MORTALITY TRENDS
• Trends in national mortality rates •
How this website came about
My name is Gary Whitlock, and I am an epidemiologist from New Zealand
(Aotearoa) working at Oxford.
In the past few years I have made more graphs and diagrams than I can count, some for work, but many more purely for pleasure: pleasure in how they look, and pleasure in what they can say.
In the hope that other people might enjoy them too, I have begun to put some of them on the web.
The material can be found on three websites: the one you are currently viewing, which shows time trends for mortality rates; another called the Graph Bank, which is a repository of my favourite graphs, maps and diagrams from any source; and a third called Football Graphs, which—for fun—depicts the historical performance levels of English and Welsh football clubs.
How to interpret the graphs on this website
Trends in mortality rates can reflect changes in disease occurrence or treatment, in which case they represent something real about a population's health. However, trends can also reflect alterations in how causes of death are defined or coded. These are artefacts, and are of little interest except simply to know they are there. If on one of these graphs you see a jagged corner, a crooked step, or a fleeting dip, very probably the excursion is an artefact, not anything of real significance. Many such artefacts are marked here by dotted lines, but most are not (there are simply too many).
A common artefact is that which results from changes in how the underlying cause of death is coded for diseases that are themselves sometimes complicated by other life-threatening conditions. For example, acute stroke is often complicated by pneumonia, and this is sometimes a fatal development. For such deaths, the underlying cause might have been coded as stroke in one period, but as pneumonia in another. On this website, mortality rates for specific causes are for what was considered to be the underlying cause at the time the death was recorded.
Not all genuine trends are born equal: some are of major public health signficance, but most are not. The key to gauging the significance of a particular trend is to measure the absolute size of any difference using the numbers shown on the vertical axis. For example, a fall in the annual mortality rate for a particular disease from 500 to 400 per 100 000 would be an even greater improvement than a fall for another disease from 10 to 1 per 100 000 (even though it is a fall of just 20% compared with 90%), as out of every 100 000 it would save 100 lives a year compared with 9.
Copying graphs
As long as your purpose is not commercial, you are welcome to copy any of the graphs on this website.
Most of the graphs were produced using data from the World Health Organization (WHO) and the United Nations Population Division ("World Population Prospects: The 2006 Revision"), but apart from providing the relevant data, these two United Nations bodies had no role either in the production of the graphs or in their interpretation on this website. For all of these graphs, an appropriate acknowledge would be:
"From www.mortality-trends.org (based on WHO and UN Population Division data)"
For the few graphs based on data from other sources (eg, the Human Mortality Database), an appropriate acknowledgement is shown below the relevant image.
Under no circumstances will permission be granted for any of these graphs (or information related to them) to be used for commercial purposes.
Info about causes of death
Further information about different causes of death is given here.
Technical details
The graphs show time trends in mortality rates. For each year, mortality rates are calculated as the number of deaths divided by the number of people in the relevant population, and then multiplied by 100 000. The relevant population is the group of people who would have been counted among the deaths had they died in that year. Thus, the population is only of males if the deaths are male, and only of those aged 60-69 years if the deaths are for that age range.
Mortality rates are standardised for age by taking the unweighted mean of the component 5-year rates. (For example, the rate for age 60-69 is the unweighted mean of the rates for 60-64 and 65-69; a slight exception is for age 1-14, where the where the age 1-4 receives only four-fifths of the usual weight.) All-cause mortality rates standardised in this way can be used to estimate the probability of death in a particular age band. The probability is 1 - e-Y.R/100 000, where Y is the number of years in the age band, and R is the standardised mortality rate. For example, the standardised mortality rate for UK men at ages 50-59 in 2005 was 593 per 100 000 (actually a smoothed rate—see below—but the principle generally holds), so the probability of a 50-year old man dying before 60 was, at 2005 rates, 6%. The corresponding probability for UK women was (based on a rate of 384 per 100 000) 4%. Half a century earlier the probabilities were, respectively, 13% and 7%.
Standardised mortality rates are then smoothed before being plotted. The smoothed rate for a particular year is a weighted average of the rate in that year (weight = 3), the preceding year (2), and the year before that (1). The smoothed rate for a particular year was not calculated if any of these three components was missing. (This is why graphs for many countries start plotting at 1952 even though mortality rates were generally available from 1950.)
Frequently asked questions (FAQ)
Q: How have these graphs been produced?
A: By a custom-made programme written in VB .Net.
Q: Why show mortality trends just for particular ages?
A: Mortality trends which are not for particular ages often hide important differences between trends at different ages.
Q: Why show mortality trends for these particular ages?
A: The age groups correspond reasonably well to key phases of life—phases that tend also to have distinctive patterns of mortality. For example, the most common causes of death in middle age tend to be the same throughout middle age, while tending also to be different from the most common causes of death in early adulthood. (Under the Choose and Random options, no information is given for deaths over age 80 years because information about cause of death is less reliable at those ages, particularly for historical trends.)
Q: Why not give mortality trends for males and females combined?
A: As with age, mortality trends sometimes differ so much between males and females that combining them conceals more than it reveals.
Q: Why do the sidebars compare rates for 1955, 1975 and 2000?
A: The choice is arbitrary. The rates are actually 3-year average rates: eg, the 1955 rate is a weighted average of the rates for the individual years 1953, 1954 and 1955. No corresponding average rate could be calculated for 1950, which is why the earliest comparison year in the sidebars is 1955 rather than 1950.
Q: How can a sidebar report that there were no deaths from a particular cause in a particular year (eg, 2005), yet the rate per 100 000 is not 0?
A: The reported number of deaths is for the specified year only( eg, 2005), but the rate is averaged across that year and the previous two (eg, 2003, 2004 and 2005). So if there were deaths in one or both of the preceding years, then the rate will not be 0. This anomaly arises for causes of death which are especially rare in the specific age group.
Q: Some sidebars say "percent of all male deaths" or "percent of all female deaths". What is meant by this?
A: What is meant is the percentage of all deaths (ie, at any age) in that year.
Q: Why do graphed lines start in the early 1950s for some countries but not for others?
A: Some countries started providing data to the World Health Organization later than others. Poland started started providing data in 1959, but its graphs start in 1967 (representing the average rates for 1965-67) because of doubts about the accuracy for particular mortality rates before 1965.
Q: How does the Random option work?
A:
It randomly selects a graph from the amongst hundreds of different graphs that can be called up individually
up under the Choose option.
However, it deliberately does not select graphs for residual categories (eg, "other vascular diseases, "other respiratory diseases")
or graphs with especially few deaths in the relevant period.
Contact
Enquiries and suggestions can be emailed to: info@mortality-trends.org
Useful links
World Health Organization (WHO): a major source of data for mortality-trends.org
United Nations Population Division: a major source of data for mortality-trends.org
Human Mortality Database: an additional source of data for mortality-trends.org
Clinical Trial Service Unit (CTSU): a research unit at Oxford University
Deaths from Smoking: the toll of tobacco deaths in different countries
National Obesity Observatory for England: useful information on obesity
Understanding Uncertainty: making sense of chance, risk and probability
Gapminder: animated graphs making a wide range of international comparisons
Graph Bank: a repository of interesting graphs, maps and diagrams
