This is the first DHS consucted in Timor Leste. A second DHS was consucted in 2009-2010 on a larger sample.
The TL 2003 DHS is the first comprehensive assessment of the demographic, health and nutrition status of the population of this newly independent state. The province of East Timor was included in the samples of the 1991 and 1997 Indonesian Demographic and Health Surveys but only a small number of households were sampled. Following the establishment of the UN transitional administration, national planning efforts were constrained in part by the paucity of credible development planning information. Although the surveys conducted as part of the 2001 Poverty Assessment of Timor Leste resulted in improved development planning data, these surveys did not focus on the collection of the demographic, social, health and nutrition data required for planning and evaluating health care services.
The TL 2003 DHS aims to fill the information gap about the demographic, health and nutrition situation in the country, and provide the planning data needed to develop health services including reproductive health, maternal and child health, child spacing and nutrition programs.
Kind of Data
Sample survey data [ssd]
The TL 2003 DHS was designed to fulfil the following specific objectives:
- Provide data concerning the basic demographic profile of the country including fertility and infant and child mortality rates needed for development planning,
- Assess the nutrition and anaemia situation of the adult and child populations to plan and prioritize future interventions,
- Provide baseline information about health and nutrition status of the population and use of health services for future evaluation of national health programs,
- Provide baseline information about fertility and contraceptive prevalence rates to evaluate future child spacing programs,
- Analyse factors associated with fertility and use of family planning methods, maternal and child health, use of health services and adult morbidity,
- Analyse factors associated with malnutrition and anaemia especially in women and children,
- Assess men's participation and utilization of health services, and their role in their family's utilization of health services,
- Create standard demographic, health and nutrition indicator information to facilitate cross-country comparisons for the program managers, policymakers, and researchers to assess the level of development in Timor Leste.
The TL 2003 DHS was required to provide estimates of key demographic, health status and health service usage indicators at the national level, and for selected policy-relevant subgroups of the population based on geographic location (west, central and east), the nature of the district (urban versus rural), and major agro-ecologic zones (lowlands and highlands).
The sample was stratified to ensure that a sufficient number of households were visited by the survey team in each of the key analytic domains. Thus in the TL 2003 DHS survey, four sampling strata were used based on location (west, central or east) and nature of district (urban versus rural). These strata included the following districts:
- Urban, including cities of Dili and Baucau, and other smaller district towns2
- Rural West (Oecussi, Bobonaro and Cova Lima districts)
- Rural Central (Aileu, Ainaro, Dili3, Emera, Liquica, Manufahi and Manatuto districts)
- Rural East (Baucau, Lautem and Viqueque districts)
The ecological zones were defined by altitude above sea level, with highland households located at more than 1000 meters above sea level. Above 1000 meters the reduced levels of oxygen in the atmosphere begin to have an impact on haemoglobin levels and growth of children. The highland households were identified from readings of altitude taken by field survey supervisors at each respondent’s home using a geographic positioning (GPS) reader (Garmin Etrex). The consistency of the altitude readings for households within each aldeia was checked and any apparent errors in readings of household altitude were replaced with the average altitude for the aldeia. There is a very low percentage of households (<5%) in the urban and rural east located in the highland areas.
About 10% of households in the rural west are located in the highlands but the highest percentage of highland households (40%) is in the rural central region.
These strata and ecological zones differ from those used in the TLSMS in that the ratio of rural to urban households is greater, and more representative of the population distribution within these areas. They also differ from the implied strata used in the TL 2002 MICS, where rural west included the districts of Emera and Liquica, which had been transferred from the rural central region. Furthermore, in both TLSMS and the TL 2002 MICS, highland communities were defined using 500 meters as the cut-off value for high altitude. Thus in the TL 2003 DHS, the highland sample is about half the size of the earlier surveys although the data exists to exactly replicate the earlier ecological zones if needed for future analyses.
