The Tanzania Demographic and Health Survey (TDHS) is part of the worldwide Demographic and Health Surveys (DHS) programme, which is designed to collect data on fertility, family planning, and maternal and child health. The primary objective of the 1999 TRCHS was to collect data at the national level (with breakdowns by urban-rural and Mainland-Zanzibar residence wherever warranted) on fertility levels and preferences, family planning use, maternal and child health, breastfeeding practices, nutritional status of young children, childhood mortality levels, knowledge and behaviour regarding HIV/AIDS, and the availability of specific health services within the community.1 Related objectives were to produce these results in a timely manner and to ensure that the data were disseminated to a wide audience of potential users in governmental and nongovernmental organisations within and outside Tanzania. The ultimate intent is to use the information to evaluate current programmes and to design new strategies for improving health and family planning services for the people of Tanzania.

Abstract

The Tanzania Demographic and Health Survey (TDHS) is part of the worldwide Demographic and Health Surveys (DHS) programme, which is designed to collect data on fertility, family planning, and maternal and child health.

The primary objective of the 1999 TRCHS was to collect data at the national level (with breakdowns by urban-rural and Mainland-Zanzibar residence wherever warranted) on fertility levels and preferences, family planning use, maternal and child health, breastfeeding practices, nutritional status of young children, childhood mortality levels, knowledge and behaviour regarding HIV/AIDS, and the availability of specific health services within the community.1 Related objectives were to produce these results in a timely manner and to ensure that the data were disseminated to a wide audience of potential users in governmental and nongovernmental organisations within and outside Tanzania. The ultimate intent is to use the information to evaluate current programmes and to design new strategies for improving health and family planning services for the people of Tanzania.

Geographic coverage

National. The sample was designed to provide estimates for the whole country, for urban and rural areas separately, and for Zanzibar and, in some cases, Unguja and Pemba separately.

Analysis unit

• Households
• Children under five years
• Women age 15-49
• Men age 15-59

Kind of data

Sample survey data

Sampling procedure

The TRCHS used a three-stage sample design. Overall, 176 census enumeration areas were selected (146 on the Mainland and 30 in Zanzibar) with probability proportional to size on an approximately self-weighting basis on the Mainland, but with oversampling of urban areas and Zanzibar. To reduce costs and maximise the ability to identify trends over time, these enumeration areas were selected from the 357 sample points that were used in the 1996 TDHS, which in turn were selected from the 1988 census frame of enumeration in a two-stage process (first wards/branches and then enumeration areas within wards/branches). Before the data collection, fieldwork teams visited the selected enumeration areas to list all the households. From these lists, households were selected to be interviewed. The sample was designed to provide estimates for the whole country, for urban and rural areas separately, and for Zanzibar and, in some cases, Unguja and Pemba separately. The health facilities component of the TRCHS involved visiting hospitals, health centres, and pharmacies located in areas around the households interviewed. In this way, the data from the two components can be linked and a richer dataset produced.

See detailed sample implementation in the APPENDIX A of the final report.

Mode of data collection

Face-to-face

Research instrument

The household survey component of the TRCHS involved three questionnaires: 1) a Household Questionnaire, 2) a Women’s Questionnaire for all individual women age 15-49 in the selected households, and 3) a Men’s Questionnaire for all men age 15-59.

The health facilities survey involved six questionnaires: 1) a Community Questionnaire administered to men and women in each selected enumeration area; 2) a Facility Questionnaire; 3) a Facility Inventory; 4) a Service Provider Questionnaire; 5) a Pharmacy Inventory Questionnaire; and 6) a questionnaire for the District Medical Officers.

All these instruments were based on model questionnaires developed for the MEASURE programme, as well as on the questionnaires used in the 1991-92 TDHS, the 1994 TKAP, and the 1996 TDHS. These model questionnaires were adapted for use in Tanzania during meetings with representatives from the Ministry of Health, the University of Dar es Salaam, the Tanzania Food and Nutrition Centre, USAID/Tanzania, UNICEF/Tanzania, UNFPA/Tanzania, and other potential data users. The questionnaires and manual were developed in English and then translated into and printed in Kiswahili.

