Periodic Reporting for period 4 - LETHE (Levels and Trends of Health Expectancy: Understanding its Measurement and Estimation Sensitivity)
Período documentado: 2022-03-01 hasta 2023-08-31
The demographic developments of European societies are characterized by aging of the populations, i.e. the absolute and relative increase of the older compared to the younger population. The extent of the consequences of this process depends strongly on the health status of the populations. Improving citizens’ health condition is therefore one of the most important and effective ways to reduce the burdens of demographic aging.
Assessing progress of respective public health programs toward their targets requires an appropriate and reliable indicator for tracking the levels and trends of population health. Several indicators have been developed for this purpose, which are referred to with the overarching term “Health Expectancy” (HE). The central aim of the project was to assess the HE indicators’ sensitivity to specific measurement (i.e. conceptual) features and specific estimation (i.e. technical) features, identifying the most important sources of possible biases and finding options for overcoming these issues.
Overall, we found that the investigated estimation features do not lead to a significant bias in most of the tested empirical applications. In other words, the HE indicators are rather robust regarding the technical aspects of their estimation. However, the opposite applies to the investigated measurement features. These have an enormous impact on the HE indicators and can lead to significant biases in empirical application. These new insights into the understanding of HE values, trends and differentials are highly relevant and valuable for all researchers, policy makers and other stakeholders who are using these indicators to track levels and trends of population health. Above all, they are helpful in reducing the risks of overinterpretation and potentially misleading conclusions.
Assessing progress of respective public health programs toward their targets requires an appropriate and reliable indicator for tracking the levels and trends of population health. Several indicators have been developed for this purpose, which are referred to with the overarching term “Health Expectancy” (HE). The central aim of the project was to assess the HE indicators’ sensitivity to specific measurement (i.e. conceptual) features and specific estimation (i.e. technical) features, identifying the most important sources of possible biases and finding options for overcoming these issues.
Overall, we found that the investigated estimation features do not lead to a significant bias in most of the tested empirical applications. In other words, the HE indicators are rather robust regarding the technical aspects of their estimation. However, the opposite applies to the investigated measurement features. These have an enormous impact on the HE indicators and can lead to significant biases in empirical application. These new insights into the understanding of HE values, trends and differentials are highly relevant and valuable for all researchers, policy makers and other stakeholders who are using these indicators to track levels and trends of population health. Above all, they are helpful in reducing the risks of overinterpretation and potentially misleading conclusions.
The main objective of the project was to test and assess the Health Expectancy (HE) indicators’ sensitivity to (1) specific measurement features and (2) specific estimation features (see graphical illustration of investigated HE sensitivities below). Our research focused primarily on HE indicators based on the three health indicators covered by the “Minimum European Health Module” (MEHM), which is included in most of the recent survey projects such as the EU-SILC or SHARE: (1) self-perceived health (SPH), (2) activity limitations (LIMIT) based on the “Global Activity Limitation Indicator” (GALI), and (3) chronic health problems (CHRON).
We found that issues mostly related to the measurement features are more relevant in explaining the sensitivity of HE indicators. The health dimension used, the choice of survey, and the interpersonal reporting heterogeneity across populations are some of the aspects that have proven to be more significant than different estimation techniques employed to compute health prevalence. On the other hand, estimation techniques have a rather negligible effect on health indicators, especially for those over 65 years old. Overall, assumptions made before the age of 15 are the most significant and have distinct effects on women and men, influencing Healthy Life Years (HLY) at birth in some countries. Also, HLY estimates using EU-SILC data are not significantly biased by health-related attrition across samples, but sample attrition increases the uncertainty in the measurement of individual countries. Lastly, we found that attrition impacts cross-sectional datasets from longitudinal SHARE samples, yielding estimates of population health that are over-optimistic.
It is important to note that our research focused on prevalence-based methods, that is, using health stock information at a given period of time and by age groups for both women and men across European countries. It is yet uncertain whether estimation techniques could impact incidence-based prevalence, or the flow of new cases of health conditions in the population. Overall, HE indicators are sensitive metrics that should be interpreted with caution, considering the health dimension employed, the source (survey) of health data, the influence of health reporting behaviour across various cultures, and the indicator's intended use.
