CAUSES OF DEATH

Information on cause of death is essential for understanding trends and inequalities in mortality. Compiling this information requires a consistent scheme for classifying causes of death and an appropriate system for registration and record-keeping. Both were developed during the nineteenth century and had become systematic in all industrialized countries at the beginning of the twentieth century. In these countries medical certification of the cause of death is routine. This is not the case, however, in most developing countries. "Verbal autopsies"–information about the symptoms and conditions which accompanied the death obtained by questioning close relatives of the deceased–can contribute some knowledge of causes of death, especially for children, but they cannot produce reliable statistics of mortality by cause.

Classification of Causes of Death

After lengthy debate, the first international classification of diseases and causes of death, largely devised by Jacques Bertillon (1851–1922), was adopted in 1893. There have been ten subsequent revisions, none of them radically changing the original structure although producing severe disruptions in time series on causes of death. The International Statistical Classification of Diseases and Related Health Problems–Tenth Revision (ICD-10), was adopted by the World Health Organization (WHO) in 1989.

Ten of the ICD-10's 21 chapters refer to a specific bodily system, such as Chapter VI, "Diseases of the nervous system," and Chapter X, "Diseases of the respiratory system." Some other chapters refer to etiological processes, like Chapter I, "Certain infectious and parasitic diseases," and Chapter II, "Neoplasms." Still others are linked to a particular period of life, like Chapter XVI, "Certain conditions originating in the prenatal period," and Chapter XV, "Pregnancy, childbirth and the puerperium."

Such a structure makes it difficult to identify homogeneous pathological processes. Trends in distinct pathologies can depend on common factors and may be influenced through appropriate intervention. To identify these processes, several authors have suggested alternative classifications, drawing on the concept of avoidable mortality. Causes of death can be divided into "avoidable" and "unavoidable." While this may be helpful in designing health policies at a particular time, it is of little value in analyzing trends as medical progress continually shifts diseases into the avoidable category.

Ideally, a useful classification should make it possible to distinguish between different etiologies. Marc d'Espine promoted this idea in the nineteenth century, in the debates surrounding the first version of the International Classification. At a time when the nature of the diseases was so little known, such an exercise would have been wholly utopian. With twenty-first-century medical science, an etiological classification could be designed–and in fact has been partially attempted, using French data. In this exercise trends in mortality from different processes (such as infectious, tumoral, or degenerative processes) could be followed more precisely. The exercise was especially useful in tracking infectious disease mortality. Although many infectious diseases are covered in ICD-10's first chapter, "Certain infectious and parasitic diseases," others are scattered through the remaining chapters. For instance, influenza falls in Chapter X, "Diseases of the respiratory system," and appendicitis in Chapter XI, "Diseases of the digestive system." Reclassifying diseases according to etiological criteria as infectious processes permits a better estimate of the weight of infection in total mortality.

Identifying Causes of Death

A death is the result of successive pathological processes that may have appeared or developed because of other preexisting conditions. Most studies on causes of death refer to only one cause. To insure some coherence in identifying this "underlying" cause, WHO recommends a model two-part medical death certificate and rather strict rules for coding. In Part I of the death certificate, the physician reports all the conditions that are directly responsible for the death in the reverse order they appeared. The first line contains the "direct" cause that immediately produced the death, and the last line the "initial" cause that induced the processes which finally led to the death. In Part II, the physician reports all other "contributory" causes that are not directly responsible for the death but which may have contributed to it. Coding rules help the physician to choose from among all these conditions the one which is considered to be the "underlying" cause of death. In most cases, this is the disease reported on the last line of Part I. However, in some specific cases the order of the pathological processes may be reconsidered by the authorities in charge of coding and another condition, reported elsewhere in the certificate, may be chosen as underlying cause of death.

The identification of only one cause of death considerably reduces the amount of information reported in the death certificate. This loss of information becomes increasingly serious under conditions of very low mortality. With very low mortality, most deaths occur at old ages to persons who may be suffering from several chronic diseases, making it difficult to choose the main cause of death. Hence efforts are being made to find ways of taking into account all the information reported in the death certificate, through multiple-cause analysis. Two approaches can be used. In the first, all mentions of a disease are noted, whatever the place they occupy in the death certificate. This approach highlights the part played in mortality by conditions like diabetes or alcoholism, which are seldom reported as the underlying cause of death but often contribute to deteriorating health. In the second approach, the most frequent associations of causes are examined, so as to identify sequences of pathological processes that are more lethal than others. Multiple-cause analysis is an important challenge for future studies of mortality and morbidity.

