DRAFT INTERIM REPORT

 Part 1 2 3 4 5 6

The socio-economic impact of HIV/AIDS on households in South Africa: Pilot study in Welkom and QwaQwa, Free State Province

Booysen, F. le R., Van Rensburg, H.C.J., Bachmann, M., O'Brien, M. & Steyn, F.

OCTOBER 2001

University of the Free State

Centre for Health Systems Research & Development

This research paper is sponsored by USAID and administered by the Joint Center for Political and Economic Studies Inc. under a subcontract agreement from Nathan Associates Inc.

EXECUTIVE SUMMARY

• The impact of HIV/AIDS on households was assessed by means of a longitudinal (cohort) study of households affected by the disease. The CHSR&D established a formal relationship with various stakeholders in the two study sites to facilitate the recruitment of affected households. Verbal informed consent was obtained from infected individuals to interview the households to which they belong. The household impact of HIV/AIDS was determined by comparing over time the observed trends in socioeconomic variables in HIV/AIDS households and a control group using statistical methods. For this purpose, a survey on the quality of life and the economics of affected and non-affected households was conducted. The results reported in this first interim report are based on a cross-section analysis of the data collected during the first phase of the project.

• Households affected by HIV predictably had a higher burden of morbidity and death. People living in affected household were 4 times as likely to have been ill in the last month, and to have infectious disease during their last illness. Affected households had an adjusted average of 2 more people ill during the last month than did unaffected households. Ill people in affected households had more severe illness, indicated by hospital admission (adjusted odds ratio (OR)=10), not recovering (OR=5) and inability to perform daily tasks (OR=3). 20% of affected households had experienced a death during the past 6 months, compared to 1% of unaffected households. 2 affected households each experienced 2 deaths.  Affected households were much more likely to have experienced a death of a family member during in the last 6 months (adjusted OR=25).

• Rural location and lower income aggravated the disease burden. The risk of death in rural QwaQwa was twice as high (adjusted OR=2) as in urban Welkom, while probability of recent illness was marginally higher (OR=1.3). Aside from affected/unaffected and urban/rural household status, other predictors of morbidity outcomes in multivariable models included lower household income, age, highest education level, and employment status. Mortality risk among households was higher if females comprising a larger proportion of a household, and was independent if other risk factors.

• People who were sick in the last month were most likely to have used primary care, that is, government clinics, followed by private doctors. A quarter of ill people were admitted to hospital during their last episode. Among deaths in affected households, the commonest final source of health care was government hospitals (55%), followed by traditional healers (18%).

• The mean cost of health care among ill household members during their last episode was estimated to be R98, with a positively skewed distribution. The mean cost of health care for household members who died was estimated to be R167, also with a skewed distribution. The median cost of funeral expenses was R4000-5000 per death. Relatively few households reported lost income due to illness or death. However this may reflect chronically ill or dying people having been unemployed for some time before their illness or death.

• Most households with ill or dying members carried a burden of caring, and this was greater in affected than in unaffected households. 75% of ill people required someone to care for them at home (relative risk among affected versus unaffected households: 1.2), while 68% required someone to accompany them to health care (RR 1.8). Relatively few carers lost income as a result. Few carers came from outside the household (7% among cases of illness and 5% among fatal cases).

• Affected households on average are slightly larger than non-affected household in terms of household size. However, the dependency ratio in affected household are higher than that in non-affected households, implying that households affected by HIV/AIDS in fact have a smaller supply of labor than non-affected households, with a larger proportion of the household consisting of children and elderly persons. Illness and death in affected households also occurred mainly among members belonging to the economically active population (age 15-49), again emphasizing the adverse effect of the epidemic on the supply of labor in affected households.

• Affected households are poorer than non-affected households, regardless of whether income and expenditure is measured at the household or individual level or in adult equivalent terms.

• Affected households are more dependent on non-employment sources of income (which consists primarily of government grants but also includes the value of own produce consumed by the household), while a smaller proportion of their income consists on employment income.

• Affected households spend less on food than non-affected households, with per capita and adult equivalent levels of expenditure on food representing between 70% and 80% of the levels of expenditure in non-affected households. In the longer run, this may contribute to malnutrition amongst household members.

