Thread by @AmJEpi

👋 hi, @LucyStats here! Kicking off some live tweeting of this exciting discussion on the future of epidemiologic methods.

@nondogmatist kicks us off, asking the panelists to discuss what they see in the future of Epi

Starting us off, @edwardsjk and @robertwplatt both talk about *data*

The difference between the “source cohort” or “dark data” and the data that we’re actually using - We often just see the tip of the iceberg 🏔, we need to develop and use methods that take this into account

@robertwplatt states that we need clear actionable questions.

❓➡️ 📈
Make the question drive the method, not the other way around.

Building on these ideas, @WhitneyEpi states that we need to be *grounded in theory* to understand what the missing data are, know the gaps and the unseen populations 🏔, specifically mentioning the importance of demography and generalizability of results

@WhitneyEpi also mentions the importance of focusing on how we build teams, who we are bringing to the table 👏👏👏

Finally, @eschisterman1 focuses on two areas, the importance of:

📁 Better data
📋 Modern study designs

We can learn from our industry partners about data quality, we need to update the way we are designing our studies to keep up with modern methods

Q from the audience: how to we balance utilizing cutting edge tools with promoting scientific rigor?

@edwardsjk tying the people with epi theory expertise with those using cutting edge methods needs to be a priority

@WhitneyEpi building interdisciplinary teams is so important!

“Team science is not just advantageous, it’s essential”

— @nondogmatist

In epi and science in general, we are often in a space of *uncertainty*, we need to learn how to effectively communicate with decision makers who ultimately need to make a choice. — @robertwplatt, embracing the Epi mantra, #itdepends

@eschisterman1 poses a question to @WhitneyEpi: What methods should we think about developing to account for exposures that tend to be huge confounders, such as socioeconomic status?

@WhitneyEpi this is an important area, there are people doing this work via indices, latent class, fixed effects designs, etc. but this could use a lot more methodological work. Many people claim to control for “SES”, but their results are often quite vulnerable to bias

@nondogmatist poses the question, what additional training should we be adding for future epidemiologists, do we need more in “epistemology”?

@robertwplatt states that he agrees, but we need to be mindful if we add something in we have to remove something-what should be removed from our current training?

Reminds me (@LucyStats) of a convo on @theeffortreport, @rdpeng is adamant that you can’t add without subtracting!

@eschisterman1: We have embraced the idea of framing questions as clinical trials, but we are conducting far fewer *actual* randomized trials. We need to focus on modern RCTs, for example adaptive designs

@WhitneyEpi I love a question that really says: “is this factor important? yes or no.” To answer this focused question, serveral study designs are possible, even a very well designed, carefully thought out case series can help answer this question

@robertwplatt: Let’s think about dependent sampling to get at an analytic cohort that is most efficient to getting back to our core question - this is bread and butter 🍞 🧈 for our survey sampling, but it’s not used as much broadly

@WhitneyEpi: an important consideration, though, is we don’t know what we don’t know 🤷‍♀️ - survey sampling is great if we assume the past is the same as the future, but we need good methodologist to help think through whether this is universally representative

Questions from students at @UNC what is your takeaway lesson for aspiring & new epidemiologists?

@WhitneyEpi: find the 🕳 gap, the thing that makes you think “someone should really be doing this” - that someone could be you!

@eschisterman1: Sometimes being stubborn has served me well; and sometimes it was very detrimental - try to predict this!

@WhitneyEpi chimes in, sometimes having good colleagues can help with this! 👯‍♀️

@robertwplatt: Learn how to learn from a variety of things - we have a lot that we can learn from outside of Epi! Sometimes you don’t know what is going to be useful, mentions:

📈 old stats class
🤔 philosophy class
💻 tech companies

All proving to be useful!

@edwardsjk echos this sentiment, states the importance of drawing on as many areas as possible!

@edwardsjk: the COVID-19 crisis has brought our skill set to the forefront, our ability to think through things like the denominator, the methods being utilized, has been so important. It had also brought the lack of public health infrastructure to the forefront

@eschisterman1 brings up a crucial point: COVID-19 forces is to consider what is the best for society versus what is best for the individual; these ethical questions are so important and hard to deal with

We have reached the end of the allotted time.

🙏 thank you to our organizer, @nondogmatist, and our panelists: @edwardsjk, @WhitneyEpi, @robertwplatt, & @eschisterman1!

I encourage everyone to keep the conversation going here in the comments!

👋,
@LucyStats