# Evaluation of Mortality Data Collected from Population Censuses

• Slides: 37

Evaluation of Mortality Data Collected from Population Censuses United Nations Statistics Division United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Outline of the presentation Some basics about life table For two items that can be used to obtain mortality statistics in census: - Survival of children ever born - Deaths in the household We discuss - Information collected - Possible quality issues related to each question - Methods of data evaluation using examples United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Some basics about life table (1) Age (x) n. Mx nqx lx n. Lx ex 0 0. 005774 0. 00579 100000 99707 79. 1561 1 0. 000284 0. 001134 99581 398184 78. 4879 5 9. 32 E-05 0. 000466 99511 497465 74. 5417 10 0. 000166 0. 000831 99475 497260 69. 5677 … … … 55 0. 007521 0. 036913 93619 462988 26. 9352 60 0. 011885 0. 057709 91576 449140 22. 4803 65 0. 021682 0. 102837 88080 429485 18. 2733 70 0. 037063 0. 1696 83714 407728 14. 096 75 0. 059397 0. 258588 79377 383688 9. 72956 80 0. 10245 0. 407803 74098 354340 5. 24461 85+ 0. 179314 1 67638 34275 0. 50674 United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Some basics about life table (2) n. Mx = period mortality rate = nqx = proportion of those people reaching their xth birthday who die before their (x+n)th birthday lx = number of person who live to their xth birthday n. Lx = number of person-years lived between exact ages x and x+n ex = life expectancy at age x (the average number of years which people have left to live when they are at age x) United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Some basics about life table (3) Model life tables • Created to estimate demographic parameters for countries with limited data • Built on empirical studies of age-specific mortality patterns in the past • Two groups of model life tables: • Coale-Demeny: based on European populations • North, South, East and west European models • United Nations: For developing countries • Latin American, Chilean, South Asian, Far Eastern, General United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Mortality statistics from population census – Introduction A group of questions can be used to obtain mortality data in a census Two distinctions: a) Level and trend of mortality vs age pattern of mortality • • b) Survival of children ever born: level and trend of mortality Household deaths: age pattern of mortality: Deaths of younger persons vs. deaths of adults • • Younger persons: survival of children ever born Adults: household deaths All approaches are to supplement death registration data, not to replace it. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – information collected Have been used for the past 50 years to collect data on infant and child mortality For every woman the following information are collected: a) the total number of female children she has borne in her lifetime. b) the total number of male children she has borne in her lifetime. c) the number of female children who are surviving d) the number of male children who are surviving United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – Use of Ever born – Surviving = Children deceased / Ever born = Proportion deceased Life table measures of infant, child and young adult mortality may be derived from the proportion of deceased. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born - Tabulation example, Turkey 2000 Age 15 - 19 20 - 24 25 - 29 30 - 34 35 - 39 Total Women CEB Total CS Proporti on of deceased children 35182 29462 28129 8 6 57 0. 045 32634 20783 19914 64 45 32 0. 042 29188 45227 43124 19 04 25 0. 047 24572 57000 53951 38 43 85 0. 053 24008 70366 65639 19 46 08 0. 067 40 - 44 19852 67070 61315 33 United 44 Nations 0. 086 Source: Tabulated using 25 data from Demographic Yearbook 45 - 49 16580 63941 57229 United Nations Sub-Regional Workshop on Census Data Evaluation 57 04 0. 105 12 Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – Brass type estimates (1) Data are used to estimate level and trend of mortality for about 20 years prior to a census or survey. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (2) Empirical findings about child mortality Age group of mother in years Age group index Proportion of children dead approximates 15 -19 1 q(1) 20 -24 2 q(2) 25 -29 3 q(3) 30 -34 4 q(5) 35 -39 5 q(10) 40 -44 6 q(15) 45 -49 7 q(20) 50 -54 8 q(25) 55 -59 9 q(30) United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born - Brass-type estimates (3) Empirical findings about child mortality - Approximation q values referring to different time period before census q(1): more recent estimates; q(20) – earlier estimates (Feeney, 1980) Under-five mortality is used more often: more robust than infant mortality However if comparing estimates with civil registration, may use infant mortality rate Feeney 1980: Estimating infant mortality trends from child survivorship data, Population Studies 34(1): 109 -128. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (4) Empirical findings about child mortality - Under-five mortality - Most commonly used - more robust than infant mortality - Upward biases from reports of younger women, usually inaccurate - More powerful results (Brass type) came from multiple data sources United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (5) An example of Mort. Pak CEBCS output United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (5) An example of Mort. Pak CEBCS output (cont. ) United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (6) How to identify the right mortality model - graphical Source: Step by step guide to the estimation of child mortality, 1990, United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (7) How to identify the right mortality model – graphical Source: Step by step guide to the estimation of child mortality, 1990, United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (8) Illustration of the relationship of mother’s age and timing of the under-5 mortality estimates Bangladesh, 1974 Retrospective Survey of Fertility and Mortality Source: Step by step guide to the estimation of child mortality, 1990, United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (9) q(5) more robust than q(1) Infant and under-five mortality, Bangladesh Source: Step by step guide to the estimation of child mortality, 1990, United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (10) Turkey example again United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born: Brass-type estimates (11) Comparison of multiple sources United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