• Affected households, in terms of the composition of household expenditure, allocate relatively MORE of their resources on food, health care and rent and LESS to education, clothing, personal items and durables when compared to non-affected households. Differences in the share of expenditure allocated to household maintenance and transport and relatively small and may not indicate significant differences in patterns of expenditure.

• Affected households save approximately 40% less than non-affected households on a monthly basis. Non-affected households have considerably higher levels of current debt than non-affected households. There are no considerable differences between the monthly repayment of debt by affected and non-affected households, which means that the servicing of current debt puts a relative larger burden on affected than non-affected households, given their lower levels of income.

• The most frequent responses of households to financial crises seem to be borrowing, followed by the utilization of savings and the sale of assets. Households were more likely to have borrowed, utilized savings or sold an asset where a larger number of deaths had occurred in the recent past, in households where expenditure on average was higher and in larger households headed by younger persons. Coping financially in one or more of these ways was also more likely in rural than in urban areas, given that rural areas are generally poorer than urban areas. Very few households experiencing a recent death had received a lump-sum payment or inheritance following the death, underscoring the few means poor households have to cope with the effect of a recent death.

• A slightly larger number of non-affected households have borrowed money in the twelve months prior to the survey compared to non-affected households. 72% and 25% of the affected household that borrowed money were respectively affected by illness and death, with only 25% and 2% of non-affected households respectively being affected by illness and death. In more than 60% of cases the money was borrowed from relatives and friends, while just more 20% of loans were obtained from money- or micro-lenders. A larger proportion of affected households indicated that the money was used to pay for funerals and medical expenses, whereas a larger proportion of non-affected households indicated that the money was used to pay for education, durables and clothing.

• A larger percentage of affected households have in the six months prior to the interview utilized savings than was the case in non-affected households. 76 and 48% of households that utilized savings were respectively affected by illness or by death, compared to 9 and 0% of non-affected households. The two purposes for utilizing savings sited most often by affected households were to pay for funerals and medical expenses, while non-affected households mainly used savings to pay for education and the maintenance of assets. The magnitude of dissaving is considerable. Affected households on average utilized twenty-one months of savings, whereas non-affected household only utilized five months of current savings. Households in which a death occurred in the six months prior to the survey utilized thirty-eight months of current savings. Households were more likely to have utilized savings when having experienced a larger number of recent deaths, when having no access to medical aid, when expenditure was higher, when headed by younger persons and where the dependency ratio was lower.

• Given that households on average owned few assets, only a small percentage of households sold assets in the twelve months prior to the survey. Households primarily sold household appliances, which represent more than 50% of the type of assets sold, and the assets sold by households were in most cases (with the exception of the sale of cattle by one household) of a non-productive nature. The reasons these assets were sold for do not outright suggest that HIV/AIDS plays an important role in causing affected households to sell assets. However, this may only indicate that affected households that do sell assets actually do so to pay for expenses they can no longer afford since having to pay for medical expenses and funerals from available resources.

• The cost of morbidity to households are relatively low where unemployment levels are very high and household members are primarily cared for by family members with no direct loss of income.

  ♥

  ♥

•  A death puts a much greater financial burden on a household than does illness. In a worst case scenario, the burden on affected households amounted to 3.4 times average monthly household income and 5.7 times average monthly household expenditure. Under alternative assumptions, the total cost represents 2.6 and 4.4 times average monthly household income and expenditure. Unlike in the case of illness, the cost of a death to households remain high even where unemployment levels are very high and household members are primarily cared for by relatives with no direct loss of income. This can be attributed to the fact the funeral costs are very high and represent the largest share of the cost of mortality.

• A relatively small percentage of children aged 7-13 were not attending school at the time of the interview (2.3%), whereas 8.9% of children aged 14-18 years were not attending schooling. In total 5.2% of children of school-going age was not attending school. A larger proportion of children in affected households not currently enrolled in school were female (60.7%). Children were less likely to still be attending school if they belong to households in which the household head is male rather than female. In the case of children aged 7-13, children belonging to households that shelter one or more orphans were amongst others less likely to still be attending school. Children aged 14-18 was more likely to not be attending school if they are older, if they are female, if they belong to households having experienced an increasing number of recent deaths, and if they belong to households that received no government grants. In general, the results indicate that a death in a household may result in older children in poorer, female-headed households being forced to interrupt their schooling.

