Conference Agenda

One key factor to construct sound measures to prevent adverse health outcomes and to allocate healthcare resources for sustainable aging populations is the possibility of identifying precise multimorbidity patterns and seizing their trajectories in time. In both developed and developing countries, understanding the structure of multimorbidity, and most ideally across time, is an urging challenge, so that groups who share the same degree of vulnerability and needs can receive assistance and intervention in a timely manner. Compared to traditional approach like factorial and clustering analysis that have been of standard practice in the literature, combining the probabilistic approach of graphical model and the intuitive visibility of network analysis is emerging quickly as powerful tool in recent years to not only efficiently explore the richness of administrative health data, but also to provide a framework with predictability. By applying these methods on reliable longitudinal data of individuals aged 50 and above residing Emilia-Romagna region (northern Italy) in 2011 and followed up to 2019 (N = 1,010,610), we study the multimorbidity patterns at older ages and their changes across time. Using hidden Markov model based on the estimated multimorbidity patterns, we construct the trajectories of multimorbidity leading to death at older ages.

Understanding the interplay of risk factors is essential in health economics and policy-making for effective interventions targeting vulnerable sub-populations. Behavioral health surveillance surveys, like the Italian PASSI survey, provide rich data on demographics, socioeconomic status, and behavioral patterns, including occupational information as textual data. However, linking this occupational information to other risk factors has been challenging and underexplored in the literature. This study proposes a novel Bayesian model integrating the Structural Topic Model (STM) to analyze textual occupational data with the multivariate probit model for binary risk factor data. STM identifies unobservable occupational macro-groups associated with varying risky behavior propensities, which a mixture of the multivariate probit regression captures. Socio-demographic covariates are incorporated into the multivariate probit model, and their effects are influenced by STM-identified topics. The two approaches are combined using Markov melding, and the estimation process is enhanced with importance sampling for improved computational efficiency.

This study explores the bio-psycho-social dimensions of frailty among Italian older adults, identifying frailty profiles based on physical, psychological, and social domains. It examines how these profiles affect access to healthcare and Public Administration (PA) services, with a focus on digital access moderated by individual digital skills.

Data is drawn from the 2022 ISTAT survey Aspects of Daily Life (AVQ), covering Italians aged 65 and above (N=10,991). Frailty is assessed using self-reported health, activity limitations, emotional states (sadness, anxiety, hopelessness), social support, and living conditions. Latent class analysis identifies four frailty profiles: non-frail (49%), physically frail (13%), psychologically frail (22%), and frail across multiple domains (16%). Logistic regression models then estimate healthcare and PA service use by frailty profile.

Results show non-frail individuals use healthcare services less often than frail individuals, especially for emergencies, but access PA services more frequently. This suggests frailty influences service use based on needs versus engagement capabilities. Future analysis will delve into digital access to healthcare to understand if it acts as a barrier or facilitator.

Traditional approaches to assess whether morbidity is compressing or expanding over time typically compare life expectancy (LE) and health-adjusted life expectancy (HALE). Changes in the HALE/LE ratio are used to support hypotheses of morbidity compression or expansion. The aim of this paper is to revisit the long-standing ‘compression vs expansion of morbidity’ debate taking advantage of recently illustrated multistate modelling techniques. Such methods allow deriving distributions estimating the number of years individuals have accumulated in good and in less-than-good health throughout their lives. Building on this approach, we propose the ‘healthy year curves’, a new tool measuring the average number of years lived in good health among those who died at a certain age. We provide an empirical application using data from the Health and Retirement Study and showing results for US women and men separately from 2000 to 2015.

In 2022, the European Survey on Income and Living Conditions (EU-SILC) included both the 3-years rolling module on lifestyles and the 6-years rolling module on quality of life. We exploit the joint availability of this information to study the association between living conditions, lifestyles and two subjective health outcomes in Italy. In particular, we use logistic regression to evaluate the association between three EU-SILC indicators of poverty or social exclusion (risk of poverty, material deprivation and housing deprivation) and the risk of both poor self-rated health (SRH) and poor mental health (MH), controlling for a full set of individual characteristics and for many lifestyle risk factors (body mass index, current smoking, alcohol consumption, healthy diet, physical activity and social relations). We show that being at risk of (relative) poverty is not significantly associated with an increased risk of poor SRH and only with a slight increase in the risk of poor MH, while material deprivation and severe housing deprivation are associated with large increases in both risks. Moreover, we find a significant impact of lifestyles on these associations, showing how observed inequalities in subjective health outcomes depend on the complex interactions between socio-economic deprivation and the adoption of healthy habits.