—from Epidemiological Bulletin, Vol. 24 No. 2, June 2003—

Techniques to Measure the Impact of Mortality:
Years of Potential Life Lost

Introduction
Mortality data represent essential elements for the quantification of health problems. Death counts and related rates are among the simplest indicators to analyze mortality. They represent a summary measure of a population’s mortality experience that may be used to establish and monitor health priorities or objectives. However, general crude or adjusted (1) mortality rates are highly influenced by the health problems of the more advanced age groups, where most deaths occur. As a result, when they are measured only with rates, causes characteristic of the oldest age groups often appear as a population’s first causes of mortality.

A main objective of public health work is to increase life expectancy in the best health conditions possible.(2) It is therefore important to identify and monitor mortality trends at all ages. This article presents a review of a technique that reflects more precisely the mortality experience of the younger age groups and gives more weight to deaths that occur at a younger age. This technique is that of the Years of Potential Life Lost (YPLL).

YPLL is a measure of the relative impact of several diseases and health problems in a society,(3) which illustrates the losses suffered as a consequence of the death of young people, or premature deaths. A death is considered premature when it occurs before a given predetermined age, for example the life expectancy at birth in the population under study. Considering the age of death rather than the mere event of death allows assigning a different weight to deaths that occur at different moments of life. The underlying assumption for the YPLL is that the more “premature” a death (i.e. the younger the person when he/she dies), the greater the loss of life. YPLL for a specific cause of death in a population are the sum, in all the persons that die of this cause, of the additional years they would have lived if they had survived to their life expectancy.

The objective of this indicator is to provide a wider view of the relative importance of the most relevant causes of premature mortality and it is used primarily in the planning and definition of health priorities.(4)

Methodological considerations
The indicator of YPLL for a specific cohort is calculated from the total number of years of life that people who die prematurely have not lived. It is the cumulative sum of the differences between the age at death and the selected age limit (superior limit). In general, an age limit of 70 years is used; however, other ages, or even the life expectancy of the population under study, can also be used. For populations with a high life expectancy, choosing a relatively low age limit can be a drawback, since in that case, age groups or causes of death that can provide important information on the state of health of the older population groups will be omitted from the calculation. For populations with lower life expectancy, it is obviously recommended to use a lower limit, for example of 65 years.

The use of life expectancy at birth as an age limit for the YPLL adjusts the calculation to the population profile of the country or area. The problem with this approximation is that the YPLL will not be comparable with that of other populations with different life expectancies. This is very important to remember to avoid making comparisons between two or more territories if the criterion used is different. The YPLL cannot be used to compare two or more situations if the criterion for calculation is not the same. In short, the final decision on the cut-off point is relatively arbitrary and depends on the objective of the analysis, if it is carried out for purposes of studying a single population or for comparisons between several populations.

In calculating YPLL in an entire population, the causes of infant mortality may represent an important weight in the indicator. However, it is recommended to include all the age groups starting at 0. In any case, if this determination is made at different age intervals (for example, adults between 25 and 65 years), this should be clearly indicated. The choice of the age range for calculation of YPLL will depend on the purpose of the study. If the YPLL is used in a study of maternal mortality, for example, the group of women between 15 and 50 years old can be included, considering causes related to maternal deaths exclusively.

YPLL is obtained by summing the products of the number of deaths at each age by the difference among this age and a set limit (See Box 1). This sum is expressed in years lost. Depending on the availability of data, the calculation of years lost can be done with individual deaths or deaths aggregated by age groups. In this case it is assumed that the deaths occur uniformly in the age group, which means that there can be some differences between the calculations using individuals and groups of individuals. However, the calculation is usually done using grouped data and it is considered that the deaths occur at the age group median. It is recommended to use 5-year or 10-year age groups so that the assumption of a uniform distribution of deaths is more realistic.(4) The result of the YPLL divided by the population (usually the population below the age limit selected) and multiplied by a factor (1,000, 10,000 or 100,000) is an index defined as Years of Potential Life Lost Index (YPLLI).

It is important to keep in mind that two populations of different sizes experiencing different mortalities may produce a similar absolute number of YPLL. To obtain a more complete panorama of the situation, it is therefore important to calculate the absolute number of YPLL along with the YPLLI.

