Haedi Thelen

PharmD, PhD Student at the University of Pennsylvania | Views my own

[ISPE Student Council @S_ISPE]

‼️Vote for your new Co-Chair in the ISPE student council‼️

Your vote is important in deciding who will represent our student community, this year as Co-Chair, and the following as Chair!

Check out the candidates and vote via the following link: https://t.co/oSqz6pllov

[ISPE Student Council @S_ISPE]

Are you an ISPE student looking to contribute to ISPE, build real-world leadership skills and gain valuable mentorship? 🤝
The ISPE Student Council is seeking a passionate student to join our team as the next Co-Chair!!

#ISPE #StudentCouncil #SISPE #ISPEAnnual2025

[Haedi Thelen @HaediThelen]

Presenting today at #ISPEAnnual2025 in DC!
📅 Aug 26 | 🕑 2:15–2:30 PM | 📍 Ballroom AB
Session: One size fits none: unravelling treatment heterogeneity
Topic: Comparing disease risk score methods to study treatment effect heterogeneity in RWD.
#Pharmacoepidemiology #HTE

[ISPE Student Council @S_ISPE]

Are you interested in participating for the 2025 Research Proposal Competition but unsure how to proceed?
Please fill out this quick survey so that we can assist you with your questions!
https://t.co/kVwCzVft3p

[Penn CREST @Penn_CREST]

Great new paper in Clinical Pharmacology & Therapeutics led by doctoral student Haedi Thelen:

Characterizing Treatment Effect Heterogeneity
Using Real-World Data

https://t.co/RyWp9mNAHv

@HaediThelen @HennessySean

[ISPE Student Council @S_ISPE]

📢 Registration is now open for the RWE Research Proposal Competition 2025! Your proposal must include the use of real-world data to address a research question.
💡The winning team will get the opportunity to present their proposal at a SISPE Webinar
https://t.co/oNUJh9qAVz

[ISPE Student Council @S_ISPE]

📢 Registration is now open for the RWE Research Proposal Competition 2025! Your proposal must include the use of real-world data to address a research question.
💡The winning team will get the opportunity to present their proposal at a SISPE Webinar
https://t.co/8r5veQT9Ue

[ISPE Student Council @S_ISPE]

The ISPE student council is excited to host the first webinar for this academic year- The Research Proposal Showcase!
Save the date and scan the QR code below to join us on the 8th of November!
Alternatively, use this link: https://t.co/DxFX4ERwp8

[ISPE Student Council @S_ISPE]

Register here: https://t.co/mCIN7rmd5V

[ISPE Student Council @S_ISPE]

The ISPE Student Council is thrilled to announce the winners of our first RWE Research Proposal Competition! We received 26 submissions from four continents, making it a huge success! Each proposal was scored by two expert reviewers from ISPE 🧵1/3

[ISPE Student Council @S_ISPE]

Congratulations to the winning team from @mcgillu @IspeMcgill Rachelle Haber, Omotayo Olaoye, and Stephen Ogbodo, for their proposal on the comparative effectiveness and safety of GIP/GLP1 dual agonists and GLP1 receptor agonists in patients with type 2 diabetes 🧵2/3

[Haedi Thelen @HaediThelen]

11/11
For more on balancing weights see these resources:
Easy: 🤗 my paper (coming soon??)
Serious: 🤔 DOI: 10.1097/EDE.0000000000001644
More serious: 🫠 DOI: 10.1201/9781003102670-16
The most serious: 🤯 DOI: 10.1002/sim.8659

[Haedi Thelen @HaediThelen]

3/ 11
Where overlap is weak, logistic regression models make the PS very large or small. This means the IPW are extreme. Extreme weights can lead to bias and wide confidence intervals❌❌

[Haedi Thelen @HaediThelen]

10/11
Better balance, less extreme weights. Balancing weights are a great alternative to logistic regression IPW, esp when overlap is weak. Use them in all the same settings, like marginal structural models and augmented IPW. What do you think? Will you use balancing weights now?

[Haedi Thelen @HaediThelen]

2/ 11
Logistic regression modeling to estimate the PS works well when groups have great overlap. But, as in most of my datasets, comparison groups are a little different. Like this:

[Haedi Thelen @HaediThelen]

1/ 11
Still using logistic regression to calculate propensity scores for your inverse probability weights?
Bet you’re tired of dealing with extreme weights and iterating that model over and over till you get balance. Time to use BALANCING WEIGHTS and get it done in one step.