Evaluation of the impact of various diseases on mortality*

RESOURCE ALLOCATION Evaluation of the impact of various diseases on mortality* KJELD KJELDSEN 1 The scarcity of resources within the health sector creates serious allocation problems. If it were legitimate to allocate the available means to the various diseases according to their impact on the length ofhuman life, how should this impact be evaluated? That is the purpose of this study, which compares and evaluates methods for measuring the importance of various diseases from the point of view of their lethal effects and significance for life expectancy. Three methods of evaluation have been used: the percentage of all deaths due to the disease (method A); the gain in life expectancy that would occur if the disease were eradicated as a cause ofdeath (method B); and the change in the ratio between productive and nonproductive groups that would result from eliminating the disease as a cause ofdeath (method C). The analyses are based on the total number ofdeaths in Denmark in 1969 and 1971. According to (A), one-third ofall deaths was caused by heart diseases, one-fourth by cancer, and one-tenth by stroke. The results ofmethod (B) were in agreement with those of method (A); the greatest gain in life expectancy was obtained by eliminating cardiac diseases; cancer ranked second, and accidents third for men and fourth for women. Method (C) yielded contrasting results. The result of eliminating most diseases would be a decrease in the proportion ofpersons ofproductive age. This was most distinct for cardiac diseases. The most important exception to this rule was accidents: if these were eliminated, the distribution of the population in the productive and nonproductive age groups would remain stable. The health sector in Denmark has for years received almost all the resources that were requested by the physicians and population. However, this situation may now be changed as a result of the so- called " perspective plan " (1). This plan, worked out in 1971, forecasts state activities until 1985. A special section deals with the health sector and indicates that in future we have to plan and assign priority for the allocation of resources to the health sector. However, no criteria were suggested for the alloca- tion of funds and a number of questions therefore arise. How much of the national product is to be allocated to the health sector? How much is to be spent on treatment, on prevention, and on research? What methods should be applied for the correct distribution of resources among the various diseases? True enough, attempts have been made, since the * From the Danish Institute for Clinical Epidemiology, Svanem0llevej 25, 2100 Copenhagen 0, Denmark. I Senior statistician. late 1950s, to establish methods by which these questions could be answered (2, 3, 4, 5, 6). The fact that these questions still are open for discussion is due mainly to the difficulties in measuring the effect of investment in health. The resources spent on various parts of the health sector can be measured rather easily, but the unsolved problem is the mea- surement of the outcome-the benefits of the cost. If it were legitimate to allocate the available means to the various diseases according to their impact on the length of human life, how should this impact be evaluated? The purpose of this study is to compare and discuss methods of measuring the relative importance of various diseases as regards their lethal effects and significance for life expectancy. MATERIALS AND METHOD At each death, a certificate is issued stating the diseases leading to death. The medical information is coded in the National Health Service, the demo- 3361 - 369 - BULL. WORLD HEALTH ORGAN., Vol. 52, 1975 K. KJELDSEN graphic data in the Statistical Department. The coding of the causes of death is carried out accord- ing to the rules given in the eighth revision of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD) (7). The disease that was the primary cause of death and any other (secondary) diseases having contributed to death are coded, but the present study is based only on the main (primary) cause. An analysis for each of the 900 diseases included in the international classifica- tion was not possible because the number of deaths