@eugeniopaglino
Data Scientist and Postdoc at @pophel.bsky.social. PhD in Demography and Sociology at UPenn, Statistics MA at Wharton. My research focuses on mortality determinants and trends. Bayesian statistics, forecasting, statistical modeling.
I am particularly excited that this is coming out in the Formal Relationships series! I've dreamed of contributing something (small) to this amazing collection 🧡 since I saw it listed as a supplementary source in the Advanced Demographic Methods course at Penn.
I also realized that the notion that the contribution of a population A to total life expectancy (LE) can be written as LE(T) - LE(B) if population A and B form the total population (T) is problematic because it depends on the choice of the reference (nerdy stuff but with practical implications).
Interestingly, the method I developed and these two great general decomposition approaches give the same result for this particular problem despite tackling the problem with very different strategies. Once again, formal demography proves to be endlessly fashinating 🤩
In developing the method, I realized that both the line-integral (Horiuchi et al., 2008) and the life table response experiment (Caswell, 1996) methods can answer the same question but it's not obvious how apply them to this problem (more details in the paper if you are curious).
It turns out the solution to this problem works with any life table metric and can be combined with other approaches 🧩to further decompose by age, cause of death, or other dimensions of interest (I am indebted to this great paper for this idea onlinelibrary.wiley.com/doi/full/10....)
This work began as a way to solve the practical problem 🧐 of how decompose the foreign-born contribution to total US life expectancy by race and ethnicity for this paper www.sciencedirect.com/science/arti...
🚨New paper out in Demographic Research🚨 and my first foray into demographic methods. It develops a new method to quantify how specific subpopulations drive demographic gaps. www.demographic-research.org/articles/vol...
@demresjournal.bsky.social @pophel.bsky.social #demography #methods
As part of this project, I will leverage the amazing Finnish register data to study the health and mortality trajectories of immigrants in Finland, how these trajectories differ by socioeconomic position, and how they contribute to national and subnational trends in health and mortality
I'm extremely grateful to the Finnish Cultural Foundation
@skr.fi for awarding me a grant to study how immigration is contributing to health and mortality trends in Finland and other Nordic countries! #skr2026 #kulttuurirahastontuella @pophel.bsky.social
📢Doctoral Defence:
Ulla Suulamo - "Long-term trends in mortality differences - Register-based studies in Finland, 1970–2020"
16 October 2025 at 12:00
Address: Porthania, PIII, Yliopistonkatu 3, Helsinki
🚨🚨 We are hiring 2 postdocs at PopHel – MaxHel. If you are interested in health inequalities and social determinants of health this is a great opportunity! Amazing working environment ✨, unique data infrastructure, and a truly interdisciplinary and international center 🌍. Deadline October 31st.
Had a great experience ⭐ at #SLLS2025! I hope we gave a good overview of what is possible with the amazing Finnish 🇫🇮 register data! @pophel.bsky.social @sllshome.bsky.social
Excess Deaths Attributable to the Los Angeles Wildfires From January 5 to February 1, 2025
An estimated 440 excess deaths were attributed to the January 2025 wildfires in Los Angeles County, underscoring indirect health effects and the need for improved mortality tracking.
ja.ma/4oFM3af #MedSky
It would be great to see more work 📄 filling these gaps and extending this approach to other events, contributing to improve our understanding of the short- and long-term impacts of different natural disasters on mortality and other health outcomes #publichealth #demography #Demography
While our study is an important first step, we could not look into which groups were more severely affected, e.g. by socioeconomic status or neighbourhood, we only looked at the first 4 weeks, and we only considered mortality rather than including other health outcomes
Demographers could play an important role in this area by leveraging their expertise on mortality data and modeling to make estimates like the ones in our paper available for other natural disasters
The contrast between 30 direct fatalities and >400 excess deaths underscores the importance of complementing cause-of-death investigation (to identify direct deaths) with techniques more suited to capture both direct and indirect mortality impacts of wildfires and other natural disasters
We find that mortality was higher 📈 than expected in the first four weeks after the wildfires (440 more deaths than expected). We performed several sensitivity analyses to rule out other mechanisms
In this work with @astokespop.bsky.social and Rafeya Raquib at @busph.bsky.social , we use mortality models to estimate expected deaths 📈 in the absence of the wildfires. We then compare observed and expected deaths to quantify how many excess deaths ☠️ are likely attributable to the wildfires🔥
New study 🚨 out in @jama.com! We find that California wildfires🔥in January may have contributed 440 excess deaths in Los Angeles County. This estimate significantly exceeds the 30 direct fatalities linked to these events jamanetwork.com/journals/jam... @pophel.bsky.social @helsinki.fi
But of course future can't explain the past, so if the slowdown appeared much earlier in some state-metro combinations, then other mechanisms have to be responsible 8/8🧵
Beyond demographic curiosity, these findings have implications for how we think about explaining the US mortality stagnation. If 2010 or 2014 are meaningful thresholds then explanations focusing on what happened shortly before (e.g. the Great Recession) seem more plausible 7/8🧵
State-level trends in metropolitan and nonmetropolitan mortality by sex in the United States (1999-2019)
For example, female mortality showed very little improvement past 2005 in Iowa and Kansas. At the same time, metropolitan counties in California, Texas, and New York show only moderate slowdowns in mortality declines over the entire period 6/8🧵
Another interesting finding is that while nationally mortality declines have slowed down starting in 2010, and stagnated after 2014, these are not obvious thresholds at the state-metro level 5/8🧵
Deviations of state mortality from the national US average for metropolitan and nonmetropolitan counties by state. The graph shows that although generally metropolitan areas increasingly experienced lower-than-average mortality, nonmetropolitan areas fell behind and experienced higher-than-average mortality. However, at the same time, significant regional variation remains.
Even in the late 2010s, 8 states had lower female and male mortality in nonmetropolitan than metropolitan areas, highlighting that national-level trends and patterns can hide significant heterogeneity 4/8🧵
While we are accustomed to think of a nonmetropolitan mortality disadvantage, we show that in the early 2000s 19 states for females and 10 for males had lower mortality in nonmetropolitan than in metropolitan areas 3/8🧵
This is a descriptive study (with publicly available data) but I think it shows that deep description of a phenomenon can provide valuable insights even if it does not directly explore explanations for the observed patterns 2/8🧵
As promised, here a summary of our newest paper on metropolitan and nonmetropolitan mortality in the US 1/8🧵
onlinelibrary.wiley.com/doi/10.1111/... @astokespop.bsky.social @pennpsc.bsky.social @pophel.bsky.social @ruralhealth.bsky.social @jruralhealth.bsky.social
Studies like the one by Jacob Bor and colleagues are of great help in quantifying the extent of misreporting and the detailed misclassification ratios they report that can be applied by other teams that do not have access to the restricted ACS-NVSS linked data
Survey data linked with mortality records are great but 1) publicly available data is limited (NHIS, NHANES, some NLMS), 2) sample sizes are small and a lot of spatial and temporal granularity is lost, and 3) linkages with vital statistics are updated with long delays (NHIS now goes to 2019)