Demographic and Health Survey 2003-2004, South Africa
South Africa, 2003 - 2004
Get Microdata
Reference ID
zaf-doh-dhs-2003-2004-v1
Producer(s)
South Africa. Department of Health (DOH), South Africa. Medical Research Council (MRC)
Collections
Created on
Oct 25, 2013
Last modified
Jun 11, 2020
Page views
185668
Downloads
2328
Identification
Survey ID Number
zaf-doh-dhs-2003-2004-v1
Title
Demographic and Health Survey 2003-2004, South Africa
Subtitle
South Africa
Country
| Name | Country code |
|---|---|
| South Africa | zaf |
Study type
Demographic and Health Survey [hh/dhs]
Abstract
The 2003-2004 South African Demographic and Health Survey is the second national health survey to be conducted by the Department of Health, following the first in 1998. Compared with the first survey, the new survey has more extensive questions around sexual behaviour and for the first time included such questions to a sample of men. Anthropometric measurements were taken on children under five years, and the adult health module has been enhanced with questions relating to physical activity and micro-nutrient intake, important risk factors associated with chronic diseases. The 2003-2004 SADHS has introduced a chapter reporting on the health, health service utilisation and living conditions of South Africa's older population (60 years or older) and how they have changed since 1998. This has been introduced because this component of the population is growing at a much higher rate than the other age groups. The chapter on adolescent health in 1998 focussed on health risk-taking behaviours of people aged 15-19 years. The chapter has been extended in the 2003-2004 SADHS to include indicators of sexual behaviour of youth aged 15-24 years.
A total of 10 214 households were targeted for inclusion in the survey and 7 756 were interviewed, reflecting an 85 percent response rate. The survey used a household schedule to capture basic information about all the members of the household, comprehensive questionnaires to all women aged 15-49, as well as anthropometric measurements of all children five years and younger. In every second household, interviews of all men 15-59 were conducted and in the alternate households, interviews and measurements of all adults 15 years and older were done including heights, weights, waist circumference, blood pressure and peak pulmonary flow. The overall response rate was 75 percent for women, 67 percent for men, 71 percent for adults, and 84 percent for children. This is slightly lower than the overall response rate for the 1998 SADHS, but varied substantially between provinces with a particularly low response rate in the Western Cape.
OBJECTIVES
In 1995 the National Health Information System of South Africa (NHIS/SA) committee identified the need for improved health information for planning services and monitoring programmes. The first South African Demographic and Health Survey (SADHS) was planned and implemented in 1998. At the time of the survey it was agreed that the survey had to be conducted every five years to enable the Department of Health to monitor trends in health services.
STUDY LIMITATIONS AND RECOMMENDATIONS
Comparison of the socio-demographic characteristics of the sample with the 2001 Population Census shows an over-representation of urban areas and the African population group, and an under-representation of whites and Indian females. It also highlights many anomalies in the ages of the sample respondents, indicating problems in the quality of the data of the 2003 survey. Careful analysis has therefore been required to distinguish the findings that can be considered more robust and can be used for decision making. This has involved considering the internal consistency in the data, and the extent to which the results are consistent with other studies.
Some of the key demographic and adult health indicators show signs of data quality problems. In particular, the prevalence of hypertension, and the related indicators of quality of care are clearly problematic and difficult to interpret. In addition, the fertility levels and the child mortality estimates are not consistent with other data sources. The data problems appear to arise from poor fieldwork, suggesting that there was inadequate training, supervision and quality control during the implementation of the survey. It is imperative that the next SADHS is implemented with stronger quality control mechanisms in place. Moreover, consideration should be given to the frequency of future surveys. It is possible that the SADHS has become overloaded - with a complex implementation required in the field. Thus it may be appropriate to consider a more frequent survey with a rotation of modules as has been suggested by the WHO.
A total of 10 214 households were targeted for inclusion in the survey and 7 756 were interviewed, reflecting an 85 percent response rate. The survey used a household schedule to capture basic information about all the members of the household, comprehensive questionnaires to all women aged 15-49, as well as anthropometric measurements of all children five years and younger. In every second household, interviews of all men 15-59 were conducted and in the alternate households, interviews and measurements of all adults 15 years and older were done including heights, weights, waist circumference, blood pressure and peak pulmonary flow. The overall response rate was 75 percent for women, 67 percent for men, 71 percent for adults, and 84 percent for children. This is slightly lower than the overall response rate for the 1998 SADHS, but varied substantially between provinces with a particularly low response rate in the Western Cape.
