Patients at Risk: The Rise of the Nurse Practitioner and Physician Assistant in Healthcare

[cara susanna]

---

Members don't see this ad.

Varies by patient and presentation of course, but I think the standard emerged because old-school psychoanalysis set an outrageously high opening bid. Amazed we have come down as much as we have from 1-2 hours 3x/week.

I often feel like my therapy sessions are maybe 15-20 minutes of actual content with another 20-30 of random chit-chat (i.e. "rapport building"), patients venting despite efforts to redirect, or me asking questions of questionable clinical utility in the vague hope of yielding grist for the mill. Maybe I'm just bad at this, but based on my conversations with others I doubt my experience is unique. Quite a few protocols I feel like could be condensed to 20-30 minutes. I actually think with technology integration we could get there. I'm not a big fan of standalone "therapy apps" despite doing some work in that area, but I think using apps (and similar tools) combined with shorter traditional sessions has substantial promise.

There will always be the occasional patient who genuinely needs more time to understand concepts and the like. I don't have the data to back it up, but I genuinely believe a good portion of a typical therapy session is just filler.

Yes, absolutely! That's why I loved PCMHI, I excelled at the 20-30 min sessions.

 

[WisNeuro]

Varies by patient and presentation of course, but I think the standard emerged because old-school psychoanalysis set an outrageously high opening bid. Amazed we have come down as much as we have from 1-2 hours 3x/week.

I often feel like my therapy sessions are maybe 15-20 minutes of actual content with another 20-30 of random chit-chat (i.e. "rapport building"), patients venting despite efforts to redirect, or me asking questions of questionable clinical utility in the vague hope of yielding grist for the mill. Maybe I'm just bad at this, but based on my conversations with others I doubt my experience is unique. Quite a few protocols I feel like could be condensed to 20-30 minutes. I actually think with technology integration we could get there. I'm not a big fan of standalone "therapy apps" despite doing some work in that area, but I think using apps (and similar tools) combined with shorter traditional sessions has substantial promise.

There will always be the occasional patient who genuinely needs more time to understand concepts and the like. I don't have the data to back it up, but I genuinely believe a good portion of a typical therapy session is just filler.

Definitely therapy specific. These days when it comes to therapy, I am pretty much only treating PTSD and panic disorder. In those sessions, I am definitely easily filling up the better part of an hour, and sometimes having to rush a little to do so. But, I would agree that in some therapy sessions, there is a lot of excess stuff going on.

 

[cara susanna]

I too have perceived interpersonal feedback on my communication style as "attacks." My postdoctoral supervisors definitely had to help me reframe that back in the day; but, it's something I still struggle with. So, while no attack was meant, I certainly can understand why your visceral response would be as such. Wishing you the best mate.

You aren't in a supervisory relationship with WisNeuro. Although we are all professionals, this isn't even a professional setting. It's exhausting enough having to worry about our tone all day without having to worry about it here, too. I sometimes disagree with how things are said here but this feels inappropriate.

 

[deleted1084759]

You aren't in a supervisory relationship with WisNeuro. Although we are all professionals, this isn't even a professional setting. It's exhausting enough having to worry about our tone all day without having to worry about it here, too. I sometimes disagree with how things are said here but this feels inappropriate.

I definitely hear what you are saying. Intepersonal feedback can come from supervisors, as well as peers and even mentees. That said though, while I'm going to sit on your feedback for awhile to consider it. I do appreciate it. Thanks!

 

[Sanman]

You aren't in a supervisory relationship with WisNeuro. Although we are all professionals, this isn't even a professional setting. It's exhausting enough having to worry about our tone all day without having to worry about it here, too. I sometimes disagree with how things are said here but this feels inappropriate.

This sort of thing happens on here from time to time. It has me wondering whether the issue is the framing of the situation (professional board vs social board vs whatever other view of this place there is) or if other folks just go around correcting their bosses communication style/tone, the guy at the grocery store, etc.

 

[Mama Bear]

or if other folks just go around correcting their bosses communication style/tone, the guy at the grocery store, etc.

You know those kiosk people at the mall who always try to hound you? To be fair, I’ve come close to yelling some feedback at them.

 

[tr]

What was disingenuous about it? You're a faculty presumably who engages in research. What you said flies in the face of scientific inquiry. I will agree with you that the area in question is far from perfect, but the answer is not to dismiss in altogether and instead rely on biased anecdotes. The answer is to engage in a scholarly way and suggest and possibly run more rigorous studies.

I'm with you on that in theory, but in practice there are ethical concerns that militate against the optimal study design. Ideally you would randomize to physician or to NP with no physician backup; but when this was tried, it was effectively unrandomized by the good sense of the clinicians involved, who systematically reassigned the most complex patients from the NP service to the physicians. I'm not sure how you propose to get around this issue.

 

[WisNeuro]

I'm with you on that in theory, but in practice there are ethical concerns that militate against the optimal study design. Ideally you would randomize to physician or to NP with no physician backup; but when this was tried, it was effectively unrandomized by the good sense of the clinicians involved, who systematically reassigned the most complex patients from the NP service to the physicians. I'm not sure how you propose to get around this issue.

Can you send me the actual paper, from the abstract it doesn't clearly delineate procedures used to unrandomize. Additionally, you can still do a comparison study in a myriad of ways. Why not step down and look at outpatient primary care outcomes for starters? To claim that this study wasn't done perfectly means that no study can be done adequately is simply not correct. There are countless ways that meaningful comparisons can be made to examine outcomes in patient care, looking at a many independent variables, provider type just being one such variable. Just because it's hard doesn't mean we should throw our hands up and rely solely on anecdote. If we do that we might as well turn over everything to the NDs.