• 27.6% of children aged fifteen years and under have lost their mother, mother or father, and both mother and father. Although a larger number of orphans are to be found in affected households, non-affected households also shelter a number of orphans. Only one orphaned child that belonged to an affected household was not attending school at the time. Households that shelter orphaned children generally are headed by females and by persons that are widowed.

• Poverty in combination with the HIV/AIDS epidemic seems to represent a major threat to the livelihood of households. Affected households have been shown to be poorer than non-affected households. The single most important predictor of poverty status is access to medical aid. Households with access to medical aid respectively were 16 and 15 times more likely to not be poor if poverty status is determined relative to household income and expenditure. In addition, households not affected by illness were more likely to not be poor, as was households sharing a larger number of years of schooling between its members having a larger number of employed members.

Background and problem statement

The HIV/AIDS epidemic poses a severe threat to the economies of developing countries, and those on the African continent in particular. South Africa, which is being affected fundamentally by the epidemic, is no exception. By the end of 1997, 2.8 million people were estimated to be living with HIV/AIDS in South Africa. By 1999, this figure had increased to 3.5 million. The estimated prevalence of HIV/AIDS among the country's adult population (11.8 per cent) is amongst the highest in the world (ILO, 2000). According to the Metropolitan-Doyle model, the number of South African living with HIV/AIDS will increase from 160 000 to almost one million between 2000 and 2010. The annual number of AIDS deaths is estimated to increase from 120 000 to between 545 and 635 thousand between 2000 and 2010 (Abt Associates, 2000: 8-9). The number of children younger than fifteen years orphaned by AIDS has been estimated to be 800 000 by 2005, rising to more than 1.95 million by 2010 (Abt Associates, 2000: 11). These infected individuals and affected children all belong to individual households and their deaths will have a significant impact on their families. Hence, the epidemic will have a considerably impact on households in South Africa.

Over the next ten to fifteen years, the epidemic has the potential to erode development gains made in past decades. As the disease takes its toll on the economically active population, production and demand are expected to decline, which will slow down economic growth and development. The disease will also have serious budgetary implications in terms of increased government expenditure on health care and social security, which will be aggravated by the decline in government revenue due to declining economic activity (Barnett and Whiteside, 1998; ILO, 2000). These effects of HIV/AIDS are not accounted for in the government's existing framework of economic policy, i.e. the Growth, Employment and Redistribution (GEAR) strategy and Reconstruction and Development Programme (RDP). In fact, GEAR currently envisages continued increases in economic growth, job creation and redistribution over the next three financial years (National Treasury, 1999). This is an unlikely scenario since the impact of HIV/AIDS is expected to become manifest during the next five to ten years. The AIDS epidemic generally lags about eight years behind the HIV epidemic, which explains why the impact of HIV+ prevalence rates currently observed will only really materialize in five to ten years’ time.

Research into the socio-economic impact of HIV/AIDS on households and communities is crucial in guiding current and future policies and intervention strategies intended to absorb this impact.  From an economic point of view, the primary impact of the disease manifests mainly among individual economic agents, i.e. individuals and households. An assessment of the socio-economic of HIV/AIDS would therefore have to start on this micro-level of analysis. Aspects of such assessment, amongst other things, will include determining how the disease affects the economic decisions and position of individuals and households over time, i.e. how they generate, save, invest and spend income in response to the disease, and how this in turn affects their quality of life. To date no comprehensive, longitudinal study of the impact of HIV/AIDS on such a micro-level of analysis has been conducted in South Africa, neither in an urban nor in a rural setting. Arndt and Lewis (2000), furthermore, have performed a preliminary assessment of the macroeconomic implications of HIV/AIDS for South Africa. Yet, their macroeconomic model still fails to allow for the effects of a number of important microeconomic impacts which are endogenous to such model, amongst others that of asset sales and investments in human capital. This failure to a large extent derives from the lack of household level economic data with which to quantify these assumptions. Work on the macroeconomic model maintained by the Department of Finance faces similar constraints (Compernolle, 2000).

OBJECTIVES

The project had the following broad objectives, which will be discussed in more detail later in this document.

