Multiple Indicator Cluster Surveys Data Dissemination - Further Analysis Workshop Mortality MICS4 Data Dissemination and Further Analysis Workshop Background • Child mortality: Probabilities of dying during the first 5 years of life, usually broken down by conventional age segments • Infant (first one year) and under-5 mortality rates (first 5 years) are the most commonly calculated probabilities Periods for Under-5 Mortality Measurement Under-5 Mortality (0-4 Years) Birth Neonatal Mortality (First month) 5 1 Post-Neonatal Mortality (1-11 months) Infant Mortality (Birth to One Year) Child Mortality (1 to 4 years) Background • MDG 4: reduce under-5 mortality by two-thirds, between 1990 and 2015 – Indicator 1.3 – Under-5 Mortality Rate – Indicator 1.4 – Infant Mortality Rate • Both indicators are measured in MICS surveys • Child mortality indicators are broad indicators of social development/health status • Used to evaluate impact of interventions based on trends Measurement of child mortality Data Sources • • • • Vital registration Population censuses Surveillance systems, sample registration systems Household surveys – Direct: Data from full birth histories, as in DHS and some MICS surveys – Indirect: Data from summary birth histories, to use “Brass methods” – Note that surveys that include birth histories can be used both for direct and indirect estimation Methods: Direct method •Based on birth histories •Required data: – Data of birth for each child (month and year) – Survival status – Date or age at death for each child who has died •Typically, synthetic cohort life table approach used to estimate rates Methods: Direct method • Rely heavily on the quality of information collected – work best in populations where dates and durations are well-known • Sources of errors: – Omission of births and deaths – Misreporting of age at death (age heaping at 12 months is common) – Birth misplacement Age heaping: child’s death at 12 months Niger DHS06 - Age at death in months 35.0 Percent of deaths 1-23 months 30.0 25.0 20.0 15.0 10.0 5.0 Linear trendline 0.0 0 2 4 6 8 10 12 14 Months 16 18 20 22 24 26 Age shifting: common issue in DHS Niger: Births by year, DHS06 3000 Questions on under-5s for all births after January 2001 2500 Numbers 2000 Living Dead Total 1500 1000 500 0 1996 1998 2000 2002 Year 2004 2006 Table CM.0: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total), 2010 Periods of analysis of 5 years Deaths Age in months Exposure Age in months Probability Age in months 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0-4 Mean 620.26 63.04 89.22 99.64 208.61 244.79 144.69 84.05 13920.14 13237.43 12969.53 12610.69 12296.12 11917.43 11361.23 10698 44.56 4.76 6.88 7.9 16.97 20.54 12.73 7.86 5-9 Mean 504.88 110.23 82.81 166.7 242.33 209.71 133.63 74.16 11978.53 11624.09 11611.1 11262.74 10425.94 9526.55 8756.18 8153.75 42.15 9.48 7.13 14.8 23.24 22.01 15.26 9.1 10-14 Mean 445.41 77.51 71.63 151.27 169.82 208.98 107.73 49.9 9302.7 8735.32 8495.96 8346.37 7958.39 7284.85 6587.17 6035.32 47.88 8.87 8.43 18.12 21.34 28.69 16.35 8.27 15-19 Mean 348.55 73.39 46.55 121.36 175.52 164.71 69.09 34.12 6793.22 6359.38 6170.02 5903.85 5335.74 4659.14 4033.44 3489.87 51.31 11.54 7.54 20.56 32.89 35.35 17.13 9.78 20-24 Mean 201.48 87.22 26.91 33.94 98.35 90.18 30.93 18.01 3910.57 3549.13 3277.48 3176.58 2964.97 2480.25 1997.63 1671.45 51.52 24.58 8.21 10.69 33.17 36.36 15.48 10.77 Check denominators for: Less than 250 cases * 250-499 cases ( ) Table