Measurement and utilization of healthy life expectancy: conceptual issues.

Reviews/Analyses Measurement and utilization of healthy life expectancy: conceptual issues J.-M. Robine,l J.-P. Michel,2 & L.G. Branch3 The periodic calculation of healthy life expectancies permits the evaluation of the impact of new health policies at a given moment, as well as the assessment of trends under changing health conditions. In spite of their apparent simplicity, the results obtained will have to be interpreted by experts. Useful ref- erence values can be provided by international comparisons. However, several choices remain to be made, such as (i) the types of morbidity and disability data to be associated with mortality data; (ii) the multiple indicators available; (iii) the type of observations to be recorded, i.e., "abilities" or 'perform- ances"; (iv) whether or not the recovery of lost functions should be considered; (v) the mode of compu- tation, i.e., life expectancy before the first morbid event or global healthy life expectancy; and (vi) the determination of thresholds based on either relative or absolute criteria. Determining population health status Numerous attributes have been attached to the measurement of healthy life expectancy (1). But in the absence of a clear distinction between the health status of individuals and that of populations, two questions arise that are connected and often con- fused with one another: (1) By living to a greater age, are we not individual- ly becoming more and more weak and decrepit? (2) As more people survive at every age level, are we not collectively becoming less strong and less effective at each increasing age? Any calculation of healthy life expectancy must take into account the health status of the population and determine whether this improves over a period of time. A calculation of cross-sectional healthy life expectancy answers the first question given above I Senior Fellow, National Institute of Health and Medical Research, Montpellier, France. 2 Head, University Geriatric Institutions, Route de Mon-ldee, 1226 Th6nex, Geneva, Switzerland. Requests for reprints should be sent to this author. 3 Director, Long-term Care Research, University Medical Center, Boston, MA, USA. Reprint No. 5345 for a fictitious cohort of individuals who would be subject, at all ages, to various health influences during a chosen period (e.g., one year) - influences such as forces that determine mortality, morbidity, disability, and cure or recovery. This calculation can- not measure (1) the health status of the individual because each member of the cohort would have experienced or will experience a different history, or (2) the health status of the population, which during the chosen period would involve multiple cohorts, but it can measure the conditions influencing the population's health during the given period. Repeat- ing the cross-sectional calculation would indicate the changes in healthy life expectancy of that same ficti- tious cohort when subjected at each age period to specific health conditions over many years in succes- sion. A comparison of the figures obtained for these successive cross-sectional life expectancies provides an answer to the second question. It is well known that in Western countries the successive cross-sectional life expectancies have increased greatly during the present century. The conclusion from this is that the conditions for a longer life are evolving favourably. But what about the healthy life expectancies in the same moment and conditions, if it had been possible to calculate them? Would they also have increased and, if so, at the same rate or more quickly or less quickly than the life expectancies? Bulletin of the World Health Organization, 70 (6): 791-800 (1992) © World Health Organization 1992 791 J.M. Robine et al. The various cases As regards each cohort exposed to different succes- sive sets of health conditions, (i) is there an absolute or relative "compression" in the number of years of bad health within the life expectancy (2), or (ii) has more or less of a "balance" been maintained between the numbers of years of good and bad health, or (iii) is there a general "expansion" in the number of years of bad health (3)? Unfortunately certain results show that under successive time periods the disability-free life expectancy will increase less quickly than life expectancy (1). In such cases, which are perhaps the most probable for the immediate future, the improve- ment in healthy life expectancy would, paradoxical- ly, be accompanied by an increase in the number of years lived in bad health within the total life expec- tancy. A favourable evolution in health conditions, as reflected in the healthy life expectancy of a fictitious cohort, does not preclude an increase in the number of years lived in bad health. In other words, in terms of survivors and under successive time periods there could be from one year to the next, for a given age, both more survivors in good health and more sur- vivors in bad health, and the increase in the number of the latter may be such that they represent a great- er proportion of the total number of survivors. Interpretation of the results Expectation of life and expectation of life in good health at birth. Just as the total increase in the num- ber of survivors reflects a favourable evolution in conditions that defer mortality, an increase in the number of survivors in good health would on the whole show a favourable evolution in health condi- tions. On the other hand, an increase in the number of survivors in poor health, examined in isolation, would not necessarily provide any information on the unfavourable development of health conditions, for such an increase could be encountered in every case, whether there is "compression", "balance" or "expansion". For similar reasons, the examination in isolation of the proportion of survivors in good or in bad health would not enable us to determine whether there had been a favourable or unfavourable evolu- tion in the prevalent health conditions. Every in- crease in the number of survivors in bad health must be examined in association with what is happening in regard to the number of survivors in good health, in order to obtain a precise picture of the evolution of the latter. In the case presented above, in which successive values