The YPLL has the advantage of being easy to calculate, since it requires only deaths by age and the total population. If deaths are available by cause of death, YPLL can be calculated for every cause. However, as for any study based on data of highly variable quality, the quality of the indicator will depend on the quality of its components. It is also important to take into account that the age structure of the population affects this indicator. Standardization techniques, which were presented in a previous issue of the Epidemiological Bulletin,(1) can be applied to the calculation of YPLL in order to control for the effect of confounding variables. However, this adjustment should not be made in place of the decision-making process by which an upper age limit is selected for the calculation of the YPLL, as mentioned in the previous paragraph.

To illustrate this concept, table 1 presents the calculation of YPLL and YPLLI for all causes of death in men between 0 and 85 years in Colombia, for the period 1995-1997. The distribution of these data is shown in Figure 1, which presents YPLLI in Colombian men between 1995 and 1997. The distribution presents three peaks: one for the youngest age, one for young adults and a third one for older adults (65 and older). Although the number of deaths is similar in the three peaks, the YPLLI are 2 to 5 times greater in the younger age group (more premature deaths). Figure 2 presents the distribution of the YPLLI in Colombian men and women for the same period. The profile of the distribution by age is similar in both sexes except in young adults, where a noticeable peak can be seen in men. In terms of absolute deaths there are 1.52 deaths for men for each death for women. On the other hand, when this information is analyzed according to the YPLLI, it can be said that for every 100 YPLL for women there are 215.52 for men, the 20-24 age group shoes the greatest difference; for every 100 YPLL for women, there are 581.52 for men, i.e. 6 times more. This indirectly measures the impact of violence among young men in this country.

Figure 1: Distribution of the YPLLI in Colombian men, 1995-1997

Figure 2: Distribution of the YPLL Index per 1,000 population in Colombian men and women, 1995-1997

Table 2 presents the calculation of YPLL using a limit of 70 years in the 29 departments of Chile. Besides showing the YPLL for each department, it also presents the population, which makes it possible to calculate the YPLL Index. The department with the least YPLL and a small population (Gral. Carlos Ibáñez) is comparable in terms of YPLLI to more populated departments like Arauco and Bio-Bio. In this case, quartiles were defined where quartile 1 (25% of the departments) corresponds to the least and quartile 4 to the most health problems.(5) This example is simple and makes it possible to order the different territories and define the departments with greater risk using this mortality indicator as a health planning tool.

To summarize, YPLL may be used in different ways: looking at the value of YPLL in each group or evaluating the total for the population; calculating it by sex or for a particular population group; or studying the value of YPLL for a specific cause. Comparisons between populations or causes can be made from these values. When analyzing the YPLL by cause, it should not be inferred that the years lost due to a cause would not have been lost if the cause had been controlled in the population. Indeed, just because a death is not due to a cause does not mean that the person could not have been exposed to other risks that could have caused death as well.(4) By observing the evolution of this indicator in time, it is also possible to compare periods and carry out trend analyses. It allows uncovering and comparing populations with occurrence of premature death.

Finally, it may be commented that this indicator is also used as methodological support in the evaluation of Unnecessarily Premature and Sanitarily Avoidable Mortality (MIPSE, for its Spanish name). This topic related to mortality assessment will be reviewed in another issue of the Epidemiological Bulletin, as well as other techniques related to mortality analysis.

References:
(1) PAHO. Standardization: A Classic Method for the Comparison of Rates. Epidemiological Bulletin 23(3):9-12; 2002
(2) PAHO. Years of Potential Life Lost - Brazil, 1980. Epidemiological Bulletin 5(5):3-6; 1986
(3) Last J. A Dictionary of Epidemiology, Fourth Edition. New York, New York: Oxford University Press. 2001
(4) PAHO, Xunta de Galicia. Ayuda del módulo de demografía. EPIDAT 3.0 [Computer program]. 2003 In print
(5) Ministerio de Salud de Chile. Indicadores Comunales para el Estudio de la Desigualdad en Salud. El Vigía. 3(11):7-13; 2000

Source: Prepared by PAHO’s Area of Health Analysis and Information Systems (AIS).

Return to Index
Epidemiological Bulletin, Vol. 24 No. 2, June 2003