CM.0: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total), 2010 Periods of analysis of 5 years Deaths Age in months Exposure Age in months Probability Age in months 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0-4 Mean 620.26 63.04 89.22 99.64 208.61 244.79 144.69 84.05 13920.14 13237.43 12969.53 12610.69 12296.12 11917.43 11361.23 10698 44.56 4.76 6.88 7.9 16.97 20.54 12.73 7.86 5-9 Mean 504.88 110.23 82.81 166.7 242.33 209.71 133.63 74.16 11978.53 11624.09 11611.1 11262.74 10425.94 9526.55 8756.18 8153.75 42.15 9.48 7.13 14.8 23.24 22.01 15.26 9.1 10-14 Mean 445.41 77.51 71.63 151.27 169.82 208.98 107.73 49.9 9302.7 8735.32 8495.96 8346.37 7958.39 7284.85 6587.17 6035.32 47.88 8.87 8.43 18.12 21.34 28.69 16.35 8.27 15-19 Mean 348.55 73.39 46.55 121.36 175.52 164.71 69.09 34.12 6793.22 6359.38 6170.02 5903.85 5335.74 4659.14 4033.44 3489.87 51.31 11.54 7.54 20.56 32.89 35.35 17.13 9.78 20-24 Mean 201.48 87.22 26.91 33.94 98.35 90.18 30.93 18.01 3910.57 3549.13 3277.48 3176.58 2964.97 2480.25 1997.63 1671.45 51.52 24.58 8.21 10.69 33.17 36.36 15.48 10.77 Neonatal mortality Table CM.0: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total), 2010 Periods of analysis of 5 years Deaths Age in months Exposure Age in months Probability Age in months 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0-4 Mean 620.26 63.04 89.22 99.64 208.61 244.79 144.69 84.05 13920.14 13237.43 12969.53 12610.69 12296.12 11917.43 11361.23 10698 44.56 4.76 6.88 7.9 16.97 20.54 12.73 7.86 5-9 Mean 504.88 110.23 82.81 166.7 242.33 209.71 133.63 74.16 11978.53 11624.09 11611.1 11262.74 10425.94 9526.55 8756.18 8153.75 42.15 9.48 7.13 14.8 23.24 22.01 15.26 9.1 10-14 Mean 445.41 77.51 71.63 151.27 169.82 208.98 107.73 49.9 9302.7 8735.32 8495.96 8346.37 7958.39 7284.85 6587.17 6035.32 47.88 8.87 8.43 18.12 21.34 28.69 16.35 8.27 15-19 Mean 348.55 73.39 46.55 121.36 175.52 164.71 69.09 34.12 6793.22 6359.38 6170.02 5903.85 5335.74 4659.14 4033.44 3489.87 51.31 11.54 7.54 20.56 32.89 35.35 17.13 9.78 20-24 Mean 201.48 87.22 26.91 33.94 98.35 90.18 30.93 18.01 3910.57 3549.13 3277.48 3176.58 2964.97 2480.25 1997.63 1671.45 51.52 24.58 8.21 10.69 33.17 36.36 15.48 10.77 Post-neonatal mortality Infant mortality Table CM.0: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total), 2010 Periods of analysis of 5 years Deaths Age in months Exposure Age in months Probability Age in months 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0-4 Mean 620.26 63.04 89.22 99.64 208.61 244.79 144.69 84.05 13920.14 13237.43 12969.53 12610.69 12296.12 11917.43 11361.23 10698 44.56 4.76 6.88 7.9 16.97 20.54 12.73 7.86 5-9 Mean 504.88 110.23 82.81 166.7 242.33 209.71 133.63 74.16 11978.53 11624.09 11611.1 11262.74 10425.94 9526.55 8756.18 8153.75 42.15 9.48 7.13 14.8 23.24 22.01 15.26 9.1 10-14 Mean 445.41 77.51 71.63 151.27 169.82 208.98 107.73 49.9 9302.7 8735.32 8495.96 8346.37 7958.39 7284.85 6587.17 6035.32 47.88 8.87 8.43 18.12 21.34 28.69 16.35 8.27 15-19 Mean 348.55 73.39 46.55 121.36 175.52 164.71 69.09 34.12 6793.22 6359.38 6170.02 5903.85 5335.74 4659.14 4033.44 3489.87 51.31 11.54 7.54 20.56 32.89 35.35 17.13 9.78 20-24 Mean 201.48 87.22 26.91 33.94 98.35 90.18 30.93 18.01 3910.57 3549.13 3277.48 3176.58 2964.97 2480.25 1997.63 1671.45 51.52 24.58 8.21 10.69 33.17 36.36 15.48 10.77 Child mortality Table CM.0: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total), 2010 Periods of analysis of 5 years Deaths Age in months Exposure Age in months Probability Age in months 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0 