§         develop and test a methodology for assessing the socio-economic impact of HIV/AIDS at the individual and household level in both an urban and a rural setting;

§         identify and capture the standard minimum criteria and indicators to be incorporated into the methods of methodologies of studies of this nature;

§         describe and evaluate the impact of different informal coping strategies and support systems adopted by individuals, households and communities, as well as that of formal HIV/AIDS-related interventions of national and provincial government departments and local authorities (TLCs), in terms of their impact over time on the quality of life of affected households living in both urban and rural areas;

§         inform economic growth analyses and studies on the macroeconomic impact of HIV/AIDS by projecting information about the microeconomic impact of the disease onto trends in labor market participation, spending, savings and investment; and

§         propose a framework for national 'best practice' for improving the quality of life of affected households in urban and rural communities based on existing macro- and micro-, as well as formal and informal responses to HIV/AIDS.

 Approach and method

(i)         Population

The impact of HIV/AIDS on individuals and households was assessed by means of a cohort study of households affected by the disease, and compared with a control group of matched households unaffected by the disease. It was conducted in two local communities in the Free State province, one urban (Welkom) and one rural (QwaQwa), in which the HIV/AIDS epidemic is particularly rife. Of the nine provinces in South Africa, the Free State has the second highest prevalence of HIV/AIDS and is also the province with the second highest rate of increase in the prevalence of HIV/AIDS (Cohen, 2000). Welkom is situated in Region C, one of six former health regions in the Free State. In 1997, Region C had the highest HIV prevalence among antenatal clinic attendees of all the six health regions in the province, i.e. 26.6 per cent. HIV prevalence in this region is the second highest in South Africa. The prevalence of HIV/AIDS in the former QwaQwa is also very high compared to other health districts. Because of high unemployment, men from this area are often employed as migrant laborers in towns and cities away from their homes. In addition, the lack of infrastructure, poor services and poor living conditions characteristic of this area further increases the vulnerability of the local population to the HIV/AIDS epidemic.

According to the report entitled Measuring Poverty published by Statistics SA early in 2000, the Welkom magisterial district is the third richest in the Free State province, with a headcount poverty ratio of 0.34 and average monthly household expenditure of R2364. The magisterial district of Witsieshoek, which is within the boundaries of the former QwaQwa, is the poorest in the Free State province and also ranks amongst the poorest in the country. The headcount poverty ratio in this district is 0.69, while average monthly household expenditure amounts to R807. Thus, the particular selection of study sites also allows one to compare the household impact of HIV/AIDS between communities that differ substantially in terms of the level of poverty (Statistics South Africa, 2000).

(ii)        Sampling

The identification of participants in the study, particularly of affected households, requires ethically meticulous research conduct. The myths and secrecy surrounding the disease, as well as the fear of stigmatization and protection of the identity of people living with HIV/AIDS, pose a real challenge for research of this nature since it complicates the identification and selection of participants. The participation of households in this research project is voluntary and based on confidentiality and informed consent and the study is introduced to respondents as such during the fieldwork. The research protocol was submitted to the Research Ethics Committee of the University of the Free State for approval in order to safeguard the rights of the participants and to ensure ethical standards of research. The committee has approved the study. Letters of approval have also been obtained from the following individuals in the Department of Health, all of which have offered their cooperation and expressed their interest in the findings of the project:

Dr. N. Simelela, Chief Director: HIV/AIDS and STDs

Prof K.C. Househam, Head of Department of Health, Free State Province

Mrs R. Sibeko, District Health Manager DC19 (QwaQwa)

Me N.J. Jolingana, District Health Manager DC18 (Welkom)

The CHSR&D established a formal relationship with various stakeholders in the two study sites to facilitate the recruitment of affected households, including the Department of Health and various NGOs and CBOs active in HIV/AIDS work. The research team met with a variety of stakeholders in each of the two areas during the initial phases of the project. These meetings had three purposes: to inform the stakeholders of the research projects and its aims and objectives, to involve the stakeholders in the recruitment of fieldwork managers and fieldworkers, and to involve the stakeholders in the recruitment of participating households. In the research team's opinion, the fact that the fieldwork was managed through and conducted by parties involved in HIV/AIDS-related work in these communities adds much value to the project. The questionnaire was also circulated to these stakeholders for comment, which is important in terms of availing them the chance to ensure that the data generated by the project is of use to them in planning and managing their activities. Through this network as many households as possible that are affected by HIV/AIDS were identified, although in practice the number did not exceeded the 100 target by far. Such approach to sampling avoids the sensitive issue of testing the members of participating households for HIV and also ensures that the selected households are indeed affected by HIV/AIDS. The manager of the fieldwork teams in each of the two study sites was responsible for coordinating this process and obtained verbal consent from each of the infected individuals belonging to the households included in the sample. The manager was also responsible for ensuring that the identified households come from a range of neighborhoods/villages in the area, thus providing the researchers with a sample that reflect differences in demographics and standards of living in the two study sites.