Over the entire project period, the team was very successful in publishing the project results in peer-reviewed journals and books. In addition, we presented the findings at several large and high impact conferences. During these presentations we noticed the strong recognition of our work. This confirmed our hypothesis that most users of the common HE indicators are not aware of the indicators’ sensitivities regarding various methodological and estimation issues which can be problematic in practice and result in misleading conclusions. This motivated us to give even more emphasis to our dissemination activities and we strengthened these in the second half project time, e.g. by producing a Shiny App and a metadata database (see below).
We found that issues mostly related to the measurement features are more relevant in explaining the sensitivity of HE indicators. The health dimension used, the choice of survey, and the interpersonal reporting heterogeneity across populations are some of the aspects that have proven to be more significant than different estimation techniques employed to compute health prevalence. On the other hand, estimation techniques have a rather negligible effect on health indicators, especially for those over 65 years old. Overall, assumptions made before the age of 15 are the most significant and have distinct effects on women and men, influencing Healthy Life Years (HLY) at birth in some countries. Also, HLY estimates using EU-SILC data are not significantly biased by health-related attrition across samples, but sample attrition increases the uncertainty in the measurement of individual countries. Lastly, we found that attrition impacts cross-sectional datasets from longitudinal SHARE samples, yielding estimates of population health that are over-optimistic.
It is important to note that our research focused on prevalence-based methods, that is, using health stock information at a given period of time and by age groups for both women and men across European countries. It is yet uncertain whether estimation techniques could impact incidence-based prevalence, or the flow of new cases of health conditions in the population. Overall, HE indicators are sensitive metrics that should be interpreted with caution, considering the health dimension employed, the source (survey) of health data, the influence of health reporting behaviour across various cultures, and the indicator's intended use.
Over the entire project period, the team was very successful in publishing the project results in peer-reviewed journals and books. In addition, we presented the findings at several large and high impact conferences. During these presentations we noticed the strong recognition of our work. This confirmed our hypothesis that most users of the common HE indicators are not aware of the indicators’ sensitivities regarding various methodological and estimation issues which can be problematic in practice and result in misleading conclusions. This motivated us to give even more emphasis to our dissemination activities and we strengthened these in the second half project time, e.g. by producing a Shiny App and a metadata database (see below).
We developed three new health indicators: HCAL, WAHE, and DIF-adjusted HLY. HCAL is the “cross-sectional average length of healthy life” and modifies the Sullivan method, the most conventional way of estimating HE, by combining the health prevalence data with the “cross-sectional average length of life” (CAL) instead of the classic period life expectancy (LE). In this sense, it captures historical mortality and health, instead of only a snapshot of the time period.
WAHE is a “well-being-adjusted health expectancy”. It combines health and mortality information into a single indicator with weights that quantify the reduction in well-being associated with decreased health. The advantage of this indicator in comparison to others is its ability to differentiate between the consequences of health limitations at various levels of severity and its transparent, simple valuation function.
DIF-adjusted HLY is an indicator where we devise an adjustment for the effect of possible bias due to interpersonal reporting heterogeneity across populations. For this, we used anchoring vignettes. Most recently, we extended the state-of-the-art also by combining the three MEHM indicators into one generic health measure.
Another important outcome of the project is a “best-practice” guideline that researchers can use when their goal is to smooth health prevalence and/or graduate age patterns of health, based on several tests using alternative methods that we employed during the last six years.
A further innovative output is a publicly available Shiny App, where interested researchers, students, policy makers, and other stakeholders can access the result of our sensitivity analyses. There, one can choose the methods, the countries, age, gender, and many other options to check how different approaches impact the results. The app allows users to easily put health and mortality into perspective in the European context.
WAHE is a “well-being-adjusted health expectancy”. It combines health and mortality information into a single indicator with weights that quantify the reduction in well-being associated with decreased health. The advantage of this indicator in comparison to others is its ability to differentiate between the consequences of health limitations at various levels of severity and its transparent, simple valuation function.
DIF-adjusted HLY is an indicator where we devise an adjustment for the effect of possible bias due to interpersonal reporting heterogeneity across populations. For this, we used anchoring vignettes. Most recently, we extended the state-of-the-art also by combining the three MEHM indicators into one generic health measure.
Another important outcome of the project is a “best-practice” guideline that researchers can use when their goal is to smooth health prevalence and/or graduate age patterns of health, based on several tests using alternative methods that we employed during the last six years.
A further innovative output is a publicly available Shiny App, where interested researchers, students, policy makers, and other stakeholders can access the result of our sensitivity analyses. There, one can choose the methods, the countries, age, gender, and many other options to check how different approaches impact the results. The app allows users to easily put health and mortality into perspective in the European context.