Colloque / Séminaire

The main objective of this study is to assess the effects of lifestyle factors (heavy alcohol drinking, tobacco smoking, and/or obesity) on the life expectancy and life expectancy without major disability of healthy 50+ adults in France. We developed a microsimulation model from 3 main National data sources in the years 2008 to 2013:
a) population census (Insee);
b) death records at the individual level (Insee);
and c) the Hospital Discharge (PMSI) database that contains all public or private hospital stays in France and detailed medical information at the individual level.

In the first part of this presentation, we will describe briefly the PMSI database as well as its benefits and possible biases. A trivial benefit is the exhaustive identification of adults affected by major diseases that do require hospital care, with possible assessment of the risks for major disability (severe dementia or bedridden state) and/or in-hospital death. An unexpected benefit is related to the high (73.5%) proportion of 50+ adults hospitalized over 6 years in France, at an increasing rate with age (≥90% after 80 years old). Therefore, the PMSI database also provides a large and representative sample on lifestyle factors, with possible assessment of the risks for major diseases.

In the second part of this presentation, we will describe the simulations techniques used to forecast the residual lifetime of an individual taking into account his health deterioration over time. This approach requires measuring the impact of lifestyle factors on the incidence rates of major diseases, and then on the survival probabilities, depending on the current heath state. We also build estimators of the lifespan (overall, with or without disability) conditionally on the characteristics of healthy 50+ adults (considering gender, age, residency, and lifestyle factors). We used regression approaches considering competing risks data with right censoring and left truncation.
Finally, possible use of the study results in different fields will be discussed.