Problems of Comparability in Time and Space

Although nearly all countries producing regular statistics of deaths by cause use the current ICD and WHO's classification rules, comparability among countries is limited because of substantial differences in medical practice and coding habits. One such problem is in use of the category "ill-defined causes." Some countries where diagnoses tend to be imprecise assign many deaths to this category. For instance, for the year 1996, almost 12 percent of deaths in Portugal were classified into Chapter XVIII, "Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified," compared to 4.5 percent in Russia, and less than 0.1 percent in Hungary. Consequently, before making international comparisons of specified causes of death, the deaths attributed to ill-defined causes must be redistributed into specified causes. If the probability for a death being recorded as having an ill-defined cause is independent of the actual cause of death, it is possible to proportionally redistribute deaths from ill-defined causes into all the specified categories. More sophisticated methods of redistribution can also be used. Beyond the general problem of illdefined causes, international comparisons are affected by systematic differences in diagnostic practice. A case in point concerns myocardial infarction and other ischemic heart diseases: some countries, such as France, prefer the first diagnosis; others, such as the United Kingdom, prefer the second. To compare the level of mortality from ischemic heart diseases, it is better to combine the two pathologies (myocar-dial infarction and other ischemic heart diseases). In the same way, in theory ICD-10 allows one to distinguish between cancer of the cervix and other cancers of the uterus. In practice the distinction is not made on the same criteria from one country to another and a comparison of deaths classified as "cancer of the cervix uteri" would lead to erroneous conclusions. In general, in any investigation that uses a detailed cause of death, it is necessary to consider at the same time all other causes that may be confused with it.

Problems of comparability are still more serious when dealing with time trends. As with cross-national comparisons, a prior redistribution of deaths from ill-defined causes is necessary. Such categorization of deaths generally decreased as diagnostic precision improved. For instance, in France, use of the category fell from 30 percent in 1925 to 6 percent in 1996. More problematic for comparisons over time are the breaks introduced in the time series by the successive revisions of the ICD. As medical knowledge expands, the contents of the ICD are revised: new disease designations are added, and others are removed. The number of items in the ICD has risen from 203 in the first classification of 1893 to more than 10,000 in ICD-10. To observe long-term trends, it is necessary to reclassify deaths using a constant medical definition of the cause. This would be relatively straightforward if registration authorities produced a double classification of deaths under the old and new classifications whenever a revision came into effect, but that is seldom done. Thus, reconstruction of long-term cause-of-death series for any country usually requires long and meticulous work to insure medical and statistical coherence. The few countries for which this has been done include France, the Netherlands, and some countries of the former Soviet Union.

Main Trends in Causes of Death

Until the 1960s in industrialized countries, the principal contribution to rising life expectancy was the reduction of mortality from infectious diseases and the subsequent decrease in infant mortality. Following this fundamental change in the pattern of causes of death, the pace of increasing life expectancy slowed under the double effect of the emergence of man-made diseases (diseases due to tobacco and alcohol, and traffic accidents) and the growing weight of chronic diseases (cardiovascular diseases, cancer). From the 1970s, life expectancies continued increasing because of successes in controlling man-made diseases and in reduction of mortality from cardiovascular diseases, especially among the elderly. However, this resumption of progress was not general. Countries of Eastern Europe (including the former Soviet Union) lagged in the control of the chronic diseases and their life expectancy, especially for males, stagnated or even decreased.

The situation in developing countries shows even greater contrasts. Some countries, such as China, South Korea, Mexico, and Tunisia, have followed the same path as the developed world and, thanks to a rapid decrease in mortality from infectious diseases, have reached high levels of life expectancy. In contrast, countries of sub-Saharan Africa have largely failed to control infectious diseases. The emergence of AIDS and the reemergence of diseases like malaria contribute to the poor health status of these populations.

BIBLIOGRAPHY

Bourgeois-Pichat, Jean. 1952. "Essai sur la mortalité biologique de l'homme." Population 7(3): 381–394.

Holland, W. W., dir. 1993. European Community Atlas of Avoidable Death, 2nd edition, volume 2. Oxford, Eng.: Oxford University Press.

Ledermann, Sully. 1955. "La répartition des décès de cause indéterminée." Revue de l'Institut international de statistique 23(1): 3.

Meslé, France. 1996. "Les causes médicales de décès." In Démographie: analyse et synthèse. Causes et conséquences des évolutions démographiques. Paris: CEPED/DSD.

——. 1999. "Classifying Causes of Death According to an Aetiological Axis." Population Studies 53(1): 97–105.

——. 1996. "Reconstructing Long-Term Series of Causes of Death." Historical Methods 29(2): 72–87.

Meslé, France, and Jacques Vallin. 1984. "The Problem of Studying Mortality Patterns By Cause over a Long Period of Time: An Example from France, 1925 to 1978." In Methodologies for Collection and Analysis of Mortality Data, ed. Jacques Vallin, John Pollard, and Larry Heligman. Liège: IUSSP, Ordina Éditions.

Omran, Abdel R. 1971. "The Epidemiologic Transition: A Theory of the Epidemiology of Population Change." Milbank Memorial Fund Quarterly 49(4): 509–538.

World Health Organization. 1992. International Statistical Classification of Diseases and Related Health Problems, 10th revision, volume 1. Geneva: World Health Organization.

Wolleswinkel-van den Bosch, Judith. 1998. The Epidemiological Transition in the Netherlands. Rotterdam: Erasmus University.

INTERNET RESOURCES.

Vallin, Jacques and France Meslé. 1998. "Les causes de décès en France depuis 1925." Institut national d'études démographiques. <http://matisse.ined.fr/~tania/causfra/data/>.

World Health Organization. 2001. <http://www.who.int/whosis>.