was too small. The diseases have therefore been concentrated into the following 7 main groups, the first 6 of which relate to clinical entities, whereas the seventh is a miscellaneous group containing a variety of conditions. (1) Infectious diseases (ICD Nos. 000-136 and 320). This group contains all epidemic diseases; an exception is influenza, which is included in group 5. (2) Cancer (ICD Nos. 140-209). (3) Diseases of the heart (ICD Nos. 390-429). This group contains heart and vascular diseases apart from stroke. Thrombosis and embolism in the veins and sclerosis of the arteries other than the cerebral and coronary arteries are not included here but are listed in group 7. (4) Stroke (ICD Nos. 430438). (5) Diseases of the respiratory system (ICD Nos. 470493). (6) Accidents (ICD Nos. E 810-999). This group includes suicides and all kinds of violence. (7) All other diseases (ICD Nos. 210-315, 340-389, 440-466, and 500-796). The importance of the diseases is measured by three different methods: A. By calculating the percentage of all deaths due to the disease. The higher the percentage, the higher the significance. B. By calculating the gain in life expectancy that would result from eliminating the disease as a cause of death. The larger the gain, the greater the importance of the disease. C. By calculating the effect of eliminating certain diseases as causes of death on the ratio between productive and nonproductive groups. The more productive the group, the more important it is to eliminate the disease from it. (If a disease is eliminated as a cause of death, this will in most cases result in a change in the age structure of the general population. Most often the effect will be a substantial increase in the number of persons in the older age groups. In other words, the elimina- tion of a disease will generally imply that there will be relatively fewer productive persons, whose duty it will be to support the unproductive.) For method A, the number of deaths during the years 1969 and 1970 (96 176) is used (8, 9). For method B, the gain in life expectancy is found by subtracting the observed number of years from that expected, which is calculated on the assumption that the disease has been eliminated as a risk of death. The observed expectation of life is taken from the official statistics and is calculated on the basis of the mortality registrations for the years 1966-1970 in the general population (10); the expected life span is calculated on the basis of a mathematical model developed by Chiang (11). The principle of the model is illustrated by an example (heart diseases). Every year about 17 000 persons die from cardiac diseases in Denmark; if heart diseases were elimi- nated as a cause of death, those persons would live longer, dying from other diseases. How much longer their lives would be is calculated by applying the new death rates that would exist when heart diseases have been eliminated. The model is based on three assumptions: first, that the risk of death from a disease is the same for all persons of the same sex and age group; secondly, that the force of mortality from a disease is independent of other diseases; and thirdly, that the ratio between the force of mortality from a specific disease and the total force of mor- tality is constant within each age group. Even though these assumptions may not be borne out in practice, the method will give a good impression of the relative importance of the various diseases. For method C, the percentage of the population of productive age (15-69 years) is calculated on the basis of the results of method B. Expressed in life table terms, the percentage is: x = (115ei5-170e70)/100000 eo where I is survival, In is the number out of 100 000 births who are alive at age n, e is life expectancy, and en is the average number of years yet to be lived by a person at age n. RESULTS Frequency of the various causes of death (method A) During the years 1969 and 1970, a total of 96 176 people died in Denmark; 52 201 (54%) were males 370 MORTALITY FROM VARIOUS DISEASES and 43 975 (46%) females. Table 1 shows that 3o X .