At a Glance
Methods, Models, and Measures in Population Health
Kieron Barclay, Christian Dudel, Roland Rau, Mikko Myrskylä, Rasmus Hoffmann, Christina Bohk-Ewald (MPIDR / University of Helsinki, Finland); in Collaboration with Hector Pifarre i Arolas (Pompeu Fabra University, Barcelona, Spain)
To study population health, it is essential to have measures that accurately reflect the key dimensions of the underlying concept or process of interest and to have methods that allow us to model the underlying data structure. In this project, we work to develop new measures, methods, and models in order to extend the existing demographic toolbox for studying population health. We summarize a few highlights of our recent work:
Rank-based measures of population health are useful for comparing health across multiple populations or groups. A new indicator is constructed based on the relative prevalence of the various health conditions within these populations and it can be readily applied to most existing health surveys. The use of this new health measure is illustrated by an application to data from the National Health Interview Survey to examine health differences across racial groups in the US. This work thus targets two substantive challenges that current health measures have yet to overcome: i) the comparability of self-reported health and symptoms across groups, and ii) differences in disease distributions across groups.
Mortality forecasts typically predict how many additional years of life people are likely to gain in the future and by how much mortality will decline or increase at each age. As many mortality forecasting approaches assume that the age profile of mortality change is constant, they are often unable to account for the dynamics in lifespan disparity and the turning points in long-term trends. We developed a new method that aims to overcome this drawback with two features: i) It forecasts mortality with varying age-specific survival improvements and ii) it optionally combines mortality trends of multiple countries. Validating the forecast accuracy with data on British and Danish women from 1991 to 2011 suggests that the new model can forecast regular and irregular mortality developments at least as well as other widely-accepted approaches do, such as the Lee-Carter model.
Other work in this project gives guidelines for constructing cardiovascular disease-risk indices from commonly used biomarkers, demonstrating an analytic trade-off between parsimonious measures that predict well relative to their complexity and complex measures that are able to identify specific high-risk groups that other methods miss.
Ageing, Mortality and Longevity, Demographic Change, Life Course, Statistics and Mathematics
Sociological Methods and Research. forthcoming. (2021)
Population Health Metrics 18:15, 1–9. (2020)
Oxford Bulletin of Economics and Statistics 82:6, 1456–1481. (2020)
Demographic Research 42:24, 713–726. (2020)
SSM-Population Health 8:100438, 1–10. (2019)
Social Indicators Research 143:3, 1219–1243. (2019)
MPIDR Working Paper WP-2019-001. (2019)
Demographic Research 38:62, 1933–2002. (2018)
Genus 73:1, 1–37. (2017)
MPIDR Working Paper WP-2017-010. (2017)