OBJECTIVES
In 1995 the National Health Information System of South Africa (NHIS/SA) committee identified the need for improved health information for planning services and monitoring programmes. The first South African Demographic and Health Survey (SADHS) was planned and implemented in 1998. At the time of the survey it was agreed that the survey had to be conducted every five years to enable the Department of Health to monitor trends in health services.
STUDY LIMITATIONS AND RECOMMENDATIONS
Comparison of the socio-demographic characteristics of the sample with the 2001 Population Census shows an over-representation of urban areas and the African population group, and an under-representation of whites and Indian females. It also highlights many anomalies in the ages of the sample respondents, indicating problems in the quality of the data of the 2003 survey. Careful analysis has therefore been required to distinguish the findings that can be considered more robust and can be used for decision making. This has involved considering the internal consistency in the data, and the extent to which the results are consistent with other studies.
Some of the key demographic and adult health indicators show signs of data quality problems. In particular, the prevalence of hypertension, and the related indicators of quality of care are clearly problematic and difficult to interpret. In addition, the fertility levels and the child mortality estimates are not consistent with other data sources. The data problems appear to arise from poor fieldwork, suggesting that there was inadequate training, supervision and quality control during the implementation of the survey. It is imperative that the next SADHS is implemented with stronger quality control mechanisms in place. Moreover, consideration should be given to the frequency of future surveys. It is possible that the SADHS has become overloaded - with a complex implementation required in the field. Thus it may be appropriate to consider a more frequent survey with a rotation of modules as has been suggested by the WHO.
Kind of Data
Sample survey data
Unit of Analysis
Households and individuals
Version
Version Description
v1: Edited, anonymised dataset for public distribution.
Version Notes
DataFirst has included 2003-2004 as the reference date for this survey, as data collection spanned both years
Scope
Notes
The South African Demographic and Health Survey 2003-2004 covers the following topics:
- Adolescent health
- Adult health: hypertension, chronic pulmonary disease and Asthma
- Antenatal, delivery, and postnatal care
- Anthropometry
- Asthma
- Breastfeeding and weaning practices
- Blood Pressure
- Characteristics of households and respondents
- Child health and immunisation
- Contraception and fertility preferences
- Domestic Violence
- Fertility
- Health of older persons
- Infant and child mortality
- Infant and child feeding
- Knowledge and use of contraceptive methods
- Knowledge of HIV and AIDS
- Malaria
- Maternal Mortality
- Marriage and recent sexual activity
- Mortality and morbidity in adults
- Maternal and child health
- Oral health
- Reproductive history
- Risk factors for chronic diseases
- Sexual behaviour, HIV and AIDS
- Utilisation of health services
- Adolescent health
- Adult health: hypertension, chronic pulmonary disease and Asthma
- Antenatal, delivery, and postnatal care
- Anthropometry
- Asthma
- Breastfeeding and weaning practices
- Blood Pressure
- Characteristics of households and respondents
- Child health and immunisation
- Contraception and fertility preferences
- Domestic Violence
- Fertility
- Health of older persons
- Infant and child mortality
- Infant and child feeding
- Knowledge and use of contraceptive methods
- Knowledge of HIV and AIDS
- Malaria
- Maternal Mortality
- Marriage and recent sexual activity
- Mortality and morbidity in adults
- Maternal and child health
- Oral health
- Reproductive history
- Risk factors for chronic diseases
- Sexual behaviour, HIV and AIDS
- Utilisation of health services
Coverage
Geographic Coverage
The SADHS sample was designed to be a nationally representative sample.
Universe
The population covered by the 2003-2004 SADHS is defined as the universe of all women age 15-49, all men 15-59 in South Africa.
Producers and sponsors
Primary investigators
| Name | Affiliation |
|---|---|
| South Africa. Department of Health (DOH) | Primary investigator |
| South Africa. Medical Research Council (MRC) | Primary investigator |
Producers
| Name | Role |
|---|---|
| Africa Strategic Research Cooperation | Technical assistance |
| ORC MACRO | Technical support in questionnaire design, sample design, field staff training, data processing and analysis |
| Statistics South Africa | Technical assistance |
| Human Sciences Research Council | Technical assistance |
Funding Agency/Sponsor
| Name | Abbreviation | Role |
|---|---|---|
| Department of Social Development | Funding | |
| Department of Health (DOH) | Funding | |
| U.S. Agency for International Development | USAID | Funding |
Sampling
Sampling Procedure
The SADHS sample was designed to be a nationally representative probability sample of approximately 10000 households. The country was stratified into the nine provinces and each province was further stratified into urban and non-urban areas.