The Household Questionnaire was used to list all the usual members and visitors in the selected households. Some basic information was collected on the characteristics of each person listed, including his/her age, sex, education, and relationship to the head of the household. The main purpose of the Household Questionnaire was to identify women and men who were eligible for individual interview and children under five who were to be weighed and measured. Information was also collected about the dwelling itself, such as the source of water, type of toilet facilities, materials used to construct the house, ownership of various consumer goods, and use of iodised salt. Finally, the Household Questionnaire was used to collect some rudimentary information about the extent of child labour.

The Women’s Questionnaire was used to collect information from women age 15-49. These women were asked questions on the following topics: · Background characteristics (age, education, religion, type of employment) · Birth history · Knowledge and use of family planning methods · Antenatal, delivery, and postnatal care · Breastfeeding and weaning practices · Vaccinations, birth registration, and health of children under age five · Marriage and recent sexual activity · Fertility preferences · Knowledge and behaviour concerning HIV/AIDS.

The Men’s Questionnaire covered most of these same issues, except that it omitted the sections on the detailed reproductive history, maternal health, and child health. The final versions of the English questionnaires are provided in Appendix E.

Before the questionnaires could be finalised, a pretest was done in July 1999 in Kibaha District to assess the viability of the questions, the flow and logical sequence of the skip pattern, and the field organisation. Modifications to the questionnaires, including wording and translations, were made based on lessons drawn from the exercise.

Response rate

In all, 3,826 households were selected for the sample, out of which 3,677 were occupied. Of the households found, 3,615 were interviewed, representing a response rate of 98 percent. The shortfall is primarily due to dwellings that were vacant or in which the inhabitants were not at home despite of several callbacks.

In the interviewed households, a total of 4,118 eligible women (i.e., women age 15-49) were identified for the individual interview, and 4,029 women were actually interviewed, yielding a response rate of 98 percent. A total of 3,792 eligible men (i.e., men age 15-59), were identified for the individual interview, of whom 3,542 were interviewed, representing a response rate of 93 percent. The principal reason for nonresponse among both eligible men and women was the failure to find them at home despite repeated visits to the household. The lower response rate among men than women was due to the more frequent and longer absences of men.

The response rates are lower in urban areas due to longer absence of respondents from their homes. One-member households are more common in urban areas and are more difficult to interview because they keep their houses locked most of the time. In urban settings, neighbours often do not know the whereabouts of such people.

Sampling error estimates

The estimates from a sample survey are affected by two types of errors: (1) non-sampling errors, and (2) sampling errors. Non-sampling errors are the results of mistakes made in implementing data collection and data processing, such as failure to locate and interview the correct household, misunderstanding of the questions on the part of either the interviewer or the respondent, and data entry errors. Although numerous efforts were made during the implementation of the TRCHS to minimise this type of error, nonsampling errors are impossible to avoid and difficult to evaluate statistically.

Sampling errors, on the other hand, can be evaluated statistically. The sample of respondents selected in the TRCHS 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 TRCHS sample is the result of a two-stage stratified design, and, consequently, it was necessary to use more complex formulae. The computer software used to calculate sampling errors for the TRCHS is the ISSA Sampling Error Module (SAMPERR). This module used the Taylor linearisation 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 rate

Note: See detailed sampling error calculation in the APPENDIX B

Abstract

The Tanzania Demographic and Health Survey (TDHS) is part of the worldwide Demographic and Health Surveys (DHS) programme, which is designed to collect data on fertility, family planning, and maternal and child health.

