Source: WHO (2004): Annex Table 2
127. Out of all of the differences that stand out from the previous listing, two have been of special concern: violence and accidents, as a major cause of male over-mortality, and maternal mortality as a cause of mortality that is specific to women. Worldwide, intentional injuries make about 750,000 more male than female victims annually; the difference with respect to unintentional injuries (accidents) is 1.2 million. Male disadvantage with respect to violent deaths is particularly evident in the countries of the former Soviet Union and in much of Latin America. Gavrilova et al. (2000) comment, for instance, on the overall rise of mortality that took place in Russia between 1991 and 1994 as a result of the tumultuous transition from a socialist to a market economy and the devastating effect that this had on male mortality rates from violent causes. Female mortality from these causes also increased, but to a lesser extent, thereby exacerbating a male-female difference which was already among the largest in the world at the time. In particular, male suicide rates increased from 47.7 in 1991 to 76.9 per 100,000 in 1994, as the corresponding female rates increased from 11.2 to 13.6. What this suggests is that men were more psychologically affected by the uncertainties surrounding the economic transition than women. Similarly, deaths due to alcohol poisoning - always a problem in the former Soviet Union (see Simpura et al., 1998, for an account on the Baltic states) - multiplied, from 19.4 to 61.2 per thousand, in the case of men, and from 4.2 to 15.8 in the case of women, whereas male homicide rates increased from 25.1 to 52.8 per thousand, as female rates went up from 6.9 to 13.6.
128. Male over-mortality from violent causes, particularly homicides, has also been a major issues in some Latin American countries, such as Brazil. In 2007, there were 45,554 registered homicides in Brazil, 92.1 per cent of which were of male victims, especially men between the ages of 15 and 40 (Isfeld, 2010). In some more developed countries (Croatia, Germany, Hungary, Japan, Republic of Korea, Slovenia, Switzerland), on the other hand, the number of male and female victims is roughly equal. There is a moderately strong positive relationship between the level of the overall homicide rate in a country and the percentage of victims that are male. In those countries in which data exist, there is also evidence that the majority (about 90 per cent globally) of perpetrators of homicides are males (UNODC, 2011: Fig. 5.12). Homicides in which both the victim and the perpetrator are female are quite rare, e.g. 2.6 per cent in the US (UNODC, 2011: 72). Whereas men are likelier to be killed in a public place, female victims are murdered mainly at home, as is the case in Europe, where half of all female victims were killed by a family member. The overwhelming majority of victims of violence committed by partners and family members are women. In Europe, for example, women accounted for almost 80 per cent of the total number of persons killed by a current or former partner in 2008. There is a general sense in the literature that the gender determinants of violent cause of death are under-studied and that they are too easily attributed to the innate aggressiveness of males. However, advancing in this area based on census data is difficult due to the fact that censuses provide no or only minimal cause-specific mortality data. Two censuses that did attempt to obtain some level of cause-specific mortality data are the 2008 census of Cambodia and the 2010 census of Zambia. The latter included the following cause categories: a) Accident; b) Injury; c) Suicide; d) Spousal violence; e) Other violence; f) Sickness/disease; g) Witchcraft; and h) Other.
129. Studying maternal mortality based on census data, while not ideal, is more viable. Globally, an estimated 287,000 maternal deaths occurred in 2010 (WHO/UNICEF/UNFPA/World Bank, 2012). Although maternal mortality is only the 20th most common cause of death for women of all ages worldwide, it is the most important cause of death for women of reproductive age (usually taken as the age range 15-49) in many developing countries. In addition, like violent causes of death, it is eminently amenable to prevention. Maternal mortality by itself is not considered a gender indicator. That does not mean that it has no linkages with gender, but rather that it is an outcome to which gender factors contribute15. One publication on UNFPA’s website states: “Preventable maternal mortality occurs where there is a failure to give effect to the rights of women to health, equality and non-discrimination. Preventable maternal mortality also often represents a violation of a woman’s right to life” (Hunt and Bueno de Mesquita, xxxx). Yet, there is little empirical evidence on the extent to which gender factors contribute to maternal mortality. A detailed discussion on this subject is beyond the scope of this manual, but one set of results may serve to illustrate the nature of the relationships.
