WHAT IS HUMAN MORTALITY? Part 1

DATA RESOURCES ABOUT HUMAN MORTALITY

Data resource basics

Over the past 100 years, human longevity has increased more than during all preceding human history combined. This remarkable change was caused by reductions in mortality; first generally at younger ages, then later progressing to older ages. The precise causes as well as the social and economic impact of this change are still unclear. What are the major patterns, driving forces and future prospects of this process? To what extent are these changes determined by period factors such as advances in medicine and economic development or by cohort effects such as early life experiences and accumulation of behavioural risks over the life course? How much further can human longevity extend?

These are central questions for scholars and policy makers. The Human Mortality Database (HMD), as a collection of detailed, consistent and high quality human mortality data, is an important resource for addressing these fundamental questions. Although there are other international databases of mortality indicators, some of which cover many more countries (eg those compiled by the United Nations Population Division, the World Health Organization, Eurostat and the Institute for Health Metrics and Evaluation), they all have limitations mostly related to quality and detail of the data. For example, most lack information on advanced ages, are restricted to short time periods or focus on periods but not cohorts; these disadvantages are addressed by the HMD.

Data resource area and population coverage

The HMD strives to include all populations for which death registration and census data are virtually complete. As a result, countries and areas in the HMD are relatively wealthy and for the most part highly industrialized. At present the database contains detailed population and mortality data for 37 developed countries and 46 populations (including sub-national groups), with series starting as early as 1751 in Sweden and covering more than a 100 years for 16 populations. The selection criteria for inclusion of a country in the HMD rely mainly on the quality of its data. In order to be included in the HMD, a country’s death registration system must be nearly complete (close to 99%). For such countries, original statistics are then collected extending as far back in time as there are age-classified census data and annual death counts in a sufficiently detailed format (at a minimum, there should be 5-year age groups with a separate category for infants and an open interval at age 80 or higher). Note that for historical data (especially for years preceding 1870–80), the quality is often lower, especially at ages 80 and above. In such situations, the HMD has well-defined and transparent methods that are applied in a consistent manner, accompanied by prominent warnings that alert users to potential problems and issues. Figure 1 shows the years for which data series are available for each population in the collection.

Figure 1.

HMD period and cohort mortality series by population and year (as of January 2015). Cohort life tables are indexed by the year of birth. Cohort life tables are unavailable in Belgium because of missing input data for the period 1914–18. Based on an earlier design by Adrien Remund (2015) 1 .

Open in new tabDownload slide

HMD period and cohort mortality series by population and year (as of January 2015). Cohort life tables are indexed by the year of birth. Cohort life tables are unavailable in Belgium because of missing input data for the period 1914–18. Based on an earlier design by Adrien Remund (2015) 1 .

For some data series, additional details are provided on specified sub-populations: (i) for populations that have suffered substantial war losses and where available data allow computation of population and mortality indicators for the total (including the military) and civilian populations separately, such as is the case for England and Wales and for France; and (ii) whenever original data series are available for important population sub-groups as collected by the official statistical agency for a country. These include separate series for West and East Germany; for England and Wales, Scotland and Northern Ireland in the United Kingdom; and for the Maori and non-Maori of New Zealand.

Figure 2 shows life expectancy trends in all HMD countries and averages for selected groups of countries since 1970 for men and women. The figure highlights the progress in most of the populations along with a broadening inter-country variation, in particular the growing disparity between the countries of the former Soviet Union and other advanced countries.

Figure 2.

Trends in life expectancy at birth by s*x from 1970 to 2010 for HMD populations and population groups. Individual populations: 36 selected HMD national or sub-national populations. Established market economies (EME): Australia, Austria, Belgium, Canada, Denmark, England and Wales, Finland, France, Iceland, Ireland, Italy, Japan, Luxembourg, The Netherlands, New Zealand, Northern Ireland, Norway, Portugal, Scotland, Spain, Sweden, Switzerland, USA, West Germany. Central and Eastern Europe (CEE): Bulgaria, Czech Republic, East Germany, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia. Former Soviet Union (FSU): Belarus, Russia, Ukraine.

Open in new tabDownload slide

Trends in life expectancy at birth by s*x from 1970 to 2010 for HMD populations and population groups. Individual populations: 36 selected HMD national or sub-national populations. Established market economies (EME): Australia, Austria, Belgium, Canada, Denmark, England and Wales, Finland, France, Iceland, Ireland, Italy, Japan, Luxembourg, The Netherlands, New Zealand, Northern Ireland, Norway, Portugal, Scotland, Spain, Sweden, Switzerland, USA, West Germany. Central and Eastern Europe (CEE): Bulgaria, Czech Republic, East Germany, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia. Former Soviet Union (FSU): Belarus, Russia, Ukraine.