The primary objective of the 1999 TRCHS was to collect data at the national level (with breakdowns by urban-rural and Mainland-Zanzibar residence wherever warranted) on fertility levels and preferences, family planning use, maternal and child health, breastfeeding practices, nutritional status of young children, childhood mortality levels, knowledge and behaviour regarding HIV/AIDS, and the availability of specific health services within the community.1 Related objectives were to produce these results in a timely manner and to ensure that the data were disseminated to a wide audience of potential users in governmental and nongovernmental organisations within and outside Tanzania. The ultimate intent is to use the information to evaluate current programmes and to design new strategies for improving health and family planning services for the people of Tanzania.

Geographic coverage

National. The sample was designed to provide estimates for the whole country, for urban and rural areas separately, and for Zanzibar and, in some cases, Unguja and Pemba separately.

Analysis unit

Households, individuals

Universe

Men and women 15-49, children under 5

Kind of data

Sample survey data

Sampling procedure

The TRCHS used a three-stage sample design. Overall, 176 census enumeration areas were selected (146 on the Mainland and 30 in Zanzibar) with probability proportional to size on an approximately self-weighting basis on the Mainland, but with oversampling of urban areas and Zanzibar. To reduce costs and maximise the ability to identify trends over time, these enumeration areas were selected from the 357 sample points that were used in the 1996 TDHS, which in turn were selected from the 1988 census frame of enumeration in a two-stage process (first wards/branches and then enumeration areas within wards/branches). Before the data collection, fieldwork teams visited the selected enumeration areas to list all the households. From these lists, households were selected to be interviewed. The sample was designed to provide estimates for the whole country, for urban and rural areas separately, and for Zanzibar and, in some cases, Unguja and Pemba separately. The health facilities component of the TRCHS involved visiting hospitals, health centres, and pharmacies located in areas around the households interviewed. In this way, the data from the two components can be linked and a richer dataset produced.

See detailed sample implementation in the APPENDIX A of the final report.

Mode of data collection

Face-to-face

Research instrument

The household survey component of the TRCHS involved three questionnaires: 1) a Household Questionnaire, 2) a Women’s Questionnaire for all individual women age 15-49 in the selected households, and 3) a Men’s Questionnaire for all men age 15-59.

The health facilities survey involved six questionnaires: 1) a Community Questionnaire administered to men and women in each selected enumeration area; 2) a Facility Questionnaire; 3) a Facility Inventory; 4) a Service Provider Questionnaire; 5) a Pharmacy Inventory Questionnaire; and 6) a questionnaire for the District Medical Officers.

All these instruments were based on model questionnaires developed for the MEASURE programme, as well as on the questionnaires used in the 1991-92 TDHS, the 1994 TKAP, and the 1996 TDHS. These model questionnaires were adapted for use in Tanzania during meetings with representatives from the Ministry of Health, the University of Dar es Salaam, the Tanzania Food and Nutrition Centre, USAID/Tanzania, UNICEF/Tanzania, UNFPA/Tanzania, and other potential data users. The questionnaires and manual were developed in English and then translated into and printed in Kiswahili.

The Household Questionnaire was used to list all the usual members and visitors in the selected households. Some basic information was collected on the characteristics of each person listed, including his/her age, sex, education, and relationship to the head of the household. The main purpose of the Household Questionnaire was to identify women and men who were eligible for individual interview and children under five who were to be weighed and measured. Information was also collected about the dwelling itself, such as the source of water, type of toilet facilities, materials used to construct the house, ownership of various consumer goods, and use of iodised salt. Finally, the Household Questionnaire was used to collect some rudimentary information about the extent of child labour.

The Women’s Questionnaire was used to collect information from women age 15-49. These women were asked questions on the following topics: · Background characteristics (age, education, religion, type of employment) · Birth history · Knowledge and use of family planning methods · Antenatal, delivery, and postnatal care · Breastfeeding and weaning practices · Vaccinations, birth registration, and health of children under age five · Marriage and recent sexual activity · Fertility preferences · Knowledge and behaviour concerning HIV/AIDS.