 

[tr]

Can you send me the actual paper, from the abstract it doesn't clearly delineate procedures used to unrandomize. Additionally, you can still do a comparison study in a myriad of ways. Why not step down and look at outpatient primary care outcomes for starters? To claim that this study wasn't done perfectly means that no study can be done adequately is simply not correct. There are countless ways that meaningful comparisons can be made to examine outcomes in patient care, looking at a many independent variables, provider type just being one such variable. Just because it's hard doesn't mean we should throw our hands up and rely solely on anecdote. If we do that we might as well turn over everything to the NDs.

" doctors wanted the flexibility to pre-empt randomization because of concerns that certain patients may be ‘too sick’ to be managed by NPs and that admitting patients to the NP ward might increase their involvement in ‘offhours’ management. NP requests for cross-overs also reflected concerns about their ability to provide adequate management for some patients, particularly patients requiring frequent ‘off-hours’ monitoring, and concerns that staffing was inadequate to accept new admissions"

 

[clausewitz2]

" doctors wanted the flexibility to pre-empt randomization because of concerns that certain patients may be ‘too sick’ to be managed by NPs and that admitting patients to the NP ward might increase their involvement in ‘offhours’ management. NP requests for cross-overs also reflected concerns about their ability to provide adequate management for some patients, particularly patients requiring frequent ‘off-hours’ monitoring, and concerns that staffing was inadequate to accept new admissions"

Interesting points:

1)NPs were only admitting during bankers' hours. As a result, a number of patients assigned to NP ward were transferred to attending ward at NP request because they would 'arrive too late on the ward for evaluation.'

Anyone who has worked as a medical resident will recognize arriving 'too late for evaluation' this is not a thing for people who are not NPs doing inpatient medicine.

2) 90 of 193 people assigned to NP service were actually admitted to attending service. 45 of these were due to NP request. Exactly one person assigned to attending service was admitted to NP ward.

3) all emergency and after hour care was responsibility of residents on call.

4)This is actually a comparison between care delivered by a team of residents with attending support and a team of NPs with attending support. If NPs were simply perma-residents I actually would have a lot fewer objections. This is however not even vaguely a test of what happens with independent NP practice or especially relevant to that question.

 

[Sanman]

" doctors wanted the flexibility to pre-empt randomization because of concerns that certain patients may be ‘too sick’ to be managed by NPs and that admitting patients to the NP ward might increase their involvement in ‘offhours’ management. NP requests for cross-overs also reflected concerns about their ability to provide adequate management for some patients, particularly patients requiring frequent ‘off-hours’ monitoring, and concerns that staffing was inadequate to accept new admissions"

I have not read the linked article, but the conversation had me curious why can a study comparing the efficacy of NPs vs Physicians cannot be done retroactively with an outpatient case review and rating system. Office visits are billed based on complexity of the patient encounter, review encounters by both providers, rate level of care on a rating scale, and compare a random sample of visits. If physicians tend to deal with more complex patients, there are established methods for treating skewed data.

Now, whether anyone has the motivation to do this and publish it is a different question as such a study could have a lot of unintended consequences.

 

[WisNeuro]

This seems rather poorly designed study for the question they are trying to address, why "randomly assign" patients if they would not actually be randomly assigned based on time of admittance or whim of physicians in the study? When things like "bed unavailability" account for a huge degree of variance as well? As Sanman and myself have stated, much easier, cheaper, and straightforward ways to address certain outcomes.

 

[clausewitz2]

I have not read the linked article, but the conversation had me curious why can a study comparing the efficacy of NPs vs Physicians cannot be done retroactively with an outpatient case review and rating system. Office visits are billed based on complexity of the patient encounter, review encounters by both providers, rate level of care on a rating scale, and compare a random sample of visits. If physicians tend to deal with more complex patients, there are established methods for treating skewed data.

Now, whether anyone has the motivation to do this and publish it is a different question as such a study could have a lot of unintended consequences.

Check out the definition of outpatient e&m codes. The complexity level of the codes is an incredibly poor proxy for actual complexity. I'm not saying it can't be done, I just don't think it is as easy as you think it is or conceptually unproblematic.

Honestly you probably need independent panel rating complexity of cases based on blind record review, and even that might be skewed by differences in what information is collected and documented and different patterns of lab/imaging ordering.

If I order an MRI for someone, it's not going to get coded as a 99213. If I did it because I don't know how to take a headache hx and order an MRI on everyone with a recurrent headache, that doesn't mean it was a complex case.

 

[Sanman]

You know those kiosk people at the mall who always try to hound you? To be fair, I’ve come close to yelling some feedback at them.

What is this mall you speak of? I vaguely recall a place where people congregated with stores and movies that was not Amazon, but I have not been to one of those since Feb 2020.

 

[Mama Bear]

What is this mall you speak of? I vaguely recall a place where people congregated with stores and movies that was not Amazon, but I have not been to one of those since Feb 2020.

Yes, that’s the place. Legend has it people would stand in these places called “food courts” with exposed platters of food they would hand out as samples. Wild times.

 

[Sanman]

Check out the definition of outpatient e&m codes. The complexity level of the codes is an incredibly poor proxy for actual complexity. I'm not saying it can't be done, I just think it is as easy as you think it is or conceptually unproblematic.

Honestly you probably need independent panel rating complexity of cases based on blind record review, and even that might be skewed by differences in what information is collected and documented and different patterns of lab/imaging ordering.

If I order an MRI for someone, it's not going to get coded as a 99213. If I did it because I don't know how to take a headache hx and order an MRI on everyone with a recurrent headache, that doesn't mean it was a complex case.