The sampling frame for the SADHS was provided by Statistics South Africa (Stats SA) based on the enumeration areas (EAs) list of approximately 86000 EAs created during the 2001 census. Since the Indian population constitutes a very small fraction of the South African population, the Census 2001 EAs were stratified into Indian and non-Indian. An EA was classified as Indian if the proportion of persons who classified themselves as Indian during Census 2001 enumeration in that EA was 80 percent or more, otherwise it was classified as Non-Indian. Within the Indian stratum, EAs were sorted descending by the proportion of persons classified as Indian. It should be noted that some provinces and non-urban areas have a very small proportion of the Indian population hence the Indian stratum could not be further stratified by province or urban/non-urban. A sample of 1000 households was allocated to the stratum. Probability proportional to size (PPS) systematic sampling was used to sample EAs and the proportion of Indian persons in an EA was the measure of size. The non-Indian stratum was stratified explicitly by province and within province by the four geo types, i.e. urban formal, urban informal, rural formal and tribal. Each province was allocated a sample of 1000 households and within province the sample was proportionally allocated to the secondary strata, i.e. geo type. For both the Indian and Non-Indian strata the sample take of households within an EA was sixteen households. The number of visited households in an EA as recorded in the Census 2001, 09 Books was used as the measure of size (MOS) in the Non-Indian stratum.
The second stage of selection involved the systematic sampling of households/stands from the selected EAs. Funds were insufficient to allow implementation of a household listing operation in selected EAs. Fortunately, most of the country is covered by aerial photographs, which Statistics SA has used to create EA-specific photos. Using these photos, ASRC identified the global positioning system (GPS) coordinates of all the stands located within the boundaries of the selected EAs and selected 16 in each EA, for a total of 10080 selected. The GPS coordinates provided a means of uniquely identifying the selected stand. As a result of the differing sample proportions, the SADHS sample is not self-weighting at the national level and weighting factors have been applied to the data in this report.
A total of 630 Primary Sampling Units (PSUs) were selected for the 2003-2004 SADHS (368 in urban areas and 262 in non-urban areas). This resulted in a total of 10214 households being selected throughout the country1. Every second household was selected for the adult health survey. In this second household, in addition to interviewing all women aged 15-49, all adults aged 15 and over were eligible to be interviewed with the adult health questionnaire. In every alternate household selected for the survey, not interviewed with the adult health questionnaire, all men aged 15-59 years were also eligible to be interviewed. It was expected that the sample would yield interviews with approximately 10000 households, 12500 women aged 15-49, 5000 adults and 5000 men.
The sampling frame for the SADHS was provided by Statistics South Africa (Stats SA) based on the enumeration areas (EAs) list of approximately 86000 EAs created during the 2001 census. Since the Indian population constitutes a very small fraction of the South African population, the Census 2001 EAs were stratified into Indian and non-Indian. An EA was classified as Indian if the proportion of persons who classified themselves as Indian during Census 2001 enumeration in that EA was 80 percent or more, otherwise it was classified as Non-Indian. Within the Indian stratum, EAs were sorted descending by the proportion of persons classified as Indian. It should be noted that some provinces and non-urban areas have a very small proportion of the Indian population hence the Indian stratum could not be further stratified by province or urban/non-urban. A sample of 1000 households was allocated to the stratum. Probability proportional to size (PPS) systematic sampling was used to sample EAs and the proportion of Indian persons in an EA was the measure of size. The non-Indian stratum was stratified explicitly by province and within province by the four geo types, i.e. urban formal, urban informal, rural formal and tribal. Each province was allocated a sample of 1000 households and within province the sample was proportionally allocated to the secondary strata, i.e. geo type. For both the Indian and Non-Indian strata the sample take of households within an EA was sixteen households. The number of visited households in an EA as recorded in the Census 2001, 09 Books was used as the measure of size (MOS) in the Non-Indian stratum.