-o cancer, heart disease, and stroke caused 68% of all deaths in both males and females. The three diseases X 0 0 0 08 had different rates for the two sexes: heart diseases caused 37% of deaths in males but only 32% in o6 4 00 o V- N O females; cancer accounted for 22% and 24% of z C LO- N-t deaths, respectively, and stroke, 9% and 12%. Acci- 0dents caused 8% of all deaths in males and 6% in 4 0 N O ° females; diseases of the respiratory system were . 0) ~ ~~cmco r c o 0 responsible for 6% and 5% of deaths, respectively. 0 X- - 0.Infectious diseases caused I1% of deaths, the per- :,<:.LO XN-OwNc centage being the same in the two sexes. The remain- @ z N C) Lo10 m LO 10 o In ~~~~-N"t c N1LO coing deaths, 17% for males and 20% for females, a) were caused by all other diseases. E CN Qo ' oQ The pattern of causes of death varies consider- @ " ably according to age. In children (up to the age of C . o e 15 years), accidents accounted for 20% of deaths in a .° boys and 18% of those in girls; cancer caused 5% 6z o N 0 o i - 2 C' le)N ')0 2-E~~~~~~~~~~~~~~and 7%, respectively, whereas heart diseases and X stroke caused less then 0.5%. In the productive age Q 4 tN group (15-69 years), heart disease was the primary <"a cause of death in males (34%), whereas cancer was 0 E e 10 10 C the first cause for females (39%). In older persons 0 Co (over 69 years), heart disease was the most frequent ' U) U) N- cause of death in both sexes, causing 42% and 38% co z 1-XN of deaths, respectively. 04 0 0u o o Table 1 gives the age distribution for each cause of XaN 0 0 death (the percentages are given in italics). The > O e) oIXX o 0 N distribution varies considerably from one disease to x2 another. Thus, among persons dying from stroke, C' LO 0 et0 o few were children (less than 0.5% for both boys and °z V" C 0o girls), about one-fifth were of productive age (25%n C~ ~0, oe ) 0for men, 18% for women), and the remainder were no0N 0| old (75% for males and 82% for females). Of those 0 8 0 N. r- 0W- CO N cowho died of accidents (9% for young boys and 8% W 8|x lN for young girls), the majority were of productive age O10 o N c o co (70% for males and 52% for females), whereas the 0_ 6 x 0X o rest were old (21 % for men and 40% for women). ' co O~~~~~~~~~~1 CV to Q,o1NN 0) toC~C 0 Gain in life expectancy (method B) 8ae 0 Fig. 1 and 2 show the gain in expectation of life 8 N N - N that might be expected if each of the 7 diseases were -E . XNco C CD X E eliminated. The gain and the significance of the a z a N N 10 o 0 different diseases vary highly according to age and, _ to some extent, also to sex. For new-born boys the a) ) V .0 0ranking order is as follows: heart diseases (4.8 E ' -, 0 .2 tyears), " all other diseases " (3.1 years), cancer (2.8 8 0e LO - v- years), accidents (1.9 years), stroke (0.9 years), dis- X -,C eases of the respiratory system (0.7 years), and infec- - z6 %_ v a. tious diseases (0.2 years). For new-born girls the order is: heart diseases (3.5 years), cancer (3.2 years), C*|- + - '-b + "all other diseases " (3.1 years), stroke (1.2 years), _ 0, o o i-0 2 371 K. KJELDSEN DOIS. OFTH HER INFETIO 5 DI E ES _ _ ,_ _ _ *.'' MALES ......... ... .... ....... . - AGE IN YEARS Fig. 1. Gain in expectation of life at a given age owing to the elimination of certain disease groups (males). INFECTIOUS D15EJS S 0l 20 30> 40 s0 60 .70 80 90 100 AGE IN YEARS Fig. 2. Gain in expectation of life at a given age owing to the elimination of certain disease groups (females). accidents (1.1 years), diseases of the respiratory system (0.5 years), and infectious diseases (0.2 years). The order and impact of the diseases were thus similar for the two sexes. With increasing age, the profile shifts as a result of the varying attack rate: those who die from accidents, for example, are mainly young. The elimination of accidents is there- fore most important for children and young adults, whereas it is of minor significance for older persons. As a result, the gain falls with age. On the other hand, stroke is an illness mainly afflicting the middle-aged and elderly; therefore, the elimination of stroke will result in the same gain for nearly all age groups. The ranking order consequently changes with age. Thus for 70-year-old males, it is: heart diseases (3.6 years), cancer (1.4 years), " all other diseases " (1.1 years), stroke (0.8 years), diseases of the respiratory system (0.4 years), accidents (0.2 years), and infectious diseases (0.0 years). For 