HMS HCP

📣Exciting news! Professor Alex Luedtke has been named a 2026 COPSS Emerging Leader Award winner! The award recognizes early-career statistical scientists whose leadership and scholarship are helping to shape the future of the field. Congratulations, Alex! 🎉 @COPSSNews

🚨We're hiring! Come join us as a Postdoctoral Fellow in Health Care Policy - Statistical Methods! (PI: Alex Luedtke) academicpositions.harvard.edu/postings/15365

Discussants include HCP's Prof. Joe Newhouse, whose leadership on the landmark Health Insurance Experiment (HIE) was central to the development of these methods. Read the preprint: rss.org.uk/training-event…

The paper is joint work with Profs. Ambarish Chattopadhyay and Carl Morris, and revisits & extends Morris's Finite Selection Model to improve randomized experiments/clinical trials.

An exciting event for anyone interested in data science & health policy! This Wednesday at 10 AM ET, HCP's Prof. Jose Zubizarreta will present his new discussion paper at the Royal Statistical Society (@RoyalStatSoc). Register to attend online: rss.org.uk/training-event…

Congratulations to @Michael_Chernew (Reinhardt Distinguished Career Award) and @FangliGeng (Outstanding Dissertation) on their @AcademyHealth awards! Your leadership and research are transforming the future of health services and policy. 🏆👏 #HealthPolicy

Hear about research at risk from Ron Kessler, who studies suicide risk in veterans: youtube.com/watch?v=VGNRHz… #Harvard #Research #HarvardMed

Pharmaceutical tariffs could raise prices and reduce innovation, says @DrWmPadula in an interview for @ABC7. Long-term, "Americans could pay by not having the availability of new innovative medications." abc7.com/post/tariff-un…

Universal annual depression screening for adolescents in primary care is cost-effective, especially if health systems improve access to telemedicine behavioral health, reduce treatment costs, or enhance treatment effectiveness. ja.ma/4jY1ebg

Two-part article in @Health_Affairs Forefront with my thoughts on improving risk adjustment Actually more of a synthesis of great work and insights from the field and a set of recommendations that follows from considering it all together. This is an area where academia/health econ/HSR – both in the US and internationally – has really shined in advancing the science. My hope is that this piece helps folks think through what is a very thorny and pressing topic. Links: Part 1 - healthaffairs.org/content/forefr… Part 2 - healthaffairs.org/content/forefr… Main points: The goal is not to maximize fit (predictive accuracy). Fit is important but does not define selection incentives alone, and there are important tradeoffs with gaming and cost control incentives. Better fit does not mean a better system. The goal is to establish a set of incentives that best advances an economic/social objective (e.g., efficiency), as opposed to a statistical one (e.g., R2). There are a ton of ideas out there. Some are better than others, but key is no single idea will get us what we want. A combination of strategies is needed. Good news is that a lot of thoughtful people have done a lot of great work, so “we have the technology” (though the tech may be less flashy than a bionic man). The inputs are critical. There are some very sensible measures that can be taken to scale back the more manipulable elements of the HCC model to mitigate coding incentives, while keeping less problematic conditions and sources of diagnoses. Don’t need to throw the baby out with the bathwater, but the baby needs a bath. Measures to scale back the HCC model will exacerbate selection incentives to some extent, motivating the addition of inputs from less manipulable data sources. Independently collected survey data hold promise, and existing surveys can be built upon. A future system in which data on a range of inputs are independently collected for a sample of a plan's or ACO’s population is a worthy aspiration (and we should want to collect those data anyway for monitoring and research purposes). Drug-based condition categories may also be helpful, and there is precedent in other systems and countries, but need to mindful of prescribing incentives (an area in need of further study). Machine learning is great. But important to apply it to candidate inputs that offer desired incentives. Reinsurance is risk adjustment’s BFF. It offers nearly free fit by capping extreme losses (and extreme profits in a “two-sided” scheme) where under/overpredictions are most severe (contributions to selection incentives potentially strongest), without eroding cost control incentives much because it only affects a very small % of enrollees. It also helps mitigate coding incentives for some nuanced reasons. Put this all together, and we could have a MUCH better risk adjustment system. The piece includes links to some great work and thinking by Tom McGuire, Rick Kronick, @SherriRose, @MikeGeruso, @richardvankleef, @Michael_Chernew, Jeah Jung, @MelindaBBunitin, @MeghanBellerose, @nirudatta, Alan Zaslavsky, Randy Ellis, Sonja Schillo, and Sherry Glied, among others It was also fun talking to Rob Lott about some of this on A Health Podyssey: healthaffairs.org/do/10.1377/hp2…

Congratulations to Josh Feng, Emma Dean, and Luca Maini on being selected for the 2025 cohort of #NIHCM Research Grantees! Read more in the release: bit.ly/ResGr25 | #NIHCMGrants2025

New editorial in @AnnalsofIM by @JMichaelMcW finds flaws in the payments of Medicare Advantage Plans-where insurers can profit from coding tactics. A solution is possible, though political will is essential. acpjournals.org/doi/10.7326/AN… #HealthPolicy #MedicareAdvantage

Congratulations to Professor José Zubizarreta on being honored with the ASA's 2025 HPSS Mid-Career Award! His pioneering work in healthcare policy and causal inference is truly inspiring. Kudos! 🏆 #Statistics #HealthPolicy @harvardmed @HarvardBiostats

Harvard physician-researchers Christopher Worsham and Anupam Jena explain why taxi drivers have a lower risk of Alzheimer's and what that means for the rest of us time.com/7206222/taxi-d…

Very interesting study from the Michigan folks. You could have predicted it was from them based on the topic and cleverness of the finding @jdimick1

Exciting DENS Seminar today! Speaker Tom Barrett shared innovative studies to accurately estimate averages in complex datasets, bridging advanced statistical theory with practical data analysis. #Statistics #DataScience #ANOVA @harvardmed @HarvardBiostats

Over 60 percent of nursing homes in US counties do not offer I-SNPs in long-term care plans. Amanda Chen of @Harvard et al investigate potential causes for low enrollment. Read the full article here: bit.ly/3Yrn52X

🎉Congratulations to STAT Wunderkind, Michael Ann Kyle! Dedicated to transforming the healthcare system from bedside care to policy research, @michaelannica is focused on alleviating patients' administrative burdens. Read more: statnews.com/meet-the-2024-… @statnews @harvardmed

Last week’s DENS Seminar featured speaker Maya Mathur, presenting new possibilities beyond traditional funnel plots! The session focused on boosting research integrity and reliability, key for healthcare policy advancement. #PValue #MetaAnalysis @harvardmed @HarvardBiostats

Research update: we have a new version of our pharma M&A paper! AND, we posted the data we collected on m&a deals. Links below, main takeaways in thread. Paper summary: lucamaini.com/working-papers… Full draft: lucamaini.com/s/feng_hwang_l… Data: lucamaini.com/data