The second stage of selection involved the systematic sampling of households/stands from the selected EAs. Funds were insufficient to allow implementation of a household listing operation in selected EAs. Fortunately, most of the country is covered by aerial photographs, which Statistics SA has used to create EA-specific photos. Using these photos, ASRC identified the global positioning system (GPS) coordinates of all the stands located within the boundaries of the selected EAs and selected 16 in each EA, for a total of 10080 selected. The GPS coordinates provided a means of uniquely identifying the selected stand. As a result of the differing sample proportions, the SADHS sample is not self-weighting at the national level and weighting factors have been applied to the data in this report.
A total of 630 Primary Sampling Units (PSUs) were selected for the 2003-2004 SADHS (368 in urban areas and 262 in non-urban areas). This resulted in a total of 10214 households being selected throughout the country1. Every second household was selected for the adult health survey. In this second household, in addition to interviewing all women aged 15-49, all adults aged 15 and over were eligible to be interviewed with the adult health questionnaire. In every alternate household selected for the survey, not interviewed with the adult health questionnaire, all men aged 15-59 years were also eligible to be interviewed. It was expected that the sample would yield interviews with approximately 10000 households, 12500 women aged 15-49, 5000 adults and 5000 men.
Response Rate
Of the total 630 PSUs that were selected, fieldwork was not implemented in nine PSUs. The data file contained information for a total of 621 PSUs. A total of 10214 households were selected for the sample and 7756 were successfully interviewed. The shortfall was primarily due to refusals and to dwellings that were vacant or in which the inhabitants had left for an extended period at the time they were visited by interviewing teams. Of the 9181 households occupied 85 percent were successfully interviewed. In these households, 7966 women were identified as eligible for the individual women's interview (15-49) and interviews were completed with 7041 or 88 percent of them. In the one half of the households that were selected for inclusion in the adult health survey 9614 eligible adults age 15 and over were identified of which 8115 or 84 percent were interviewed. In the other half of the households that were selected for the men's questionnaire to be completed 3930 eligible men aged 15-59 were identified of which 3118 or 79 percent were interviewed. The principal reason for non-response among eligible women and men was the failure to find them at home despite repeated visits to the household.
Data Collection
Dates of Data Collection
| Start | End |
|---|---|
| 2003-10 | 2004-08 |
Data Collection Mode
Face-to-face [f2f]
Supervision
In the course of the fieldwork quality control measures were instituted at three levels. Firstly, field team leaders and editors were trained to identify the enumerator areas included in the sample and guide interviewers in the selection of dwellings for interviews. Secondly, a team consisting of staff from the HSRC carried out independent quality control visits to check questionnaires for errors, quality of identification and interviews at the enumerator area and dwelling levels. A team of staff members from the NDoH also carried out independent quality control visits to check questionnaires for errors, quality of identification and interviews at the enumerator area and dwelling levels. An independent consultant was appointed by the NDoH in January 2004 to assist ASRC with the implementation and fieldwork management after problems in this regard were identified.
Data Collection Notes
A tender for the implementation of the field work for the survey was issued by the National Department of Health. The contract for the field work was awarded to Africa Strategic Research Corporation (ASRC), a private firm based in Johannesburg. ASRC organised a 2-week training course from September 15-30, 2003 at a centre outside of Pretoria. The training of field workers was conducted by personnel from the MRC, the HSRC, National Department of Health and ORC Macro as well as staff and consultants appointed by ASRC. Training consisted of plenary sessions on more general issues like interviewing techniques, survey administration, and explaining the questionnaire and how to complete it, as well as smaller sessions to practice the anthropometric measurements and interviewing in local languages. The training included mock interviews between participants and two written tests. A practice session was arranged one Saturday to give trainees experience with interviewing actual households living around Pretoria and Johannesburg. ASRC was unable to recruit a sufficient number of interviewers of the required racial and gender groups for the first training. Consequently, a second training for an additional 49 trainees was arranged for October 6-11. In order to further balance the ethnic group and gender composition of the teams as well as to make up for attrition of field staff, some additional fieldworkers were trained in February 2004. 192 candidates were recruited for field work. The fieldworkers were organised into teams consisting of varying numbers of female and male interviewers and headed by a supervisor.
