Data Science x Public Health
This podcast discusses the concepts of data science and public health, and then delves into their intersection, exploring the connection between the two fields in greater detail.
Data Science x Public Health
This Is Why Competing Risks Don’t Work (And Nobody Talks About It)
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Competing risks methods are often presented as a more realistic way to analyze time-to-event data in epidemiology and public health. They promise to handle situations where other events prevent the outcome of interest from ever occurring. But what if the method becomes more sophisticated while the interpretation becomes less clear?
In this episode, we break down why competing risks analyses are often overtrusted, how the choice of estimand quietly changes what the result means, and why better methods do not remove the need for sharper scientific thinking.
👉 Enjoyed the episode? Follow the show to get new episodes automatically.
If you found the content helpful, consider leaving a rating or review—it helps support the podcast.
For business and sponsorship inquiries, email us at: 📧 contact@bjanalytics.com
Youtube: https://www.youtube.com/@BJANALYTICS
Instagram: https://www.instagram.com/bjanalyticsconsulting/
Twitter/X: https://x.com/BJANALYTICS
Imagine you buy a high-end GPS. It calculates traffic, weather, road closures, but then you accidentally type in the wrong destination. So all that brilliant math is just getting you lost with incredible precision. Welcome to the deep dive. Today we are looking at an article called The Illusion of Realism in Competing Risks Analysis. Our mission here is to figure out why a highly advanced statistical method in medical research often creates this false sense of security. Like it actually tricks scientists into answering the completely wrong questions. You might think better math always equals better science, but well, today we'll uncover why technical sophistication can actually become a trap.
SPEAKER_01Exactly. And to understand that trap, we really first have to understand the math, right? Which it was originally designed to fix a very real flaw in how we study diseases.
SPEAKER_00Right. The standard models.
SPEAKER_01Historically, if researchers were studying, say, the risk of getting a hospital complication, standard models just assumed every single patient remained eligible to get that complication forever.
SPEAKER_00Which is obviously not how reality works.
SPEAKER_01No, real life just doesn't work that way. Like a patient dying from something completely different prevents that hospital complication from ever happening at all.
SPEAKER_00Yeah, it's like waiting at a bus stop forever, just completely unaware that the route was actually canceled. Like standard models just keep you waiting at the stop, ticking up the theoretical odds that the bus will eventually arrive. But competing risks analysis actually acknowledges the cancellation.
SPEAKER_01Yeah, and that cancellation completely changes the math. Cancer research or chronic disease, acknowledging that one event can block another from happening is just a massive step forward. And it models the world much closer to how it actually operates in reality.
SPEAKER_00But because this method feels so much closer to reality, it creates a dangerous blind spot. Researchers start suffering from what the authors actually call conceptual complacency. So they use this advanced math, the paper looks impressively modern, and everyone assumes the underlying problem is just solved. But wait, I'm stuck here. If we're studying strokes and the math accurately accounts for the fact that some patients died of other causes, shouldn't the stroke risk it spits out just be automatically correct?
SPEAKER_01Well, you would definitely think so.
SPEAKER_00So how does being perfectly accurate lead to a blind spot?
SPEAKER_01Because the math only processes numbers, but doesn't choose the meaning behind them. The analyst still has to decide what specific version of reality they're actually modeling.
SPEAKER_00Okay. What do you mean by version of reality?
SPEAKER_01Well, are you trying to isolate the underlying biological tendency towards strokes, sort of pretending death from other causes simply didn't exist? Or are you looking to measure the actual real-world burden of strokes in a specific population? Because those are two entirely different scientific questions.
SPEAKER_00But they look similar on paper.
SPEAKER_01Exactly. And researchers often run the math without clarifying which one of those questions they are actually asking.
SPEAKER_00Okay, so if a researcher accidentally calculates the biological tendency instead of the real-world burden, how does that actually impact a patient walking into a clinic tomorrow?
SPEAKER_01Aaron Powell Well, think about hospital planning. Let's say administrators use this math to plan how many stroke recovery beds they'll need over the next decade.
SPEAKER_00Which requires knowing the real world burden.
SPEAKER_01But if the researcher accidentally gave them the model that pretends death from other causes doesn't exist, they are going to vastly overestimate how many beds they actually need.
SPEAKER_00Oh wow. So millions of dollars just get wasted.
SPEAKER_01Yeah, millions. The math was executed flawlessly, but it answered the wrong question. It essentially treated a theoretical biological scenario as if it were a real-world prediction.
SPEAKER_00So the method gave them a highly precise answer to a question they didn't even know they were asking. So how do researchers actually fix this?
SPEAKER_01Well, they have to define the target question, what statisticians call the estimand, and they have to do this before they even touch the data. They must explicitly state what specific world their estimate lives in. Like good epidemiology shouldn't just claim a competing risk was handled.
SPEAKER_00It needs to explain what handling it actually means for the real-world interpretation. So the core lesson here isn't that competing risks fail because of bad math. They fail because analysts and readers treat methodological sophistication as a substitute for conceptual clarity.
SPEAKER_01It really all comes back to your ultra-realistic GPS analogy.
SPEAKER_00Right. It doesn't matter how flawless the routing algorithm is if you're driving to the wrong city. Which really leaves you with this to think about. If advanced statistical models and epidemiology can provide such false confidence by obscuring the true question, how many other data heavy fields in our lives are currently suffering from their own illusion of realism, like financial forecasting or artificial intelligence, elegantly providing us with exactly the wrong answers?