Data Analysis and Report Writing for Civil Registration: Life Table Analysis and Functions

MODULE 9 Online course on data analysis and report writing for civil registration based vital statistics MODEL LIFE TABLES AND LIFE TABLE 12 July to 10 September 2021 ANALYSIS

WHAT WILL WE COVER IN THIS MODULE Life table analysis and review of life table functions Model life tables Decomposition of life expectancy Re-cap on preparing a vital statistics report

LIFE TABLES FOLLOW UP Any questions?

LIFE TABLE ANALYSIS Probability of surviving from age x to age y: = ly / lx

LIFE TABLE ANALYSIS Oceania survival curves, 1950-55 and 2015-20 100000 80000 60000 lx 40000 20000 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 2015-20 1950-55

Probability of dying between ages x and y: = 1 [ly/lx]

LIFE TABLE ANALYSIS Oceania probability of death curves, 1950-55 and 2015-20 1.0000 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 0.1000 qx 0.0100 0.0010 1950-55 2015-20

Number of people dying between ages x and y: = lx ly

Oceania deaths , 1950-55 and 2015-20 18000 16000 14000 12000 10000 dx 8000 6000 4000 2000 0 0 20 40 60 80 100 1950-55 2015-20

Number of person years lived between ages x and y: = Tx Ty

LIFE TABLE ANALYSIS Oceania survival curves, 1950-55 and 2015-20 100000 80000 60000 lx 40000 20000 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 2015-20 1950-55

LIFE TABLE ANALYSIS Oceania life expectancy, 1950-55 and 2015-20 70 60 50 Life expectancy 40 30 20 10 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 1950-55 2015-20

OTHER LIFE TABLE APPLICATIONS Mortality (time from birth until death) Nuptiality (time from birth until first marriage) Contraceptive failure (time from initiating contraceptive use until conception) Fertility (time from birth until first child, time from birth until second child) Sexual debut (time from birth until first sex)

DECOMPOSING A DIFFERENCE IN LIFE EXP A change in life expectancy does not imply age-specific mortality rates change in the same magnitude and direction Contributions of various causes of death in the changes in expectations of life will not be of the same magnitude and direction. Some age-cause-specific death rates will increase in the two groups compared, thus contributing to a decrease in expectation of life, while others may decrease in some age intervals, thus increasing expectation of life. Disentangling these contribution by age-specific (and age-cause-specific) death rates to the difference between two expectations of life helps interpret the dynamics of changes in the mortality. Two main methods: Arriaga and Pollard (we will look at Arriaga).

ARRIAGA METHOD 1 1 ??? 2 1 2 1 2 ??+? ??+? 1 ? ? =?? 2 ??? 1+??+? ?? ?? 1 2 ?0 ?? ?0 ?? Direct effect of a change in mortality rates between ages x and x+n, i.e. the effect that a change in the number of years lived between ages x and x+n produces on life exp. at birth Sum of the indirect and interaction effects, i.e. the contribution resulting from the person-years to be added because the additional survivors at age x+n are exposed to new mortality conditions Effect of a difference in mortality rates between ages x and x+n

EXAMPLE: US FEMALES 1935 & 1995 Total difference: ?0 79.00-63.32 = 15.68 years ? (1995) - ?0 ? (1935) = How much did the change in infant mortality contribute to this decline? How does this compare to other ages?

WHY DO WE NEED MODEL LIFE TABLES Creating life tables requires good age-specific mortality rates, ideally from CRVS systems. Where death registration is not universal, model life tables may be useful as a comparator or to provide estimates. Can leverage partial information, even just one estimate of IMR. Beginning in the 1950s, repeated attempts have been made to generate systems of life tables that might be of use in mortality investigations across the world.

WHAT IS A MODEL LIFE TABLE? High correlations between sets of death rates drawn from different pops. Widely observed in populations with good data. For e.g. when death rates are high at ages 1-4. they tend to be high at ages 40-44 and 80-84. Therefore possible to create model life tables depending on different relationships.

KEY COMPONENTS TO EXPLAIN VARIATION 1. Overall level of mortality 2. Balance between child and adult mortality (relates to the steepness of the slope of the lx curve 3. Variations in mortality level specific to young ages (child mortality) 4. Variations in mortality level specific to older ages (old age mortality) 5. Gender differences Collectively explain over 95% of the variation in mortality patterns.

WHY DIFFERENCES IN AGE PATTERNS OF MORTALITY EXIST? If countries have the same life expectancy, why might the age pattern of mortality differ? Broad environmental factors (climate, natural disasters, crop productivity, etc.). Disease vectors (tuberculosis, malaria, HIV, etc.) and other causes of death. Public health spending and interventions may benefit certain demographic groups.

REPRESENTATIONS OF MORTALITY Model life tables seek to define typical and reliable age patterns of mortality. Reliable and documented mortality experiences grouped together to create typical mortality pattens showing relative magnitudes of infant, child and adult mortality.