The manner in which the participating households were sampled to a large extent ensures that affected households are indeed affected by HIV/AIDS. However, many infected individuals have not disclosed their status to their families, which means that the study could not be introduced to respondents as an HIV/AIDS study and therefore inadvertently reveal the identity of the infected person to other household members. Households interviewed as controls may also be discouraged to participate in the study if directly introduced as an HIV/AIDS impact study, with particular significant problems being experienced if the household become affected in later phases of the project. Hence, the study was introduced to respondents as 'a study of the impact of morbidity and mortality on households in the Free State province'. The research team found the issue of disclosure to be an important obstacle in the recruitment process and other researchers involved in similar projects are encouraged to find innovative solutions to this problem. Possible ways to perhaps deal with this problem are using the infected individual rather than the household as unit of analysis OR allowing more time for recruitment to actually facilitate a process of disclosure and involve the entire household in the data collection process.

In order to control for the effect on households of socioeconomic changes not related to HIV/AIDS, a control group of 100 households that are not affected by HIV/AIDS was recruited to voluntarily partake in the study. These households were recruited in the following manner. For each affected household that the fieldworker visited for interview purposes, the fieldworker also interviewed a household living in close proximity to the affected household, e.g. a neighboring household. In order to ensure that this household is not affected by HIV/AIDS the fieldworker first asked the respondent a few key questions, i.e. whether someone in the household is being treated for TB or whether someone has been hospitalized with pneumonia in the past six months. Initially, a direct question about whether someone in the household has HIV/AIDS was included in the set of key questions. However, this question was dropped once it became clear during the practice interviews that this question caused respondents to refuse to participate, possibly because of the stigmatization that still surrounds the epidemic. If the respondent answered any of these questions in the affirmative (with a 'YES'), the fieldworker moved to the next household until they found a household for which none of the key questions were answered in the affirmative. Hence, it meant that the fieldworker often had to visit a number of households before they successfully identified a control for each affected household. Fieldworkers were trained to take appropriate care in allowing time for this activity when conducting their interviews. Fieldworkers were also trained to take particular care in recording the address and details of this household. This is crucial for the purposes of revisiting this household six months later during the second wave of the data collection phase of the project.

In order to keep track with interviewed households, all respondents were supplied with a paid, self-addressed postcard on which any change in address can be recorded and mailed to the research team.

Households in the control group that are affected by HIV/AIDS over the three year study period will become part of the sample of affected households. In case the increasing spread of HIV/AIDS and rising AIDS deaths threaten the sustainability of the control group in later phases of the longitudinal study, new respondents will be sampled from the selected communities to act as controls. Since the research will require the continued participation of those households that originally agree to become part of the study, the payment of a minimal participation fee (R100 per household per survey visit) to those households is expected to ensure sustainability of the sample over the three years. In the interest of sustaining the livelihood of severely affected households, it is envisaged that this payment may be made in kind, e.g. in the form of basic foodstuffs of an equal value.

(iii)       Data collection and analysis

The impact of HIV/AIDS on households was assessed by means of a longitudinal (cohort) study of households affected by the disease. The household impact of HIV/AIDS was determined by comparing over time the observed trends in socioeconomic variables in HIV/AIDS households and a control group using statistical methods. For this purpose, a six-monthly survey on the quality of life and the economics of affected and non-affected households was conducted. Interviews were conducted with one respondent only, namely the "person responsible for the daily organization of the household, including household finances". The results reported in this first interim report are based on a cross-section analysis of the data collected during the first phase of the project. Subsequent analysis will focus on a time-series analysis of the data collected during the six-monthly interviews conducted with the sampled households.

The instrument used for this purpose explores the issue mainly in quantitative terms. The instrument explores the economic impact of the disease on, amongst other things, household income and expenditure patterns. It also explores the experiences of households affected by HIV/AIDS with regard to their response to it with regard to caring for affected household members, utilizing certain services, and coping with the impact on their socioeconomic circumstances. The design of the instrument was informed by a literature review of the methodology of household impact studies, focus group sessions with key informants, and the piloting and revision of the draft instrument. For the purposes of comparative analysis, the instrument used for data collection in affected households is the same as that employed in collecting data from unaffected households, although certain sections of the questionnaire (notably that on morbidity and mortality) did not always apply to these households.