of disability-free life expectancy increased less quickly than life expectancy, the correct inter- pretation of the data should mention the following four points: (i) conditions that defer mortality are evolving favourably, as is shown by the increase in cross-sectional life expectancy; (ii) health conditions (morbidity, disability, etc.) are also evolving favour- ably, as is likewise shown by the evolution in healthy life expectancy; (iii) however, since healthy life expectancy increases less quickly than life expectancy, the cross-sectional expectancy of life in poor health is increasing to a relatively greater extent; and consequently, (iv) the numbers of per- sons in bad health among the survivors as a whole are also increasing. Besides the above analysis of life expectancy at birth and life expectancy of a group initially in good health, in future it will also be necessary to know how to interpret the evolution of life expectancy and healthy life expectancy at any age and particularly at advanced ages. Expectation of life and expectation of life in good health at 75 years. At advanced ages, 75 years for example, the evolution of healthy life expectancy in the survivors as a whole is clearly the evolution of a weighted mean of two expectancies: the healthy life expectancy of the survivors in good health at 75 years and the healthy life expectancy of the survivors in poor health at 75 years, adjusted for the different weighting of these two groups of survivors to 75 years of age under successive prevailing numbers. Thus, a decrease in successive cross-sectional measures of healthy life expectancy at 75 years of age does not necessarily reflect a deterioration in health conditions. The decrease may be due to an in- crease in the proportion of survivors in bad health among the total who survive to 75 years. It may conceal both an increase in healthy life expectancy for survivors in good health at 75 years and an increase in the proportion of those survivors in good health at 75 years, in relation to the initial number of people (from birth) in the successive fictitious cohorts. The decrease in healthy life expectancy at an advanced age (in relation to the initial age of the cohort under consideration) may therefore conceal a favourable evolution in health conditions. In this example and under the successive cross-sectional measures, the number of survivors in good health at a given age is higher, and beyond that age these survivors live in good health for a longer period on average. Until more thorough consideration has been given to ways of interpreting the results at a particu- lar age, it should be noted that the evolution of cross- sectional healthy life expectancy in the survivors as a whole: - should not be examined in isolation; - should be examined in association with the WHO Bulletin OMS. Vol 70 1992792 Measurement and utilization of healthy life expectancy healthy life expectancy of those survivors initially in good health at the age under consideration; and should be examined in association with the evo- lution of the proportion of the survivors in good health at the age under consideration in relation to the initial cohort. Assessing health conditions during a given period Calculation of healthy life expectancy would make it possible to assess the health conditions encountered by the population at a given time, in order to find out whether those conditions improve over time and to determine the consequences of an improvement or differential deterioration in the different mortality and morbidity conditions. Theoretically, the aim is to assess the health sta- tus of the population and to find out whether it is improving over time. In practice, the calculation of healthy life expectancy would make it possible to assess the influence of health conditions encountered by that population at a given period and the evolu- tion of those conditions over time. Since it is often more valuable to assess the health conditions encountered by a population at a given time than to determine its "health status", this difference does not matter much. At first glance the health status of a population seems an attractive concept, but it is a very ambi- guous one. Modem definitions of health are very dynamic. Health, for example, can be considered as the "capacity to fall ill and get well", or the capacity to adapt, etc.a Increasingly these definitions are based on movement (the occurrence of a particular event, entering upon or coming out of a particular state) and less and less on the presence or absence of a particular sign. Thus, in the long run it may be considered that people in good health are all those whose "status", improves (or is maintained) during a period whatever their initial "status", while people in bad health are all those whose "status" deteriorates in any way, whatever the initial "status" (bearing in mind all the usual reservations on normal evolution in relation to age). A population is fixed in a given space-time situation and is composed of individuals or cohorts of individuals. The modem concepts of health are therefore difficult to tie in with the classical notion of population. The "health status" of a population during a given period is liable to be quickly reduced a According to the Constitution of the World Health Organiza- tion, "health is a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity." to the presence within it of "stocks" (e.g., of depen- dents, insane person, etc.). These stocks of a given period have very little to do with population health today. They result from a past history which differs greatly from one country to another, or even from one region to another, and which for certain cohorts has already lasted almost a hundred years whereas for others it has barely begun. Assessment of these stocks is not without interest in itself, for they provide information on the "health needs" of the population, which is not our concern here. The health conditions encountered by a popula- tion during a given period (forces of mortality, mor- bidity, disability, cure and recovery, etc.) are much more closely connected with population health today, even though they will never be completely independent of history. Thus, the effectiveness of the various present health programmes must be assessed through their effects on the conditions