Physicians and by extension, providers do not understand stats....

[physasst]

Great piece here.

 

http://www.bbc.com/news/magazine-28166019

 

But it's not just that doctors and dentists can't reel off the relevant stats for every treatment option. Even when the information is placed in front of them, Gigerenzer says, they often can't make sense of it.

In 2006 and 2007 Gigerenzer gave a series of statistics workshops to more than 1,000 practising gynaecologists, and kicked off every session with the same question:

A 50-year-old woman, no symptoms, participates in routine mammography screening. She tests positive, is alarmed, and wants to know from you whether she has breast cancer for certain or what the chances are. Apart from the screening results, you know nothing else about this woman. How many women who test positive actually have breast cancer? What is the best answer?

• nine in 10
• eight in 10
• one in 10
• one in 100

Gigerenzer then supplied the assembled doctors with some data about Western women of this age to help them answer his question. (His figures were based on US studies from the 1990s, rounded up or down for simplicity - current stats from Britain's National Health Service are slightly different).

1. The probability that a woman has breast cancer is 1% ("prevalence")
2. If a woman has breast cancer, the probability that she tests positive is 90% ("sensitivity")
3. If a woman does not have breast cancer, the probability that she nevertheless tests positive is 9% ("false alarm rate")

In one session, almost half the group of 160 gynaecologists responded that the woman's chance of having cancer was nine in 10. Only 21% said that the figure was one in 10 - which is the correct answer. That's a worse result than if the doctors had been answering at random.

The fact that 90% of women with breast cancer get a positive result from a mammogram doesn't mean that 90% of women with positive results have breast cancer. The high false alarm rate, combined with the disease's prevalence of 1%, means that roughly nine out of 10 women with a worrying mammogram don't actually have breast cancer.

 

In my area in research, shared decision making, this has substantial implications. I would add that this extends to statistics in general, not just probability.

 

While providers know more than the general population regarding stats, that isn't necessarily saying much. I know that my PA program gave me barely any preparation for evaluating evidence, besides saying...."you need to keep up with current evidence"....There was little education in stats. When I finished my Masters, I had a better understanding, but still not a great one.

 

Now, I know that most PA programs and Medical Schools have courses in Evidence Based Practice/Medicine, and most know how to weight evidence (level I, II, IIa, IIb, etc.), but I still don't think we necessarily really understand the statistical analyses used in big clinical studies.

 

I do now, completing my Doctorate forced me to learn a heckuva lot more, and then I furthered my education with some postdoc work in stats at my current institution.

 

It's not just PAs, talking with many of my physician partners, they will readily admit that they had little education in either medical school or residency to really evaluate numbers and statistics, unless they did a separate PhD, or MS in clinical research, etc.

 

The point is, we need to be aware of this. How can we appropriately counsel patients, how can we provide a patient accurate information in order to participate in a true shared encounter, if many providers struggle with understanding the data they are trying to share, and may present inaccuracies?

 

Glyn Elwyn, who is a monster in this field, does give some suggestions....

 

I don't propose any answers, but this should be valuable food for thought for ALL providers, regardless of pedigree.

[Guest JMPA]

your example is poor at best. if a woman tests positive by mammography then her percentage should not be of concern. Rather the importance of rule out and definitive diagnosis should be stressed. In this case scenario using stats for comfort can be detrimental to her mental well being. On another note, your point at interpreting stats may just well be correct.

[Contrarian]

Apparently neither does the OP.....

[physasst]

your example is poor at best. if a woman tests positive by mammography then her percentage should not be of concern. Rather the importance of rule out and definitive diagnosis should be stressed. In this case scenario using stats for comfort can be detrimental to her mental well being. On another note, your point at interpreting stats may just well be correct.

 

 

Not really. If we want to move to a system of truly interactive, shared decision making, then an understanding of the statistics plays a huge role.

 

In this example, there are several options. She can have a biopsy immediately, she can have more definitive imaging (diagnostic mammogram), or she can also simply wait and see, and have a mammogram in 6-12 months to see if this was real, or a false alarm, there is no absolutely correct answer.

 

To put it another way, if you mistakenly tell a patient that they have a 90% chance of having cancer versus a 10% chance, their decision making might be very different.

 

Understanding probabilities can help the patient make that decision. I use a couple of decision aids with patients at work to help them understand probabilities as well, and I am developing some to help the providers during these discussions. I am working on one right now to address cervical myelopathy.

 

Mike

[Planteater]

We are looking for the PPV, right?

PPV= (sensitivity x prevalence)/ [sensitivity x prevalence + (1- specificity) x (1-prevalence)], right?

