The impact of adult mortality on household dissolution and migration in rural South Africa

Abstract

Objective: 

To investigate the effect of adult death on household dissolution and migration.

Design: 

Demographic surveillance of the population in a rural area of northern KwaZulu Natal, South Africa.

Methods: 

Data on households resident in the surveillance area on 1 January 2000 were used to examine the effect of adult mortality and household risk factors on household dissolution and mobility between January 2000 and October 2002. Cox regression models were used to assess the risk of household dissolution and migration, controlling for multiple risk factors including causes of death, household composition and household assets.

Results: 

By October 2002, 238 households (2%) had dissolved and 874 (8%) migrated out of the area; 21% (2179) of all households had at least one adult death (18 years and older). Households where one or more adult members died during the follow-up period were four times more likely to dissolve, after controlling for household and community level risk factors [4.3; 95% confidence interval, (CI), 3.3–5.7]. The risk of dissolution was significantly higher in households with multiple deaths (2.3; 95% CI, 1.3–4.3). There were no significant differential risks associated with cause of death, age or sex of the deceased. Adult mortality in the household was not associated with migration.

Conclusions: 

Poorer households, as measured by asset ownership, and households trying to cope with adult deaths are vulnerable to dissolution. The dramatic increase in adult mortality attributable to AIDS will increase the number of households that do not survive as a functional and cohesive social group.

Introduction

South Africa has experienced a rapidly growing HIV epidemic, resulting in substantial increases in adult mortality, since the mid-1990s [¹]. The death of an adult may have many short- and long-term consequences for households and their surviving members, including reduced economic status [^2,3], increased household mobility [⁴] and household dissolution [⁵]. There have been very few empirical studies of the sociodemographic impact of AIDS on households and their communities in sub-Saharan Africa. One reason for the paucity of such studies is the scarcity of longitudinal household data with detailed information about the level and causes of mortality, household composition and survival, changes in socioeconomic status and migration [^6–8].

This study uses demographic surveillance system data to describe the dissolution and migration of 10 612 households over 2½ year period between 2000 and 2002 in a rural area of KwaZulu Natal and to examine the association between the experience of adult deaths and household dissolution and migration.

Methods

Study area

The study area is part of the rural district of Umkhanyakude located about 250 km north of the provincial capital of Durban. The area includes both land under tribal authority that was designated as a Zulu ‘homeland’ under South Africa's former apartheid policy and a township under municipal authority. Infrastructure is poor. In 2001, only 13% of households had access to piped or public tapped water and 50% of households had no electricity supply. Although this is a rural area, there is little subsistence agriculture and most households rely on waged income and pensions. Unemployment is high: in 2001, 25% of those aged 15–65 years reported that they were unemployed and actively seeking work [⁹].

KwaZulu Natal is the province of South Africa with the highest HIV prevalence rate among antenatal clinic attendees. An antenatal survey conducted in the study area in 1998 found that 41% [95% confidence interval (CI), 34.7–47.9] of pregnant women were infected with HIV [¹⁰]. Mortality in the study area rose sharply in the late-1990s. By 2000, the probability of dying between the ages of 15 and 60 years was 58% for women and 75% for men. AIDS, with or without tuberculosis, was the leading cause of death in adulthood (48%) [¹¹].

Data collection

The Africa Centre Demographic Information System (ACDIS) started data collection on 1 January 2000. The demographic surveillance area (DSA) was mapped and all households registered. The study population included all individuals reported by household informants to be household members, both resident and non-resident. Demographic and health information was collected every 4 months for all registered households and individuals. It included reports of births, deaths, migration and changes in household membership. Cause-specific mortality and socioeconomic data collected by ACDIS was used in this analysis [¹¹].

The period of follow-up was from 1 January 2000 to 31 October 2002. Cause-specific mortality data were available for the first 2 years of the follow-up period. At the beginning of 2001, a socioeconomic survey of all households and their members was conducted. Therefore, these data were not available for households that dissolved or outmigrated in 2000.

Household mobility

The legacy of the Apartheid Group Areas Act and the labour migration system, together with a lack of local employment opportunities, has meant that many members of rural households reside elsewhere [^12–14]. Approximately 35% of adult (18 years or older) female household members and 40% of adult males reside outside the area [¹⁵].

A household is eligible for registration if its residence is within the DSA and, at least one member is resident. A household can only have one place of residence with the DSA at any given time. Households may migrate within the DSA or to/from a place outside the DSA. Routine surveillance of a household ends if it migrates out of the area, referred to as ‘outmigration’ in this paper.

When a household migrates, the social group continues to function, they merely have a new place as their main physical residence. When the household dissolves, however, the household (i.e., the social structure to which individuals belong by self-definition) has ceased to exist. For example, after the death of a husband, the surviving partner may join the household of one of her children.

Survival analysis

Cox regression models were used to examine the relationship of multiple risk factors to two outcomes, household dissolution and household outmigration, in the period from 1 January 2000 to 31 October 2002. Multivariate models were used to examine the effect of adult mortality on the risk of household outcomes, controlling for those household factors found to be associated significantly in the univariate models. A separate indicator was included to control for missing socioeconomic data. In order to consider the effect of short follow-up periods, all models were re-estimated excluding households dissolving or migrating in 2000 and deaths occurring in 2000. All models were also estimated excluding single-person households.

Results

Residential and survival data were available for 10 612 households resident in the DSA on 1 January 2000. These households had 80 497 members. By the 31 October 2002, 238 households (2%) had dissolved and 874 (8%) had migrated out of the area. The mean interval from 1 January 2000 to dissolution and outmigration were 15.6 months and 15.1 months, respectively. Households that dissolved or outmigrated were significantly different (P < 0.01) from the other households in all the variables presented in Table 1.