A first draft of the questionnaire was completed in early April 2001. Before finalizing the questionnaire and having it translated, a first draft was circulated for comment amongst stakeholders from government departments, NGOs, and CBOs, as well as other academics, which was integrated into the final instrument with issues raised in the pre-testing of the questionnaire. The socioeconomic questions/sections in the questionnaire was standardized in accordance with the recommendations put forward following a meeting between the researchers from different AIDS research projects in Johannesburg toward the end of April 2001. The questionnaire was translated into Sesotho and Afrikaans, which together with English presents the major languages spoken by the population residing in the two study sites, after which final changes were made following problems arising from the pre-testing of the questionnaire in Bloemfontein. A training manual was compiled for the fieldworkers, editors and fieldwork managers following the finalization of the questionnaire.

A common characteristic of household impact studies is to also collect data from other stakeholders, using techniques other than household interviews. To this end the research team also embarked on the following data collection efforts. In terms of qualitative methods, six focus group discussions with women were conducted in Welkom and QwaQwa (three in each site) to obtain additional information on coping and support from the general population in the two study sites. The focus groups were conducted by two female, junior researchers attached to the CHSR&D, namely Tanja Arntz and Dibolelo Molehe. Tanja Arntz will employ this information in combination with data from the household survey in her dissertation for her Masters in Development Studies, which focuses on coping and support mechanisms adopted by affected families. Jacob Molelekoa, a black master student in the Department of Economics, conducted an investigation into the cost of home-based care in Welkom and QwaQwa, which to some extent will inform policy proposals about the extension of home-based care to affected families. He is conducting this research as part of his research for his master's dissertation. These research efforts contributed to building capacity amongst black and female researchers at the University. The main findings from these related research activities will be incorporated into the final report due for release in March/April 2002.

Following an interview process, a fieldwork team consisting of a manager, editor and five fieldworkers was recruited in each of the two study sites, mainly from amongst persons working as volunteers in HIV/AIDS programs. On completion of the training, each member of the research teams signed a contract that stipulates the conditions of services and other project regulations. The research teams recruited for each of the two areas consist of the following individuals (for purposes of capacity building and the involvement of previously disadvantaged persons in the project, please note that all the recruited persons are of PDI status):

Welkom

Fieldwork manager                                           Mr J. Molefi

Editor                                                               Ms K.D. Rankhakile

Fieldworkers                                                    Ms E. Van Rooi

                                                                        Mr D.T. Tlali

                                                                        Ms D. Chabeli

                                                                        Ms. G. Moeti

                                                                        Mr J. Moholobela

QwaQwa

Fieldwork manager                                           Mr N. Khoapa

Editor                                                               Ms K.R. Mofutsanyana

Fieldworkers                                                    Ms M. Maduna

                                                                        Mr T. Motaung

                                                                        Mr L. Mosia

                                                                        Ms D. Masindwa

                                                                        Ms M. Masisi

All members of the two fieldwork teams had received the basic HIV/AIDS training provided to AIDS counselors and volunteer workers by ATTIC by the time the fieldwork commenced. A team of four researchers conducted three-day training sessions in QwaQwa (28-30 May 2001) and Welkom (5-7 June 2001) with the two fieldwork teams. The training consisted of classroom training, scenarios and practice interviews. A researcher spent two more days with the fieldwork team when the fieldwork commenced to further guide the fieldwork team in the data collection process and manage the logistics and administration. The researcher paid regular visits to the two study sites to continuously liaise with the fieldwork manager on the progress of the data collection and to administer payments made to the fieldwork team.

The following has also become apparent during the first weeks of data collection. One fieldworker in Welkom was relieved of his duties after his work was found to be of poor quality, a provision allowed for in the contract signed by the members of the fieldwork team during the training. His interviews were shared between the other fieldworkers and the fieldwork manager. Furthermore, the research team in their efforts to also employ fieldworker training as a tool for capacity building put much effort into guiding the fieldwork teams during the data collection process. A researcher paid regular visits to the area to perform quality control checks, to assist the editor with the editing of questionnaires, and to ensure that the process is on track. The data collection in the two study sites was completed by the end of July 2001. The months of August and September 2001 was taken up by the coding and computerization of the data, while data analysis and report writing commenced in October 2001.