1-2 3-5 6-11 12-23 24-35 36-47 48-59 0-4 Mean 620.26 63.04 89.22 99.64 208.61 244.79 144.69 84.05 13920.14 13237.43 12969.53 12610.69 12296.12 11917.43 11361.23 10698 44.56 4.76 6.88 7.9 16.97 20.54 12.73 7.86 5-9 Mean 504.88 110.23 82.81 166.7 242.33 209.71 133.63 74.16 11978.53 11624.09 11611.1 11262.74 10425.94 9526.55 8756.18 8153.75 42.15 9.48 7.13 14.8 23.24 22.01 15.26 9.1 10-14 Mean 445.41 77.51 71.63 151.27 169.82 208.98 107.73 49.9 9302.7 8735.32 8495.96 8346.37 7958.39 7284.85 6587.17 6035.32 47.88 8.87 8.43 18.12 21.34 28.69 16.35 8.27 15-19 Mean 348.55 73.39 46.55 121.36 175.52 164.71 69.09 34.12 6793.22 6359.38 6170.02 5903.85 5335.74 4659.14 4033.44 3489.87 51.31 11.54 7.54 20.56 32.89 35.35 17.13 9.78 20-24 Mean 201.48 87.22 26.91 33.94 98.35 90.18 30.93 18.01 3910.57 3549.13 3277.48 3176.58 2964.97 2480.25 1997.63 1671.45 51.52 24.58 8.21 10.69 33.17 36.36 15.48 10.77 Under-5 mortality Estimates from direct method Table CM.1: Early childhood mortality rates Neonatal, post-neonatal, Infant, child and under-five mortality rates for five year periods preceding the survey, (Total) PostNeonatal neonatal Infant Child Under five mortality mortality mortality mortality mortality rate [1] rate [2] rate [3] rate [4] rate [5] Periods of analysis 0-4 of 5 years 5-9 44.56 18.55 63.11 56.89 116.41 42.15 29.79 71.94 67.88 134.94 10-14 47.88 33.36 81.24 72.69 148.03 15-19 51.31 37.15 88.46 92.03 172.35 20-24 51.52 40.71 92.23 92.63 176.32 C1 MICS Survey Direct estimates of U5MR 250 200 150 100 50 0 1985.00 1990.00 1995.00 2000.00 2005.00 2010.00 Methods: Indirect method •Required data – Age of women – The total number of children she has ever borne, and – The number of those children who have died (or, the number who are still alive) •Require relatively fewer information than direct method Indirect method Table CM.1: Children ever born, children surviving and proportion dead Mean and total numbers of children ever born, children surviving and proportion dead by age of women, Country, 2010 Mean Total Mean Total number of number of number number of children children children children Proportion Number of ever born ever born surviving surviving dead women Age 15-19 20-24 25-29 30-34 35-39 40-44 45-49 Total .286 1316 .267 1229 .082 4601 1.255 4732 1.107 4175 .123 3770 2.522 8287 2.160 7100 .149 3286 3.743 8339 3.202 7132 .145 2228 5.026 10654 4.105 8703 .185 2120 5.772 8421 4.669 6812 .192 1459 6.407 8138 5.016 6372 .217 1270 2.663 49887 2.216 41523 .170 18734 Methods: Indirect method • Distributes children ever born to women retrospectively over time using models • Assumes – Little or no change in fertility levels and age patterns – No change or a linear decline in mortality – A pattern of mortality by age that conforms to known model life table “families” which basically derived from European experience Methods: Indirect method (3) Converts proportion dead of children ever born (D(i)) reported by women in age groups 15-19, 20-24, etc. into estimates of probability of dying before attaining certain exact childhood ages, q(x): q(x) = K(i)*D(i) where the multiplier K(i) is meant to adjust for non mortality factors determining the value of D(i) MICS4 Workshop Methods: Indirect method •The age pattern of child mortality --- select the right model life table – Coale-Demeny patterns by region: • East, North, South, and West – United Nations patterns by region: • Latin America, Chilean, South Asian, Far Eastern, and General Select the right model