The Men’s Questionnaire covered most of these same issues, except that it omitted the sections on the detailed reproductive history, maternal health, and child health. The final versions of the English questionnaires are provided in Appendix E.

Before the questionnaires could be finalised, a pretest was done in July 1999 in Kibaha District to assess the viability of the questions, the flow and logical sequence of the skip pattern, and the field organisation. Modifications to the questionnaires, including wording and translations, were made based on lessons drawn from the exercise.

Response rate

In all, 3,826 households were selected for the sample, out of which 3,677 were occupied. Of the households found, 3,615 were interviewed, representing a response rate of 98 percent. The shortfall is primarily due to dwellings that were vacant or in which the inhabitants were not at home despite of several callbacks.

In the interviewed households, a total of 4,118 eligible women (i.e., women age 15-49) were identified for the individual interview, and 4,029 women were actually interviewed, yielding a response rate of 98 percent. A total of 3,792 eligible men (i.e., men age 15-59), were identified for the individual interview, of whom 3,542 were interviewed, representing a response rate of 93 percent. The principal reason for nonresponse among both eligible men and women was the failure to find them at home despite repeated visits to the household. The lower response rate among men than women was due to the more frequent and longer absences of men.

The response rates are lower in urban areas due to longer absence of respondents from their homes. One-member households are more common in urban areas and are more difficult to interview because they keep their houses locked most of the time. In urban settings, neighbours often do not know the whereabouts of such people.

Sampling error estimates

The estimates from a sample survey are affected by two types of errors: (1) non-sampling errors, and (2) sampling errors. Non-sampling errors are the results of mistakes made in implementing data collection and data processing, such as failure to locate and interview the correct household, misunderstanding of the questions on the part of either the interviewer or the respondent, and data entry errors. Although numerous efforts were made during the implementation of the TRCHS to minimise this type of error, nonsampling errors are impossible to avoid and difficult to evaluate statistically.

Sampling errors, on the other hand, can be evaluated statistically. The sample of respondents selected in the TRCHS 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 TRCHS sample is the result of a two-stage stratified design, and, consequently, it was necessary to use more complex formulae. The computer software used to calculate sampling errors for the TRCHS is the ISSA Sampling Error Module (SAMPERR). This module used the Taylor linearisation 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 rate

Note: See detailed sampling error

Abstract

The Tanzania Demographic and Health Survey (TDHS) is part of the worldwide Demographic and Health Surveys (DHS) programme, which is designed to collect data on fertility, family planning, and maternal and child health.

The general objectives of the 1996 TDHS are to: - Provide national-level data that will allow the calculation of demographic rates, particularly fertility and childhood mortality rates - Analyze the direct and indirect factors which determine the level and trends of fertility - Measure the level of contraceptive knowledge and practice (of both women and men) by method, by urban-rural residence, and by region - Collect reliable data on maternal and child health indicators; immunization, prevalence, and treatment of diarrhea and other diseases among children under age five; antenatal visits; assistance at delivery; and breastfeeding - Assess the nutritional status of children under age five and their mothers by means of anthropometric measurements (weight and height), and child feeding practices - Assess among women and men the prevailing level of specific knowledge and attitudes regarding AIDS and evaluate patterns of recent behavior regarding condom use - Measure maternal mortality and collect data on female circumcision.

Geographic coverage

The survey was designed to provide estimates (based on the results of the Woman's Questionnaire) for the whole country, for urban and rural areas in the country, and groups of regions (zones). In addition, the sample provides certain estimates for each of the 20 regions in the mainland and 2 subgroups in Zanzibar: Pemba Island and Ungaja.

In most regions, one in every four households was selected for the men's survey, and in six regions (Dares Salaam, Dodoma, Iringa, Kilimanjaro, Morogoro, and Shinyanga), men in every second household were selected for the interview. The sample of men was designed to provide estimates for the country as a whole and for urban and rural areas.