occurring during the period (flows, probabilities of entering or leaving a particular state) and not on stocks. It would be both erroneous and unjust to try to assess the effectiveness of a policy of maintenance at home on the basis of the prevailing number of individuals in institutions (the stock), for even if it is very effective (i.e., if the probabilities of entering an institution have been greatly reduced), such a policy cannot for a long time have anything more than a marginal effect on the stock. On the other hand, a cross- sectional calculation of non-institutionalized life ex- pectancy over the successive periods would clearly measure the reduction in the probabilities of institu- tionalization. Knowing what to measure Healthy life expectancy is theoretically a tool for checking "what is occurring" or "what is being done" at present, because it makes possible a de- tailed evaluation of present conditions, i.e., it charac- terizes a precise period (a year, for example). This tool should make it possible to compare the health con- ditions in different countries during the same period or to follow the evolution of health conditions in a single country in different periods. Calculation of healthy life expectancy does not provide information on the "needs of the population", which are more in the nature of consequences, including the con- sequences of past conditions. It is not therefore a tool for social and health planning or programming. On the other hand, health objectives can be fixed in terms of healthy life expectancies. Measuring incidences To be able to compare healthy life expectancies be- tween neighbouring countries, the most accurate esti- WHO Bulletin OMS. Vol 70 1992 793 J.M. Robine et al. mates possible must be available. This entails rejec- tion of the cross-sectional method of Sullivan (4), which provides a gross estimate of the value sought with an unknown degree of precision. This method, which is simple to use, has provided us with the first estimates of the duration of healthy life expectancy in western countries (5). The experimental work illustrates the very limits of non-standardized calculations: none of the values found are strict- ly comparable. Taken together, they give an indi- cation (a range with upper and lower limits) of the duration of healthy life expectancy in the developed countries (6 ). By combining prevalences of disability with a cross-sectional life-table, Sullivan mixes flows and present mortality conditions with the present stock of the disabled; which in itself is largely inherited from past disease conditions, i.e., every country's own particular history. It is impossible to use such a sta- tistical construction for geographical comparisons except, for example, if we wish to make a crude comparison between a few developed countries and a few developing countries. At the outset, a method should be selected that makes healthy life expectan- cy truly a period indicator and only combines period probabilities, probabilities of death, and probabilities of health. It is therefore necessary to measure the incidence. Longitudinal surveys and the observation or simulation of incidences have reversed many opi- nions derived from cross-sectional studies on the health status of elderly populations whether in regard to: (i) excess incident disability in old women (7, 8); (ii) the prevalence of years of disability in years lived in excess by women (9); and (iii) the irreversibility of acquired disability (8, 10, 11, 13). Checking the quality of the years gained at advanced ages Let us now consider the first question which was posed at the beginning of this article. By living to a greater age are we not becoming more and more weak? In other words, we are first of all worried by the consequences of the decrease in mortality at advanced ages and we should like to estimate the quality of the years gained at very advanced ages. It is therefore necessary, first and foremost, to evaluate the health conditions encountered by old people, and to do that it is essential to measure the incidence of morbid events that indicate or reflect a decrease in the quality of life of the old. Incidentally, it can be observed that the quality of years added before death, or more generally the quality of total years of life, is an issue that can be raised at any age.b Just as the calculation of life expectancy takes into account the decrease in mortality at all ages in life and summarizes the totality of present mortality conditions, calculations of healthy life expectancies would summarize the totality of health conditions. It is therefore necessary also to evaluate the health conditions encountered by the population at all ages-in infancy, during adolescence, etc. In develop- ing countries, it is probable that a combination of mortality and disability data would be of particular relevance for the adult population. Replying to questions concerning the population as a whole It is important to examine the questions for the popu- lation as a whole and not only those formulated by a particular group of individuals, such as the managers of the health system. Cross-sectional life expectancy is really a checking instrument because the data used in the calculation are clearly defined and known to all. To be able to check on present health conditions, healthy life expectancy must reflect clearly defined and precisely formulated data. For that reason it is essential to begin drawing up a list of the questions that are asked and to reflect on the source of the questions. Are adults, old people and gerontologists asking the same questions about the quality of the years lived at advanced ages? Is the way of replying to these questions the same? There is no prima facie reason: (i) for eliminating a particu- lar question; (ii) for paying more attention to a parti- cular answer; or (iii) for deciding theoretically what should be measured-complaints, disorders, symp- toms, diseases, impairments, disabilities, abilities, performances, etc. (12), particularly since in the first instances it will not be possible to reply in terms of healthy life expectancies to many of the questions posed. It is essential, therefore, to consider the ques- tions and to determine above all which of them can be answered. We believe that it is reasonable to collect incidences of clearly defined events which, once aggregated with mortality probabilities, would provide unambiguous answers to the questions. Taking into account the recovery of lost functions In the first two experimental calculations of healthy life expectancies carried out solely with present b The Associations of parents of children suffering from neuro- muscular diseases, for instance, are more and more interested in the concept of healthy life expectancy. 