Each province had 1 or 2 fieldwork supervisors and at least one editor who were responsible for the logistics and first round of checking of questionnaires. Each province had at least one team of interviewers consisting of different numbers of female and male members. This allowed for the teams to interview different members in households simultaneously, e.g. whilst the woman was interviewed by the female team member, the male team member interviewed the men. Due to political sensitivity and language problems teams were constructed in such a manner to be sensitive for the demographics of a specific area. This resulted in a team of white interviewers who where circulated between different provinces to do interviews in predominantly white areas. In each province there was a provincial manager who was an overall supervisor of the fieldwork operations. Staff from HSRC and the DoH conducted periodic quality control visits during fieldwork. Fieldwork commenced in mid-October 2003 and was completed in August 2004.
Each province had 1 or 2 fieldwork supervisors and at least one editor who were responsible for the logistics and first round of checking of questionnaires. Each province had at least one team of interviewers consisting of different numbers of female and male members. This allowed for the teams to interview different members in households simultaneously, e.g. whilst the woman was interviewed by the female team member, the male team member interviewed the men. Due to political sensitivity and language problems teams were constructed in such a manner to be sensitive for the demographics of a specific area. This resulted in a team of white interviewers who where circulated between different provinces to do interviews in predominantly white areas. In each province there was a provincial manager who was an overall supervisor of the fieldwork operations. Staff from HSRC and the DoH conducted periodic quality control visits during fieldwork. Fieldwork commenced in mid-October 2003 and was completed in August 2004.
Data Collectors
| Name |
|---|
| Africa Strategic Research Cooperation |
Questionnaires
Questionnaires
The questionnaire for each DHS can be found as an appendix in the final report for each study.
The survey utilised five questionnaires: a Household Questionnaire, a Women's Questionnaire, a Men's Questionnaire, an Adult Health Questionnaire and an Additional Children Questionnaire. The contents of the first three questionnaires were based on the DHS Model Questionnaires. These model questionnaires were adapted for use in South Africa during a series of meetings with a Project Team that consisted of representatives from the National Department of Health, the Medical Research Council, the Human Sciences Research Council, Statistics South Africa, National Department of Social Development and ORCMacro. Draft questionnaires were circulated to other interested groups, e.g. such as academic institutions. The Additional Children and Men's Questionnaires were developed to address information needs identified by stakeholders, e.g. information on children who were not staying with their biological mothers. All questionnaires were developed in English and then translated in all 11 official languages in South Africa (English, Afrikaans, isiXhosa, isiZulu, Sesotho, Setswana, Sepedi, SiSwati, Tshivenda, Xitsonga and isiNdebele).
a) The Household Questionnaire was used to list all the usual members and visitors in the selected households. Basic information was collected on the characteristics of each person listed, including age, sex, education and relationship to the head of the household. Information was collected about social grants, work status and injuries experienced in the last month. An important purpose of the Household Questionnaire was to identify women, men and adults who were eligible for individual interviews. In addition information was collected about the dwelling itself, such as the source of water, type of toilet facilities, material used to construct the house and ownership of various consumer goods.
b) The Women's Questionnaire was used to collect information from women aged 15-49 in all households. These women were asked questions on the following topics:
- Background characteristics (age, education, race, residence, marital status, etc.)
- Reproductive history
- Knowledge and use of contraceptive methods
- Antenatal, delivery, and postnatal care
- Breastfeeding and weaning practices
- Child health and immunisation
- Marriage and recent sexual activity
- Fertility preferences
- Adult and maternal mortality
- Knowledge of HIV and AIDS
- Husband's background and respondent's work
c) In every second household, all men and women aged 15 and above were eligible to be interviewed with the Adult Health Questionnaire. The respondents were asked questions on:
- Recent utilisation of health services, family medical history,
- Clinical conditions
- Dental health
- Occupational health
- Medications taken
- Habits and lifestyles
- Anthropometric measurements, and
- Blood pressure and lung function test.
d) In every second household in addition to the women, all men aged 15-59 were eligible to be interviewed. The Men's Questionnaire collected similar information contained in the Woman's Questionnaire but was shorter because it did not contain questions on reproductive history, maternal and child health, nutrition, and maternal mortality. Men were asked questions on the following topics:
- Husband's background and respondent's work
- Knowledge and use of contraceptive methods
- Antenatal, delivery, and postnatal care
- Breastfeeding and weaning practices
- Marriage and recent sexual activity
- Fertility preferences
- Adult mortality
- Knowledge of HIV and AIDS
e) In households in which there was a child under six years of age whose biological mother was either not alive or did not live in the household, information about the child was collected from a guardian using the Additional Child's Questionnaire. The level of child fostering is relatively high in South Africa and data on children's health collected only from biological mothers might be incomplete.