70-year-old females, the order is: heart diseases (3.3 years), " all other dis- eases" (1.4 years), cancer (1.2 years), stroke (1.1 years), diseases of the respiratory system (0.8 years), accidents (0.3 years), and infectious diseases (0.0 years). Compared with the profile among children, heart disease thus still ranks first for men, but cancer, stroke, and diseases of the respiratory system have climbed up one step and " other diseases " have moved down one step-accidents have even re- treated two steps. For females, heart disease is still number one; " other diseases" and diseases of the respiratory system have moved up one step, and cancer and accidents have moved down one step. Percentage of the population in the productive age (method C) On the basis of the death rates during the period 1966-70, it is forecast that a new-born boy will be of productive age (15-69 years) for 70% of his life and of nonproductive age (childhood or old age) for 30 %; for females, the corresponding percentages are 68 and 32 (Table 2). Elimination of any of the 7 diseases would have a decreasing effect on the percentage of productive years. There are only two exceptions: accidents in males and infectious diseases in females. In these cases, the distribution in productive and nonproduc- tive years would remain the same as at present. If heart diseases were eliminated, the productive part of life would fall from 70% to 67% for males from 68% to 65% for females. In other words, the years spent in the unproductive period would in- crease from 30% to 33 % for men and from 32% to 35% for women. With the elimination of cancer, the proportion of persons of productive age would also fall and the proportion of those in the nonproductive period would increase, but the changes are smaller than for cardiac disease (1.3% for men and 1.4% for women). Whereas the decrease in the productive group is higher for males than for females as regards heart diseases and cancer, the opposite applies to stroke. The change from productive to nonproduc- tive age is 0.6% for males and 0.8% for females. As regards diseases of the respiratory system, the change is small: 0.4% for men and 0.2% for women. IMPACT OF DISEASE GROUPS MEASURED BY METHODS A, B, AND C In order to compare the significance of the differ- ent diseases according to methods A, B, and C, they have been matched, one by one, against heart disease (cf. Fig. 3). The set-up of this figure can be illus- 372 MORTALITY FROM VARIOUS DISEASES Table 2. Distribution of the population in the productive and nonproductive age groups after elimination of various causes of death Males Females Status productive non- decrase productive G non- decrease a ae productive in prdutie prdcie nage age productive age age productive % age group % age group present 69.9 30.1 . 67.9 32.1 after elimination of: infectious diseases 69.8 30.2 0.1 67.9 32.1 - cancer 68.5 31.5 1.4 66.6 33.4 1.3 diseases of the heart 66.8 33.2 3.1 65.3 34.7 2.6 stroke 69.3 30.7 0.6 67.1 32.9 0.8 diseases of the respiratory system 69.5 30.5 0.4 67.7 32.3 0.2 accidents 69.9 30.1 - 67.8 32.2 0.1 all others 69.0 31.0 0.9 66.8 33.2 1.1 a Difference between present status (productive age) and status after elimination of cause of death(productive age) trated by an example: according to method A, acci- dents accounted for 8% and heart diseases for 37 % of all deaths in males (Table 1). The ratio of acci- dents to heart disease is therefore 8: 37 = 22%. According to method B, the elimination of accidents would add 1.9 years to the life expectancy of the new-born; similarly, 4.8 years would be added by eliminating heart disease (Fig. 1); the ratio of acci- dents to heart disease is thus 1.9 : 4.8 = 39%. The impact of accidents according to method C was 0% and of heart diseases -3.1% (Table 2); the ratio here is 0: -3.1 = 00%. Fig. 3 clearly shows that no other disease had as high a score as heart disease, according to meth- ods A and B. On the other hand, the elimination of cardiac disease also had the highest negative score when method C was used, causing the greatest change in the population's age structure. It is also seen that the significance of the various diseases differs according to the method: stroke and diseases of the respiratory system are ranked high with method