Demographic

Of the 25,594 siblings reported in the sibling histories of SDHS respondents, survival status wasnot reported for 32 (less than 0.1 percent). Among surviving siblings, current ages (used to estimateexposure to death) were reported for virtually all surviving siblings (99.5 percent). Among deceasedsiblings, complete reporting of age at death and years since death was also nearly universal. For 96percent of deceased siblings, both age at death and years since the death (or year of death) were reported.Both age at death and the years since death were missing for only around 1 percent of deceased siblings.Rather than exclude siblings with missing data from further analysis, information on the birth order ofsiblings in conjunction with other information was used to impute the missing data. 1 The siblingsurvivorship data, including cases with imputed values, were used in the direct estimation of adult andmaternal mortality.15.2 DIRECT ESTIMATES OF ADULT MORTALITYThe direct approach to estimating adult mortalityemploys information on the age of surviving siblings, the age atdeath of siblings who died, and the number of years since thesibling died. This approach allows the data to be aggregated todetermine the number of person-years of exposure to mortalityrisk and the number of sibling deaths occurring in definedcalendar periods. Adult mortality rates are obtained by dividingfemale or male adult deaths in the given calendar periods byperson-years of exposure to death.Table 15.2 shows age-specific mortality rates for menand women age 15-49 for the period 0-6 years before the 2006-07 SDHS. These results allow an assessment of the recent levelof mortality in the population of reproductive age in Swaziland.Because the number of deaths on which the age-specific ratesare based is not very large (between 75 and 340 unweighteddeaths per age group for the total population), the estimatedage-specific rates are subject to considerable sampling variation.The results in Table 15.2 indicate that overall mortalityis virtually identical among women and men in the reproductive-agepopulation (14.4 deaths per 1,000 years of exposure,respectively). Considering the pattern within age groups,female mortality is substantially higher than male mortalityamong those under age 30. The pattern is reversed for oldermen and women, with male mortality levels being higher thanfemale mortality, particularly in the 40-49 age group (Figure 15.1).Table 15.2 Adult mortality ratesAge-specific mortality rates for women and menage 15-49 based on the survivorship of sistersand brothers of survey respondents for the period0-6 years preceding the survey, Swaziland2006-07Age Deaths ExposureWOMENMortalityrates15-19 49 11,972 4.120-24 146 12,335 11.925-29 218 10,651 20.530-34 155 8,644 18.035-39 139 6,383 21.840-44 80 4,074 19.745-49 54 2,399 22.5Total 841 56,458 14.4 aMEN15-19 26 11,960 2.220-24 72 12,688 5.725-29 127 10,481 12.130-34 187 8,399 22.235-39 151 6,230 24.240-44 127 3,889 32.545-49 83 2,231 37.015-49 771 55,878 14.4 aaRates are age-standardised.1 The imputation procedure is based on the assumption that the reported birth ordering of siblings in the history iscorrect. The first step is to calculate birth dates. For each living sibling with a reported age and each dead siblingwith complete information on both age at death and years since death, the birth date was calculated. For a siblingmissing these data, a birth date was imputed within the range defined by the birth dates of the bracketing siblings. Inthe case of living siblings, an age was then calculated from the imputed birth date. In the case of dead siblings, ifeither the age at death or years since death was reported, that information was combined with the birth date toproduce the information missing. If both pieces of information were missing, the distribution of the ages at death forsiblings for whom the years since death was unreported, but age at death was reported, was used as a basis forimputing the age at death.240 | Adult and Maternal Mortality