I am aware that it is a poor proxy for case complexity, but it is a place to start and as you alluded to, cases can be ranked independently.

I am not saying it is easy, but if we can rate parent-child attachment via video tape interactions...

 

[PsyDr]

Yes, that’s the place. Legend has it people would stand in these places called “food courts” with exposed platters of food they would hand out as samples. Wild times.

Ohhhh, you mean Costco.

 

[tr]

This seems rather poorly designed study for the question they are trying to address, why "randomly assign" patients if they would not actually be randomly assigned based on time of admittance or whim of physicians in the study? When things like "bed unavailability" account for a huge degree of variance as well? As Sanman and myself have stated, much easier, cheaper, and straightforward ways to address certain outcomes.

Not really, epidemiological analysis of existing datasets is even worse because the data are hopelessly confounded by nonrandom distribution of patients and by physician backup of NPs, both explicit and undeclared.

The design of the study is good (much better than trying to mine existing health care datasets that were not designed to answer this question), but the conditions of the real world would not permit it to be carried out as intended.

You ask why patients could not be randomly assigned. *It's because the necessity of safeguarding patient welfare would not permit it.* This is why we can't have nice data.

 

[WisNeuro]

Not really, epidemiological analysis of existing datasets is even worse because the data are hopelessly confounded by nonrandom distribution of patients and by physician backup of NPs, both explicit and undeclared.

The design of the study is good (much better than trying to mine existing health care datasets that were not designed to answer this question), but the conditions of the real world would not permit it to be carried out as intended.

You ask why patients could not be randomly assigned. *It's because the necessity of safeguarding patient welfare would not permit it.* This is why we can't have nice data.

But they can easily be in different settings, such as outpatient settings in states where NPs practice independently.

 

[tr]

But they can easily be in different settings, such as outpatient settings in states where NPs practice independently.

As stated above, those places are not offering up their data for analysis. The people/institutions who have the wherewithal and motivation to conduct outcomes research also have well supported and defined NP services. Private practice independent unsupervised NPs aren't rushing to offer up their results for outcomes research.

I'm not saying this research is impossible to do (although there are many barriers). I'm saying that the research that has been done so far is not particularly enlightening. Therefore, between a bunch of poorly designed studies with useless endpoints and my own personal experience, for now I'm going with my experience. You can disdain that as 'anecdote' if you like but the fact is that just because someone managed to publish something doesn't mean the information is actually useful.

 

[WisNeuro]

I'm not saying this research is impossible to do (although there are many barriers). I'm saying that the research that has been done so far is not particularly enlightening. Therefore, between a bunch of poorly designed studies with useless endpoints and my own personal experience, for now I'm going with my experience. You can disdain that as 'anecdote' if you like but the fact is that just because someone managed to publish something doesn't mean the information is actually useful.

I would surmise that many people are holding this body of literature to a much higher standard than many other areas in their clinical work. Anecdotes can be useful, but not as a justification for general hypocrisy.

 

[aim-agm]

No, you should advocate for additional research that addresses limitations of previous research from a methodological point of view. What you should not do is disregard research purely because you do not like its assumptions. I am not a fan of midlevels, but I still have to accept the notion that it is possible that meaningful patient outcomes may be similar. It's how research is supposed to work.

I would surmise that many people are holding this body of literature to a much higher standard than many other areas in their clinical work. Anecdotes can be useful, but not as a justification for general hypocrisy.

You are making a great deal of assumptions about the scientific rigor and intellectual honesty of not just the physicians in general, but also the specific individuals in this discussion. I assume you have a body of literature and evidence to support this characterization 😉

Let's steelman the debate framed in the scientific method:
Null Hypothesis: Quality of clinical treatment significantly correlates with the teaching and training of the clinician. Significant differences in quality of care are therefore expected when comparing physicians to nurse practitioners.
Hypothesis: The correlation of quality to training is asymptotic, and the training of nurse practitioners is sufficient to reach that portion of the curve. While training to the doctoral level would result in better outcomes in theory, in practice the difference is small enough to be overshadowed by other factors and is not significant.

Experiment: Reliable results in favor of the hypothesis could not be obtained, usually due to flawed study design.

Conclusion: The null hypothesis cannot be rejected at this time. Interventions based on the hypothesis are not supported.

If I wanted to take this a step further, it would not be unreasonable to infer the following (as I usually do when assessing Pharma studies for example):
Meta-Conclusion: The inability of highly-motivated and incentivized individuals to reject the null hypothesis despite herculean effort can be inferred as strong support of the null hypothesis.

 

[WisNeuro]

We don't really suggest null hypothesis testing as the primary analysis in any study, far too problematic, especially in the above case. Further, I wouldn't say that the null hypothesis in the proposed case has not been rejected. So far, at least in this thread, there has been one study, which definitely had problematic design, which still found no real differences in their outcome variables. This is a fairly large body of literature, with many studies on one side, but nothing besides criticisms on the other side for the most part. From a scientific point of view, I remain unconvinced that there are any real differences in outcomes.

 

[Sanman]

You are making a great deal of assumptions about the scientific rigor and intellectual honesty of not just the physicians in general, but also the specific individuals in this discussion. I assume you have a body of literature and evidence to support this characterization 😉

Let's steelman the debate framed in the scientific method:
Null Hypothesis: Quality of clinical treatment significantly correlates with the teaching and training of the clinician. Significant differences in quality of care are therefore expected when comparing physicians to nurse practitioners.
Hypothesis: The correlation of quality to training is asymptotic, and the training of nurse practitioners is sufficient to reach that portion of the curve. While training to the doctoral level would result in better outcomes in theory, in practice the difference is small enough to be overshadowed by other factors and is not significant.