Measures

The HMD contains uniform death rates and complete (by single year of age up to 110+) and abridged (ages 0, 1–4, 5–9, 10–14… 110+) period life tables for extended periods of time, with regular updates at 2- to 3-year intervals in general, though more frequently for some of the largest countries. All standard life table quantities are provided, including age-specific death rates, probabilities of dying, numbers of survivors, person-years lived, person-years remaining and remaining life expectancy.

In addition to complete series of period life tables, cohort life tables are provided where the observation period is sufficiently long to include at least one cohort observed from birth until extinction (around age 110). In that case, life tables are presented for all extinct cohorts as well as for almost-extinct cohorts (i.e. those that have reached 90 years of age or older).

The HMD also includes the original raw data collected from reliable sources (mostly as published or distributed by national statistical offices). These raw or ‘input’ data include the birth, death and census counts as well as official population estimates, from which all mortality estimates and life tables are derived. The data are accompanied by detailed documentation for the database as a whole (with both a complete methods protocol and a summary protocol) and for each particular country (see Figure 3 for a snapshot of a typical HMD country page).

Figure 3.

Screen shot of a typical HMD country page.

Open in new tabDownload slide

Screen shot of a typical HMD country page.

Methods

The HMD was launched online in 2002. It is a resource developed and maintained jointly by the Department of Demography at the University of California, Berkeley (USA) and the Laboratory of Demographic Data at the Max Planck Institute for Demographic Research (MPIDR) in Rostock (Germany). The French Institute for Demographic Studies (INED) has been involved more recently. Information about the background, history and methodology of the HMD can be found at [ www.mortality.org ].

Within the HMD team, country specialists are responsible for a specific country or set of countries. Country specialists are in charge of establishing contact with in-country experts (usually at the national statistical office or in the academic community), carrying out all systematic steps involved in assessing the reliability of input data, computing mortality rates and life tables and updating existing data series. The process of constructing the mortality series for each country is described in detail in the HMD methods protocol and in a brief summary available on the website [ www.mortality.org ]. 2 Major steps include the following.

Processing birth counts: annual counts of live births by s*x are collected for each population over the longest time period available. These counts are used mainly for estimating the size of individual cohorts from their birth.

Processing death counts: death counts are collected by s*x, completed age, year of death and year of birth (whenever available). Deaths of unknown age are distributed proportionately across the age range and aggregated deaths are split into finer age classes.

Estimating population size: census data and, where available, annual population estimates by s*x and age are collected from statistical offices. If reliable annual population estimates are not available, estimates of population size are derived using an intercensal survival method. To adjust for lower data quality at ages above 80, population estimates are derived by the extinct generation method and the survivor ratio method.

Death rate calculations: For both periods and cohorts, death rates are computed as the ratio of deaths to exposure-to-risk in matched intervals of age. Some adjustments for changes in territorial boundaries are introduced where necessary to ensure that death counts refer to the population exposed to risk.

Construction of life tables by single calendar year and single year of age: Period death rates are converted to probabilities of dying by a standard method. Cohort probabilities of dying are computed directly from estimates of deaths and exposure and are related to cohort death rates in a consistent way.

Construction of multi-year and abridged life tables: life tables are systematically provided by single calendar year and single year of age, no matter the format of the original data. They are also provided by 5-year and 10-year periods and for 5-year age groups by pooling detailed data over years, ages and cohorts.

In addition to these general methods, customized procedures are developed to address country-specific issues, such as gaps in the input data for certain years (for example, 1914–18 in Belgium). These special methods are described in a separate document available as an appendix to the methods protocol (available at www.mortality.org ).

Before we include data for an additional country or extend the data series for an existing country, extensive checks are performed to assess the reliability of input data. Country experts are invited to share their understanding of the local statistical and demographic situation and they are consulted whenever specific issues arise. In addition to country-specific knowledge and verifications, a set of standard data quality indicators is computed (including: indices of attraction to particular ages in the reporting of deaths and population; plausibility of mortality patterns; and s*x ratios at birth). Diagnostic charts are produced automatically to guide country specialists in their analyses. An internal report is written by the country specialist and discussed with the HMD team before a decision is reached about whether or not to include an additional country or to extend an existing series to an earlier period. The experience accumulated over the past 12 years suggests that there are few remaining countries that can be added to the database in its current form, due to the strict requirements in terms of data availability and quality. However, the existing HMD data series continue to be updated regularly as new data become available for each country.

Data resource use

Over 30 000 unique users have accessed the HMD at various times since 2002. As of December 2014, a non-exhaustive list of known publications that cite the HMD includes over 1000 journal articles (many published in this journal), 140 books or book chapters, 58 dissertations and theses, 51 official reports and 293 technical reports or working papers: see [ h***:://www.mortality.org/Public/HMD-Publist.pdf ] for the detailed list. The HMD has been and continues to be a major reference.

