Projecting life expectancies for real populations

In this project, which is both methodological and applied, at the crossroads of several disciplines such as epidemiology, demography and biostatistics, we are taking a critical look at some commonly used indicators, such as period life expectancy or a standardized mortality rate. In particular, we question their relevance to provide reliable information on the impact on mortality of an exceptional event, such as the recent COVID-19 pandemic, and to support decision-making in the fields of insurance, social protection, and healthcare costs management. We shall thus define and study new concepts which will have the advantages of a) referring to a real population, rather than to hypothetical cohorts, as is the case with the classical concept of period life expectancy, and b) weighting deaths differently according to the age at which they occur, contrary to what is done with a classical standardized mortality rate. To achieve this, we will 1) estimate the remaining life expectancies of non-extinct cohorts by opportunely modeling and projecting age-specific mortality rates; 2) combine these estimates to introduce a new concept of population life expectancy, defined as the expected age at death of individuals belonging to a real population. This will be applied to the data of the Human Mortality Database (HMD), with a particular (although not exclusive) focus on Switzerland. Two further objectives of the project will be: 3) to extend the notion of a population life expectancy to that of a good health population life expectancy, by supplementing mortality data from the HMD with data from the Swiss Health Interview Surveys (SHIS) conducted every 5 years since 1992 by the Swiss Federal Statistical Office (FSO), and; 4) generalize the notion of standardized mortality rate by weighting each death with an importance weight inversely proportional to the age at which the person dies, yielding a new concept of weighted standardized mortality rate. Our project has a wide range of potential impacts at individual, societal and scientific levels. In particular, it will provide answers to essential questions that everyone may ask, namely the expected time remaining to live, respectively to live in good health, depending on age and gender. Furthermore, such reliable estimations, that apply to the real (not hypothetical) population, will have great potential to improve and facilitate decision-making in many fields, such as insurance or retirement. Our new concepts will also enable a more relevant assessment of the impact of the recent Covid-19 pandemic on mortality and longevity in a real population, a topic that has been one of the applicants' main research interests over the past three years, as well as the impact of possible future exceptional events like pandemics, energy supply problems, or heat waves, and structural changes, such as climate change. Our projections of mortality by age will also be directly useful, along with some assumptions about fertility and migration trends, in making projections of future changes in population size and age structure, that are questions of great interest to statistical offices such as the FSO for Switzerland. Finally, our modelling strategy, already tested by the applicants in the field of cancer incidence prediction, and which will also be duly evaluated as part of the present project, could be of interest in fields other than mortality.