Experiment: Reliable results in favor of the hypothesis could not be obtained, usually due to flawed study design.

Conclusion: The null hypothesis cannot be rejected at this time. Interventions based on the hypothesis are not supported.

If I wanted to take this a step further, it would not be unreasonable to infer the following (as I usually do when assessing Pharma studies for example):
Meta-Conclusion: The inability of highly-motivated and incentivized individuals to reject the null hypothesis despite herculean effort can be inferred as strong support of the null hypothesis.

Null hypothesis should be that clinical treatment quality has no correlation to education or training level. Beyond that, name your variables any way you please.

 

[clausewitz2]

Null hypothesis should be that clinical treatment quality has no correlation to education or training level. Beyond that, name your variables any way you please.

Let's be bayesian about this and get real explicit on our priors. You are saying that in the absence of any evidence to the contrary, you think that it is more likely than not that the amount, kind or rigor of training clinicians receive is in fact irrelevant to quality of treatment delivered. Is this actually your position?

Where is the line on this? Can particularly bright high school students practice medicine or perform psychotherapy? Any old random humanities PhD?

Is there direct evidence that the training of people performing the various forms of psychological assessments y'all do has an impact on any significant outcomes for anyone? Sure, having no idea how to do it properly might invalidate the assumptions made by the literature establishing the validity of the measures, but neither the patient, assessor, or test materials actually burst into flame. Can we just have anyone with basic numeracy run testing practices? Can we demonstrate that doing that would make a difference in outcomes that matter to anyone who hasn't taken graduate statistics courses?

I imagine you can immediately think of all kinds of reasons that's not a supportable idea. To which I say, 'yeah, no ****'

 

[WisNeuro]

Let's be bayesian about this and get real explicit on our priors. You are saying that in the absence of any evidence to the contrary, you think that it is more likely than not that the amount, kind or rigor of training clinicians receive is in fact irrelevant to quality of treatment delivered. Is this actually your position?

Where is the line on this? Can particularly bright high school students practice medicine or perform psychotherapy? Any old random humanities PhD?

Is there direct evidence that the training of people performing the various forms of psychological assessments y'all do has an impact on any significant outcomes for anyone? Sure, having no idea how to do it properly might invalidate the assumptions made by the literature establishing the validity of the measures, but neither the patient, assessor, or test materials actually burst into flame. Can we just have anyone with basic numeracy run testing practices? Can we demonstrate that doing that would make a difference in outcomes that matter to anyone who hasn't taken graduate statistics courses?

I imagine you can immediately think of all kinds of reasons that's not a supportable idea. To which I say, 'yeah, no ****'

This is a strawman, as there is a training model that is being compared. At no point has anyone claimed that training is wholly irrelevant. The question being what is necessary and sufficient for a certain level of outcome.

 

[Sanman]

Let's be bayesian about this and get real explicit on our priors. You are saying that in the absence of any evidence to the contrary, you think that it is more likely than not that the amount, kind or rigor of training clinicians receive is in fact irrelevant to quality of treatment delivered. Is this actually your position?

Google the term null hypothesis, basic research methods teaches us that the null hypothesis would suggest no difference between treatment groups (educational levels in this case)

Where is the line on this? Can particularly bright high school students practice medicine or perform psychotherapy? Any old random humanities PhD?

Obviously, the answer to this is any licensed professional allowed to prescribe medications. If you want to have an IRB approve HS students to prescribe meds, good luck.

Is there direct evidence that the training of people performing the various forms of psychological assessments y'all do has an impact on any significant outcomes for anyone? Sure, having no idea how to do it properly might invalidate the assumptions made by the literature establishing the validity of the measures, but neither the patient, assessor, or test materials actually burst into flame. Can we just have anyone with basic numeracy run testing practices? Can we demonstrate that doing that would make a difference in outcomes that matter to anyone who hasn't taken graduate statistics courses?

This is a straw man, the hypothesis you are testing has no impact on how to perform a research study properly. If we looked at psych assessment the null hypothesis would still be no difference between treatment groups. That is always the null hypothesis and the basis of proper scientific study. You never assume between group differences without established scientific literature to back it up.

I imagine you can immediately think of all kinds of reasons that's not a supportable idea. To which I say, 'yeah, no ****'

Nope, again this is about conducting sound research.

 

[clausewitz2]

Google the term null hypothesis, basic research methods teaches us that the null hypothesis would suggest no difference between treatment groups (educational levels in this case)

Yeah, I am not talking about null hypothesis testing because it is a garbage method for reaching interesting conclusions. I fully take and agree with the point that it should not be central to well-designed research.

Obviously, the answer to this is any licensed professional allowed to prescribe medications. If you want to have an IRB approve HS students to prescribe meds, good luck.

This is a straw man, the hypothesis you are testing has no impact on how to perform a research study properly. If we looked at psych assessment the null hypothesis would still be no difference between treatment groups. That is always the null hypothesis and the basis of proper scientific study. You never assume between group differences without established scientific literature to back it up.

I see that you are consistent which is laudable but I think you are mistaking the distinction between 'no formal published well-designed study suggesting' and 'doesn't exist.' this is what I mean when I stress a Bayesian perspective with informative priors. I think it is totally legitimate to weigh different kinds of evidence differently and update priors in a differential way. Fine and dandy, no objections there. But to simply not update then at all as a result of broad categories of information is not a good way to iteratively approximate a good model of reality.

Clinical experience isn't reliable and reproducible in the same way a well-designed study is. Sure, no dispute. The informational value of it in summation is also not zero, and likewise a good updating function will not simply exclude it from consideration.