Analysis unit

• Households
• Children under five years
• Women age 15-49
• Men age 15-59

Kind of data

Sample survey data

Sampling procedure

The TDHS sample was a three-stage design consisting of the same 357 enumeration areas (EAs) that were used in the 1991-92 TDHS (262 EAs in rural and 95 EAs in urban areas). The selection of EAs was made in two stages: first, wards/branches and then EAs within wards/branches were selected. Lists of all households were prepared for the selected EAs and, at the third sampling stage; households were selected from these lists. The TDHS was designed to provide estimates (based on the results of the Woman's Questionnaire) for the whole country, for urban and rural areas in the country, and groups of regions (zones). In addition, the sample will provide certain estimates for each of the 20 regions in the mainland and 2 subgroups in Zanzibar: Pemba Island and Ungaja. In most regions, one in every four households was selected for the men's survey, and in six regions (Dares Salaam, Dodoma, Iringa, Kilimanjaro, Morogoro, and Shinyanga), men in every second household were selected for the interview. The sample of men was designed to provide estimates for the country as a whole and for urban and rural areas.

Unlike most other DHS surveys, households in Tanzania were selected from the household listing for each ward (or branch) on the basis of contiguity, beginning with a randomly selected start number. This selection process was used to minimize the difficulty encountered in moving from one selected household to another given the scattered nature of households.

See detailed sample design information in the APPENDIX A of the final 1996 Tanzania Demographic and Health Survey report.

Mode of data collection

Face-to-face

Research instrument

Three types of questionnaires were used during the survey. The Household Questionnaire was used to list the names of the household members and certain individual characteristics of all usual members of the household and visitors who had spent the previous night in the household. Certain basic information was collected on characteristics of each person listed, including relationship, age, sex, education, and place of residence. Furthermore, the Household Questionnaire collected information on characteristics relating to the household. These included the source of water, type of toilet facilities, materials used for the floor of the house, and ownership of various durable goods. However, the main purpose of the Household Questionnaire was to identify women and men who were eligible for the individual interview.

The Female Questionnaire was used to collect information from eligible women age 15-49. The topics covered in this questionnaire included the following: - Background characteristics of the woman including age, education, residential history - Reproductive history - Knowledge and use of family planning methods - Fertility preferences and attitudes about family planning - Antenatal and delivery care - Breastfeeding and weaning practices - Vaccinations and health status of children under age five - Marriage and sexual activity - Husband's occupation and education - Woman's employment, occupation, and earnings - Awareness and behavior regarding AIDS and other sexually transmitted diseases - Maternal mortality - Female circumcision - Height and weight of children under five years and their mothers.

The Male Questionnaire was used to collect information from a subsample of men age 15-59, namely, those living in every fourth household except in Dares Salaam, Dodoma, Kilimanjaro, Morogoro, Shinyanga, and Iringa regions where every second household was selected for the male interview. The Male Questionnaire collected much of the same information found in the Women's Questionnaire, but was shorter because it did not contain questions on reproductive history and maternal and child health. All questionnaires were translated and printed in Kiswahili.

Before the design of the questionnaires could be finalized, a pretest was done in May-June, 1996 to assess the viability of the questions, the flow and logical sequence of the skip pattern, and the field organization. It covered an area outside Dares Salaam and took about a week to complete. Modifications to the questionnaires were then made based on lessons drawn from the exercise.

Response rate

In all, 8,900 households were selected, out of which 8,141 were occupied. Of the households found, 7,969 were interviewed, representing a response rate of 98 percent. The shortfall between the selected and the interviewed households was largely because many dwellings were either vacant or no competent respondents were present at the time of the visit.

In the interviewed households, 8,501 eligible women (i.e. women age 15- 49) were identified for the individual interview, and 8,120 women were actually interviewed, yielding a response rate of 96 percent. In the subsample of households selected for the male interview, 2,658 eligible men (i.e., men age 15-59) were identified, 2,256 were interviewed, representing a response rate of 85 percent. The principal reason for nonresponse among both eligible men and women was the failure to find them at home despite repeated visits to the household. The lower response rates among men than women were due to the more frequent and longer absences of men.