794 WHO Bulletin OMS. Vol 70 1992 Measurement and utilization of healthy life expectancy probabilities, i.e., active life expectancy at 65 years in Massachusetts (USA) (13) and life expectancy without confinement to the home at 65 years in Upper Normandy (France)c, acquired disabilities were considered to be definitive, without any pos- sible recovery of lost functions (7). This approach was adopted not because it is easier to calculate healthy life expectancies with two rather than three series of probabilities, or more convenient to com- bine mortality probabilities and probabilities of be- coming disabled, while neglecting the probabilities of recovering from the disability, but because in the absence of sufficient data on the incidence of recov- eries, it seemed the most reasonable. In both these studies, the absence of usable data on the recovery of lost functions (by sex and age) was linked to the size of the cohorts of the old people followed up and, in particular, the small number of subjects who were initially dependent (in the case of Massachusetts) or initially confined to their homes (in the case of Upper Normandy). More recently, several surveys of specific populations have emphasized that the pro- portion of disabled persons who regain lost functions is much larger than was generally believed (8-11). Thus, the image of an inexorable decline of health and functional capacities with age is false and out of date (10), and the idea of an irreversible devel- opment of chronic disabling diseases in old people must be revised,d since otherwise these conventional views can only serve as an excuse for discriminatory practices against old people or those with disabili- ties. Beyond experimental calculations, it would not be realistic to propose a model that neglected these recoveries, particularly as the improvement in health conditions, which is assessed with the help of preva- lent healthy life expectancies, depends as much on therapeutic actions that increase the probabilities of recovery as on preventive actions that diminish the probabilities of becoming disabled. It is essential therefore to measure not only the incidences of mor- bid events that indicate or reflect a deterioration in the quality of life of the old, as stated earlier, but also the incidences of events that indicate an im- provement in the quality of life. Attempts to simulate the incidence of confine- ment to the home on the basis of probabilities of death and probabilities of survival observed in Upper c Brouard, N. & Robine, J.M. Modelling the confinement of the elderly to the home. Paper presented at the Conference on Theoretical and Medical Biology, Angers, 16-18 September 1986. d Haan, M.N. Compression of morbidity, the Kaiser Permanente Study of Oldest Old. Paper presented at REVES, Quebec City, 11-12 September 1989. Normandy suggested that the incidence of confine- ment to the home is almost comparable in men and women. Calculation and comparison of life expec- tancy and life expectancy without confinement to the home at 65 years of age indicated that 84% of the years lived by men and only 72% of the years lived by women were free of confinement to the home.e These results were similar to those obtained in Mas- sachusetts in the active life expectancy study (13). On the basis of these results, derived from a model that neglected the forces of recovery of lost func- tions, the hypothesis was formulated that the excess disability observed in women could be explained by the difference in survival time once the disability observed had been acquired (7). Analysis of National Long-Term Care Surveys (NLTCS) confirmed that the incidence of becoming disabled is approximately the same in men and women and the same hypoth- esis was adopted to explain excess disability in women (8). The new calculations of active life expectancy incorporating the probabilities of recov- ery of lost functions show, on the contrary, that the proportion of the years lived without dependence is no higher in men than in women. This result sug- gests that the extra years lived by women, under present conditions, show the same qualities of vigour and disability as the years lived by men (9). The proportion of years lived in good health among the total number of years lived beyond a certain age depend on the three forces-mortality, disability, and recovery. However, whether the hypotheses formulated to explain the difference between the sexes in regard to these forces are jointly compatible with this new "balanced" result needs to be verified. Assessing the quality of life Because functional capacities and independence play an important role in the quality of life, it is often considered that a calculation of disability-free life expectancy or active life expectancy would assess this quality correctly. In fact, the matter is much more complicated because the quality of life is more closely connected with autonomy than with indepen- dence (14), and there is wide interindividual and intercommunity variability in the concept of "quality of life". Independent people are by definition autono- mous and active life expectancy implies a duration of life without loss of autonomy, but there is no indi- cation of how the autonomy of old people or depen- dent disabled people is measured. It would seem eSee footnote c. WHO Bulletin OMS. Vol 701992 795 J.M. Robine et al. necessary, therefore, to be able to measure the inci- dences of events that indicate a deterioration or an improvement in the autonomy of dependent persons. For example, by providing meals-on-wheels services to an old lady who finds it difficult to go out shop- ping, we may not only be creating a dependence and encouraging the development of disability, which could have been avoided by therapeutic solutions (14), but we may also be reducing the autonomy of the persons concemed, if by having no more shop- ping to do she loses the right