Relative vs Absolute risks: Why Relative Risks Are Misleading, and How To Communicate Absolute Risks
ฝัง
- เผยแพร่เมื่อ 11 มิ.ย. 2024
- We see relative risks is the headlines all the time. "Eating X DOUBLES the risk of Y." But they are often misleading: to understand whether a relative risk is actually worth caring about, you need to know the change in absolute risk.
A change from 0.001% to 0.002% is after all, a doubling of risk, but so is a change from 5% to 10%. The first one hardly matters, the second really does!
David explains the difference between the two with some real life examples and provides some best practice tips to communicate absolute risks.
***
David Spiegelhalter is a statistician, author and broadcaster. His book THE ART OF STATISTICS is a worldwide bestseller. He was the Winton Professor for the Public Understanding of Risk at the University of Cambridge for many years, and served a term as the President of the Royal Statistical Society.
🔗🔗 Links🔗🔗
RealRisk: a tool for science communicators 👉👉 realrisk.wintoncentre.uk
The Winton Centre's website: 🌐 wintoncentre.maths.cam.ac.uk 🌐
Risky Talk the podcast with David Spiegelhalter: 🗣️🎙️ riskytalk.libsyn.com/ - วิทยาศาสตร์และเทคโนโลยี
2:03 "18% of WHAT?"
2:50 "expected frequencies: what it means for 100 people" (absolute risk, AR).
This is the perfect video! Thankyou!
Thank you so much Mr. David. This one helped a lot to comprehend the difference and the importance of communicating the same in a specified manner. Really helpful 🙏
Absolutely beautiful explanation, thank you!
Excellent work.
Great explanation. Thank you!
Well done! Thank you!
Great examples!
much appreciated!
1% increase and the research was done using epidemiology / food frequency questionnaire.. But the media release goes "Bacon will kill you" 🤦🏼♂
Great. Now I understand.
This was a really useful video - but one aspect that was missing, and I felt would have been really helpful to detail a bit more, was the actual calculation/formula's for context, but also to speak a bit more to the 'baseline risk' aspect. This seems a bit vague and I envisage is still not fully grasped by many. For example, in an epidemiology cohort study looking at say levels of physical activity (e.g. mins/day) and CVD events, you might compare different levels of activity (30, 60, 120 min/day) to a reference of say 0 or 10 mins/day using a cubic spline and computing hazard ratios (e.g. HR for 30 vs. 10 mins/day = HR 0.75, 60 vs. 10 = HR 0.50...). How would you then determine absolute risk(s) in this context for better interpretation? What is the 'unexposed' group here in order to determine 'baseline risk' and thus absolute risk? It's easy when thinking about a control vs. intervention group, but less so when thinking about a certain level of exposure relative to a chosen reference. I've had a look at RealRisk, but still feeling rather uncertain. Thanks so much in advance.
To answer your question: How would you then determine absolute risk(s) in this context for better interpretation? You would need to know the number of CVD events for your reference group. For example, if we know that a cohort (n=500) that only exercises 10 mins/day has a rate of 4 heart attacks per year, and the hazard ratio for 30 mins/day with respect to 10 mins/day is 0.75, then we expect to see only 25% of the reference rate of CVD, which is 1 heart attack per year. That is an absolute reduction of 3 heart attacks per year. Hope this helps.
Nice. Subscribed
That is good to know.
Is Absolute risk the same as Attributable risk? In the First Aid textbook for the USMLE step one exam they only mentioned Attributable risk and Absolute risk reduction but there is no Absolute risk, thank you
Can you ple explain and compare the RR and OR together, I actually can differentiate between them in meaning 😢.
I need you to come live with me and calm me the fuck down😂😂
Ha ha. Spiegelhalter is a legend, to calm down just listen or read most of his TV appearances or in the press.
Great video, thanks! The RealRisk link you provide in your notes no longer works. Could you update it? Or did you discontinue the tool?
Odd - should be fixed now!? The backslash seemed to be throwing it off...
so absolute risk is over all prevalence
Yes, basically!
@@TheWintonCentre that is more intuitive. After taking biostat and and epidemiology your 4 minute video is more magic.
@@TheWintonCentre 0.18*6 =1.08, so the 6+1.08=7
@@TheWintonCentre Can we also use same calculations for odds ratio?
What you think abt vaccine of corona
'I do feel it is for the good of our society that as many are vaccinated as possible.'
www.standard.co.uk/comment/david-spiegelhalter-risks-benefits-astrazeneca-vaccine-blood-clots-b928717.html
@@TheWintonCentre I mean did you checked it's ARR???
@@rallyworld3417 By ARR are you referring to 'absolute risk reduction'!? The most useful statistics for how much of a difference vaccines make are efficacy and effectiveness (slightly different!). Efficacy is the relative risk reduction in a trial, comparing the unvaccinated group with vaccinated group. Effectiveness is the same idea but in the real world, where conditions may be slightly different than in a trial. All the major vaccines have impressive efficacy & effectiveness, and have already saved many thousands of lives!
It's worth noting that vaccines are one area where relative risk reduction (i.e. their efficacy) are actually more useful and appropriate than absolute risk figures - because the relative reduction can be used across different contexts, while the absolute figure is specific to one context.
@@TheWintonCentre no arr is efficacy compared to ppl who got vaccinated and placebo. And rrr is efficacy relative to other vac
Enjoy your sandwich while others are scared of eating meats and proteins
I don't know anyone that's afraid of ingesting protein. Sounds like a contrived and vacuous point here.
There are go reasons why the industrialized pharmaceutical corporations use relative risk in the marketing campaigns and downplay absolute risk. Mostly it is because of perversion and greed.
Surprisingly MDs like parroting drug companies marketing strategies. Sad really.
Great video but please skip the bacon sandwich!!
I no longer believe anything you say. Fired….you have been bought and hopefully you will acknowledge this and leave public life.
Why? Because he is explaining basic statistics? I'm vegan, but he's right. Notice that he didn't promote eating extra bacon. He was just pointing out the personal health impacts. Of course, if he were to mention the actual numbers per million people that eat extra bacon and extra pork sandwiches, it would be intuitive but he is directly addressing the statistical differences between two often confused terms. Both figures are significant and to reduce overall risk, humans shouldn't be eating bacon anyway. Studies that look at vegetarians versus processed meat eaters of course reveal far more distribution between the means, but people that are already unhealthy show small gains in being more unhealthy over time. No surprise. Intervention studies are amazing in my view as they can quickly show changes in metabolism and overall health very quickly, lending credence to changing one's current diet, meat-based or otherwise.
Ive always been curious - if carnivores think that every study showing meat is even a tiny bit bad for you is just funded by vegans and is untrustworthy, where do you get your information from? how do you trust any source out there? considering that every carnivore influencer is selling you a product and wants you to believe them as well?