Methodological work

Methodological work

N Brouard, E Cambois, A Désesquelles, JM Robine

In addition to their work on health, incapacity and disability, INED’s researchers contribute to methodological work and participate in the production of data. They look at what the indicators used in the surveys actually measure, as well as their limitations and biases. They also develop techniques for analysing the data collected.

Designing population surveys

INED has been involved in the design of population surveys on health and disability for many years. Through our participation in the working groups that design the surveys, we can communicate our research experience, methodological expertise and knowledge of appropriate definition and measurement instruments. INED has thus been active in the design of the HID survey, ten-year health surveys and disability and health surveys (enquêtes Handicap Santé). We are currently involved in monitoring mortality after those surveys, which enable us to analyse differential mortality depending on states of health and dependency.

As part of these activities, the researchers are also seeking to understand how the measurement instruments work in order to adjust interpretations for any biases and limitations: perceived health, activity restrictions or mental health.

Probabilities of transitions between different states of health: IMaCh

Some years ago INED developed expertise in calculating the probabilities of transition between different states of health. Agnès Lièvre, supervised by Nicolas Brouard and working with Christopher Heathcote, Professor of Statistics at the Australian National University in Canberra, developed a software program called IMaCh, which calculates probabilities of transition based on longitudinal data, making it possible to calculate disability-free life expectancies from incidence. The method and the software seem to meet a need, judging by the strong interest they have attracted, especially in Asia, where the method has already been taught on a number of occasions (UNFPA, Bangkok - April 2001), (Beijing University - December 2001) (UNFPA/Nihon University, Tokyo - April 2002).

The software needs to undergo some modifications, such as to enable the analysis of mortality using models other than Gompertz. A version of IMaCh should also meet the growing demand from users for estimating models where a lost function cannot be recovered. Indeed, while the main advantage of the model is that it can handle the complexity of people moving back and forth between different states of health, it can sometimes be useful to model permanent changes in state of health such as a heart attack or cancer. The general idea would be to reduce the model, by excluding transitions wherever the similarity is too flat. Indeed, a conference organised by the NIDI in May 2010 attended by a large number of authors of multi-state model software, demonstrated that an autonomous software system like IMaCH, which is not integrated or only partly integrated into a system like R, would retain the necessary speed of execution.

Decomposition techniques

Decomposition methods pinpoint the role of diseases and causes of death in trends and divergences in life expectancy and disability-free life expectancy. While these methods are already widely used at the unit to analyse trends in life expectancy, we have not yet used them to analyse trends and differences in disability-free life expectancy.

For our research, we will draw on a collaboration with W. Nusselder to apply the tool she developed (see Nusselder W. et al. 2004, "Decomposition of differences in health expectancy by cause", Demography, 41 (2), pp. 315-34) to French disability-free life expectancy data. The project will initially determine the diseases and causes of death that contribute to gender inequality in disability-free life expectancy.

Mortality projections

1) Basellini and Camarda: "Modelling and Forecasting Adult Age-at-death Distributions".

Résume: "Age-at-death distributions provide an informative description of the mortality pattern of a population, but they have generally been neglected for modelling and forecasting mortality. In this article, we use the distribution of deaths to model and forecast adult mortality. In particular, we introduce a relational model that relates a fixed "standard’’ to a series of observed distributions by a transformation of the age axis. The proposed STAD model is parsimonious and efficient: using only three parameters, it captures and disentangles mortality developments in terms of shifting and compression dynamics of mortality. In addition, mortality forecasts can be derived from parameters’ extrapolation using time series models. We illustrate and compare our methodology with the Lee-Carter model and its variants for females in four high-longevity countries. We show that the STAD fits very well the observed mortality pattern, and that its forecasts are more accurate and optimistic than the Lee-Carter variants. "

2) Basellini and Camarda: "A Three-Component Lee-Carter approach to decompose and forecast human mortality".

Résume: "The Lee-Carter model is an elegant and powerful methodology to model and forecast mortality based on a log-bilinear form for the hazard function. We propose a novel extension of the model that overcomes its drawback of a fixed age-specific rate of mortality improvements. This new approach improves goodness-of-fit and offers an innovative perspective in terms of forecasting by decomposing mortality into childhood, early-adulthood and senescent components. The three components are estimated via an iterative series of penalized composite link models. We illustrate the approach on Swiss male mortality data. "

Some references

Cambois Emmanuelle, Robine Jean-Marie, Mormiche Pierre. 2007. Une forte baisse de l'incapacité en France dans les années 1990? Discussion autour des questions de l'enquête santé. Population, 62 (2), p. 363-386

Désesquelles Aline. 2005. Handicaps, disabilities, and dependency in French prisons - Lessons from a special survey. Courrier des Statistiques, English series, (11), p. 23-34

Désesquelles Aline, Egidi Viviana., Salvatore Michele Antonio. 2009. Why do Italian people rate their health worse than French people ? An exploration of cross-country differentials of subjective health. Social Science and Medicine 68 (6), p. 1124-8

Lièvre A, Brouard Nicolas, Heathcote C 2003. The estimation of health expectancies from cross-longitudinal surveys. . Math. Pop. Studies 10 (4), p. 211-248

Lièvre Agnès, Jusot Florence, Barnay Thomas, Sermet Catherine, Brouard Nicolas, Robine Jean-Marie, Brieu Marie-Anne, Forette Françoise. 2007. Healthy working life expectancies at age 50 in Europe : a new indicator. J Nutr Health aging, 11 (6), p. 508-514

Basellini and Camarda: "Modelling and Forecasting Adult Age-at-death Distributions"

Basellini and Camarda: "A Three-Component Lee-Carter approach to decompose and forecast human mortality"