to choose what she is going to eat (15). Close attention must therefore be paid to choosing what activities we can undertake and determining how appropriate they are by making more frequent checks on the quality of the services provided for dependent old people in regard to their autonomy. The daily activities of old people who may be having difficulties in carrying them out should be carefully assessed without favouring the creation of too much dependence. While the restoration or main- tenance of individual abilities is preferable to the provision of home help (14), there is a price to be paid for independent daily functioning. Older per- sons with increasing functional limitations over a period of years may take steps to perform satisfac- torily their daily activities! The value of the activi- ties carried out must be compared with theoretical abilities as well as the quality of life with regard to the suffering of pain, anxiety, fatigue, etc. Thus, the assessment of the quality of life during a given period requires many more control instru- ments to compare the real gains and losses. A list of such instruments, without considering their feasibili- ty, would include: (i) life expectancy without loss of autonomy; (ii) life expectancy without loss of the ability to carry out daily activities (independent life expectancy); (iii) life expectancy without loss of abilities; (iv) life expectancy without feeling suffering; These four items correspond to the following basic questions: how long can one expect to live, i.e., to function, without suffering pain, tiredness and anxiety; without loss of abilities; without deteriora- tion in performance and dependence for carrying out basic actions; and without loss of autonomy? How long can we expect to live in the community? In practice the questions asked are much less theor- etical and much more practical, e.g., how long can f See footnote d, page 795. we expect to live without being placed in an institu- tion or long-term hospital establishment? An assessment of the present conditions for in- stitutionalization, e.g., admission to and discharge from inpatient facilities, when more and more pro- grammes for support in the home are being establish- ed, is clearly required. In theory, non-institutional life expectancy is the best measure of the effective- ness of policies designed to enable old people to remain in their own homes. Its repeated calculation should show whether the conditions in favour of institutionalization in the successive periods are diminishing. When a chronological series of non-institutional life expectancies have been established for several countries, an international comparison could usefully be carried out. This could also explain why two com- parable countries with very similar cross-sectional life expectancies should show an important differ- ence in their non-institutional life expectancies. The determination of cross-sectional non-institutional life expectancy would combine the probabilities of death with the present probabilities of admission to and discharge from institutions. For international com- parisons one should make sure that the definitions of institutions are compatible between the countries concemed. But how do we define an institution? Especially now when (i) a variety of home-care services (nurs- ing, physiotherapy, etc.) for maintaining old people in their own homes are being developed, which previously could only be given in institutions; (ii) policies are being defined for making long-stay hospitals and nursing homes more like an ordinary home (for example, do the persons in the establish- ment have a personal telephone with a direct line or one going through a switchboard); and (iii) old people's homes have been developing therapeutic sections and medical beds and are more and more frequently providing care on the spot. Where can a line be drawn between ordinary homes, which are becoming more and more treat- ment-orientated and to which the home-care service often has the keys, and establishments that are increasingly respecting one's privacy so that the inhabitants have their own possessions and their own keys to their rooms? In such situations there can be no identical definitions of "institution" between one country and another. However, if comparisons are to be made of non-institutional life expectancies an agreement must first be reached on a unified defini- tion of "institution". Should it not be based more on excluding the community and geriatric enclaves than on concepts of services? As in the case of dependent persons it would also be of value to be able to assess the autonomy 796 WHO Bulletin OMS. Vol 70 1992 Measurement and utilization of healthy life expectancy and social isolation of persons placed in institutions in order to determine the quality of their life under the conditions of the given period. It is possible, therefore, from the outset to add two new com- ponents of healthy life expectancies to the above- mentioned list: - non-institutional life expectancy; - life expectancy without social isolation. Assessing the maintenance of quality of life in the community. At what price is it possible to maintain a person in his or her home? This brings up a second series of questions on the quality of the life lived in the community. Is confinement to one's home the price to be paid for maintenance there? Is going back to live with their children the price to be paid for maintenance in the community? Two more healthy life expectancies can now be defined specific to each of these: - life expectancy without confinement to one's home; - life expectancy under one's own roof. The principle is always the same: to refine the assessment of health conditions in a given period by using more control instruments. In this case, the decrease in the probabilities of having to enter an institution, for example, should not be at the expense of an increase in the probabilities of being confined to the home, and going back to live with one's child- ren is not a substitute for being placed in an institu- tion or vice versa. However, the difficulties of intemational com- parisons when using such instruments are obvious. There is a climatic dimension to confinement to the home which would interfere, for example, with direct comparisons between France and Canada. There is also a cultural