life table: India Indirect method Table CM.1: Children ever born, children surviving and proportion dead Mean and total numbers of children ever born, children surviving and proportion dead by age of women, Country, 2010 Mean Total Mean Total number of number of number number of children children children children Proportion Number of ever born ever born surviving surviving dead women Age 15-19 20-24 25-29 30-34 35-39 40-44 45-49 Total .286 1.255 2.522 3.743 5.026 5.772 6.407 2.663 1316 4732 8287 8339 10654 8421 8138 49887 .267 1.107 2.160 3.202 4.105 4.669 5.016 2.216 Check denominators 1229 4175 7100 7132 8703 6812 6372 41523 .082 .123 .149 .145 .185 .192 .217 .170 4601 3770 3286 2228 2120 1459 1270 18734 Indirect method Coale-Demeny Models (Trussel equations) Proportion Mean children children dead of ever born born Age i Q(i) North t(i) North Q(i) South t(i) South Age group 15-19 .286 .082 1 .071 1.4 .068 1.3 20-24 1.255 .123 2 .116 2.7 .122 2.7 25-29 2.522 .149 3 .140 4.4 .150 4.5 30-34 3.743 .145 5 .144 6.4 .149 6.6 35-39 5.026 .185 10 .196 8.5 .194 8.9 40-44 5.772 .192 15 .201 10.9 .197 11.5 45-49 6.407 .217 20 .223 13.7 .220 14.5 Indirect method Under-five Mortality Rate (Male) Under-five Under-five Under-five Reference Mortality Reference Mortality Reference Mortality Reference date North Rate North date South Rate South date East Rate East date West Age 15-19 group 20-24 25-29 30-34 35-39 40-44 45-49 2009.0 .111 2009.0 .087 2009.0 .097 2009.0 2007.7 .150 2007.7 .142 2007.6 .140 2007.7 2006.0 .161 2005.9 .161 2005.8 .158 2005.8 2004.0 .144 2003.8 .149 2003.6 .147 2003.7 2001.8 .169 2001.5 .181 2001.3 .178 2001.5 1999.4 .162 1998.9 .178 1998.7 .175 1999.0 1996.6 .166 1995.9 .187 1995.5 .186 1996.1 MICS Survey Indirect estimates 180 160 140 120 100 80 60 40 20 0 1985.00 1990.00 1995.00 2000.00 2005.00 2010.00 Final estimates • As the “final” or “most recent” estimate, we use an average of estimates based on women age 25-29 and 30-34 • Ignore estimates based on women age 15-19 and 20-24: selection bias C1: “Final” estimates Under-five Infant Mortality Mortality Rate Rate [1] [2] Sexo Masculino 111 162 Feminino 98 146 Region SAB 87 119 Leste 130 207 Northe 104 153 Sul 83 112 Area de Urbano 93 131 residência Rural 110 167 Quintil de Mais rico 114 174 riqueza Segundo 116 179 Meio 104 154 Quarto 102 149 Mais 69 89 pobre Total 105 155 Data quality issues • Main errors in data on children ever born and children dead/surviving – Omission of deaths – Misreporting of women’s age • Other drawbacks – Violation of assumptions – Use model life tables to adjust the data for the age pattern of mortality in the general population --- Inappropriate model life table may results in mis-estimation of trends. Checking quality of mortality estimates • Compare child mortality across sub-groups • Expected patterns by sex, background characteristics • Check estimates from successive data sources Compare CEB, CS, CD data Age in 2000 Mean Number of Children Ever Born 2000 2006 2010 0.2860 0.6542 1.2551 5-9 10-14 Mean Number of Children Surviving 2000 2006 2010 0.2671 0.5396 1.1073 Mean Number of Children Deceased 2000 2006 2010 0.0189 0.1145 0.1478 15-19 0.5028 1.7176 2.5216 0.4255 1.4191 2.1605 0.0773 0.2985 0.3612 20-24 1.7158 3.3253 3.7432 1.4058 2.6190 3.2016 0.3100 0.7063 0.5416 25-29 3.3431 4.6799 5.0258 2.6323 3.6527 4.1055 0.7108 1.0272 0.9203 30-34 4.9895 5.8593 5.7723 3.8199 4.3312 4.6691 1.1696 1.5281 1.1032 35-39 6.1893 6.4765 6.4072 4.5381 4.7859 5.0164 1.6512 1.6906 