 

[clausewitz2]

This is a strawman, as there is a training model that is being compared. At no point has anyone claimed that training is wholly irrelevant. The question being what is necessary and sufficient for a certain level of outcome.

I don't know, 'clinical treatment quality has no correlation with training or education level' sounds a lot like 'training is irrelevant to clinical treatment quality.' I mean, I have been accused of being literal-minded occasionally but I feel like equating 'not correlated' with 'irrelevant' is not such a massive leap.

And I actually want to push back on the idea of it being a straw man. I think it's actually a pretty close analogy. We all know there are midlevels therapists who use standardized instruments that they don't entirely understand and draw conclusions from results that would horrify the test creators. No doubt there are some psychologists who do so too. Where are the studies demonstrating harms in functional outcomes from people who have not been adequately trained administering these things? Do we conclude from the lack of such studies that the most likely explanation is that it's all fine and dandy?

You're right the situations aren't literally isomorphic. But I think they are pretty closely parallel, not least in that I think expertise or training matters to some significant extent despite there not being a clear and undisputed RCT 'proving' such in both cases.

 

[Sanman]

I don't know, 'clinical treatment quality has no correlation with training or education level' sounds a lot like 'training is irrelevant to clinical treatment quality.' I mean, I have been accused of being literal-minded occasionally but I feel like equating 'not correlated' with 'irrelevant' is not such a massive leap.

And I actually want to push back on the idea of it being a straw man. I think it's actually a pretty close analogy. We all know there are midlevels therapists who use standardized instruments that they don't entirely understand and draw conclusions from results that would horrify the test creators. No doubt there are some psychologists who do so too. Where are the studies demonstrating harms in functional outcomes from people who have not been adequately trained administering these things? Do we conclude from the lack of such studies that the most likely explanation is that it's all fine and dandy?

You're right the situations aren't literally isomorphic. But I think they are pretty closely parallel, not least in that I think expertise or training matters to some significant extent despite there not being a clear and undisputed RCT 'proving' such in both cases.

The problem here is with the assumption that "clinical treatment quality" is a static defined or that "standardized instrument" is a defined measure.

I would argue a mid-level provider has no deficit in clinical decision making compared to a physician in a case there a person with no prior history of treatment is a being prescribed a low dose of a first line medication for x issue. Similarly, you do not need to be a neuropsychologist to score and interpret an MMSE with relative accuracy. The arguments come into play with more complex cases and much lower levels of training (could a HS student with 10 hrs training accurately interpret an MMSE?). I think you would need to establish where that line is before assuming any probable differences.

 

[tr]

I would argue a mid-level provider has no deficit in clinical decision making compared to a physician in a case there a person with no prior history of treatment is a being prescribed a low dose of a first line medication for x issue.

You mean like giving a painkiller to a healthy young person with no medical history who comes in with a headache?
That was my first case on my first clinical rotation in medical school (3rd year), when I had already had as many years of medical education as a graduating NP.
I was like, uh, tension headache?
Nope!! Meningitis!!!


The first presentation for a given complaint is always the hairiest because the diagnosis could be anything. No prior history is the worst. It's much more reasonable to hand patients to NP for follow-up after initial workup and diagnosis are completed, when the scope of the issue and options for treatment are more defined.

 

[WisNeuro]

I don't know, 'clinical treatment quality has no correlation with training or education level' sounds a lot like 'training is irrelevant to clinical treatment quality.' I mean, I have been accused of being literal-minded occasionally but I feel like equating 'not correlated' with 'irrelevant' is not such a massive leap.

And I actually want to push back on the idea of it being a straw man. I think it's actually a pretty close analogy. We all know there are midlevels therapists who use standardized instruments that they don't entirely understand and draw conclusions from results that would horrify the test creators. No doubt there are some psychologists who do so too. Where are the studies demonstrating harms in functional outcomes from people who have not been adequately trained administering these things? Do we conclude from the lack of such studies that the most likely explanation is that it's all fine and dandy?

You're right the situations aren't literally isomorphic. But I think they are pretty closely parallel, not least in that I think expertise or training matters to some significant extent despite there not being a clear and undisputed RCT 'proving' such in both cases.

This is the fallacy, you are suggesting that people are saying that zero clinical training is equal to physician level training, no one has said that except for you. The question at hand is what level of clinical training is sufficient to provide generally equivalent patient outcomes in certain settings That's where the research comes in handy. I'd rather have well-designed studies than solely rely on someone merely saying "this is the way it needs to be because I said so" when they have a vested monetary interest, on either side of the debate. Until then, it's all just a turf war with money on the line. That's fine, just be honest about it.

 

[clausewitz2]

The problem here is with the assumption that "clinical treatment quality" is a static defined or that "standardized instrument" is a defined measure.

I would argue a mid-level provider has no deficit in clinical decision making compared to a physician in a case there a person with no prior history of treatment is a being prescribed a low dose of a first line medication for x issue. Similarly, you do not need to be a neuropsychologist to score and interpret an MMSE with relative accuracy. The arguments come into play with more complex cases and much lower levels of training (could a HS student with 10 hrs training accurately interpret an MMSE?). I think you would need to establish where that line is before assuming any probable differences.

See that strikes me as perfectly plausible. Basic algorithmic clinical approaches to any well-defined, not hyperacute problem are not actually that complicated regardless of the domain. You could teach a particularly motivated orangutan to do an appendectomy quite honestly. It seems I misunderstood your position; this version I actually more or less agree with.

 

[clausewitz2]

This is the fallacy, you are suggesting that people are saying that zero clinical training is equal to physician level training, no one has said that except for you. The question at hand is what level of clinical training is sufficient to provide generally equivalent patient outcomes in certain settings That's where the research comes in handy. I'd rather have well-designed studies than solely rely on someone merely saying "this is the way it needs to be because I said so" when they have a vested monetary interest, on either side of the debate. Until then, it's all just a turf war with money on the line. That's fine, just be honest about it.