dimension that would bias any comparison with strongly Islamic countries. The cultural dimension is still more obvious in the mode of cohabitation. It is certainly valuable to compare such instruments on an international basis but their use must be limited to very similar countries with closely harmonized data collections. Measuring and comparing abilities or performances The only way of removing biases due to climatic, cultural, religious or other factors would be to study abilities or theoretical capacities, rather than every- day performances or capacities used, which are too closely connected with the environment. Answers can be sought therefore to the following questions: (i) For how long is a person capable of getting about outside his or her home? (ii) For how long is a person capable of living alone in an ordinary dwelling? (iii) For how long is a person able to maintain social activities in the community? Performances are capacities used in a specific physical and social environment: nobody lives in an institution if there is no institution. A particular environment, climate, pollution, health policy, etc. is precisely what is characteristic of a given period and what defines the health conditions encountered by the population during that period. Abilities, on the other hand, are theoretical capacities that are much more closely connected with the history of the various generations than with the given period. By calculating cross-sectional values the research- er is largely freed from considering each country's or each generation's own particular history. In using mortality data, the researcher is not interested in those who are already dead but in those who are going to die during the given period (in a given year, for example). When using performance data, such as confinement to the home, the researcher would be interested not in those who are already confined to their homes (prevalence stock), but in those who are going to be so confined or released from such con- finement during the given period (incidence flows). When using ability data, such as the ability to leave one's home, the researcher would be interested not in those who have already lost that ability but in those who are going to lose it or recover it during the given period. What difference is there between the two period probabilities: the probability of being confined to one's home and the probability of losing the ability to leave one's home? The question is a complicated one. In intemational mortality comparisons, no attempt is made to eliminate the effect of environ- ment on mortality, e.g., by discarding a particular cause of death because it is too closely connected with the environment. Both physical and social environmental conditions influence the mortality en- countered by the population. Thus, a comparison of cross-sectional life expectancy between, for example, France and a central African country does not compare the health status of the two populations but the conditions influencing the mortality encoun- tered by these two populations. Therefore one cannot eliminate performances in daily functioning from calculations of cross-sectional healthy life expectan- cies because they are intimately connected with the environment in any given period. There is also a close relationship between mortality data and per- formance data in daily functioning, since sur- vival depends on one's performance in a particular environment. Calculations of healthy life expectancies with data on loss and recovery of abilities could be com- pared from one country to another. These, however, WHO Bulletin OMS. Vol 70 1992 797 J.M. Robine et al. would supply another theoretical assessment of the health conditions encountered by a population by combining survival performances in a particular environment with capabilities of functioning. Until conclusions are available from a more thorough study of the advantages and shortcomings of the various types of data that can be aggregated with mortality data, caution must be exercised as to the combinations selected. It is doubtless better to de- velop a greater number of simple instruments, some of them associating performances in functioning and other associating abilities with mortality, rather than to carry out complex statistical aggregations that are very difficult to grasp. Assessing more severe conditions While a general assessment is needed of the duration and quality of life in the community, answers are also needed to more specific questions corresponding to anxieties regarding the deterioration of the aged that are independent of the type of dwelling occupied and the mode of cohabitation. An assessment must be made of the most severe conditions and the proba- bilities of recovery from them. An initial list of such conditions would comprise mental deterioration, social isolation, loss of independence in carrying out certain private activities of daily living (ADL, such as going to the lavatory, washing, and changing clothes), incontinence, loss of independence for eating, confinement to bed (bedridden), and finally loss of autonomy. Healthy life expectancies based on data on the loss/recovery of function in regard to these health conditions would supply information on: - the duration of life without mental deterioration; - the duration of life without social isolation; - the duration of life without loss of dignity (loss of independence in carrying out private ADL as mentioned above); - the duration of life without suffering from incon- tinence; - the duration of life without confinement to bed (bedridden); and - the duration of life without loss of autonomy. Some of these measurements would clearly cor- respond to present objectives in public health, such as preventing dependence and keeping old people free for as long as possible to do by themselves all the basic activities of daily living. They would also make it possible to check on the effectiveness of health activities and to quantify the objective, e.g., a 10% reduction over a period of ten years in the pro- portion of years of dependence among the years lived under the conditions of the given period. Other measurements would correspond to clear- ly displayed anxieties, such as fears regarding the increase in the number of persons with mental disor- ders. To have available at each age (for persons