Table 8: Strength and significance of trends calculated using polynomial regression analysis for all variables in the study
Source: McAlister and Baskett (2006)
130. The element to note in the above table is the relatively poor performance of “pure” gender indicators as predictors of maternal mortality, as compared to indicators that reflect overall level of development. Of particular note is the finding that the Human Development Index scores very high, with an R2 of 81.2, and that this improves to 82.9 with the Gender-related Development Index. Thus, gender is shown to be a dimension of maternal mortality, but not the principal one.
131. A different perspective, but resulting in similar conclusions, is provided through the ”Three Delays” model. This model proposes that pregnancy-related mortality is overwhelmingly due to delays in:
1) Deciding to seek appropriate medical help for an obstetric emergency;
2) Reaching an appropriate obstetric facility; and
3) Receiving adequate care when a facility is reached.
1) Their number of Children Ever Born alive (CEB);
2) Children born during the past 12 months before the census; and
3) Survival of Children Ever Born alive.
133. As was indicated in the previous chapter, most countries ask for this information disaggregated by sex of the child, but there are still a few countries where this information is not available. In countries that disaggregate the basic fertility and mortality data by sex, important information can be obtained about the sex ratio at birth and on differential mortality between young girls and boys. This issue, although directly related to fertility, will be discussed in the next sub-chapter. Typically, the information from questions 1) and 2), disaggregated by the age of the mother, is combined to estimate fertility, whereas 1) and 3) (more rarely 2) are combined for the purpose of mortality estimation. In addition, some censuses ask about the survival of the last child born or children born in the past 12 months.
134. A limitation of this method is that it can only provide information for mortality levels up to age 15 or 20. That means that mortality levels at higher ages (including the life expectancy) have to be estimated based on extrapolations, using typical relations between the mortality under age 20 and at higher ages. Such extrapolations contain a good deal of uncertainty and consequently the life expectancy estimates for many developing countries (including the sex differential) need to be treated with caution.
135. Some censuses have additional questions that serve primarily to complement the information on early mortality by adult mortality estimates. One such question is the orphanhood question, which asks members of the household whether their mother, father or both are still alive. Based on the age of the respondent and typical fertility patterns in the country, this allows the estimation of probabilities of death for the parents. A limitation of this method is that the estimates obtained in this manner refer to deaths that occurred at any time during the birth of the respondent and the present. Especially in the case of older respondents, these estimates can be quite distinct from current mortality levels. There is also the possibility that parents live in unspecified areas different from the current residence of the respondent, thereby making it difficult to use the information for sub-national mortality estimates. This limitation also applies to the infant and child mortality estimates of the previous paragraph, but the potential bias is more serious in the case of adult mortality. For all of these reasons, the questions on orphanhood are generally not considered very effective and only about 25 countries currently include them in their censuses.
136. Rather than asking about the parents, another option is to ask about the survival of sisters of adult members of the household. There are two variants if this method. In the direct sisterhood method, which is the standard method used in the DHS, the detection of deaths of sisters is followed up by more detailed questions about the year in which the date occurred and the age of the sister at the time. This method, however, is too laborious for most censuses which use the indirect sisterhood method, in which only the age of the respondent is used and the remaining information is attributed based on averages. This makes the indirect variant much less efficient than the direct variant. Although the sisterhood method can be used to estimate adult (female) mortality in general, its more typical use is the estimation of maternal mortality, in which it has to be combined with follow-up questions about the likely cause of death. However, as will be explained below, its use in censuses for this purpose is generally not recommended.
137. The other major type of question that can be used to measure adult mortality is the one that asks about the age and sex of members of the household that died during the past 12 months or another appropriate reference period. The most common problem with this question is that it tends to systematically under or (more rarely) over-estimate mortality due to factors such as the following:
Confusion about the reference period (e.g. current calendar year, rather than past 12 months);
Confusion about the meaning of “household”, as opposed to “family” or “community”; or
Confusion about the meaning of “belonging to this household”, especially in the case of prolonged hospitalization prior to death.