Obviously it is in my financial interest for this to shake out a certain way, no getting away from that. I'll admit that up front. I don't think it is the entirety of my motivation but of course I'd say that, wouldn't I? I wonder if we could have this discussion without resorting to impugning the true motivations of participants.

I agree with you in many respects. Well-designed studies would be great. I think it is going to be way harder than you think it is, mainly because the data sources you have suggested so far have the massive problems that I don't think are correctable that we have already discussed.

At the same time it is an error to equate a)reasoning based on practical considerations, direct experience in the clinical fields in question and the trivially true observation that in the limit experience and training impact clinical decision-making to an important extent with b) the ex cathedra pronouncements of the great and good. The later is of minimal persuasive impact to right-thinking people, as it should be. The former is not and I think if you are unwilling to credit it in any way you are making a mistake.

 

[WisNeuro]

Obviously it is in my financial interest for this to shake out a certain way, no getting away from that. I'll admit that up front. I don't think it is the entirety of my motivation but of course I'd say that, wouldn't I? I wonder if we could have this discussion without resorting to impugning the true motivations of participants.

I agree with you in many respects. Well-designed studies would be great. I think it is going to be way harder than you think it is, mainly because the data sources you have suggested so far have the massive problems that I don't think are correctable that we have already discussed.

At the same time it is an error to equate a)reasoning based on practical considerations, direct experience in the clinical fields in question and the trivially true observation that in the limit experience and training impact clinical decision-making to an important extent with b) the ex cathedra pronouncements of the great and good. The later is of minimal persuasive impact to right-thinking people, as it should be. The former is not and I think if you are unwilling to credit it in any way you are making a mistake.

We could dispense with the impuging, sure, but we can agree that financial interests lie heavily in this no matter how we frame it.

And yes, we can credit direct experience in clinical fields, but I would caution not over crediting it in some ways as we know various biases exist, as well as errors in clinical decision making as opposed to more actuarial based decision making.

I would have to disagree on the "uncorrectable" sentiment. Merely a failure of ingenuity in study design. Aside from that, there are exceedingly easy ways to study meaningful outcomes from a design point of view, though anything large scale would require a fairly large grant.

 

[tr]

This is the fallacy, you are suggesting that people are saying that zero clinical training is equal to physician level training, no one has said that except for you. The question at hand is what level of clinical training is sufficient to provide generally equivalent patient outcomes in certain settings That's where the research comes in handy. I'd rather have well-designed studies than solely rely on someone merely saying "this is the way it needs to be because I said so" when they have a vested monetary interest, on either side of the debate. Until then, it's all just a turf war with money on the line. That's fine, just be honest about it.

The 'in certain settings' is key here. In some settings (ones with well-defined areas of practice where expert backup is available), it's completely reasonable to expect outcomes to be equivalent. In other settings (broad scope of practice, no available backup), it would be shocking if they were. Nobody is studying the latter though; the studies are mostly coming out of the former setting.

I don't personally have any vested monetary interest in this. At this point in my career I'm working exclusively in a specialized subarea and the combination of specialty care, medical education, and research isn't commonly done by NPs (never mind that the academic salary wouldn't be particularly attractive to them).

I do have an interest in cutting down on the number of train-wreck referrals I get who were previously seeing NPs though.

 

[WisNeuro]

The 'in certain settings' is key here. In some settings (ones with well-defined areas of practice where expert backup is available), it's completely reasonable to expect outcomes to be equivalent. In other settings (broad scope of practice, no available backup), it would be shocking if they were. Nobody is studying the latter though; the studies are mostly coming out of the former setting.

I don't personally have any vested monetary interest in this. At this point in my career I'm working exclusively in a specialized subarea and the combination of specialty care, medical education, and research isn't commonly done by NPs (never mind that the academic salary wouldn't be particularly attractive to them).

I do have an interest in cutting down on the number of train-wreck referrals I get who were previously seeing NPs though.

This is where the bias comes in, though. In my most recent position, the absolute worst consistetly bad medication regimens came from 2 of the psychiatrists in our system. I've seen bad in both NPs, psychiatrists, FMs, etc. Proportion wise in my area, no real difference. I realize that is anecdotal though, which is where better research would come in handy. Are there areas where midlevels can practice independently with no real difference in meaningful outcomes? Most likely. Are there areas where outcomes would suffer, also most likely. I'd be interested in finding where these lines are, empirically informed.

 

[tr]

This is where the bias comes in, though. In my most recent position, the absolute worst consistetly bad medication regimens came from 2 of the psychiatrists in our system. I've seen bad in both NPs, psychiatrists, FMs, etc. Proportion wise in my area, no real difference.

Beg pardon, but how can you tell? I couldn't look at a cardiologist's selection of medication for a patient with arrhythmias and decide whether it was 'bad' or 'good.' I'm wondering how you are able to do this for an area that you haven't been trained in?

 

[WisNeuro]

Beg pardon, but how can you tell? I couldn't look at a cardiologist's selection of medication for a patient with arrhythmias and decide whether it was 'bad' or 'good.' I'm wondering how you are able to do this for an area that you haven't been trained in?

But I have been trained in psychopharmacology, as well as in medications that have significant CNS side effects.

 

[AbnormalPsych]

Yeah, I am not talking about null hypothesis testing because it is a garbage method for reaching interesting conclusions. I fully take and agree with the point that it should not be central to well-designed research.

Say what? Go on...


(I teach this stuff)

 

[clausewitz2]

Say what? Go on...