SOCIO-ECONOMIC PROFILE OF SAMPLE POPULATION

In order to determine to what population the results described in this report about the socioeconomic impact of HIV/AIDS on households can be generalized, it is important to explore the main demographic and socioeconomic circumstances of the households included in the sample. These details are reported in Tables A to F in the Appendix. In essence, the group of households included in this household impact study can be described as follows:

·        The households are mainly African and Colored (88.4 and 11.3 % respectively of the total sample), while only one White household was interviewed. Nationally the African and Colored populations respectively represent 76.7 and 8.9 % of the country's population.

·        A slightly larger proportion of households is headed by females (53.7% compared to 46.3% headed by males).

·        The persons heading these households are aged 40-49 years (25.4%), 30-39 years (24%), 50-59 years (19.3%) and 60-69 years (13.3%), which represents a relatively normal distribution.

·        A fairly large proportion of persons heading households are widows/widowers (30.8%, while 40.4% are married (civil or traditional), 14.5% are divorced/separated, and 10.8% have never been married.

·        A larger share of household members is female (57.5% compared to 42.5% of household members that are male). According to the 1996 population census, 48.1 and 51.9 % of the population were respectively male and female.

·        Households on average have just more than 30 years of schooling amongst them, with the largest proportion of households having 20-39 years of education. Given the average household size of nearly five, this means that one is looking at relatively poorly educated households.

·        Few households have access to medical aid and only 15% of households include a member with access to medical aid. Fewer affected households had access to medical aid compared to non-affected households (9.9 versus 20.1 %), which as explained below are mainly due to the sampling design.

·        Most households indicated that they feel very safe (50.7%) or rather safe (27.8%) living in the areas where they reside, while only a very small proportion of households indicated that they feel very unsafe (4.4%) or not safe at all (3%).

·        The majority of households live in one dwelling (76.6%), while 19.4% of households indicated that they live in more than one separate dwelling. Only a small percentage of households (3.9%) shared a dwelling with another household, more so in the urban setting (Welkom) than in the rural setting (QwaQwa).

·        Nearly 80% of households lives in a main dwelling on a separate stand or yard, while 13% live in some kind of informal dwelling (informal dwelling in backyard or informal settlement). A small proportion of households (4.9%) live in traditional dwellings, representing households in the rural sample (QwaQwa).

·        The main dwellings in which households live on average consist of four rooms of which two are used for sleeping.

·        Just more than 90% of households own the dwelling in which they reside. Households living in dwellings not owned by the household mainly live in dwellings owned by a private owner renting out their property.

·        The majority of respondents (89.2%) indicated that they have not lived at their current place of residence since birth. Of respondents who had before resided in a different place than where they were born, 64.6% previously resided in urban areas, 23.1% in rural areas and 11.1% on commercial farms. The main reasons these respondents sited for having changed their place of residence were moving to a new house (42.1%), work-related reasons (32%) and marriage-related reasons (20.9%). In terms of their place of birth, 45.7% of respondents were born in urban areas, 28.7% in rural areas and 24.8% on commercial farms. The reason respondents moved from their place of birth was mainly having moved to a new house alone or with their family (59.1%) or because of reasons related to work (18.5%) or marriage (13.5%).

·        Only in a relatively small proportion of households do someone own a cellular phone (29.6%) or does the household have a telephone in their dwelling (27.3%). Only 12% of households have access to either a cellular phone or a telephone at home.

·        Almost all households have access to sanitation, with 39.9% of households having access to a flush toilet in their dwelling, 31.4% having access to a pit latrine on site and 26 % having access to a flush toilet on site.

·        Just more than 75 % of households have access to piped water, be it in the dwelling (47.2%), on site (23.2%) or at neighbors (4.7%). However, nearly 25% of households were dependent on a public tap for their supply of water.

·        In terms of refuse removal, 70.2% of households had their refuse removed by their local authority are least once a week, while 18% of households had an own refuse dump and 8.6% of households had no refuse removal.

