6%	12.7%
employment status of those aged 18 years and above
employed	30.7%	21.5%	53.4%	29.9%
unemployed, not searching	10.3%	33.4%	14.4%	22.6%
unemployed, searching	58.9%	45.1%	32.2%	47.6%
receive any government grant
no	66.0%	51.8%	86.5%	62.3%
yes	34.0%	48.2%	13.5%	37.7%
household size
1	53.7%	23.7%	18.1%	32.7%
2	20.3%	18.1%	16.2%	18.5%
3	5.4%	15.9%	17.4%	12.7%
4	2.3%	10.7%	14.4%	8.5%
>=5	18.2%	31.6%	33.9%	27.5%
socioeconomic status (total number of assets)
0–2	65.3%	0.8%	0.0%	22.2%
3–4	30.8%	9.9%	4.3%	16.0%
5–6	3.9%	52.9%	49.9%	36.1%
7–9	0.0%	36.3%	45.8%	25.8%
number of medical conditions
=0	74.9%	74.8%	38.4%	68.8%
=1	19.3%	18.8%	43.3%	23.0%
=2	3.7%	5.4%	14.9%	6.4%
>=3	2.0%	1.0%	3.4%	1.7%
victim of crime
no	94.9%	89.2%	82.5%	90.0%
yes	5.1%	10.8%	17.5%	10.0%
partner is violent
no	47.7%	54.2%	71.6%	54.9%
yes	52.3%	45.8%	28.4%	45.1%
Mental health and other changes during COVID variables
changes in community violence
less	28.7%	14.4%	18.7%	19.8%
same	66.2%	63.3%	38.5%	60.2%
more	5.1%	22.3%	42.8%	20.0%
self-reported changes in isolation
Better	14.2%	16.4%	27.7%	17.5%
Same	68.1%	63.8%	47.1%	62.5%
Worse	17.7%	19.8%	25.2%	20.0%
self-reported changes in anxiety
Better	14.7%	14.7%	24.5%	18.4%
Same	65.0%	51.8%	45.4%	57.7%
Worse	20.3%	33.5%	30.1%	23.9%
self-reported changes in depressive symptoms
Better	14.9%	18.4%	25.6%	18.4%
Same	67.3%	54.0%	49.7%	57.7%
Worse	17.8%	27.6%	24.7%	23.9%
better in all mental health domains
no	92.2%	88.8%	85.6%	89.4%
yes	7.8%	11.2%	14.4%	10.6%
worse in all mental health domains
no	87.7%	84.5%	90.5%	86.6%
yes	12.3%	15.5%	9.5%	13.4%
better in at least two mental health domains
no	87.9%	85.8%	77.2%	85.1%
yes	12.1%	14.2%	22.8%	14.9%
worse in at least two mental health domains
no	85.0%	71.8%	74.0%	76.6%
yes	15.0%	28.2%	26.0%	23.4%

Note: Assets include television, radio, mobile phone, landline telephone, computer/laptop, access to internet, good toilet, access to electricity, and access to clean water.

Table 2

Proportion of people reporting worries during the pandemic (weighted analysis)
	Melusi	Atteridgeville	Hillbrow	Total
Variable	(N = 298)	(N = 620)	(N = 414)	(N = 1332)
worry about having a job
very or extremely worried	57.0%	57.5%	63.3%	58.3%
neutral	12.1%	4.5%	10.6%	8.0%
not or a little worried	30.9%	37.9%	26.1%	33.6%
worry about being alone
very or extremely worried	28.7%	47.0%	35.7%	39.0%
neutral	25.2%	2.6%	11.8%	11.7%
not or a little worried	46.1%	50.3%	52.4%	49.3%
worry about crime
very or extremely worried	36.4%	66.2%	64.1%	55.9%
neutral	27.4%	3.6%	5.6%	11.9%
not or a little worried	36.2%	30.2%	30.3%	32.2%
worry about children's education
very or extremely worried	40.6%	60.7%	65.7%	54.8%
neutral	25.8%	5.9%	13.5%	13.8%
not or a little worried	33.7%	33.4%	20.8%	31.4%
worry about basic expenses
very or extremely worried	40.9%	68.1%	65.9%	58.7%
neutral	18.6%	2.5%	7.8%	8.7%
not or a little worried	40.5%	29.4%	26.4%	32.6%
worry about catching COVID-19
very or extremely worried	50.4%	68.6%	69.3%	62.7%
neutral	20.9%	2.5%	8.6%	9.6%
not or a little worried	28.7%	28.9%	22.1%	27.7%

