EUROREVES: Research design and calculation methods

The subcommittee on ``Research Design and Calculation Methods'' has in charge to raise most on the problems encountered for the calculation of an ``Index of Health Expectancy''. This index calculated for a particular population and calendar year should be comparable over time and over other populations or sub-populations of the European Union. The sub-committee, as any others Euro-REVES sub-committee, is promoting Health Expectancy calculations amongst European countries which are not yet involved in such calculation process. To promote Health Expectancy calculations and harmonization of the calculation process, the sub-committee uses a World-Wide Web server (http://euroreves.ined.fr/euroreves/) which collects and provides information on Health Expectancies. The REVES International database is also partly hosted by the server. During the first year, the sub-committee collects information on the data required for the Health Expectancy calculations: population by age, mortality rates by age, prevalence of incapacity by age, incidences of disability and recovery, etc. Each kind of data (population estimates, mortality rates, etc) is a source of problems and discrepancies between countries, but problems encountered in the computation of Health Expectancy can be divided into two main parts according to two different kind of index:

Indices computed from cross-sectional prevalences of disability by age. These indices belong to the family of the Index of Sullivan, which, in our point of view is not really an Expectancy in the common sense used in statistics, as the index is based on prevalence ratios and not on incidence rates. Therefore it describes the actual situation, like a census does, but cannot be used to simulate the future "expectations" of a fictious cohort which would, at each age, have the actual prevalence. Prevalence ratios are the results of in and out flows and therefore cannot be supposed to remain constant over time even if "nothing changes". It can be the case only if the flow rates are consistent with the stock, ie if the incidence rates are consistent with the prevalence. This is rarely the case if prevalence is itself changing.
By combining a set of cross-sectional age specific prevalence with a period life table (which corresponds to a real projection of the survival function in the future if the mortality rates remain constant over time), the Sullivan method is hybrid and is only a first method to understand the concept of Health Expectancy, even if it is, most of time, the only index which can be computed.
You can download a spreadsheet (Excel) sulliv2.xls which will help you to compute a Sullivan index from a period life table and prevalence rates estimates from a disability survey. The computation of the standard deviation of the index due to a small sample is also included. It will correspond to a forthcoming version of a Practical Guide.
An earlier version of this Practical Guide, written by Eric Hauet, has also a spreadsheet, . sulliv1.xls. It takes the example of France in 1981 and 1991 for females, including the information brought by the 1990 census on the institutionalized population, because the french Health surveys did not sampled institutions.
Indices computed from transition rates between health statuses. These indices are real expectancies because they correspond to what a young boy or girl could expect to live in the future if all the transition rates observed at a particular date remain the same in the future. Such indices are not related to the past history of the cohorts involved in the cross-sectional survey.

In the first case, data required to compute such an index consist in a single cross-sectional health survey. In the second case, at least a second pass is required to estimate incidences. Only a few countries have such incidence data.
In the first case, the sub-committee will focus on the harmonization of age groups, age range, sample size, etc.
In the second case, the sub-committee will focus on the interval between passes, the sample size, age group, incidence estimation. The scientific literature is rich of texts on the subject. One of the aim of this sub-committe is also to write basic documents on these subjects.