in good mental health and those with mental disorders) figures for life expectancy, with or without mental deterioration, and to compare all the instruments concemed on an international basis, would certainly upset many accepted ideas at a time when we are not even sure that the population of persons with mental disorders has benefited from the general improve- ment in present mortality conditions as much as the rest of the population. Other measurements would also help to over- come taboos such as those surrounding problems of urinary incontinence. Whereas the few studies on the subject reveal that the prevalence of incontinence in old women is considerable, nothing has been done for the great majority of those concerned. Through consistent health campaigns in this area, life expec- tancy with no problem of incontinence could, by summing up the successive cross-sectional condi- tions in terms of preventing incontinence or doing something about it, become an instrument for check- ing on the effectiveness of the actions undertaken, even if the campaign had little influence for the time being on the numbers affected (prevalence stock). Finally, other measures would strengthen the great principles such as the right to autonomy which most democratic countries recognize for their citizens including the older ones. Obtaining the necessary data Types of cross-sectional calculations Information solely in one direction (e.g., loss of a function) is sometimes collected on the main cohort without observing, with the same precision, what goes on in the other direction in a specific cohort (e.g., recovery of the lost function). In this case cal- culations are made of the type, "How long can one expect to live without experiencing a particular event?"-the first event of its type, by definition. The collection of information for this type of calculation is relatively simple, which is important in the case of a project for standardizing procedures. For international comparisons the protocol should not rule out such calculations, even if a priori it is thought desirable to be able to take into account the recovery of the lost functions. In fact, the protocol should help to better define the two types of possible calculation-with or without recovery of the lost functions-clarifying the specific application of each instrument. An active life expectancy calculated solely with the probability of becoming ADL-dependent would WHO Bulletin OMS. Vol 70 1992798 Measurement and utilization of healthy life expectancy provide information on the average length of time during which individuals in a fictitious cohort can expect to live under current conditions without experiencing a first episode of dependence. Such an indicator, which totally neglects the rehabilitation and therapeutic action aspects of current health conditions, is very much directed towards preven- tion. It would be valuable to carry out at the same time: (i) a calculation for the whole of the popula- tion, (ii) a calculation limited to the initially healthy population, and (iii) a calculation limited to the indi- viduals who experience the event under study during the period. This last calculation, carried out regular- ly, could provide direct information on the possible postponement of the average age of onset of a parti- cular morbid event under the prevailing conditions. An active life expectancy calculated with the probabilities of becoming ADL-dependent and the probabilities of recovering ADL performances would provide information on the average length of time for which individuals in a fictitious cohort can hope to live under the prevailing conditions without being ADL-dependent. To put it more simply, the arith- metical difference between the two types of active life expectancy would sum up the effectiveness, at a given time, of rehabilitation and therapeutic activi- ties. There again, several calculations can be made- one for the whole population, one limited to the ini- tially healthy population, and a third limited to those individuals who are experiencing the event under study under the prevailing conditions. This last cal- culation would provide information on the duration of dependence of ADL-dependent persons under pre- vailing conditions. Disability thresholds There are two types of criteria for distinguishing among healthy persons, individuals who acquire a disability and those who recover lost functions: - absolute criteria of the "threshold" type common to all individuals; - relative criteria applicable to each individual, taking into account his or her initial state. For example, with current mobility conditions and with criteria of the threshold type, it would be possible to calculate a life expectancy without con- finement to the home, whereas with relative criteria it would be possible to calculate a life expectancy without notable reduction in mobility, whatever the initial level of mobility of each individual may be. The difference between these two types of healthy life expectancy is still more apparent in the example of cognitive capacities. With criteria of the threshold type it would be possible to calculate life expectancy without any sign of mental disorder, and with relative criteria a life expectancy without a notable reduction in intellectual capacities, whatever the initial level of the capacities of each individual. In terms of functioning, some individuals are initially very far from the threshold selected, while others are very close to it. On an individual basis the second calculation seems more interesting-how much time can one hope to live without a notable reduction in one's intellectual capacities?