Table 3

Logistic regression models (adjusted ORs)
Variables	Better in all 3 mental health domains				Worse in all 3 mental health domains
	All (1)	Mel (2)	Att (3)	Hil (4)	All (5)	Mel (6)	Att (7)	Hil (8)
male	0.57**	0.44	0.52*	1.24	0.75	0.24***	1.29	0.39*
	(0.36–0.91)	(0.14–1.34)	(0.26–1.03)	(0.60–2.54)	(0.45–1.26)	(0.09–0.63)	(0.49–3.41)	(0.14–1.06)
age group
16–24	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
25–34	2.97**	7.34	4.98*	0.87	1.66	1.53	2.37	1.01
	(1.28–6.87)	(0.64–84.48)	(0.94–26.26)	(0.31–2.43)	(0.61–4.51)	(0.46–5.13)	(0.22–25.34)	(0.35–2.89)
35–44	1.46	2.96	1.68	1.33	3.61***	3.14*	10.45*	0.71
	(0.60–3.56)	(0.22–39.08)	(0.27–10.54)	(0.44–4.06)	(1.38–9.43)	(0.90–10.95)	(0.94–116.33)	(0.23–2.12)
45–54	1.17	0.92	1.24	1.23	2.62*	6.00**	4.00	0.87
	(0.43–3.20)	(0.01–60.69)	(0.19–8.25)	(0.30–5.10)	(0.93–7.40)	(1.23–29.22)	(0.37–42.87)	(0.23–3.31)
55–64	6.21***	25.09**	9.25**	4.59*	1.59	2.38	2.54	0.50
	(2.29–16.81)	(1.67–377.92)	(1.56–54.89)	(0.94–22.52)	(0.53–4.80)	(0.27–20.81)	(0.32–19.91)	(0.08–3.23)
>=65	7.96***	–	15.66***	5.99	3.19**	–	4.32	7.02
	(2.14–29.66)		(2.67–91.85)	(0.32–110.87)	(1.00–10.12)		(0.52–36.15)	(0.49–100.00)
education level
none	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
primary	0.31	0.22	1.05	0.10	1.00	1.20	0.80	1.27
	(0.06–1.70)	(0.01–4.18)	(0.01–112.14)	(0.00–3.51)	(0.11–9.45)	(0.11–13.61)	(0.16–3.94)	(0.25–6.41)
some secondary	0.32	0.19	0.31	0.13	1.59	2.62	2.14	1.05
	(0.07–1.45)	(0.01–4.60)	(0.00–28.66)	(0.00–4.02)	(0.19–13.39)	(0.25–27.12)	(0.75–6.11)	(0.31–3.61)
matric	0.37	0.17	0.42	0.15	1.04	2.58	0.81	0.85
	(0.08–1.60)	(0.01–4.19)	(0.00–40.11)	(0.00–4.44)	(0.11–9.40)	(0.24–27.34)	(0.27–2.43)	(0.29–2.54)
more	0.34	0.86	0.27	0.12	0.88	0.22	–	–
	(0.07–1.76)	(0.04–19.21)	(0.00–27.18)	(0.00–4.39)	(0.09–8.33)	(0.00–41.99)
employment status
employed	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
unemployed, not searching	0.40**	0.09	0.23***	1.85	1.06	0.45	2.58*	0.39
	(0.20–0.82)	(0.00–2.51)	(0.08–0.62)	(0.68–5.08)	(0.50–2.25)	(0.03–6.03)	(0.90–7.40)	(0.07–2.17)
unemployed, searching	0.52**	0.62	0.26***	0.69	1.35	0.68	3.20**	0.81
	(0.30–0.92)	(0.22–1.76)	(0.10–0.69)	(0.31–1.56)	(0.80–2.30)	(0.26–1.78)	(1.25–8.19)	(0.37–1.76)
receive any government grant	0.82	1.46	0.65	1.06	0.54**	0.28**	0.71	0.87
	(0.48–1.42)	(0.53–4.06)	(0.30–1.45)	(0.39–2.90)	(0.32–0.90)	(0.09–0.87)	(0.34–1.50)	(0.29–2.62)