The SADHS questionnaires were pre-tested (in two languages) in July 2003, using the “behind the glass”2 technique. The questionnaires were then adapted to take into account the suggested changes for questions that were misunderstood or were not clear. Subsequently four teams of interviewers (one for each of four main language groups) were formed; the household, male, female and adult health questionnaires were tested in 4 identified areas. The lessons learnt from the two exercises were used to finalise the survey instruments. The questions were translated and produced in all official languages in South Africa (English, Afrikaans, isiXhosa, isiZulu, Sesotho, Setswana, Sepedi, SiSwati, TshiVenda, Xitsonga and isiNdebele)
The survey utilised five questionnaires: a Household Questionnaire, a Women's Questionnaire, a Men's Questionnaire, an Adult Health Questionnaire and an Additional Children Questionnaire. The contents of the first three questionnaires were based on the DHS Model Questionnaires. These model questionnaires were adapted for use in South Africa during a series of meetings with a Project Team that consisted of representatives from the National Department of Health, the Medical Research Council, the Human Sciences Research Council, Statistics South Africa, National Department of Social Development and ORCMacro. Draft questionnaires were circulated to other interested groups, e.g. such as academic institutions. The Additional Children and Men's Questionnaires were developed to address information needs identified by stakeholders, e.g. information on children who were not staying with their biological mothers. All questionnaires were developed in English and then translated in all 11 official languages in South Africa (English, Afrikaans, isiXhosa, isiZulu, Sesotho, Setswana, Sepedi, SiSwati, Tshivenda, Xitsonga and isiNdebele).
a) The Household Questionnaire was used to list all the usual members and visitors in the selected households. Basic information was collected on the characteristics of each person listed, including age, sex, education and relationship to the head of the household. Information was collected about social grants, work status and injuries experienced in the last month. An important purpose of the Household Questionnaire was to identify women, men and adults who were eligible for individual interviews. In addition information was collected about the dwelling itself, such as the source of water, type of toilet facilities, material used to construct the house and ownership of various consumer goods.
b) The Women's Questionnaire was used to collect information from women aged 15-49 in all households. These women were asked questions on the following topics:
- Background characteristics (age, education, race, residence, marital status, etc.)
- Reproductive history
- Knowledge and use of contraceptive methods
- Antenatal, delivery, and postnatal care
- Breastfeeding and weaning practices
- Child health and immunisation
- Marriage and recent sexual activity
- Fertility preferences
- Adult and maternal mortality
- Knowledge of HIV and AIDS
- Husband's background and respondent's work
c) In every second household, all men and women aged 15 and above were eligible to be interviewed with the Adult Health Questionnaire. The respondents were asked questions on:
- Recent utilisation of health services, family medical history,
- Clinical conditions
- Dental health
- Occupational health
- Medications taken
- Habits and lifestyles
- Anthropometric measurements, and
- Blood pressure and lung function test.
d) In every second household in addition to the women, all men aged 15-59 were eligible to be interviewed. The Men's Questionnaire collected similar information contained in the Woman's Questionnaire but was shorter because it did not contain questions on reproductive history, maternal and child health, nutrition, and maternal mortality. Men were asked questions on the following topics:
- Husband's background and respondent's work
- Knowledge and use of contraceptive methods
- Antenatal, delivery, and postnatal care
- Breastfeeding and weaning practices
- Marriage and recent sexual activity
- Fertility preferences
- Adult mortality
- Knowledge of HIV and AIDS
e) In households in which there was a child under six years of age whose biological mother was either not alive or did not live in the household, information about the child was collected from a guardian using the Additional Child's Questionnaire. The level of child fostering is relatively high in South Africa and data on children's health collected only from biological mothers might be incomplete.
The SADHS questionnaires were pre-tested (in two languages) in July 2003, using the “behind the glass”2 technique. The questionnaires were then adapted to take into account the suggested changes for questions that were misunderstood or were not clear. Subsequently four teams of interviewers (one for each of four main language groups) were formed; the household, male, female and adult health questionnaires were tested in 4 identified areas. The lessons learnt from the two exercises were used to finalise the survey instruments. The questions were translated and produced in all official languages in South Africa (English, Afrikaans, isiXhosa, isiZulu, Sesotho, Setswana, Sepedi, SiSwati, TshiVenda, Xitsonga and isiNdebele)
Data Processing
Data Editing
A preliminary round of data processing of the SADHS questionnaires was started in November 2003 so as to provide some feedback to field teams. The actual data processing did not start until January 2004, after a contract was arranged with the HSRC in Pretoria. Completed questionnaires were returned periodically from the field to ASRC, which in turn submitted them to HSRC, where they were entered and edited by data processing personnel specially trained for this task. Data were entered using programmes written in CSPro by ORC Macro. All data were entered twice (100 percent verification). The data processing of the survey was completed in October 2004.
