View previous topic | View next topic

Communicating risk

Page 1 of 1


How would you prefer this risk reduction to be communicated to you by e.g. your doctor?
This drug will reduce your current risk of heart attack by 20%.
76%
 76%  [ 10 ]
If we treated 100 people like you with this medication, 4 people would benefit by not having a heart attack.
7%
 7%  [ 1 ]
The likelihood of you benefiting from taking this medication is 4%.
15%
 15%  [ 2 ]
Twenty five people would need to take this drug in order for 1 person to benefit.
0%
 0%  [ 0 ]
There is a 96% chance you will not benefit from taking this drug.
0%
 0%  [ 0 ]
Total Votes : 13

smiley_face
1231649.  Thu Mar 23, 2017 5:53 pm Reply with quote

Just a quick poll about how you would like risk to be communicated to you if you were being offered a medication (one tablet, once a day, long term, hypothetically no side effects) to reduce your heart attack risk.

There are several options - which one seems clearest to you?

This is something I find to be Quite Interesting....

A brief explanation of the stats in white below (highlight it to see it after answering the poll)

You currently have a 20% chance of having a heart attack in the next ten years. You can take a tablet that will reduce that risk to 16%.
Option 1 uses relative risk, that is, how your risk changes relative to the original risk.
Option 2 uses number needed to treat as a natural frequency (i.e. out of 100).
Option 3 present the absolute risk reduction.
Option 4 give you the number needed to treat.
Option 5 present the absolute risk reduction, but framed negatively.

 
suze
1231658.  Thu Mar 23, 2017 7:13 pm Reply with quote

I chose Option 1, since that is the way that I'd prefer to have the advantage of the medication presented to me. I know there's a maximum length for a poll answer on these forums and that may have constrained you, but I'd prefer Option 2 to continue "that they would otherwise have had".


But let us now introduce a downside, and in practice there will almost always be one. Let us suppose that the drug did indeed reduce your risk of heart attack by 4%, but at the same time it increased your risk of contracting lupus (other unpleasant medical conditions are available, but it's never lupus ...) by 1%. (I know that that in practice this drug would never be approved if that figure were so high, but let's keep the numbers simple.)

I think I'd want that second information communicated to me in the style of Option 3, so "The likelihood of you getting lupus as a result of taking this drug, when you wouldn't have done otherwise, is 1%".

Is that because I want the good news to be a big number but the bad news to be a small number? Probably, and I know that it doesn't really make sense.

 
dr.bob
1231711.  Fri Mar 24, 2017 7:08 am Reply with quote

I avoided option 1 since relative risk tells you nothing without other information.

This is a tactic that most tabloids exploit when you see headlines that scream "<random-thing> increases your risk of cancer by 200%!!!!!" It's only when you read to the final paragraph of the article that you find your risk is increased from 0.0000001% to 0.0000003%.

 
smiley_face
1231781.  Fri Mar 24, 2017 4:48 pm Reply with quote

Indeed, dr.bob, relative risk reduction is not terribly helpful in real life, although it is probably the most commonly used.

Relative risk reduction is the usual value reported in clinical trials, but is not terribly helpful when trying to explain the benefit of a drug to an actual patient.

I quite like using natural frequencies:

If you take 100 patients with a 20% risk of MI in the next 10 years and give them a medication that reduces their risk by 20%...

80 of those patients would never have an MI anyway.
16 of those patients will have an MI regardless.
4 patients will actually benefit from the drug (that is, they would have had an MI without the drug, but by taking it, the MI is prevented).

 
Jenny
1231892.  Sat Mar 25, 2017 6:07 pm Reply with quote

But option one is the most persuasive. I don't care how many other people will benefit, or that I may not benefit at all. What I want to know is what my chances are. If you want to sell me a lottery ticket and tell me that my chance of winning is 20%, I'm more likely to buy it than if you tell me my chance of losing is 80%. Yes, I know it's the same but it highlights the advantage to me.

 
Leith
1231896.  Sat Mar 25, 2017 6:32 pm Reply with quote

I think I'd just prefer to be given the full picture - i.e. what is the absolute risk before and after. All of the above options are missing some part of that information.

If it had to be one of the above, I'd say option 3 provides the more useful subset of relevent information, and expresses it in the terms most specifically relevent to my needs.

 
dr.bob
1232024.  Mon Mar 27, 2017 4:38 am Reply with quote

Maybe we can push for a standardised method of communicating risk, where everything has to be expressed in Micromorts.

 
smiley_face
1232083.  Mon Mar 27, 2017 3:06 pm Reply with quote

Jenny wrote:
If you want to sell me a lottery ticket and tell me that my chance of winning is 20%, I'm more likely to buy it than if you tell me my chance of losing is 80%. Yes, I know it's the same but it highlights the advantage to me.

This is known as the framing effect, whereby people respond differently to potential losses and gains, in all sorts of circumstances, not just medical risks/benefits. Such as:

Quote:
93% of PhD students registered early when a penalty fee for late registration was emphasized, with only 67% doing so when this was presented as a discount for earlier registration.

From here.

 
RLDavies
1233488.  Sun Apr 09, 2017 10:10 am Reply with quote

I'm with Leith. The best and most ethical way of presenting the information is: "You currently have a 20% chance of having a heart attack in the next ten years. You can take a tablet that will reduce that risk to 16%." This is enough data to begin to make an informed choice.

In reality, the doctor and patient would also need to take into account precisely why the patient is at risk of heart attack, and how the drug works. If the patient has a faulty pacemaking system and the drug lowers cholesterol, then it would be useless.

"Will reduce your risk by 20%" is biased in favour of persuading the patient to take the drug. The other four choices are all biased in favour of persuading the patient not to take it.

"25 people need to take it for 1 person to benefit" has the odd implication that the 25 people are somehow working together for the good of the one.

 
RLDavies
1233716.  Tue Apr 11, 2017 9:01 am Reply with quote

Forgot to add that, of course, in the real world there's still more risk balancing to take into account. Medicine, being a branch of biology, gets messy like that.

You have a 20% chance of having a heart attack in the next ten years...
Based on statistics of the general population sharing your sex, age, weight, and general lifestyle habits. And assuming you don't change your lifestyle in the meantime.

This tablet will reduce that risk to 16%...
Assuming it adequately addresses the reason why you have the risk in the first place.

Meanwhile, the tablet has a 20% chance of causing side effects...
Again based on general statistics.

...of which 5% will cause enough discomfort that you'll prefer to stop taking the tablet...
Depending on your level of tolerance to discomfort.

...and 0.05% will be actually dangerous to your life.
Mainly depending on your genetic profile, which is unknown, so we'll just have to see what happens.

There's also a 0.5% chance that you might develop an allergy to it.
Assuming you're not particularly allergy-prone or already allergic to related drugs, in which case it's probably not worth the risk.

Drug X blocks the action of this new one, so if you're on X we'll have to change it or rethink the new one.
Drug Y boosts the effect of the new one, so if you're on Y we'll have to consider dosages carefully.
Drug Z interacts with the new one to cause a whole different set of potential side effects, including a greater risk of dangerous ones.

 

Page 1 of 1

All times are GMT - 5 Hours


Display posts from previous:   

Search Search Forums

Powered by phpBB © 2001, 2002 phpBB Group