DHS Maternal Mortality ModuleTool Name: DHS Maternal Mortality Module A technique for the direct calculation of maternal mortality estimates Origin / Source: The module is an extension of the Indirect Sisterhood Method for estimating levels of maternal mortality (Graham, et. al., 1989). Data collection is more detailed than is necessary for estimation with the Indirect Sis terhood Method. The more detailed data allow the direct estimation of maternal mortality levels for well-defined time periods. Technical Area: Maternal Health Source: Demographic and Health Survey Project Basic Description: The module is a set questions designed for use as an add-on to a household survey. Respondents can be adult females, adult males or both in the age range from about 15 to 49 or 59. The module has primarily been used with female respondents in selected DHS surveys. The selected surveys have been national in scope, large in size (5,000 to 10,000 respondents) and pertain to populations experiencing high levels of fertility and maternal mortality. The module obtains information about the sex and survivorship of all of a respondent's siblings.(1) Information is also collected about their current age, age at death, number of years since death and, in the case of deceased sisters, if the death occurred during pregnancy, childbirth or within two months of childbirth. From this information, the number of pregnancy related deaths are obtained and those deaths are converted to various maternal mortality indicators (e.g., the maternal mortality rate or the maternal mortality ratio) applicable to a specific number of years prior to a survey. Maternal mortality rates can be estimated for specific time periods preceding the survey and frequently have been for the periods 0-6 and 7-13 years prior to the survey. The estimates suffer from imprecision due to both sampling and nonsampling error. In most of applications of the module, the estimate for the earlier time period is lower than that for the recent time period, presumably due to underreporting of events, and is considered particularly unreliable. Country Applications:
Languages Available: English, French and Spanish Technical Scope/Purpose: DHS currently employs the module primarily to produce a single national-level estimate for a time period immediately preceding a survey, usually 0-6 years before the survey. Early descriptions of the module suggested that it might be useful for documenting time trends and differentials in maternal mortality (Rutenburg and Sullivan, 1990). Subsequent experience has shown that, due to large sampling error, those expectations were overly optimistic (Stanton, et. al., 2000). Design: The module is meant to be administered in a cross-sectional household survey in which fieldwork typically covers a period of 3 to 5 months. Maternal mortality indicators are published at the time of the final survey report which is usually nine months after the end of fieldwork. Thus, results are available about a year after the start of a survey. Method: quantitative Frequency of Administration: Because the pace of change in maternal mortality in a country is generally very slow and because the indicators produced by the module are imprecise, it is recommended that the module should be applied in a particular population no more frequently than every ten years (WHO/UNICEF, 1997 and Stanton, 2000). Key Users of the Information: National policymakers, MOH program managers, family planning service providers. Objectives and Scope of the Tool: The module has the limited objective of producing an estimate for a single time period, say 0-6 years immediately preceding a survey. However, even those suffer from the problem of relatively large sampling error, on average about 15 percent of the estimated rate. This converts to confidence intervals that are, on average, plus or minus 30 percent of the estimated rate. (Stanton, 1997 and 2000). Of course, sampling imprecision could be mitigated by samples larger than the 5,000-10,000 female respondents typically included in DHS surveys. But that would be costly. Moreover, in some surveys there is evidence suggesting underrporting of maternal deaths by respondents. The combination of imprecision and possible bias from event underreporting precludes using the estimates for evaluating the impact of interventions or monitoring time trends in maternal mortality. There is a consensus that the most appropriate use of the estimates is for advocacy (i.e., to raise awareness of the problem of maternal mortality in a country). When considering adding the module to a survey, it is also important to recognize what the module does not provide. The module does not provide the type of detailed information which is required to
a) identify the causes of maternal mortality, If the identification of problem areas in the delivery of reproductive health services or the evaluation of interventions is desired, studies yielding indicators of the quality of those services should be pursued. Key Indicators: The module produces an estimated number of pregnancy-related deaths to the sisters of respondents and women-years lived during the childbearing years. Those data are easily transformed into the following maternal mortality indicators.