The response rates are lower in urban areas. One-member households are more common in urban areas and are more difficult to interview because they keep their houses locked up most of the time. In urban settings, neighbors often do not know the whereabouts of such people.

Sampling error estimates

The estimates from a sample survey are affected by two types of errors: non-sampling errors, and sampling errors. Non-sampling errors are the results of mistakes made in implementing data collection and data processing, such as failure to locate and interview the correct household, misunderstanding the questions on the part of either the interviewer or the respondent, and data entry errors. Although numerous efforts were made during the implementation of the 1996 TDHS to minimize this type of error, non-sampling errors are impossible to avoid and difficult to evaluate statistically.

Sampling errors, on the other hand, can be evaluated statistically. The sample of respondents selected in the TDHS 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, straightforward formulae for calculating sampling errors could have been used. However, the TDHS sample is the result of a two-stage stratified design, and, consequently, it was necessary to use more complex formulae. The computer software that calculated sampling errors

Abstract

The Tanzania Demographic and Health Survey (TDHS) is part of the worldwide Demographic and Health Surveys (DHS) programme, which is designed to collect data on fertility, family planning, and maternal and child health.

The general objectives of the 1996 TDHS are to: - Provide national-level data that will allow the calculation of demographic rates, particularly fertility and childhood mortality rates - Analyze the direct and indirect factors which determine the level and trends of fertility - Measure the level of contraceptive knowledge and practice (of both women and men) by method, by urban-rural residence, and by region - Collect reliable data on maternal and child health indicators; immunization, prevalence, and treatment of diarrhea and other diseases among children under age five; antenatal visits; assistance at delivery; and breastfeeding - Assess the nutritional status of children under age five and their mothers by means of anthropometric measurements (weight and height), and child feeding practices - Assess among women and men the prevailing level of specific knowledge and attitudes regarding AIDS and evaluate patterns of recent behavior regarding condom use - Measure maternal mortality and collect data on female circumcision.

Geographic coverage

The survey was designed to provide estimates (based on the results of the Woman's Questionnaire) for the whole country, for urban and rural areas in the country, and groups of regions (zones). In addition, the sample provides certain estimates for each of the 20 regions in the mainland and 2 subgroups in Zanzibar: Pemba Island and Ungaja.

In most regions, one in every four households was selected for the men's survey, and in six regions (Dares Salaam, Dodoma, Iringa, Kilimanjaro, Morogoro, and Shinyanga), men in every second household were selected for the interview. The sample of men was designed to provide estimates for the country as a whole and for urban and rural areas.

Analysis unit

Households, individuals

Universe

Men and women between the ages of 15-49, children under 5

Kind of data

Sample survey data

Sampling procedure

The TDHS sample was a three-stage design consisting of the same 357 enumeration areas (EAs) that were used in the 1991-92 TDHS (262 EAs in rural and 95 EAs in urban areas). The selection of EAs was made in two stages: first, wards/branches and then EAs within wards/branches were selected. Lists of all households were prepared for the selected EAs and, at the third sampling stage; households were selected from these lists. The TDHS was designed to provide estimates (based on the results of the Woman's Questionnaire) for the whole country, for urban and rural areas in the country, and groups of regions (zones). In addition, the sample will provide certain estimates for each of the 20 regions in the mainland and 2 subgroups in Zanzibar: Pemba Island and Ungaja. In most regions, one in every four households was selected for the men's survey, and in six regions (Dares Salaam, Dodoma, Iringa, Kilimanjaro, Morogoro, and Shinyanga), men in every second household were selected for the interview. The sample of men was designed to provide estimates for the country as a whole and for urban and rural areas.