Producers and sponsors
Ministry of Health and National Statistics Office
ACIL Australia Pty Ltd
Implementation of the survey
University of Newcastle
Technical support / analysis
Australian National University
Technical support / analysis
A cross sectional household survey was conducted over a period of four months from May to August 2003 in which 4320 households were sampled from four different geographic areas in Timor Leste using a cluster sampling method.
To facilitate comparisons with other household surveys conducted in Timor Leste, the TL 2003 DHS sampling design used the same sampling frame and similar analytical domains to those developed in two of the surveys of the 2001 Poverty Assessment of Timor Loro Sa'e. The first was an administrative census, the SSTL Survey of Sucos in Timor Loro Sa'e (East Timor Transitional Administration, 2001a), which was conducted in all local administrative units across Timor Leste between February and April of 2001, and provided a sampling frame for national household surveys. The second was the TLSMS Timor Loro Sa'e Household Survey (East Timor Transitional Administration, 2001b), a household expenditure survey of a nationally representative sample of 1800 households, in which basic analytical domains were defined for surveys in Timor Leste.
Timor Leste is divided into 13 major units called distritos (districts), which are further subdivided into 67 postos (sub-districts), 498 sucos (villages) and 2336 aldeias (subvillages). The SSTL was an administrative census in which data were obtained on the number of households at suco and aldeia level throughout Timor Leste, based on reports of suco chiefs or other key local officials. The limitations of this sampling frame were reported in the TL 2002 MIC survey where a "tendency toward overstatement of both population and households in many areas" especially in Dili was noted. However, in the absence of census data the population estimates for suco and aldeia reported in the Suco Survey were used as the sampling frame for the TL 2003 DHS.
A stratified cluster sampling procedure was used to select the sample of households for the TL 2003 DHS. Within each of the four strata, a three stage sampling procedure was used, which was similar to the scheme employed in the TLSMS survey. In stage one, 40 sucos were selected within each strata, using a probability proportional to size (PPS) method. Three aldeias were then selected from each suco, again using a PPS method. In the third stage 10 households were selected from each aldeia, using the list of households for each aldeia obtained from the suco office as the frame and simple random sampling. With this procedure each household had a similar probability of selection, and within each stratum the sample was approximately self-weighted.
The lack of recent census data and well defined census tracks for Timor Leste provided limitations for the sampling procedure. The sampling procedure used the administrative units of aldeia as the final sampling segments but it is usually recommended (DHS Sampling Manual) that segments of approximately equal size are used. However, in the TL 2003 DHS sampling procedure no attempt was made to adjust for the variation in size of the aldeia.
Based on experience with other DHS surveys (DHS Sampling Manual), it was estimated that a sample size of at least 1000 women aged 15-49 years within each strata was required to provide strata specific estimates of key demographic and health outcomes, including (but not limited to) fertility, maternal and child mortality and morbidity, and health service utilisation.
The response rates for individual interviews with women (98.4%) were very high and similar across each stratum. The response rates for individual interviews with men (95.3%) were slightly lower than for women and varied by strata with lower response rates in the rural east (91.9%) and the urban (93.8%) regions.
As expected, the response rates for anthropometric and haemoglobin examinations in adults were slightly lower than for interviews. In children less than 60 months of age the response rate for anthropometry was higher (96.9%) than for haemoglobin examinations (92.8%). For both these types of examinations the response rates were lower in the rural east region, but especially for the haemoglobin examination (83.4%). Most of this lower level of participation in the haemoglobin examination occurred in a single district, Bacau, where the response rate for the child haemoglobin examination was only 75% suggesting local problems with engaging this community in this part of the survey.
The overall high survey response rates reflect the high level of community support in Timor Leste for activities seen to contribute to national development. However, the field procedures also contributed to these high response rates. The interviewers were required to repeatedly revisit sampled households where potential survey participants were temporarily absent, until the interviews were secured. No substitution of sampled households was allowed if the household members were not at home on the first visit. In some instances the field team needed to return to a sampled aldeia to finally secure the interviews. It appears that the addition of the biological measurements to the survey had no impact on the participation of men and women in the interviews.