(I teach this stuff)

Andrew Gelman and co summarize this far better than I could so will leave it to them as a representative sample of the statistical discourse on this issue:

https://www.google.com/url?sa=t&sou...FjAAegQIARAB&usg=AOvVaw319YH9rnKl5L687zLLal_U

[/URL]

For those not clicking through, the piece is entitled 'abandon statistical significance'

 

[PsyDr]

Andrew Gelman and co summarize this far better than I could so will leave it to them as a representative sample of the statistical discourse on this issue:

https://www.google.com/url?sa=t&sou...FjAAegQIARAB&usg=AOvVaw319YH9rnKl5L687zLLal_U

[/URL]

For those not clicking through, the piece is entitled 'abandon statistical significance'

Uhhhhhh, that's an interesting take on Gelman's work and that paper. You might want to see his proposition in your chosen paper.

"....we propose that the p-value be demoted from its threshold screening role and instead, treated continuously, be considered along with the neglected factors as just one among many pieces of evidence. "

 

[clausewitz2]

Uhhhhhh, that's an interesting take on Gelman's work and that paper. You might want to see his proposition in your chosen paper.

"....we propose that the p-value be demoted from its threshold screening role and instead, treated continuously, be considered along with the neglected factors as just one among many pieces of evidence. "

A lot of his blog is dedicated to why null hypothesis significance testing should not be seriously pursued anymore, the superiority of bayesian modeling approaches, and the ways in which frequentist mathematics is never going to license the conclusions you would generally like to be able to draw from a given study. So while I am happy to concede he has never used that particular phrasing, it's not actually that far off from theme of many things he has written and published over the last decade or so.

That paper also says: 'it seldom makes sense to calibrate evidence as a function of p-value.' maybe you read this as meaning something other than 'don't draw conclusions based mainly on NHST' but I'll confess I don't have any idea what that interpretation might be

Edit: paragraph I pulled quote from:

"While we agree treating the p-value continuously rather than in a thresholded manner constitutes an improvement, we go further and argue that it seldom makes sense to calibrate evidence
as a function of the p-value. We hold this for at least three reasons. First, and in our view the most important, it seldom makes
sense because the p-value is, in the overwhelming majority of
applications, defned relative to the generally implausible and uninteresting sharp point null hypothesis of zero efect and zero systematic error. Second, because it is a poor measure of the evidence for or against a statistical hypothesis (Edwards, Lindman,
and Savage 1963; Berger and Sellke 1987; Cohen 1994; Hubbard and Lindsay 2008). Third, because it tests the hypothesis that
one or more model parameters equal the tested values—but only given all other model assumptions. These other assumptions—
in particular, zero systematic error—seldom hold (or are at least far from given) in the biomedical and social sciences.
Consequently, “a small p-value only signals that there may be a problem with at least one assumption, without saying which one.
Asymmetrically, a large p-value only means that this particular test did not detect a problem—perhaps because there is none, or
because the test is insensitive to the problems, or because biases and random errors largely canceled each other out” (Greenland
2017). We note similar considerations hold for other purely statistical measures."

 

[PsyDr]

A lot of his blog is dedicated to why null hypothesis significance testing should not be seriously pursued anymore, the superiority of bayesian modeling approaches, and the ways in which frequentist mathematics is never going to license the conclusions you would generally like to be able to draw from a given study. So while I am happy to concede he has never used that particular phrasing, it's not actually that far off from theme of many things he has written and published over the last decade or so.

That paper also says: 'it seldom makes sense to calibrate evidence as a function of p-value.' maybe you read this as meaning something other than 'don't draw conclusions based mainly on NHST' but I'll confess I don't have any idea what that interpretation might be

Edit: paragraph I pulled quote from:

"While we agree treating the p-value continuously rather than in a thresholded manner constitutes an improvement, we go further and argue that it seldom makes sense to calibrate evidence
as a function of the p-value. We hold this for at least three reasons. First, and in our view the most important, it seldom makes
sense because the p-value is, in the overwhelming majority of
applications, defned relative to the generally implausible and uninteresting sharp point null hypothesis of zero efect and zero systematic error. Second, because it is a poor measure of the evidence for or against a statistical hypothesis (Edwards, Lindman,
and Savage 1963; Berger and Sellke 1987; Cohen 1994; Hubbard and Lindsay 2008). Third, because it tests the hypothesis that
one or more model parameters equal the tested values—but only given all other model assumptions. These other assumptions—
in particular, zero systematic error—seldom hold (or are at least far from given) in the biomedical and social sciences.
Consequently, “a small p-value only signals that there may be a problem with at least one assumption, without saying which one.
Asymmetrically, a large p-value only means that this particular test did not detect a problem—perhaps because there is none, or
because the test is insensitive to the problems, or because biases and random errors largely canceled each other out” (Greenland
2017). We note similar considerations hold for other purely statistical measures."

So with all due respect, you’re moving the goalposts. Your argument (to my understanding) went from a rejection of the null hypothesis standard to a rejection of the predominant statistical methods involved in publication of null hypothesis testing to whatever your current argument is. While I do not think this is malicious or intentional, we both know the necessity of sticking to strict definitions when having higher order discussions.

 

[tr]

I think @clausewitz2 is just trying to make the point that posterior probabilities are important, and that the whole concept of structuring a research question around the null hypothesis pretty much constitutes a studied choice to reject their importance. I don't find it unclear.

 

[clausewitz2]

I think @clausewitz2 is just trying to make the point that posterior probabilities are important, and that the whole concept of structuring a research question around the null hypothesis pretty much constitutes a studied choice to reject their importance. I don't find it unclear.