In our case, sensitivity = 0.9; prevalence = 0.01; specificity = 0.91

So, 0.9 x 0.01/ (0.9 x 0.01 + (1-0.91) x (1-0.01))

So PPV = 0.009/0.009 + 0.0891

PPV = 0.009/0.0981

PPV = 0.0917...

So wouldn't the answer be 9 out of 100? Did I calculate specificity correctly?

 

What am I missing?

[DissentingVoice]

your example is poor at best. if a woman tests positive by mammography then her percentage should not be of concern. Rather the importance of rule out and definitive diagnosis should be stressed. In this case scenario using stats for comfort can be detrimental to her mental well being. On another note, your point at interpreting stats may just well be correct.

The thing with life is this:  we ALWAYS have a chance of something bad happening.  I may be injured by a spacecraft tomorrow (ICD code v95.41).  Unlikely but possible.  By understanding statistics, you can have a discussion with patients to determine what probability is acceptable to them.  The percentage is very much a concern.

 

 

We are looking for the PPV, right?

PPV= (sensitivity x prevalence)/ [sensitivity x prevalence + (1- specificity) x (1-prevalence)], right?

In our case, sensitivity = 0.9; prevalence = 0.01; specificity = 0.91

So, 0.9 x 0.01/ (0.9 x 0.01 + (1-0.91) x (1-0.01))

So PPV = 0.009/0.009 + 0.0891

PPV = 0.009/0.0981

PPV = 0.0917...

So wouldn't the answer be 9 out of 100? Did I calculate specificity correctly?

 

What am I missing?

 

Yup.  No physician is actually going to take the time to figure that out, but it can be easily estimated with simple logic.

Use 1000 people.

10 have cancer. (1%)

990 are cancer free (99%)

90 cancer-free people test positive (9% false positive)

9 cancer patients test positive (90% sensitivity)

99 of 1000 people test positive (False positives + True positives)

9 of those 99 actually have cancer (approximately 1 in 10)

 

Make sense?

 

IMO, a lot of physicians don't get statistics well.  

[Guest Paula]

Seriously?  There's an ICD code for getting hurt by a spacecraft? You don't mean an alien spacecraft, do you? LOL! What would be the odds?

[DissentingVoice]

More fun codes:

V91.07 Burn due to water-skis on fire

W22.02 Walked into lamppost

 

W5609XA: Other contact with dolphin, initial encounter

S1087XA: Other superficial bite of other specified part of neck, initial encounter. (Hickey)

T7501XD, or, shock due to being struck by lightning, subsequent encounter

 

There's also a "Struck by turtle subsequent encounter" in there somewhere...

[db_pavnp]

This was a great article.  Another one I read a bit ago, residents were given a statistics quiz and did poorly despite having had coursework on the topic a couple years earlier.

[Guest Paula]

So a waterskiier whose skis are lit on fire by a hickey and is rescued by a dolphin is covered in the code book.  I'm so relieved. 

[medic25]

More fun codes:

V91.07 Burn due to water-skis on fire

W22.02 Walked into lamppost

 

W5609XA: Other contact with dolphin, initial encounter

S1087XA: Other superficial bite of other specified part of neck, initial encounter. (Hickey)

T7501XD, or, shock due to being struck by lightning, subsequent encounter

 

There's also a "Struck by turtle subsequent encounter" in there somewhere...

This was the best thing about switching to EPIC for EMR.  You could sit and enter the most random weird chief complaints and try to find the matching ICD-9 code.

[EMEDPA]

epic has a dx of "ill defined disorder" as well. 

[tunafish]

Working in cancer research for 15 years prior to coming to PA profession I yet to understand the value of stats and probability for a patient. Would you change your attitude towards your disease if I tell you that you have 60% chance of negative outcome vs. 40%. It is a huge difference, but does it matter to YOU? For a person have something, it is always 100%. What difference does it mean for you that pheo is a rare cause of HTN if you have it? It is important for providers to prescribe treatment or look for causes, but for patients... And even for providers: lets say there is chemo that works magic and 5% of patients and you yet to know how to find those 5% prior to treatment... you do it and hope? You are not doing it since it is expensive and it will not work for 95 out of 100???

We all became the slaves of cost cutting administrators and our patients are just numbers and parts of the statistics.

[physasst]

Working in cancer research for 15 years prior to coming to PA profession I yet to understand the value of stats and probability for a patient. Would you change your attitude towards your disease if I tell you that you have 60% chance of negative outcome vs. 40%. It is a huge difference, but does it matter to YOU? For a person have something, it is always 100%. What difference does it mean for you that pheo is a rare cause of HTN if you have it? It is important for providers to prescribe treatment or look for causes, but for patients... And even for providers: lets say there is chemo that works magic and 5% of patients and you yet to know how to find those 5% prior to treatment... you do it and hope? You are not doing it since it is expensive and it will not work for 95 out of 100???