Dates of Data Collection
Data Collection Mode
Data Collection Notes
The TL 2003 DHS was organized through a partnership of the Ministry of Health and the National Statistics Office, Ministry of Planning and Finance. The government of Timor Leste provided the funds for the survey through a grant from the World Bank as part of the Timor Leste, Health Sector Rehabilitation and Development Project. A steering committee was establish under the leadership of the Ministry of Health and included representatives from the Ministry of Health, the National Statistics Office, the World Health Organization, UNICEF, UNFPA, representatives of bilateral donor organizations including European Union and AusAID, and technical staff of the ACILmanaged TL 2003 DHS project team from the University of Newcastle, Newcastle, Australia and the Australian National University, Canberra, Australia. This committee reviewed technical issues related to the survey including implementation plans and all questionnaires prior to the start of the survey. The questionnaires were developed by the TL 2003 DHS project team through consultation with members of the staff of the Ministry of Health, the Steering Committee and other non-government organizations involved in health and population in Timor Leste.
The National Statistics Office together with the TL 2003 DHS project team implemented the survey data collection and data processing. A special survey team was recruited and trained and their work was supervised by the project team and staff of the National Statistics Office.
The TL 2003 DHS survey data were collected by six teams of field workers. Each field team consisted of one supervisor, one deputy supervisor, five male and five female interviewers and two drivers (all field survey team members are listed in Appendix E). In total, the field survey teams consisted of 84 people, of whom 35 women and 37 men participated in the survey as interviewers, physical examiners or field supervisors. The full survey team was trained over a two-week period from 7 April to 18 April 2003. The field supervisors received additional training in supervision, field data editing techniques, and use of GPS readers, and the anthropometrists and blood collectors received additional training in biological measurements from 19 April to 28 April 2003. Monitoring of fieldwork was conducted by six staff from the National Statistics Office and the project field manager.
Data collection took four months to complete starting 29 April 2003 and finishing 27 August 2003. Because of the difficulties in accessing the survey locations two four-wheel drive vehicles were required for each team. Approximately 10% of the sampled suco could not be reached by vehicles and the survey team had to walk in over two to three hours. About 25% of the sampled aldeia could only be accessed on foot or on horse back taking between 30 minutes and 2 hours. To facilitate the field team in reaching the correct locations, local guides were often provided by the suco chief. Field data collection teams stayed in the field during the data collection period, but returned to Dili every two weeks for debriefings with supervisors and to gather additional supplies and questionnaires. When in the field, the survey teams started by contacting the District Administrator and the Head of the District Health Office to inform them of their survey activities in that district. Prior to the arrival of the full survey team, the deputy team leader contacted the head of each suco to prepare accommodation for the survey team and to list all households in the selected aldeia. The local officials contacted the sampled households to inform them of the pending survey. On the day of the survey the full team worked in the aldeia and completed the interviews and biological measurements in one day. This allowed ample time for return visits if eligible household members were not available for interview or data errors needed correcting before the team left the suco.
The TL 2003 DHS used four questionnaires: the Household Questionnaire, the Women’s Questionnaire for ever-married women 15-49 years old, the Men’s Questionnaire for evermarried men 15-54 years old; and the Nutrition Measurements form. The survey instruments were based on the standard Demographic and Health Surveys (DHS) Model A Questionnaire for High Contraceptive Prevalence Countries4, the Timor Loro Sa’e Living Standards Survey (TLSMS) (World Bank 2002), and standard WHO recommended nutritional status indicators for women and pre-school aged children (WHO 1995).
Initially the survey team reviewed existing DHS survey instruments including the Indonesian 2002/2003 DHS questionnaires, the TLSMS questionnaire and other instruments used in recent surveys in Timor Leste. The standard DHS questionnaires were modified to respond to the unique social, economic, historical and epidemiological situation in East Timor. The draft questionnaires were adjusted to reflect the health system in Timor Leste, and cover the key health and family planning issues through a process of consultation with members of the TL 2003 DHS Steering Committee, staff of the Ministry of Health and the National Statistics Office, and other stakeholders from organizations likely to use the survey results.