1.3907 40-44 6.9033 6.7208 4.9685 4.6518 1.9349 2.0690 45-49 7.2666 5.1520 2.1146 Quality check: sample size • Sample size needs to be sufficiently large to produce statistically reliable estimates of infant and under-five mortality • Mortality data may carry wide confidence intervals • Number of births and deaths for children of women aged 1519 is often very small, thus have effects on the parity ratio and on the regression used to derive estimation equations, therefore may bias the indirect estimates For further analysis • Compare estimates from different sources • Analyze mortality by coverage indicators • Check age patterns of mortality (from direct method), compare with model patterns • Multivariate analyses The IGME Work Members of the IGME • UN Inter-agency Group for Child Mortality Estimation (IGME) was formed in 2004, led by UNICEF, WHO, and includes members of UN Population Division and The World Bank • Technical Advisory Group (TAG) of the IGME – Independent – Composed of leading experts in demography and biostatistics – Provide technical guidance on estimation methods, technical issues and strategies for data analysis and data quality assessment Objectives of the IGME • Objectives of the IGME – Harmonize estimates within the UN system – Improve methods for child mortality estimation – Produce consistent estimates of child mortality worldwide for reporting on progress towards MDG 4 – Enhance the capacity of countries to produce timely estimates of child mortality: regional workshops and country visits The IGME method to estimate child mortality • Update estimates annually – Compile all nationally representative data for each country – Check data quality – Fit a regression line to all data points that meet data quality standards established by the IGME and extrapolate to a common reference year – Additional adjustment applied to countries with high HIV/AIDS prevalence • The IGME Estimates are based on national data from surveys, census, vital registrations, etc, but may differ from these data Why is it necessary to produce interagency child mortality estimates • • • • • No single, high quality source in most countries Multiple data sources often inconsistent Project estimates Important to estimate since 1990 Consistent methodology Example: Data rich and consistency countries Mali The available data sources cluster over a narrow band and show considerable consistency The estimate line is fitted to all the data Example: Data rich countries with wide variations in mortality level from different sources Nigeria Has one of the widest spreads of source data, with a range from 120 to 240 deaths per 1,000 live birth In driving the estimate line, all sources with dotted lines are rated of lower quality and are not used. Discrepancies between national and interagency estimates • National estimates: often use data directly from censuses, surveys, or vital registration systems • IGME estimates: use national data from censuses, surveys, or vital registration systems as underlying data to generate estimates by fitting a tend line to these data • For countries with consistent data, national estimates and interagency estimates are similar. • For countries with inconsistent or messy data, differences might be large Direct and indirect estimates 300 250 200 150 100 50 0 1985.00 1990.00 1995.00 MICS 2000 (Indirect) MICS 2006 (Indirect) 2000.00 RHS/MICS 2010 (Direct) 2005.00 RHS/MICS 2010 (Indirect) 2010.00 CMEInfo The IGME’s Child Mortality Database: www.childmortality.org