Table 1: Mean characteristics of households resident in the demographic surveillance area on 1 January 2000 by household survival status on 31 October 2002.

There was at least one adult death in 2179 households (21%) during follow-up, and 349 households (3%) had more than one adult death. In the first 2 years of follow-up, for which cause-specific mortality data were available, 880 (8%) households experienced at least one death of an adult from AIDS, and 133 (1%) experienced at least one death of an adult from homicide, suicide or accident.

Household dissolution

Table 2 presents Cox regression models of household and mortality effects on household dissolution. After controlling for household risk factors, deaths of young adults (18–59 years) were associated with a threefold increased risk (3.1; 95% CI, 2.3–4.3) of household dissolution compared with households without such deaths. In the 836 households that experienced at least one adult AIDS death and dissolved, the average time between the first AIDS death and dissolution was 4.8 months. Child mortality was not associated with household dissolution.

Table 2: Risk of household dissolution and outmigration according to household socioeconomic factors and mortality experience for households resident in the demographic surveillance area on 1 January 2000 ^a.

In households that experienced an adult death (lower section of Table 2), the age or sex of the deceased was not associated with household dissolution. However, multiple deaths were associated with an increased risk of dissolution (2.3; 95% CI, 1.3–4.3) compared with households experiencing a single death. The risk of dissolution did not differ significantly between AIDS and non-AIDS deaths. Violent and accidental adult deaths in a household increased the risk of dissolving 2.4 times (95% CI, 1.4–4,2), after controlling for household risk factors and the number of deaths, compared with households experiencing adult deaths from any other causes.

Repeating the analyses excluding 446 households that dissolved in the first year of follow-up or excluding the 159 (1%) single-person households did not appreciably alter the hazard associated with mortality.

Household migration

The rate of household migration is high: 1140 (11%) of all households migrated either within or out of the DSA during the follow-up period, with 874 (8%) of them ending their last residency in the DSA by migrating outside the DSA (Table 2). Adult mortality was not associated with the probability of household outmigration after controlling for household risk factors (0.9; 95% CI, 0.7–1.1). The results were similar in re-estimated models that excluded households migrating in the first year of follow-up and single person households.

Discussion

The level of household mobility in this area is high. Eleven per cent of the households moved at least once during the 2½ years of the study. Household dissolution affected 238 households (2%) and 8% of households experienced an adult AIDS death.

The death of adult members is strongly associated with household dissolution after controlling for other risk factors such as household size and economic status. These findings are similar to those of a study of 8399 households in eastern Zimbabwe [⁵], where 10% of households had dissolved and 24% had moved at least once between 2000 and 2003. The risk of dissolution was higher for households in which an adult had died [⁵]. Comparison with a similar study of HIV/AIDS and household mobility in rural Tanzania is limited by the definition of household dissolution used in that study; a limitation noted by the authors themselves [⁴]. In a study of 1422 rural Kenyan households interviewed in 1997 and 2000, only 9 (< 1%) had dissolved by 2000 [³].

In this study, the increased risk of household dissolution was associated with the experience of adult death per se. It was not a function of the age and sex of the deceased or the cause of death. The repeated experience of adult death increased the vulnerability of households to dissolution. The probability of dissolution did not differ significantly between households with an AIDS death and those with a non-AIDS death. However, as approximately 50% of all adult deaths were from AIDS in this period, rising AIDS mortality will produce higher rates of household dissolution. Mushati et al. [⁵] also found no difference between AIDS and non-AIDS deaths in Zimbabwe.

Sudden adult death through violence or accidents put households at twice the risk of dissolution compared with other adult deaths. Violent and accidental mortality rates were high in the study area: 17% of all deaths in men aged 15–44, and 4% in women in the same age group, were from homicide and road traffic accidents [¹¹]. Households may be affected more severely by such deaths than those preceded by chronic illness because no opportunity exists for forward planning by survivors.

Adult mortality has no effect on the risk of household migration after controlling for other household risk factors. Why might this be? The migration of some members, rather than the whole household, is likely to be a more common response to crisis. Households may cope in the short term by sending dependents to be cared for in other households or by sending adults to find work to replace lost income. While households do migrate in negative circumstances (e.g., defaulting on rent), household migration may often be a response to advantageous ‘pull’ factors at the new residence, such as employment, marriage or a better house. Households unable to cope in situ may also be unable to migrate successfully and go on to dissolve instead.

Financial security is an important determinant of the long-term viability of households. However, it was not possible with only cross-sectional asset data to examine whether household resources had declined as a consequence of adult illness and death. In the long term, the impact of HIV on the social cohesion of families in HIV-affected communities will be severe. Future research needs to examine the social and residential arrangements of surviving members after households dissolve, and the impact on their long-term well-being.

Acknowledgements

The authors are part of the Africa Centre Population Studies Group and acknowledge the contribution of other group members to ACDIS. The authors thank Patrick Heuveline and other participants at the IUSSP Durban 2003 meeting for their comments. Anna-Maria Vanneste and Isolde Birdthistle also made helpful contributions to this work.

Note: An earlier version of the paper was presented at the International Union for the Scientific Study of Population Scientific Meeting on the Empirical Evidence for the Demographic and Socio-economic Impact of AIDS. Durban, South Africa, March 2003.

Sponsorship: This work was supported by the Wellcome Trust UK through grants to ACDIS (65377), the Africa Centre for Health and Population Studies (50534) and Timæus/Hosegood (61145).