Epidemiologists and public health specialists have used the HMD to evaluate the representativeness of population samples with respect to mortality and for assessment of healthy life expectancy, years of life (or working life) lost and other composite indicators. For example, Popham et al.3 examined cross-national variations in 37 high-income countries in relation to their welfare systems and found that Nordic countries exhibit smaller inequalities in male mortality but not in female mortality. Berry et al.4 assessed the cost-effectiveness of certain medical interventions based on a combination of age-specific death rates for all causes from the HMD, supplemented with information about breast cancer mortality, prevention and treatment from other sources; they concluded that screening and therapy jointly contributed to breast cancer mortality trends in the USA from 1975 to 2000. In addition, several epidemiological studies have used the HMD as a baseline against which to compare survivorship of longitudinal study respondents or clinical trial participants. 5–8

The reliability of HMD indicators at advanced ages has allowed gerontologists and population biologists to monitor survival and mortality trends up to the very end of the life cycle. Together with other life and social science researchers, they have relied on the HMD collection to study ageing, population dynamics and mortality change. For example, Oeppen and Vaupel 9 demonstrated the near-linear increase in record life expectancy at birth (+ 0.24 years per calendar year for women, or 6 h a day) from 1840 to the present. Across cohorts born from 1870 to 1920, progress was even steeper: + 0.43 life-years per year of birth, or 10 h a day. 10,11 The rise of the maximum age at death has been slower: for Sweden from 1969 until 1999, following an acceleration from the slow increase of the previous century, the maximum age at death rose at around 0.11 years per calendar year for both men and women, as shown by Wilmoth et al . 12 using HMD data.

Investigating whether increased life expectancy translated into a parallel increase in years spent in disability, Christensen et al.13 concluded that, overall, healthy life expectancies increased faster than overall life expectancies in high-income countries, thus resulting in morbidity compression. This result was confirmed by a later study, also using HMD data, that contrasted survival rates and health status (measured by survey data) in two Danish cohorts, which determined that cognitive function and the ability to perform various activities of daily living were significantly better in the 1915 than in the 1905 cohort. 14

Because of the standardized methods implemented to produce life tables, HMD data have also been extensively used by demographers and statisticians to model mortality and survival and to forecast population and longevity. Such studies have identified the geographical areas and age patterns responsible for diverging mortality trends in Europe, particularly Eastern Europe. Comparing Russia, Lithuania and Belarus, Grigoriev et al.15 found that although all three countries experienced a decline in mortality over the 1990s, the speed of recovery was strongly and inversely related to the severity of the transition from a centralized to a market economy. Diverging trends within the broad group of industrial nations have also been investigated using HMD data. Such analyses have improved our understanding of the degradation in the US international ranking on life expectancy since the 1980s. In particular, Wilmoth et al.16 showed that this degradation is only partly attributable to increasing geographical inequalities within the USA over the past three decades.

HMD users also include: university teachers and students, who experiment with newly learned demographic and statistical techniques; policy makers, who simulate the economic impact of changes in retirement systems or of immigration reforms given various demographic scenarios; international organizations such as the United Nations or the World Health Organization, which use the HMD as a complement to their own databases; as well as private-sector organizations (insurance companies and actuaries in particular) who need reliable and harmonized life table estimates.

Strengths and weaknesses

The main strengths of the HMD result from its four guiding principles:

comparability of the reconstructed series through the use of standardized methods to format and process input data;

flexibility in the format provided to users, with six alternative output data series for each population;

accessibility through the internet to all registered users; and

reproducibility via extensive documentation and provision of original input data, which in many cases are unpublished and difficult to access by other means (especially for historical information).

In addition to regular updates, HMD data series are extended backward in time whenever additional data are identified.

Users are encouraged to contact the HMD staff when they feel the need, whether because of difficulties in accessing the data or because of particular questions or concerns about specific country series or data points. The Frequently Asked Questions section contains information on matters of general interest.

One of the main weaknesses of the HMD is a consequence of its strict data requirements in terms of both format and quality. These requirements have limited the number of national populations that could be included in the database.

Another weakness of the database is the lack of more detailed information. In particular, the database does not include any information about the deceased’s characteristics except for age and s*x. Thus, HMD mortality estimates are not available by socioeconomic status, because the necessary data are not available in a consistent manner over time and across countries. Nor is cause-of-death information available at this time. For various reasons, data series in the HMD are not generally available at a finer geographical level than national. However, as indicated below, we and our collaborators are in the process of adding sub-national data series for selected countries.