household size
1	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
2	1.27	0.38	3.06**	1.18	1.76*	4.13***	1.32	1.03
	(0.68–2.35)	(0.10–1.44)	(1.02–9.24)	(0.38–3.64)	(0.93–3.35)	(1.42–11.98)	(0.53–3.26)	(0.20–5.26)
3	1.83*	2.05	4.38***	0.51	0.99	6.92**	0.28	2.84
	(0.91–3.69)	(0.40–10.55)	(1.43–13.42)	(0.13–2.00)	(0.44–2.21)	(1.53–31.35)	(0.06–1.27)	(0.72–11.18)
4	1.25	2.56	1.83	0.46	0.47	1.74	0.21*	0.85
	(0.49–3.16)	(0.53–12.52)	(0.44–7.70)	(0.13–1.66)	(0.17–1.29)	(0.07–40.25)	(0.03–1.24)	(0.17–4.15)
>=5	1.52	1.07	2.54	1.25	1.19	8.23***	0.62	1.44
	(0.84–2.76)	(0.27–4.30)	(0.81–7.91)	(0.47–3.36)	(0.64–2.24)	(2.53–26.77)	(0.22–1.76)	(0.37–5.56)
socioeconomic status (number of assets)
0–2	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
3–4	1.05	1.02	1.34	–	5.91***	5.73***	0.81	–
	(0.43–2.59)	(0.31–3.29)	(0.05–37.88)		(2.77–12.59)	(2.29–14.37)	(0.28–2.41)
5–6	2.47	2.79	2.32	1.17	4.06**	44.99***	0.34***	1.18
	(0.71–8.59)	(0.48–16.28)	(0.12–43.16)	(0.16–8.37)	(1.21–13.58)	(7.24–279.55)	(0.16–0.72)	(0.30–4.58)
7–9	2.15	–	1.84	1.03	7.43***	–	–	0.59
	(0.57–8.04)		(0.10–34.97)	(0.14–7.62)	(2.05–26.99)			(0.14–2.41)
number of diagnosed medical conditions	0.63**	0.27*	0.44**	0.91	1.46**	1.84	1.14	1.42*
	(0.44–0.91)	(0.07–1.12)	(0.23–0.82)	(0.57–1.44)	(1.07–2.00)	(0.85–4.00)	(0.71–1.83)	(0.95–2.14)
changes in community violence
less	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
same	0.24***	0.24***	0.14***	0.66	0.52**	0.38**	0.37	0.38
	(0.14–0.41)	(0.09–0.64)	(0.05–0.37)	(0.28–1.60)	(0.29–0.92)	(0.16–0.91)	(0.11–1.23)	(0.09–1.65)
more	0.24***	0.61	0.13***	0.48*	1.11	1.13	0.95	2.44
	(0.12–0.48)	(0.09–4.33)	(0.05–0.37)	(0.20–1.14)	(0.55–2.23)	(0.23–5.49)	(0.23–3.94)	(0.77–7.71)
victim of crime	0.43**	–	0.25*	1.17	1.34	1.03	1.01	1.42
	(0.19–0.98)		(0.05–1.23)	(0.39–3.52)	(0.59–3.06)	(0.18–6.05)	(0.22–4.54)	(0.52–3.88)
partner is violent	0.68	1.36	0.33***	1.44	1.74**	0.76	3.35***	0.29**
	(0.40–1.13)	(0.51–3.62)	(0.15–0.76)	(0.67–3.10)	(1.13–2.69)	(0.35–1.65)	(1.61–6.97)	(0.11–0.79)
residence
Melusi	Ref				Ref
Atteridgeville	0.99				0.64
	(0.31–3.16)				(0.22–1.91)
Hillbrow	1.10				0.22**
	(0.32–3.71)				(0.07–0.71)
Constant	0.56	0.41	0.80	1.11	0.02***	0.02**	0.04*	0.10*
	(0.09–3.30)	(0.01–32.72)	(0.00–219.75)	(0.02–60.84)	(0.00–0.27)	(0.00–0.41)	(0.00–1.36)	(0.01–1.21)
Observations	1,283	274	591	403	1,283	288	587	401