However, to the extent that these errors affect all age groups more or less equally, the results can still be used to determine a mortality pattern. In addition, there are methods (see Hill et al., 2011) to estimate correction factors, based on the observed population sizes by age and sex, to correct for the systematic errors in estimated mortality levels. By asking appropriate follow-up questions (see below), this question can also be used to measure maternal mortality. In the 2010 census round, this method for measuring maternal mortality has been followed in more than 30 countries that do not have reliable registration data.
138. Measurement of maternal mortality through a population census is recommended for countries where other sources of maternal mortality information such as the vital registration system are deficient. In practice this recommendation only applies to countries with at least 500,000 population because of the need to have sufficiently large denominators to reliably measure this event. In this context it is important to realize that maternal deaths are relatively rare events and in order to measure them through a sample survey the sample size needs to be very large, often resulting in prohibitive costs.
139. Questions on maternal mortality in a census typically result in information on pregnancy related deaths, which is not the same as maternal deaths (see also the definition in an earlier paragraph). Pregnancy related deaths include deaths from any cause, occurring while a woman was pregnant or within 42 days after delivery. Using this data for analysis of maternal mortality results in a measure called the Pregnancy-Related Mortality Rate (PRMR). Comparisons of census-based estimates of the PRMR with survey-based estimates of MMR found that approximately 85% of pregnancy related deaths are maternal deaths. It is believed that the correspondence between PRMR and MMR is quite close since the number of pregnancy related deaths tends to be under-reported in censuses (Hill, 2009; NIPORT; ORC Macro; Johns Hopkins and ICDDR.B, 2001). Nevertheless, the results from census-based maternal mortality questions should not be taken at face value and should ideally be followed up by a survey among the reported pregnancy related deaths to empirically establish the proportion of pregnancy related deaths that are maternal.
Recommended census questions to estimate maternal deaths:
Q1: Have any residents of this household died during the last 12 months?
For each deceased:
Q2: Sex of the deceased;
Q3: Age of the deceased;
Q4: Date of death;
For female deceased between the ages of 15 and 49:
Q5: Was the deceased pregnant at the time of death or did the death occur within 42 days after delivery
140. The recommended questions to measure maternal mortality in a census are placed in the household module, and extend the “standard” questions on deaths (by age and sex) in the household over the past 12 months by one additional question: whether the woman was pregnant at the time of death, or the death occurred within 42 days after delivery.
141. Some countries (e.g. Lesotho, Malawi, Swaziland) do not use this format, but instead ask about the survival of the sisters of the respondent. A variant (direct sisterhood) of this so-called sisterhood method is also used in the DHS, but the difference is that the DHS asks for additional information on ages and times of occurrence, making the resulting information much more accurate. The census data, however, is used for indirect estimation of maternal mortality using the indirect sisterhood method (see above). This method results in estimates of maternal mortality that refer to approximately 10-15 years before the date of the census. Its validity is contested by WHO and others as it relies on too many assumptions and the reference period is too long in the past.
144. Of the health and mortality-related Minimum Set of Gender Indicators approved by the Statistical Commission in February of 2012, the following can be computed from census data:
1. Under-5 mortality rate by sex;
2. Maternal mortality ratio (in censuses that ask the appropriate question);
3. Life expectancy at age 60, by sex; and
4. Adult mortality by age group (but not by cause).
The following indicators, which are related to health, rather than mortality, cannot usually be computed from census data:
1. Smoking prevalence among persons aged 15 and over, by sex;
2. Women's share of the population aged 15-49 living with HIV/AIDS; and
3. Proportion of adults who are obese, by sex.
Bourne and Walker (1991) show for the case of India that, while increased education of mothers generally favours child survival, the effect is larger for girls than for boys.
Download 3.92 Mb.
Share with your friends:
The database is protected by copyright ©sckool.org 2020