Unlike most other DHS surveys, households in Tanzania were selected from the household listing for each ward (or branch) on the basis of contiguity, beginning with a randomly selected start number. This selection process was used to minimize the difficulty encountered in moving from one selected household to another given the scattered nature of households.

See detailed sample design information in the APPENDIX A of the final 1996 Tanzania Demographic and Health Survey report.

Mode of data collection

Face-to-face

Research instrument

Three types of questionnaires were used during the survey. The Household Questionnaire was used to list the names of the household members and certain individual characteristics of all usual members of the household and visitors who had spent the previous night in the household. Certain basic information was collected on characteristics of each person listed, including relationship, age, sex, education, and place of residence. Furthermore, the Household Questionnaire collected information on characteristics relating to the household. These included the source of water, type of toilet facilities, materials used for the floor of the house, and ownership of various durable goods. However, the main purpose of the Household Questionnaire was to identify women and men who were eligible for the individual interview.

The Female Questionnaire was used to collect information from eligible women age 15-49. The topics covered in this questionnaire included the following: - Background characteristics of the woman including age, education, residential history - Reproductive history - Knowledge and use of family planning methods - Fertility preferences and attitudes about family planning - Antenatal and delivery care - Breastfeeding and weaning practices - Vaccinations and health status of children under age five - Marriage and sexual activity - Husband's occupation and education - Woman's employment, occupation, and earnings - Awareness and behavior regarding AIDS and other sexually transmitted diseases - Maternal mortality - Female circumcision - Height and weight of children under five years and their mothers.

The Male Questionnaire was used to collect information from a subsample of men age 15-59, namely, those living in every fourth household except in Dares Salaam, Dodoma, Kilimanjaro, Morogoro, Shinyanga, and Iringa regions where every second household was selected for the male interview. The Male Questionnaire collected much of the same information found in the Women's Questionnaire, but was shorter because it did not contain questions on reproductive history and maternal and child health. All questionnaires were translated and printed in Kiswahili.

Before the design of the questionnaires could be finalized, a pretest was done in May-June, 1996 to assess the viability of the questions, the flow and logical sequence of the skip pattern, and the field organization. It covered an area outside Dares Salaam and took about a week to complete. Modifications to the questionnaires were then made based on lessons drawn from the exercise.

Response rate

In all, 8,900 households were selected, out of which 8,141 were occupied. Of the households found, 7,969 were interviewed, representing a response rate of 98 percent. The shortfall between the selected and the interviewed households was largely because many dwellings were either vacant or no competent respondents were present at the time of the visit.

In the interviewed households, 8,501 eligible women (i.e. women age 15- 49) were identified for the individual interview, and 8,120 women were actually interviewed, yielding a response rate of 96 percent. In the subsample of households selected for the male interview, 2,658 eligible men (i.e., men age 15-59) were identified, 2,256 were interviewed, representing a response rate of 85 percent. The principal reason for nonresponse among both eligible men and women was the failure to find them at home despite repeated visits to the household. The lower response rates among men than women were due to the more frequent and longer absences of men.

The response rates are lower in urban areas. One-member households are more common in urban areas and are more difficult to interview because they keep their houses locked up most of the time. In urban settings, neighbors often do not know the whereabouts of such people.

Sampling error estimates

The estimates from a sample survey are affected by two types of errors: non-sampling errors, and sampling errors. Non-sampling errors are the results of mistakes made in implementing data collection and data processing, such as failure to locate and interview the correct household, misunderstanding the questions on the part of either the interviewer or the respondent, and data entry errors. Although numerous efforts were made during the implementation of the 1996 TDHS to minimize this type of error, non-sampling errors are impossible to avoid and difficult to evaluate statistically.

Sampling errors, on the other hand, can be evaluated statistically. The sample of respondents selected in the TDHS 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, straightforward formulae for calculating sampling errors could have been used. However, the TDHS sample is the result of a two-stage stratified design, and, consequently, it was necessary to use more complex formulae.