·        The source of energy for lighting was mainly electricity (77.6%), while 19% of households used candles (mainly in rural areas) and 3.4% used paraffin. Electricity was again the main source of energy for heating (47.1%), while 23.1% of households, again mainly rural households, used coal and 17.1% used paraffin for heating as fuel source for heating. The source of energy for cooking reflects a similar picture, with 62.3, 24.1 and 10.6 % of households respectively using electricity, paraffin and coal as energy source. Those households that used coal as energy source for cooking all live in rural areas (QwaQwa).

Evident from the above is that although the sample in certain instances closely reflects the socioeconomic profile of the national population (e.g. male/female distribution of the population), it in most cases differs distinctly from the general South African population. The profile of the sample of households included in this impact study can largely be attributed to the sampling design. Given that affected households were sampled from networks and/or organizations involved in counseling, home-based care and public health care and mainly in poorer communities, the sample does not include affected households that mainly utilize private health care services. Moreover, the study was conducted in one specific province (Free State) and in two selected sites only (Welkom and QwaQwa). However, the fact that South Africa's poor, predominantly Black population face relatively high HIV prevalence rates and are particularly vulnerable to the epidemic and therefore dependent on support from the public service sphere, means that the findings and policy recommendations put forward in this report are especially relevant to informing government's responses to HIV/AIDS.

 KEY CONCEPTS FOR COMPARATIVE ANALYSIS

The results presented in the subsequent pages of this documents draws comparisons between households in terms of the socioeconomic impact of HIV/AIDS based on four stratifications of the data. These concepts and terminology can be defined as follows.

·        HOUSEHOLD: Households were defined in terms of the standard definition employed by Statistics South Africa in the October Household Survey, i.e. "a person or a group of persons who live together at least four nights a week at the same address, eat together and share resources".

·        URBAN versus RURAL comparisons: This refers to the distinction between households living in Welkom and households living in QwaQwa. Welkom is a relatively large urban settlement in the Goldfields in the Eastern Free State. QwaQwa is a former homeland, which is still governed mainly by traditional leadership in an area where communities reside in 42 smaller villages. The distinction therefore between urban/rural is based on the nature of governance structures in the two areas rather than the physical housing infrastructure characteristic of these areas. In QwaQwa for example the majority of the population reside in formal dwellings (refer page elsewhere), yet the community remains a predominantly rural one.

·        AFFECTED versus NON-AFFECTED comparisons: This refers to the distinction between interviewed households in which at least one person is known to be HIV-positive as opposed to interviewed households residing in close proximity in the affected households which was sampled as controls (see discussion elsewhere). The former households were recruited purposively from established networks and/or organizations in the two areas involved in HIV/AIDS. In the case of the latter households no one in these households is known to be HIV-positive insofar testing could not be conducted, nor was any member of these households presently treated for tuberculosis or hospitalized for pneumonia in the month before the interview.

·        ILLNESS versus NO ILLNESS comparisons: This refers to the distinction between households in which one or more members had been continuously ill in the month preceding the interview as opposed to households where no member had been continuously ill in the month preceding the interview.

·        DEATH versus NO DEATH comparisons: This refers to the distinction between households in which one or more members had died in the six month preceding the interview as opposed to households where no member had died in the six month preceding the interview.

In the subsequent pages, the results and main findings of the project are elaborated on. Section A focuses on health outcomes, which is important in establishing whether affected and non-affected households actually represent a foundation for determining the impact of HIV/AIDS and for informing certain aspects of health policies related to coping with the HIV/AIDS epidemic. Section B focuses on various aspects of the socioeconomic impact of HIV/AIDS on households, e.g. the supply of labor at the household level, expenditure patterns, financial coping strategies, and issues related to the impact on children. The conclusions are discussed in the final part of the report.

 METHODS

Proportions of households (or household members) were compared between affected and unaffected households, and between Welkom and QwaQwa, using Pearson c² or exact tests. Outcomes were where possible compared at both individual and household levels.

Multiple logistic regression analysis was used to determine the independent influences of certain explanatory variables on selected outcomes related to morbidity, mortality and the socioeconomic impact of HIV/AIDS, adjusting for influential personal, household and area characteristics. Variables were retained in each model if they significantly improved the respective model.

Logistic regression models with individual level outcomes were adjusted for clustering of outcomes at household level, using Stata statistical software. Intra-household correlation of each outcome was expressed as an intra-cluster correlation coefficient (ICC). The ICC is the proportion of the outcome’s total variance accounted for by inter-household (as opposed to inter-individual) differences.

 Part 1 2 3 4 5 6