Data Appraisal
Estimates of Sampling Error
The sample of respondents selected in the 2003-2004 SADHS is only one of many samples that could have been selected from the same population, using the same design and expected size. Each of these samples would yield results that differ somewhat from the results of the actual sample selected. Sampling errors are a measure of the variability between all possible samples. Although the degree of variability is not known exactly, it can be estimated from the survey results.
A sampling error is usually measured in terms of the standard error for a particular statistic (mean, percentage, etc.), which is the square root of the variance. The standard error can be used to calculate confidence intervals within which the true value for the population can reasonably be assumed to fall. For example, for any given statistic calculated from a sample survey, the value of that statistic will fall within a range of plus or minus two times the standard error of that statistic in 95 percent of all possible samples of identical size and design.
If the sample of respondents had been selected as a simple random sample, it would have been possible to use straightforward formulas for calculating sampling errors. However, the SADHS sample is the result of a multi-stage stratified design, and, consequently, it was necessary to use more complex formulae. The computer software used to calculate sampling errors for the 2003-2004 SADHS is the ISSA Sampling Error Module. This module used the Taylor linearization method of variance estimation for survey estimates that are means or proportions. The Jackknife repeated replication method is used for variance estimation of more complex statistics such as fertility and mortality rates.
The Jackknife repeated replication method derives estimates of complex rates from each of several replications of the parent sample, and calculates standard errors for these estimates using simple formulae. Each replication considers all but one cluster in the calculation of the estimates. Pseudo-independent replications are thus created. In the SADHS, there were 621 non-empty clusters. Hence, 621 replications were created.
In addition to the standard error, ISSA computes the design effect (DEFT) for each estimate, which is defined as the ratio between the standard error using the given sample design and the standard error that would result if a simple random sample had been used. A DEFT value of 1.0 indicates that the sample design is as efficient as a simple random sample, while a value greater than 1.0 indicates the increase in the sampling error due to the use of a more complex and less statistically efficient design. ISSA also computes the relative error and confidence limits for the estimates. In the case of indicators from the adult health module, SAS has been used to calculate these parameters.
Sampling errors for the 2003-2004 SADHS are calculated for a few selected variables considered to be of primary interest. The results are presented in an appendix to the Final Report for the country as a whole and for urban and rural areas and for each province. For each variable, the type of statistic (mean, proportion, or rate) and the base population are given in Table B.1 of the Final Report. Tables B.2 to B.13 present the value of the statistic (R), its standard error (SE), the number of unweighted (N) and weighted (WN) cases, the design effect (DEFT), the relative standard error (SE/R), and the 95 percent confidence limits (R±2SE), for each variable. The DEFT is considered undefined when the standard error considering simple random sample is zero (when the estimate is close to 0 or 1).
The confidence interval (e.g., as calculated for had an HIV test and received results in the 12 months preceding the survey) can be interpreted as follows: the overall average from the national sample is 8.5 percent for women and its standard error is 0.005. Therefore, to obtain the 95 percent confidence limits, one adds and subtracts twice the standard error to the sample estimate, i.e., 0.085±2×0.005. There is a high probability (95 percent) that the true proportion of women age 15-49 in South Africa who had an HIV test and received the results in the 12 months prior to the survey is between 7.5 and 9.5 percent.
The relative standard errors (SE/R) for women at the national level range between 0.5 percent and 12.5 percent; the highest relative standard errors are for estimates of very low values (e.g., Had two or more sexual partners in last 12 months). In general, the relative standard error for most estimates for the country as a whole is small, except for estimates of very small proportions. Sampling errors are higher for subpopulations, e.g., urban and rural, than they are for the national population as a whole.
For the total sample, the value of the design effect (DEFT) for women at the national level, averaged over all variables is 1.36, which means that, due to multi-stage clustering of the sample, the average standard error is increased by a factor of 1.36 over that in an equivalent simple random sample.