We all became the slaves of cost cutting administrators and our patients are just numbers and parts of the statistics.

 

Some people might. This where patient centered care comes in. Many patients want everything done, some do not. What if you were a patient who didn't have insurance? That patients attitude regarding odds might change dramatically.

 

I see international patients all the time. Many of them are paying full out of pocket price for their care. They question things far, far more than most American patients do, although that is changing.

 

The point is, you need to be able to have a shared encounter with your patient, where you explain the odds, options for further diagnosis or treatment, probabilities, and allow THE PATIENT to make the decision. My point is, that they cannot make an accurate decision if the information they get is inaccurate.

 

One of the greatest crimes we commit as human beings, and yes, providers are human beings, is assuming that because we think a certain way....everyone else should think that way too. Not everyone is wired the same, not everyone has the same beliefs or fears, and there is a huge variance in the way people process information and make decisions informed by personal and cultural beliefs, fears, knowledge, etc.. We need to be able to tailor our discussions, and allow the patients preferences to be front and center. Yet they cannot do that, if they are making decisions based on erroneous facts. This is where we have to be accurate in what we are telling them.

[EMEDPA]

also people may make different decisions at different points in their lives.

tell a 90 yr old with cancer that a tx has a 10% success rate and they may say no, but the same pt at 50 might have said yes.

[ventana]

here is another way to look at it.....

 

100% chance of getting sued if the patient delays getting Dx Biopsy......  Stats are great, but when it comes to sitting down and talking to a 50 yr old female about a + mamo - the only thing to do is recommend Bx - seriously - I never want to sit in a witness stand, and say that the PPV was only x% and that is why I didn't Bx....

 

 

I know I normally preach following guidelines, and limiting interventions, but some times the writings on the wall.......

[physasst]

here is another way to look at it.....

 

100% chance of getting sued if the patient delays getting Dx Biopsy......  Stats are great, but when it comes to sitting down and talking to a 50 yr old female about a + mamo - the only thing to do is recommend Bx - seriously - I never want to sit in a witness stand, and say that the PPV was only x% and that is why I didn't Bx....

 

 

I know I normally preach following guidelines, and limiting interventions, but some times the writings on the wall.......

 

That's not patient centered care. I have discussions similar to this all the time regarding cervical spondylotic myelopathy. Even in younger patients, I often just watch them, unless they want a surgical referral, or, if they have severe symptoms. Many don't. Many have a mild myelopathic presentation. 

 

I explain the percentages, what the probability is that they might have a catastrophic neurologic event or progression of their myelopathy....(approximately 30% progress, while 30% improve, and 30% remain stable), I tell them that we don't really know the natural history of CSM that well, but that major, catastrophic events are rare.

 

I have about 15 of them right now in my panel (at least 3 that I can think of that are under the age of 55)..that I simply see every six months and re-examine. They're comfortable with that.....I'm comfortable with that. If they get worse, I will refer them over to our surgeons.

 

I'd also add, that patient preference is key. I also have some CSM patients that don't like those odds and demand to see a surgeon, and I will set that up.

 

It's about the patient, and their needs, desires, and preferences....our fears shouldn't play into that.

 

Mike

[GetMeOuttaThisMess]

Jerry Hoffman, M.D. out of UCLA will drive you nuts with statistical data in his monthly abstracts reviews, but IMO he's dead on.  I love to give percentages, but in a way that the patient can relate to (my prior posted example of statistical benefit from statins in primary CVD risk prevention for example).  This is where I get my "who gives a rip about strep" rant as well.

[physasst]

Yep, personally, I like the pictograms. A sheet showing 100 figures all grayed out. Then a number of them reddened to show risk.

 

I use one for acute low back pain, that shows 100 figures, with 7 of them reddened. This represents the probability that their acute back pain represents something serious such as a stenosis, fx, or other problem.

 

Then I show them a second one....it also has a 100 figures on it. This time, 70 of them are reddened. This represents the percentage of people without symptoms or pain walking around who will have abnormalities on MRI imaging of the spine.

 

I use them to explain why I would recommend against MRI imaging in acute low back pain (without extenuating circumstances-severe trauma, objective neurologic findings..etc.etc.)

 

Patients seem to understand these better. I think it is because we aren't just throwing numbers at them, but also showing them in visual depiction, what that means.