A few notes about Brass type estimates • Almost smooth due to modeling • If see rough and unsmooth data, indicates quality issues • The last increase of q(5) does not mean increasing mortality, but rather biases generated from mother of young age groups (15 -19) • There is violation of assumptions about age patterns in the method, i. e. , child death depends on children’s age only. But children born to very young mothers tend to be disadvantaged United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality (1) Experience has shown that it is possible to get high quality responses to this kind of questions in any data collection exercise, including censuses. If both CEB and CS are understated, some cancellation of errors will occur. But in practice, reporting of CS is more likely to be complete than reporting of CEB => calculated proportions of deceased children are likely to be too low. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality (2) Other influences on the accuracy of estimates derived from these data: Assumptions about the age pattern of mortality: mortality of child relies only on their own age (which will fail at young age of mothers, i. e. , the 1 st or 2 nd age groups of mothers) In the ideal case, data on CEB and CS will be available from two or more data collection exercises, at different points in time. This will allow comparison, providing a powerful test of the quality of the estimates. United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (1) - Initial assessment: - Any missing values in children surviving data? - Missing values for any relevant variables: age of mother, sex of those who died - Plausibility of data - Children survival data; age distribution - Distribution of women with socio-economic characteristics United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (2) Example: missing or implausible values of CEB and CS data “… systematic failure in data collection…” Source: Estimation of mortality using the South African Census 2001 data, Dorrington, Moultrie and Timæus, Centre of Actuarial Research, University of Cape Town, 2001 United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (5) Comparing age patterns of proportion deceased children Source: Graph produced based on data collected by the United Nations Demographic Yearbook and Measure DHS country report United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (6) A rapid assessment: Burundi, 1990 census: CS and CEB data Age Total women Average CEB Average CS CS/CEB 15 + 1483895 3. 82 2. 93 0. 77 15 - 19 250329 0. 07 0. 89 20 - 24 229655 1. 02 0. 86 0. 84 25 - 29 214467 2. 75 2. 28 0. 83 30 - 34 187348 4. 49 3. 63 0. 81 35 - 39 135551 5. 62 4. 51 0. 80 40 - 44 97537 6. 10 4. 85 0. 80 45 - 49 75526 6. 23 4. 89 0. 79 50 - 54 76100 6. 21 4. 70 0. 76 55 - 59 50817 6. 22 4. 58 0. 74 60 - 64 53775 6. 12 4. 28 0. 70 110062 6. 07 3. 82 0. 63 2728 4. 47 3. 37 0. 76 65 + Unknown Source: Graph produced based on data collected by the United Nations Demographic Yearbook United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (7) A rapid assessment of CEB and CS data - (1 -0. 81)=0. 19 for the 30 -34 age group: the proportion of deceased among all children born to mother of 30 -34 years of age ≈ q(5), the proportion of children born who die before their 5 th birthday 7 years prior to census - Compare with other estimates, e. g. , UN Population Division estimates of under-5 mortality - 1990 census estimates of under-5 child mortality = 190 per 1000 for 1983 - UN Pop Division estimates for the period 1980 -1985: 196 per 1000 - Slightly underestimates Method: Rapid Assessment of Census Data on Children Born and Surviving, Griffith Feeney, 2009. http: //www. demographer. com/rapid-assessment-of-ceb-and-cs-data/ United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (8) Comparing with UN Population Division under-five mortality estimates Source: World Population Prospects: The 2010 Revision United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Survival of children ever born – quality assessment (9) Existing external sources - UN population division (World Population Prospect) - UNICEF child mortality website (www. childmortality. org) United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (1) - Direct estimates of current death rates can be obtained, however, with substantial errors - Under-reporting, especially for child deaths and older age deaths - Reference period errors in reporting of deaths (versus the usual 12 months reference period) - Death question omitted by interviewers - Household breaking up due to the death of a senior household member - Age-heaping and age exaggeration - The method is mainly used for adult mortality United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (2) Initial assessment Tabulation of enumerated deaths with associated variables, e. g. , year/month of death • Quality of age reporting for the deceased United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (4): Comparing age-specific death rates Source: Graph produced based on data collected by the United Nations Demographic Yearbook and Measure DHS country report United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (3) Assessment: death distribution methods General Growth Balance (GGB), assumes • constant coverage of household deaths and population across all ages (this would not work for children deaths) • Negligible migration • Stable population (constant births and deaths) • Accurate reporting of age for both population and deaths Synthetic Extinct Generations method (SEG), assumes • All the above, except for stable population assumption was relaxed in later version • Constant coverage of population across time (may be relaxed if use a “combined GGB-SEG approach”) United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (4) Assessment: example of GGB method United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011

Household deaths in the last 12 months – adult mortality (5) Assessment: example of GGB method f: slope of the fitted line (1/f)*100% = 41. 2% only 41. 2% of the deaths were being reported United Nations Sub-Regional Workshop on Census Data Evaluation Phnom Penh, Cambodia, 14 -17 November 2011