A, whereas infectious diseases, accidents, and " other diseases " have the highest score accord- ing to method B. According to method C, the diseases with the highest score are accidents in males and infectious diseases in females. If these causes were eliminated, the age composition of the total population would remain unaltered, whereas the |'50°EM~~~~MAES tu i:0\ INFECI CANCER 01SIoF THE STrROKE DIS. SYSTTHE CIDENTS1 ALLR -C HeAR requen as cause of death W Gain in eapecotian of life -lOX b . _ _~~~~~~Population in productive age oreaps * Di oeses of the hseri.IOO% Fig. 3. Comparison of three methods of evaluating the impact of a disease group. elimination of any other disease would have a negative effect. DISCUSSION In order to evaluate the severity of a disease from a public health point of view, it is necessary to know how many people suffer from the disease; how long they are sick and need treatment; how high the relapse rate is; and how lethal the disease-i.e., to what extent is the patient's life expectancy shortened by the disease? These questions, although simple 373 K. KJELDSEN and fundamental, remain unanswered for most dis- eases. Only for tuberculosis are well-documented data available (12). Therefore, a burning need exists for producing such data, so that the above-men- tioned parameters may be estimated. These can then be included in a more comprehensive evaluation of the various diseases. The present study is limited to lethal diseases, and only their impact on mortality is considered. But a considerable number of nonlethal diseases, such as deafness, skin diseases, and periodontosis, burden mankind. They can substantially reduce the quality of life; hence they are also of high significance for public health evaluation. A large survey has been undertaken to develop methods for the evaluation of such diseases and conditions (13, 14, 15), but so far such methods have remained inapplicable in prac- tice. As long as the burden of different diseases cannot be measured by a health index, their impact must be evaluated indirectly by other parameters, such as mortality, the expectation of life, or changes in the age distribution of the population. The present study showed that the effect of eliminating a disease, according to method C, would have been negative in many instances, since more and more persons would have been of nonproductive age. In the industrial- ized countries, this " side-effect " could probably be solved economically. As the present results refer to the situation in an industrialized country, they may not apply to the developing countries, where the disease pattern is different and where younger age groups are frequently attacked by severe diseases. In those countries, it seems most likely that almost any eradication programme will cause an increase of the population of productive age-a change with human as well as economic repercussions. Even though there are shortcomings in this study as a result of the bias in the data, it may still be used as an indicator of the significance of the various diseases. It can therefore be part of the basis needed for giving priorities to research, hospital services, and prophylaxis. In this context, two other factors should also be considered-i.e., what is possible as regards therapy, and what the cost will be. Many would be tempted to adhere to existing possibili- ties-i.e., to take action only when effective treat- ment exists or where it can be carried out at a reasonable cost. But even if the possibilities of today are small or nonexistent, it would be challenging to invest research in areas where the results would be the most beneficial from the general point of view. ACKNOWLEDGEMENTS The study was supported in part by a WHO fellowship. The valuable assistance of Mrs J. Tonacia, Department of Epidemiology, School of Public Health, Johns Hopkins University, Baltimore, MD, USA, is gratefully acknowledged. R1iSUMt tVALUATION DE L'INCIDENCE DES DIFFRENTES MALADIES SUR LA MORTALITt Les ressources etant limitees, les questions suivantes se posent: Quelle part du produit national faut-il affecter au secteur de la sante? Combien faut-il d6penser pour le traitement, la pr6vention et la recherche? Et quelles methodes convient-il d'appliquer pour proceder a une repartition