 

A researcher, prominent in this field from Denver, was working on decision aids for patients to decide on LVAD in heart failure patients. When he designed the pictogram which showed virtually an identical severe complication rate to the mortality rate without the device, he said the cardiologists refused to use it. "No one would ever elect to have one of these if they saw this".....so he was able to keep the pictogram for the mortality rate, but he had to change the complication rate to a description only, that is, if he wanted the cardiologists to use it. The problem is, that automatically biases any shared encounter. A shared encounter is one in which the patient is presented information on the diagnosis and options and probabilities for each treatment WITHOUT any bias, allowing the patient to evaluate the data, and make an informed decision. It was a eye opener when he presented to us. It simply means we need to do more than educate patients....we need to educate providers too.

 

For those of you in Emergency Medicine, we did this a while ago with acute chest pain presentations in the ED. Conventional treatments would indicate serial troponins with stress testing to r/o disease. What we found was, the majority of these patients that we admitted to our ED observation unit, had negative testing. So, was there a better way? We developed a chest pain rule, and then using that, would have a shared encounter with the patient, and using a pictogram, generated on the spot by the rule, we would explain that based on their history, ECG, and SINGLE troponin, this was the risk of them having a serious coronary event in the next 48 hours.

 

We would then explain that there were two options. 1. We could proceed as we always have, and admit to our observation unit for serial troponins and stress testing. 2. We could let them go home, sleep in their own bed, and have next day follow up with our cardiologists.

 

What was interesting? Many patients chose to go home. Some didn't. Some said, "I am not comfortable with that, I need to know, and I would feel better being here". Some would say...."That's it? I figured my odds would be higher....I'll go home and save my money, as long as I can get it checked out tomorrow".....

 

There was no right or wrong answer....both were acceptable outcomes.

[whoRyou]

The thing with life is this:  we ALWAYS have a chance of something bad happening.  I may be injured by a spacecraft tomorrow (ICD code v95.41).  Unlikely but possible.  By understanding statistics, you can have a discussion with patients to determine what probability is acceptable to them.  The percentage is very much a concern.

 

 

 

 

 

I would love to learn more about stats, but my spacecraft is about to leave without me ... 

 

 

2014 ICD-10-CM Diagnosis Codes > External causes of morbidity V00-Y99 > Air and space transport accidents V95-V97 > Accident to powered aircraft causing injury to occupant V95-

2014 ICD-10-CM Diagnosis Code V95.41 Spacecraft crash injuring occupant

• V95.41 is not a billable ICD-10-CM diagnosis code and cannot be used to indicate a medical diagnosis as there are 3 codes below V95.41 that describe this diagnosis in greater detail.
• ICD-10-CM codes become active beginning October 1, 2015, therefore, this and all ICD-10-CM diagnosis codes should only be used for training or planning purposes until then.

Source: icd10data.com

 

The statistics on sanity are that one out of every four Americans is suffering from some form of mental illness. Think of your three best friends. If they're okay, then it's you.

 

Lies, damned lies, and statistics

[db_pavnp]

physasst,

 

Great thread.  Do you have a background in stats that makes you feel more confident generating solutions like your troponin workout?

[EMEDPA]

physasst,

 

Great thread.  Do you have a background in stats that makes you feel more confident generating solutions like your troponin workout?

he is a recent DHSc grad. We have required coursework in stats and research methodology. one of the more painful courses I have had to take. not something I will focus on once done with school...

[EMEDPA]

 

The statistics on sanity are that one out of every four Americans is suffering from some form of mental illness. Think of your three best friends. If they're okay, then it's you.

 

 

2 of them are pretty sketchy in the psych dept so I guess I'm ok....:)

[physasst]

physasst,

 

Great thread.  Do you have a background in stats that makes you feel more confident generating solutions like your troponin workout?

 

Yes, I was an economist (undergrad) and we had a fair amount of stats/modeling work. Mine was a BS in Labor Econ. Then I also have a MS in addition to my PA Masters, and then finally, as Emed noted, I finished a DHSc in Organizational Behavior.

 

I also have done 4 post doc courses in stats (1. Biostatistics....2. Secondary dataset analysis....3. Health economic analysis/Bayesian modeling...and finally, 4. Survey statistics)...

 

There are a lot of things that I don't know very well, but I feel fairly confident with most statistical analyses, although, that being said, the really complicated, more advanced biostatistics I still leave to the PhD statisticians, or I go ask them how to do it if I don't have money..:)

[whoRyou]

2 of them are pretty sketchy in the psych dept so I guess I'm ok....:)

Hey I resemble that remark! LOL Wait, what does that mean about me having a BS degree in psych? Oh yeah, a bunch of Bull Crap LOL