The questionnaires were drafted in English and Indonesian. The Indonesian version was used in field work with interviewers translating it into Tetum or other local languages as needed. The layout and formatting of the questionnaires followed the usual DHS standard but was optimized to facilitate use of the instruments by the field team during interviewing and field data editing in remote locations.
The Household Questionnaire was used to list and collect information on all the usual household members and visitors who stayed the night before the survey as well as their shared household level characteristics. The basic information collected for each person listed included: age, sex, education, and relationship to the head of the household. Also recorded were household facilities and assets, such as the source of water, type of toilet facilities, construction materials used for the floor and outer walls of the house, and ownership of various durable goods, which reflect the household’s socioeconomic status. Additional information was gathered on household food security and household access to and use of health services. All of this information was solicited from the head of household or other responsible adult who usually resided in the household. A further function of the Household Questionnaire was to identify the adults eligible for individual interview, and the women, men and children eligible for measurement of anthropometry and haemoglobin.
The Women’s Questionnaire was used to collect information from all ever-married women aged 15-49 in the sampled households. These women were asked questions on the following topics: age, language use, religion, education, and media exposure in the section ‘Respondent’s Background’; a full birth history in the section on ‘Reproduction’; marital status and recent sexual activity in the section ‘Marriage and Sexual Activity’; knowledge and practice of family planning methods in the section ‘Knowledge and Practice of Child Spacing’; the women’s experiences in pregnancy, delivery and postnatal care and her infant feeding practices for all births since 1998 in the section ‘Antenatal, Postnatal Care and Breastfeeding’; details on immunizations and disease history and recent food intake for all births since 1998 in the section ‘Immunization, Child Health and Nutrition’; desire for more children and plans for use of family planning methods in the section ‘Fertility Preferences’; her recent work history and type of income in section on ‘Woman’s Work’; awareness and behaviour regarding AIDS and other sexually transmitted infections in the section ‘AIDS and Other Sexually Transmitted Diseases’; and recent functional morbidity experience and symptoms of tuberculosis in the section ‘Adult Morbidity’. In the societies of Timor, marriage is often a multi-stage process, and thus details of when cohabitation, legal marriage, church marriage, and traditional marriage celebrations took place (if at all) were included in the section ‘Marriage and Sexual Activity’.
The Men’s Questionnaire was used to collect information, similar to that collected in the Women’s Questionnaire but with less detail, from all ever married men aged 15-54 in every household. The men were asked questions on the following topics: age, language use, religion, education, and media exposure in the section ‘Respondent’s Background’; a brief history of the man’s reproductive history in the section on ‘Reproduction’; marital status and attitudes to women in the section ‘Marriage and Attitudes to Women’; knowledge and practice of family planning methods in the section ‘Knowledge and Attitude on Family Planning’; knowledge and participation in the pregnancy and healthseeking practices for their youngest living child born since 1998 in the section ‘Pregnancy, Postnatal Care, Breastfeeding and Child Health’; desire for more children and plans for use of family planning methods in the section ‘Fertility Preferences’; awareness and behaviour regarding AIDS and other sexually transmitted infections in the section ‘AIDS and Other Sexually Transmitted Diseases’; and recent functional morbidity experience and symptoms of tuberculosis in the section ‘Adult Morbidity’.
The Nutrition Measurements form was used to record the anthropometric and haemoglobin measurements of adults and children. For adults the following items were recorded: consent for the measurement, age in years, weight, height, left upper arm circumference, haemoglobin (for women), and referral information. For children the following items were recorded: parental consent for the measurement, date of birth, weight, height, left upper arm circumference, haemoglobin, and referral information.