@tr sums it up well. I absolutely do recject null hypothesis testing as a primary means of licensing useful inferences about the world. I don't claim to speak for Andrew Gelman but I would wager a large sum of money he would too as it happens.

 

[VA Hopeful Dr]

Comparison of the quality of patient referrals from physicians, physician assistants, and nurse practitioners - PubMed

The quality of referrals to an academic medical center was higher for physicians than for NPs and PAs regarding the clarity of the referral question, understanding of pathophysiology, and adequate prereferral evaluation and documentation.

pubmed.ncbi.nlm.nih.gov

Biopsy Use in Skin Cancer Diagnosis

Histopathologic evaluation is the criterion standard for diagnosis of skin cancer. Underuse of biopsies may promote misdiagnosis, and overuse will increase cost and morbidity. There is no benchmark with which to quantitatively compare health care professionals’ diagnostic accuracy and biopsy use....

jamanetwork.com

Opioid Prescribing by Primary Care Providers: a Cross-Sectional Analysis of Nurse Practitioner, Physician Assistant, and Physician Prescribing Patterns - Journal of General Internal Medicine

Background Prescription opioid overprescribing is a focal point for legislators, but little is known about opioid prescribing patterns of primary care nurse practitioners (NPs) and physician assistants (PAs). Objective To identify prescription opioid overprescribers by comparing prescribing...

link.springer.com

Factors influencing unexpected disposition after orthopedic ambulatory surgery - PubMed

The decreased risk for unexpected disposition associated with more recent data and with freestanding versus hospital-based facilities may represent improvements in efficiency, while the decreased odds for such disposition status associated with the use of other than general or regional...

pubmed.ncbi.nlm.nih.gov

Outpatient Antibiotic Prescribing Among United States Nurse Practitioners and Physician Assistants - PMC

We examined US nurse practitioner (NP) and physician assistant (PA) outpatient antibiotic prescribing. Antibiotics were more frequently prescribed during visits involving NP/PA visits compared with physician-only visits, including overall visits ...

www.ncbi.nlm.nih.gov

That a decent start?

I think @Matthew9Thirtyfive may have more.

 

[tr]

Thank you for dropping those @VA Hopeful Dr . Those are smart outcome choices.

 

[Matthew9Thirtyfive]

Comparison of the quality of patient referrals from physicians, physician assistants, and nurse practitioners - PubMed

The quality of referrals to an academic medical center was higher for physicians than for NPs and PAs regarding the clarity of the referral question, understanding of pathophysiology, and adequate prereferral evaluation and documentation.

pubmed.ncbi.nlm.nih.gov

Biopsy Use in Skin Cancer Diagnosis

Histopathologic evaluation is the criterion standard for diagnosis of skin cancer. Underuse of biopsies may promote misdiagnosis, and overuse will increase cost and morbidity. There is no benchmark with which to quantitatively compare health care professionals’ diagnostic accuracy and biopsy use....

jamanetwork.com

Opioid Prescribing by Primary Care Providers: a Cross-Sectional Analysis of Nurse Practitioner, Physician Assistant, and Physician Prescribing Patterns - Journal of General Internal Medicine

Background Prescription opioid overprescribing is a focal point for legislators, but little is known about opioid prescribing patterns of primary care nurse practitioners (NPs) and physician assistants (PAs). Objective To identify prescription opioid overprescribers by comparing prescribing...

link.springer.com

Factors influencing unexpected disposition after orthopedic ambulatory surgery - PubMed

The decreased risk for unexpected disposition associated with more recent data and with freestanding versus hospital-based facilities may represent improvements in efficiency, while the decreased odds for such disposition status associated with the use of other than general or regional...

pubmed.ncbi.nlm.nih.gov

Outpatient Antibiotic Prescribing Among United States Nurse Practitioners and Physician Assistants - PMC

We examined US nurse practitioner (NP) and physician assistant (PA) outpatient antibiotic prescribing. Antibiotics were more frequently prescribed during visits involving NP/PA visits compared with physician-only visits, including overall visits ...

www.ncbi.nlm.nih.gov

That a decent start?

I think @Matthew9Thirtyfive may have more.

Yes, I dropped about 15 in another thread. There are many of them.

Here’s a link to a bunch:

 

[PsyDr]

I think @clausewitz2 is just trying to make the point that posterior probabilities are important, and that the whole concept of structuring a research question around the null hypothesis pretty much constitutes a studied choice to reject their importance. I don't find it unclear.

That is still conflating an argument about:

1) the prevailing statistical methods used in null hypothesis testing
WITH
2) the general concept of the null hypothesis.

Let me try to put forth an analogy. To my understanding, the initial vitamin D literature was skewed because of some mathematical errors. Then it was corrected. That's a statistical error. If I said that vitamin D is basically a non-issue, and cited that error, you'd probably express frustration that I am conflating the importance of a concept with an apt statistical methodological problem. That's what I'm trying to point out.

@tr sums it up well. I absolutely do recject null hypothesis testing as a primary means of licensing useful inferences about the world. I don't claim to speak for Andrew Gelman but I would wager a large sum of money he would too as it happens.

Okay, follow up:

1) Define which theoretical setting the research question falls within. Is this set theory, model theory, etc? This changes everything. Closed set? Then Friedman would agree with you, but you'd have a lot of other methodological problems.
2) Is the actual argument that the CONCEPT of the null hypothesis is wrong? If so, then why was Gelman cited?
3) Is the actual argument that the prevailing statistical methods in null hypothesis significance testing is wrong? If so, then reconcile the stated opinion about the work cited with Gelman's statements that the prevailing methods should be included, but not retain the same importance.


4) Or is this all just a debate style? I'm fine if it is.