-but this calculation does not apply to any individual in parti- cular, or to any real cohort or population. In both cases there is the question of calculating healthy life expectancy for individuals in a fictitious cohort. It is a matter of two indications on current mental health conditions: (i) what proportion of the cohort is crossing the threshold of mental disorder; and (ii) what proportion of the cohort is experiencing significant mental deterioration. These two propor- tions may be very different. In regard to a common protocol, the use of cri- teria of the threshold type seems to be simpler at the outset. This option that has been selected for the experimental calculations of cross-sectional healthy life expectancies (9, 11, 13).9 It is, of course, not easy to take into account the recovery of lost func- tions by using relative variations in functioning as the criteria of deterioration in health. Supposing that a method is found of quantifying mobility and that a loss or an improvement of 10% in mobility is con- sidered as significant, an individual who, after having lost 10% of his mobility (recognized deterio- ration), experiences a slight improvement which is not recognized (<10%), may no longer meet the conditions for significant loss of mobility: he has notably lost part of his mobility, or he has not significantly improved his mobility, or he is no longer in a state of significant loss of mobility compared with his initial situation. Resume Mesure et utilisation de 1'esperance de vie en sant6: aspects conceptuels L'article examine toute une s6rie de problemes conceptuels et th6oriques liis au calcul et a l'utili- sation des esp6rances de sante dans les pays developpes ou l'esperance de vie a fortement augmente au cours du XXbme siecle. Aurait-on observe la meme evolution pour "I'esp6rance de vie en bonne sante" si on avait e capable de la 9 See footnote c on page 795. WHO Bulletin OMS. Vol 70 1992 799 J.M. Robine et al. mesurer? Des r6sultats recents semblent indiquer que "I'esperance de vie sans incapacite" augmen- te moins vite que l'esperance de vie totale. De tels r6sultats indiqueraient a la fois (i) une am6lio- ration des conditions de mortalit6, et (ii) de morbi- dit6 de la p6riode, mais aussi (iii) une augmenta- tion relativement plus grande des ann6es v6cues en 6tat d'incapacit6, et partant (iv) un accroisse- ment de la pr6valence de l'incapacit6 dans la population totale. Aux ages 6eeves 1'esperance de vie en bonne sant6 est clairement la moyenne pond6r6e de deux esp6rances, celle des sujets en bonne sant6 et celle des sujets en mauvaise sante. Dans une s6rie de calculs aux ages elev6s, une plus grande survie des sujets en mauvaise sant6, aux ages plus jeunes, peut en theorie annuler 1'effet d'une am6lioration des conditions de mortalite et de morbidit6 de la periode, aux ages 6lev6s. L'6tat de sant6 actuel de la population, meme s'il n'est pas totalement ind6pendant de l'histoire, est tres lie aux conditions de sant6 de la p6riode et il est 6vident que l'efficacite des programmes de sante doit etre mesure a travers leurs effets sur les conditions du moment (probabilit6s d'entrer ou de quitter un 6tat de sant6) et non pas sur le stock des personnes en mauvaise sant6. Pour cela les esp6rances de sante doivent vrai- ment devenir des indicateurs de la p6riode (i.e. conjoncturels). Cette approche a modifi6 sensible- ment nos connaissances sur la sur-incapacit6 feminine en termes d'incidence et de pr6valence ainsi que sur la reversibilite de l'incapacite. 11 est grand temps de reconnaitre que l'id6e d'un d6ve- loppement irr6versible de l'incapacit6, au cours du vieillissement, doit etre revis6e. On doit multiplier les indicateurs afin de veri- fier que ce qui est gagne sur une dimension de la sante n'est pas reperdu sur une autre: par exemple v6rifier la qualite des services apport6s aux personnes agees d6pendantes en mesurant leur autonomie. On doit aussi distinguer entre les aptitudes th6oriques des individus et leurs perfor- mances dans la vie quotidienne et mesurer les conditions les plus s6veres pour les diff6rentes dimensions de la sant6 (d6pendance pour les actes el6mentaires de la vie quotidienne, d6terio- ration mentale, isolement social, perte d'autono- mie). Les calculs opposent les esperances de vie avant la premiere entr6e dans l'6tat d'incapacit6 s6lectionn6 (par exemple, 1'esp6rance de vie avant un premier 6pisode de confinement au domicile) des esperances de vie sans l'etat d'inca- pacit6 s6lectionn6 (par exemple, 1'esp6rance de vie sans confinement au domicile). 11 reste a choi- sir le type de seuil, relatif ou absolu, pour d6finir 1'entree en incapacite et la r6cup6ration des fonc- tions perdues. Acknowledgements We are indebted to Karen Ritchie, Denis Bucquet and Bernard Grab for their advice in the preparation of certain sections of this paper, and to Michele-Claude Seroussi, Dorindo Maio and Patricia Capperon for their technical assistance. References 1. Robine, J.M. Disability-free life expectancy. Quebec, Conseil des Affaires Sociales et de la Famille, 1986. 2. Fries, J.F. The compression of morbidity: near or far? Milbank Memoral Fund quarterly, 67: 208-232 (1989). 3. Kramer, M. The rising pandemic of mental disorders and associated chronic diseases and disabilities. Acta psychiatr. scand., 62 (Suppl. 285): 382-397 (1980). 4. Sullivan, D.F. A single index of mortality and morbi- dity. HSMHA Health Reports, 86: 347-354 (1971). 5. Robine, J.M. Estimation de la valeur de l'esperance de vie sans incapacit6 (EVS) pour les pays occiden- taux au cours de la derniere decennie. Quel peut etre l'utilit6 de ce nouvel indicateur de l'etat de sante? Wld hlth statist. qrtly, 42: 141-150 (1989). 6. Robine, J.M. & Ritchie, K. Healthy life expectancy: evaluation of a new global indicator of change in population health. Br. med. j., 302: 457-460 (1991). 7. Robine, J.M. et al. Les indicateurs d'esperance de vie sans incapacite (EVSI). Des indicateurs globaux de l'etat de sante des populations. Rev. 6pid6miol. et sant6 publ., 35: 206-224 (1987). 8. Manton, K.G. A longitudinal study of functional change and mortality in the United States. J. g6ron- tol., 43: S153-161 (1988). 9. Branch, L.G. et al. Active life expectancy for 10 000 Caucasian men and women in three commu- nities. J. gerontol., 46: M 145-150 (1991). 10. Chirikos, T.N. & Nestel, G. Longitudinal analysis of functional disabilities in older men. J. gerontol., 40 (4): 426-433 (1988). 11. Rogers, A. et al. Longer life but worse health? Measurement and dynamics. Gerontologist, 30: 640-649 (1990). 12. World Health Organization. International classifica- tion of Impairments, Disabilities, and Handicaps. Geneva, 1980. 13. Katz, S. et al. Active life expectancy. New Engl. j. med., 309: 1218-1224 (1983). 14. Grimley Evans, J. Prevention of age-associated loss of autonomy: epidemiological approaches. J. chron. dis., 37: 353-363 (1984). 15. Robine, J.M. La mesure des incapacit6s. Paris, Ins- titut des Sciences Politiques, memoire DEA, 1982. 800 WHO Bulletin OMS. Vol 70 1992