MODEL LIFE TABLE SETS Two main sets relevant for today Coale Demeney Set New UN set Both sets available on the UN Population Division website: http://esa.un.org/unpd/wpp/Model-Life-Tables/download- page.html

COALE DEMENEY MODEL LIFE TABLES Developed in 1960 s but still widely used today. Recognises regional variation in relationship between level and age- pattern of mortality. 326 pairs of male and female empirical mortality schedules collected (based on registered deaths and census derived estimates). Therefore, possible to create model life tables depending on different relationships.

THE 4 SETS OF THE COALE DEMENY MODELS Classify life tables into 4 different sets: West, East, North and South. Dependent on the patterns of mortality in rough regions of Europe from the original data. North: low IMR and low mortality above 50 (9 tables from Sweden, Iceland and Norway). South: high CMR and high mortality above 65 (22 tables from Spain, Portugal and S Italy). East: high IMR and increasingly high mortality above age 50 (31 tables from Austria, Germany, N Italy, Hungary and Poland). West: free of substantial deviations (130 tables from W Europe, overseas European pops, Japan and Taiwan Province of China) generic patten if the others don t fit.

THE 4 SETS OF THE COALE DEMENY MODELS In each of these sets, the life expectancy at age 10 was correlated with the probability of dying at different ages, and these correlations provided the basis for estimating a series of nested life tables at different overall levels of e(x) but with different age patterns of mortality.

LEVELS OF COALE DEMENY LIFE TABLES Constructed for each family by sex Level 1 = life expectancy at birth of 20 years Then 2.5 years increments Up to level 25 = life expectancy at birth of 80 years

IMR FOR FEMALES BY LIFE EXPECTANCY AT BIRTH AND REGION (PER THOUSAND) Model Life expectancy at birth 30.00 40.00 50.00 60.00 70.00 West 256.11 178.22 118.79 71.16 31.16 North 224.30 156.92 106.02 66.28 32.64 East 306.50 216.83 147.40 89.70 40.96 South 228.81 172.52 130.97 94.91 59.11

ISSUES WITH COALE DEMENEY LIFE TABLES Very sensitive choice of regional model and need some information on mortality pattens (can use similar populations). Very high or low levels of mortality outside the range of the empirical data, extrapolation performed poorly. Experience based on developed countries but applications to developing countries with incomplete systems. Also, relationship between child and adult mortality changed.

THE NEW UN SET UN produced a new set of model life tables in 1980s. Based on 36 life tables from developing countries. Far fewer due to data quality and availability. As with Coale Demeny, regional clustering of mortality patterns identified. Tables prepared with increments of life expectancy.

THE NEW UN SET: SPECIFIC MORTALITY PATTERNS South Asian (based on South and West Asian countries and Tunisia) High mortality at extremes of age, below 15 and above 60 Far Eastern (also Malaysia, Guyana and Trinidad and Tobago) High mortality at older ages, possibly due in part to a high incidence of death from TB or Hep B Latin American (includes also the Philippines, Sri Lanka and Thailand) High mortality during infancy, childhood and young adulthood, lower at older ages Chilean (just data from Chile) Extremely high infant mortality (possibly due to respiratory disease) General Remaining life tables, default table like West But use data to drive choice, not actual region!

DEVIATIONS OF UN MLT FROM COALE-DEMENY WEST

QUESTION? If the choice of life table is based on IMR, how might the COVID19 pandemic affect the relevance of life tables? How do you think the estimate of life expectancy could be biased as a result?

DATA TO CONSIDER WHEN VALIDATING MORTALITY PATTERNS FROM CR DATA Census data Deaths in the last 12 months Demographic and Health Surveys/Multiple Indicator Cluster Surveys Infant and child mortality Sibling/orphanhood data Other survey data, if available Demographic surveillance data, if available

SOFTWARE PACKAGE TO HELP WITH LIFE TABLES AND MORTALITY ESTIMATION MortPak https://www.un.org/en/development/desa/population/publications/mor tality/mortpak.asp Uses Excel-like worksheets

PREPARING A VITAL STATISTICS REPORT Why is it important? Creates a single source of information about demographic patterns within a country Provides insight into the CRVS system, how data is collected, its completeness and reliability Turns complex data into easily digestible statistics which can be used to infer the general health of a population Supports policy makers to make evidence-based decisions

HOW TO USE THE REPORT TEMPLATE Breaks report down into sections Each section explains what you should consider writing about and what each section should contain Poses questions that you can ask yourself as you write, to ensure you have considered all the important points Contains sample text that could be used to initiate the writing process Facilitators are also available to support you in this process.

PRINCIPLES FOR GETTING STARTED Break up the work into digestible sections and write as you go along while the analysis is stilll fresh in your mind Start with short bullet points: Describe the overall patterns Do you observe any changes over time? What do your results mean? Are there potential explanations for these patterns/trends? As you interpret your analysis, reflect on the quality of your data Are there any factors that may impact the quality of your data? (ie. missing data, insufficient coverage) Just start writing!

Thank You!