All forms were numbered and packaged according to cluster before the survey teams went to the field and this method of organizing the forms was maintained until data entry and archiving of the forms. A key element of the data management system was the process of field data editing, because the very difficult terrain precluded the field team from re-interviewing respondents once they left the sampled suco. To facilitate field editing, checklists were prepared to guide the interviewers and field team leaders as to how to edit the forms. These daily checks aimed to detect missing values, range errors, invalid values, errors with skips and internal inconsistencies.
The Census and Survey Processing System (CSPro) version 2.2 was used for entry, editing, and dissemination of census and survey data. The common CSPro procedure language was used to create data dictionaries, data entry screens with error checks, and batch and error check programs. External programs, linked to the standard CSPro program and data, computed the anthropometric indices and exported the dual language data dictionaries. Data entry was completed by four data entry staff and one supervisor who were trained over a five day period. The survey data has a hierarchical data structure. The first level is the household questionnaire data and the biological measurements and the second level is the data from the women’s and men’s interview. A single questionnaire was completed for each household in the TL 2003 DHS sample, but the number of women’s or men’s questionnaires completed depended on the number of eligible women or men listed in the household questionnaire. Thus, for each household questionnaire there may be none or
several of the individual interview questionnaires.
Within each level, there were one or more different types of records. In the first level, there were single household records (eg household characteristics) or multiple household records (eg household members listing, anthropometric and haemoglobin measurements). In the second level in the women’s data, there were single records (eg the woman’s individual characteristics or health behaviours) or multiple records (eg the listing of her births).
The CSPro data entry program followed this hierarchical data structure. Data entry started with the household questionnaire, and then followed with the other forms at the second level. This approach maintained the integrity of the data structure and ensured accurate linkage between levels and records. The data files produced were a mirror of the paper questionnaires. Each section of the questionnaire was defined as a record in the data file. The data files were stored as ASCII text files. Data dictionaries in English and Indonesian described the data files, and in the data screens the documentation appeared in both languages. Special purpose programs were developed to export the data dictionaries in English and Indonesian. Also, at data entry, a special purpose program calculated anthropometric indices for children using the WHO international growth reference and computer subroutines provided by the US Center for Disease Control (CDC). After data cleaning and correcting for anthropometric measurements and the child’s age, these indices were recalculated. The data were read into the Nutrition module of Epi Info program and the indices were then calculated for both 1978 WHO and 2000 CDC growth references.
Estimates of Sampling Error
Based on the planned sample size, the precision of various outcomes within strata can be estimated ignoring the effects of the cluster sampling design. For example, for dichotomous outcomes, the precision of estimates will vary between 1% and 3% for outcomes for women 15-49 years, between 2-3% for children under 5 years of age and between 1.5% and 4% for children under 3 years of age, depending on the expected prevalence. For continuous outcomes, the precision will be approximately 0.06 of a standard deviation for outcomes for women 15-49 years and for children under 5 years of age and approximately 0.07 of a standard deviation for children under 3 years.
However these precision estimates change if we take into account the likely affect of the cluster sampling design. For example with a design effect of 2, the precision of dichotomous outcomes will vary between approximately 2.5% and 4% for outcomes for women 15-49 years, between 2.2% and 5% for children under 5 years of age and between 3.3% and 5% for children under 3 years of age. For continuous outcomes, the precision will be approximately 0.08 of a standard deviation for outcomes for women 15-49 years and for children under 5 years of age and approximately 0.10 of a standard deviation for children under 3 years.
The estimated sample size for TL 2003 DHS has adequate power to detect important differences between strata for a variety of survey outcomes. For example, a sample size of 543 children in each stratum would be sufficient to detect a 10% difference in low height for age between two strata if the higher prevalence was 55%. The actual number of children under 3 years of age is more than sufficient to detect this difference.
Use of the dataset must be acknowledged using a citation which would include:
- the Identification of the Primary Investigator
- the title of the survey (including country, acronym and year of implementation)
- the survey reference number
- the source and date of download
Disclaimer and copyrights
The user of the data acknowledges that the original collector of the data, the authorized distributor of the data, and the relevant funding agency bear no responsibility for use of the data or for interpretations or inferences based upon such uses.