Note: 95% Confidence Interval in parentheses; *** p<0.01, ** p<0.05, * p<0.1; Ref stands for the reference category; and ORs are adjusted odds ratios. Att – Atteridgeville; Hil – Hillbrow; and Mel – Melusi. Better mental health = 1 if a respondent reports better in all three mental health domains, and 0 otherwise. Worse mental health = 1 if a respondent reports worse in all three mental health domains, and 0 otherwise

Table 4

Logistic regression results (adjusted ORs) by sex
	Better in all 3 mental health domains		Better in all 3 mental health domains
Variables	Female (1)	Male (2)	Female (3)	Male (4)
age group
16–24	Ref	Ref	Ref	Ref
25–34	4.16***	1.18	1.29	5.65*
	(1.46–11.85)	(0.28–5.01)	(0.41–4.05)	(0.88–36.19)
35–44	2.77*	0.29	3.06**	16.90***
	(0.93–8.30)	(0.05–1.63)	(1.01–9.26)	(2.81–101.61)
45–54	1.89	0.28	2.86*	5.84*
	(0.49–7.28)	(0.04–1.89)	(0.92–8.88)	(0.84–40.41)
55–64	11.50***	2.36	0.69	7.37**
	(3.60–36.72)	(0.47–11.87)	(0.16–3.00)	(1.06–51.04)
>=65	7.33***	8.01**	5.19***	4.48
	(2.00–26.89)	(1.14–56.36)	(1.50–18.02)	(0.55–36.31)
education level
none	Ref	Ref	Ref	Ref
primary	0.11*	1.56	1.48	0.50
	(0.01–1.03)	(0.14–17.86)	(0.14–16.03)	(0.04–5.63)
some secondary	0.32	0.65	1.25	1.48
	(0.05–2.07)	(0.07–6.47)	(0.14–11.21)	(0.31–7.02)
matric	0.52	0.44	0.64	1.51
	(0.08–3.23)	(0.04–4.89)	(0.07–5.81)	(0.44–5.20)
more	0.45	0.42	0.67	–
	(0.06–3.34)	(0.03–6.72)	(0.07–6.66)
employment status
employed	Ref	Ref	Ref	Ref
unemployed, not searching	0.45*	0.36	0.87	1.43
	(0.20–1.02)	(0.09–1.36)	(0.33–2.26)	(0.42–4.91)
unemployed, searching	0.40***	0.80	1.45	1.48
	(0.21–0.76)	(0.26–2.45)	(0.77–2.73)	(0.58–3.78)
receive any government grant	1.17	0.28**	0.41***	0.75
	(0.64–2.14)	(0.10–0.80)	(0.22–0.77)	(0.33–1.68)
household size
1	Ref	Ref	Ref	Ref
2	1.27	1.19	1.38	2.72*
	(0.57–2.84)	(0.47–3.03)	(0.63–3.03)	(0.97–7.57)
3	2.21*	1.27	1.86	0.38
	(0.93–5.22)	(0.39–4.22)	(0.69–5.06)	(0.09–1.59)
4	1.96	0.44	0.30*	0.72
	(0.72–5.35)	(0.09–2.13)	(0.08–1.18)	(0.16–3.14)
>=5	2.12*	0.86	1.01	1.48
	(0.99–4.54)	(0.31–2.40)	(0.46–2.21)	(0.54–4.04)
socioeconomic status (number of assets)
0–2	Ref	Ref	Ref	Ref
3–4	0.47	4.41	8.52***	5.17**
	(0.15–1.50)	(0.62–31.59)	(3.41–21.29)	(1.28–20.83)
5–6	0.81	12.79*	2.30	10.68**
	(0.21–3.12)	(0.67–244.43)	(0.71–7.48)	(1.09–104.76)
7–9	0.76	9.23	8.01***	10.46*
	(0.19–3.09)	(0.43–197.81)	(2.17–29.60)	(0.93–117.96)
number of diagnosed medical conditions	0.50***	0.85	1.40**	1.85**
	(0.35–0.71)	(0.47–1.55)	(1.01–1.93)	(1.11–3.09)
changes in community violence
less	Ref	Ref	Ref	Ref
same	0.28***	0.14***	0.44**	0.67
	(0.15–0.52)	(0.05–0.38)	(0.21–0.94)	(0.30–1.50)
more	0.25***	0.19***	1.06	1.51
	(0.12–0.53)	(0.07–0.53)	(0.46–2.44)	(0.51–4.53)
victim of crime	0.28**	0.65	0.57	2.29
	(0.09–0.91)	(0.17–2.44)	(0.21–1.54)	(0.74–7.16)
partner is violent	0.45***	1.10	1.97**	1.64
	(0.25–0.80)	(0.43–2.78)	(1.16–3.35)	(0.75–3.59)
residence
Melusi	Ref	Ref	Ref	Ref
Atteridgeville	1.15	1.30	0.60	0.61
	(0.34–3.83)	(0.10–17.10)	(0.22–1.66)	(0.07–5.72)
Hillbrow	1.19	1.40	0.31**	0.09**
	(0.30–4.73)	(0.13–15.04)	(0.10–0.96)	(0.01–0.85)
Constant	0.62	0.14	0.05**	0.00***
	(0.07–5.74)	(0.00–4.98)	(0.00–0.65)	(0.00–0.03)
Observations	853	430	853	425
95% Confidence Interval (Eform) in parentheses
* p < 0.01, p < 0.05, * p < 0.1

Note: Ref stands for the reference category, and ORs are adjusted odds ratios. In Models 1-3, our dependent variable (better mental health) is binary and is = 1 if a respondent reports better in all three mental health domains, and 0 otherwise. In Models 4-6, our dependent variable (worse mental health) is binary and is = 1 if a respondent reports worse in all three mental health domains, and 0 otherwise.

AppendixA.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Variations in self-reported mental health in low-income urban areas during COVID-19: Insights from South Africa

Status:

Version 1

Abstract

Figures

Introduction

Background

Methods

Results

Discussion

Conclusion

Declarations

References

Tables

Supplementary Files

Status:

Version 1