judicieuse des ressources entre les differentes maladies? Depuis la fin des annees 50, on a tent6 de definir des m6thodes permettant de repondre a ces ques- tions. Toutefois, le debat reste ouvert. Pour repondre a ces questions, on s'est servi de trois methodes permettant de mesurer l'importance des diffe- rentes maladies: A. le pourcentage des d6ces dus a la maladie; B. l'augmentation de l'esperance de vie conse- cutive a l'elimination de la maladie en tant que cause de deces; et C. Ia modification du rapport entre les pro- ductifs et les non productifs consecutive 'a l'elimination de la maladie en tant que cause de deces. Selon la m6thode A, un tiers des deces sont dus a des cardiopathies, un quart au cancer et un dixieme a des accidents vasculaires cer6braux. La methode B a montre que l'esperance de vie augmenterait de 5 ans pour les hommes et de 3,5 ans pour les femmes si l'on parvenait i eliminer les cardiopathies, alors que pour le cancer, les accidents vasculaires cer6braux et les accidents, les chiffres seraient respectivement de 2,8 et 3,2 ans, 0,8 et 1,2 ans et 1,8 et 1 an. Pour les maladies de l'appareil respiratoire et les maladies infectieuses, l'augmentation serait inf6rieure a un an. La m6thode C a montre que le changement le plus important aurait trait aux cardiopathies. En effet, si 374 MORTALITY FROM VARIOUS DISEASES 375 celles-ci etaient eliminees, la proportion des productifs passerait de 70 a 67% pour les hommes et de 68 a 65% pour les femmes. Pour le cancer, la proportion passerait de 70 it 68% pour les hommes et de 68 a 67% pour les femmes. Pour les autres maladies et les accidents, la modi- fication serait inferieure a 1% ou meme nulle. En definitive, l'incidence des differentes maladies sur la mortalite se mesure en fonction de l'importance des cardiopathies. En effet, ce sont elles qui ont la plus forte incidence lorsqu'on les mesure 'a I'aide des methodes A et B, et 1'effet le plus negatif lorsqu'on utilise la methode C. REFERENCES 1. Perspektivplanlcegning [Perspective planning] 1970-1985. Copenhagen, Schultz, 1971. 2. FEIN, R. Health programmes and economic develop- ment. In: Proceedings of the Conference on the Economics of Health and Medical Care, Ann Arbor, MI, 10-12 May 1962. Ann Arbor, University of Michigan, 1964. 3. RicE, D. Estimating the cost of illness. Washington, US Governiment Printing Office, 1966 (Health eco- nomics series, No. 6; PHS publication No. 947-6). 4. GROSSE, R. M. Cost-benefit analysis in disease control programmes. In: Kendall, M. G., ed. Cost- benefit analysis. Proceedings of a symposium beld in The Hague, July 1969, under the aegis of the NATO Scientific Affairs Committee. London, 1971, pp. 17-34. 5. HIMATSINGANI, C. Approaches to health and per- sonal social services planning in the National Health Service and the place of health indices. International journal of epidemiology, 2: 15-21 (1973). 6. PoLE, J. D. The use of outcomes measures in health service planning. International journal of epidemiol- ogy, 2: 23-30 (1973). 7. WORLD HEALTH ORGANIZATION. Manual of the In- ternational Statistical Classification of Diseases, Injuries, and Causes of Death. Geneva, 1967. 8. SUNDHEDSSTYRELSEN [Danish Health Administra- ion]. D0dsArsagerne i Danmark [Causes of death in Denmark], 1969. Copenhagen, 1972, p. 187. 9. SUNDEHDSSTYRELSEN [Danish Health Administra- ion]. D0dsArsagerne i Danmark [Causes of death in Denmark], 1970. Copenhagen, 1973, p. 179. 10. DANMARKs STATSTK [Danish Statistical Office]. Sta- tistical yearbook. Copenhagen, 1973. 11. CHIANG, C. L. Introduction to stochastic processes in biostatistics. New York, Wiley, 1968, p. 313. 12. HoRwrrz, 0. Disease, cure and death: epidemiologic and clinical parameter for chronic diseases illustrated by a model-tuberculosis. American journal of epi- demiology, 97: 148-159 (1973). 13. FANSHEL, S. & BUSH, J. W. A health-status index and its application to health services outcomes. Opera- tions research, 18: 1021-1066 (1970). 14. SULLIVAN, D. F. Conceptual problem in developing an index of health. Washington, DC, US Govem- ment Printing Office, 1966 (PHS Publication No. 100, Series 2, No. 17). 15. CHIANG, C. L. & COHEN, R. D. How to measure health: a stochastic model for an index of health. International journal of epidemiology, 2: 7-13 (1973).