Tool Design: The module requires that respondents report a sibling history, including; sex, current age (living siblings), the age at death and years since death (deceased siblings). For deceased sisters, additional questions are asked to determine if death occurred during pregnancy, childbirth or within two months postpartum. There is a possibility of over counting the number of maternal deaths with this "time-of-death" definition but that tendency is counterbalanced, albeit to an unknown extent, by event underreporting (e.g., maternal deaths following induced abortion). NOTE: The denominator of the MMRatio, which is the most frequently cited indicator of maternal mortality, is live births. When using this module in a DHS-type questionnaire, the data on live births are derived from the birth history section of the DHS questionnaire and not from the maternal mortality module. Implementation and Training: All DHS interviewers are trained to administer the module. Training requires about half a day, inclusive of time for practice interviewing. The DHS Women's questionnaire generally requires about an hour of interviewing time. The module usually requires two to three minutes interviewing time for each of the respondent's siblings. Thus, the module can add 8 to 10 minutes of interviewing time per respondent. Manuals and Guidelines: Rutenburg N. and Sullivan, J. M. 1991. Direct and Indirect Estimates of Maternal Mortality from the Sisterhood Method. Proceedings of the Demographic and Health Surveys World Conference, August 5-7 1991, Washington, D.C., Vol. III/1669-1697. IRD/MACRO International Inc., Columbia , Maryland. Marckwardt, A. M. 1993. Illustrative Analysis: Maternal mortality in Peru. DHS manuscript. Measure/DHS+ Project, MACRO International Inc., Calverton, Maryland. WHO/UNICEF, 1997. The Sisterhood Method for Estimating Maternal Mortality: Guidance notes for potential users. WHO Division of Reproductive Health (technical Support), Family and Reproductive Health (WHO/RHT/97.28 and UNICEF/EPP/(97.1). Stanton, C. Abderrahim, N. and Hill, K. 1997. DHS maternal mortality indicators: An assessment of data quality and implications for data use. DHS Analytical Reports No. 4. Calverton, Maryland: Macro International Inc. Stanton, C. Abderrahim, N. and Hill, K. 2000. An Assessment of DHS Maternal Mortality Indicators. Studies in Family Planning31/2:111-123. Data Processing and Analysis: The preparation of the data for calculating mortality rates is a complex and tedious task and as the interpretation of the results of analysis is also demanding and requires judgement on the part of the analyst. The first step is the calculation the month and year of birth for each of a respondent's siblings from the information on current age, age at death and number of years ago of death. When all the data on age and years ago of death are present, this is involves setting narrow boundary values for the month and year of birth. When information is missing, relatively wide boundary values are set utilizing information about the dates of birth of other siblings. All boundary values for a sibship are then adjusted, if necessary to avoid overlapping ranges. Final birth dates for sibship members are imputed as the midpoint of the adjusted boundary values. For deceased siblings, month and year of death is imputed after setting possible boundary values from information on age at death or years since death. The statistic derived from the data on maternal deaths is the MMRate. The data on sisters are organized to provide, for specific time periods, the number of maternal deaths and sister-years lived in the seven five-year age intervals of the reproductive years. Age-specific maternal mortality rates are compute and are combined into an overall rate by weighting (Marckwardt, 1993). The weighting distribution is typically taken to be the distribution of the females in the DHS Household Schedule or the Wo men's Questionnaire. Other indicators can also be obtained (e.g. MMRatio = MMRate/General Fertility Rate; Lifetime Risk of Maternal Death and Proportion of Adult Female Deaths due to Maternal Causes). Reporting and Dissemination of Results: The estimates of maternal mortality indicators are published in the DHS Final Survey Reports. Results are highlighted in the executive summary of those reports and at the national seminars for each country. Care should be taken when reporting results to indicate that the estimates are subject to large sampling error. When possible an indication of the magnitude of sampling error and/or the confidence interval associated with an estimate should be presented. Cost: The overall cost of a DHS survey is increased by approximately 15 percent by adding this module. However, costs other than monetary must be considered before including adding this module in a survey. The additional interviewing time per respondent and the difficulty of obtaining some information (e.g., age at death of deceased siblings) may frust rate interviewers and result a general deterioration of the quality of all survey data. On the other hand, in addition to a maternal mortality estimate the module produces overall adult mortality estimates for a population. In countries where little is known about adult mortality and it is a priority issue, this may be an consideration in favor of using the module. Lessons from Experience: Results obtained to date from this module have not been as useful as other DHS statistics, say estimates of infant and child mortality. This is also true for other techniques for producing maternal mortality estimates. A serious constraint for this module has been the level of imprecision of the estimates (arising from both sampling variability and nonsampling error). The level of imprecision is large both in an absolute sense and relative to the amount of change one can plausibly expect to see at a national level over periods of 5 or 10 years. This situation is one of the reasons why maternal mortality indicators are not useful for program monitoring. On the other hand, this module has provided a substantial number of countries, which had no national information on maternal mortality, with their first empirically-based estimate. This method is recommended for countries facing that situation. MMRatios are very valuable advocacy tools, and have been shown to be effective at starting a national dialogue on the issue of maternal survival. As mentioned above, the problem lies in monitoring trends in maternal mortality. Given the constraints of this method, the use of the module to generate an approximate estimate of maternal mortality indicators at most at intervals of 10 years appears justified. Contact: 1. As well as estimates of maternal mortality, the module also yields estimates of adult mortality; an issue not further discussed here. |