A sampling error is usually measured in terms of the standard error for a particular statistic (mean, percentage, etc.), which is the square root of the variance. The standard error can be used to calculate confidence intervals within which the true value for the population can reasonably be assumed to fall. For example, for any given statistic calculated from a sample survey, the value of that statistic will fall within a range of plus or minus two times the standard error of that statistic in 95 percent of all possible samples of identical size and design.
If the sample of respondents had been selected as a simple random sample, it would have been possible to use straightforward formulas for calculating sampling errors. However, the SADHS sample is the result of a multi-stage stratified design, and, consequently, it was necessary to use more complex formulae. The computer software used to calculate sampling errors for the 2003-2004 SADHS is the ISSA Sampling Error Module. This module used the Taylor linearization method of variance estimation for survey estimates that are means or proportions. The Jackknife repeated replication method is used for variance estimation of more complex statistics such as fertility and mortality rates.
The Jackknife repeated replication method derives estimates of complex rates from each of several replications of the parent sample, and calculates standard errors for these estimates using simple formulae. Each replication considers all but one cluster in the calculation of the estimates. Pseudo-independent replications are thus created. In the SADHS, there were 621 non-empty clusters. Hence, 621 replications were created.
In addition to the standard error, ISSA computes the design effect (DEFT) for each estimate, which is defined as the ratio between the standard error using the given sample design and the standard error that would result if a simple random sample had been used. A DEFT value of 1.0 indicates that the sample design is as efficient as a simple random sample, while a value greater than 1.0 indicates the increase in the sampling error due to the use of a more complex and less statistically efficient design. ISSA also computes the relative error and confidence limits for the estimates. In the case of indicators from the adult health module, SAS has been used to calculate these parameters.
Sampling errors for the 2003-2004 SADHS are calculated for a few selected variables considered to be of primary interest. The results are presented in an appendix to the Final Report for the country as a whole and for urban and rural areas and for each province. For each variable, the type of statistic (mean, proportion, or rate) and the base population are given in Table B.1 of the Final Report. Tables B.2 to B.13 present the value of the statistic (R), its standard error (SE), the number of unweighted (N) and weighted (WN) cases, the design effect (DEFT), the relative standard error (SE/R), and the 95 percent confidence limits (R±2SE), for each variable. The DEFT is considered undefined when the standard error considering simple random sample is zero (when the estimate is close to 0 or 1).
The confidence interval (e.g., as calculated for had an HIV test and received results in the 12 months preceding the survey) can be interpreted as follows: the overall average from the national sample is 8.5 percent for women and its standard error is 0.005. Therefore, to obtain the 95 percent confidence limits, one adds and subtracts twice the standard error to the sample estimate, i.e., 0.085±2×0.005. There is a high probability (95 percent) that the true proportion of women age 15-49 in South Africa who had an HIV test and received the results in the 12 months prior to the survey is between 7.5 and 9.5 percent.
The relative standard errors (SE/R) for women at the national level range between 0.5 percent and 12.5 percent; the highest relative standard errors are for estimates of very low values (e.g., Had two or more sexual partners in last 12 months). In general, the relative standard error for most estimates for the country as a whole is small, except for estimates of very small proportions. Sampling errors are higher for subpopulations, e.g., urban and rural, than they are for the national population as a whole.
For the total sample, the value of the design effect (DEFT) for women at the national level, averaged over all variables is 1.36, which means that, due to multi-stage clustering of the sample, the average standard error is increased by a factor of 1.36 over that in an equivalent simple random sample.
Access policy
Contacts
| Name | URL | |
|---|---|---|
| DataFirst Support | support@data1st | Link |
Access conditions
Public use files, available to all
Citation requirements
Department of Health & Medical Research Council. Demographic and Health Survey 2003-2004 [dataset]. Pretoria: DOH & MRC [producer]. Cape Town: DataFirst [distributor], 2020.
Access authority
| Name | Affiliation | URL |
|---|---|---|
| South African Data Archive | National Research Foundation | www.nrf.ac.za/sada |
Metadata production
DDI Document ID
ddi-zaf-datafirst-dhs-2003-2004-v2
Producers
| Name | Affiliation | Role |
|---|---|---|
| DataFirst | University of Cape Town | Metadata Producer |
Date of Metadata Production
2020-05-06
DDI Document version
Version 5