Name:
Clinical Prediction Rules: Interview With Dr Thomas G. McGinn
Description:
Clinical Prediction Rules: Interview With Dr Thomas G. McGinn
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/ca5163b1-8980-4cea-87a0-8ff570f7b162/thumbnails/ca5163b1-8980-4cea-87a0-8ff570f7b162.jpg?sv=2019-02-02&sr=c&sig=02ljND%2Fmy5E3vZdxP%2FscXleXuOh1hAn%2FzjQXtcu20PY%3D&st=2024-12-22T06%3A20%3A40Z&se=2024-12-22T10%3A25%3A40Z&sp=r
Duration:
T00H06M55S
Embed URL:
https://stream.cadmore.media/player/ca5163b1-8980-4cea-87a0-8ff570f7b162
Content URL:
https://cadmoreoriginalmedia.blob.core.windows.net/ca5163b1-8980-4cea-87a0-8ff570f7b162/11164727.mp3?sv=2019-02-02&sr=c&sig=%2BREJIusQskrx2RaEEWMsIyyW0n6TkMoo1AWnmHeOHV0%3D&st=2024-12-22T06%3A20%3A40Z&se=2024-12-22T08%3A25%3A40Z&sp=r
Upload Date:
2022-02-28T00:00:00.0000000
Transcript:
Language: EN.
Segment:0 .
>> This is Ed Livingston, Deputy Editor of Clinical Reviews and Education at JAMA. Today we are going to discuss clinical prediction rules, a topic covered in the Users' Guide to the Medical Literature. I'm speaking with the chapter's author Thomas McGinn who is Chairman of Medicine at the North Shore Long Island Jewish Medical Center. Dr. McGinn is a health services researcher who has expertise in the development of clinical prediction rules. Clinical prediction rules enable physicians to avoid unnecessary testing by incorporating various findings in a clinical examination to provide the likelihood that a particular disease entity exists.
An example of a well-known clinical prediction rule is the Ottawa Ankle Index. This index is used to determine the probability of having an ankle fracture when a patient presents with ankle trauma. Use of the clinical prediction rule can avoid the use of radiology when examining patients with this very common clinical problem. Dr. McGinn, what is your definition of a clinical prediction rule? >> Well, broadly defined, it's a decision aid that can be used at the point of care to help physicians make quick decisions, what we like to call frontline decisions, sometimes a prognostic or treatment decision or maybe a diagnostic test decision.
It usually incorporates elements of the history, the physical exam, some labs that are readily available that can be put together in a predictive model that can give a probabilistic outcome. Usually high, medium, low probability of some outcome. >> These rules are developed by three major steps Determination, validation, and impact. Could you tell us about those steps? >> Usually, the rule is derived off of a single data set or group of patients.
And then once the rule is created you have to then take that rule and validate it on another population. And from there usually it's validated on two or three different groups of patients. The third step, which is actually not as frequently performed, you'd like to know when you apply that rule what is the impact it has on the community of patients. So it's a derivation, it's the creation of the rule. The validation is usually performed in a few different areas.
And then the impact is done to determine what the effect of applying that rule in practice would be. It might be helpful to give a brief example. So the Ottawa Ankle Rule which was used to determine whether somebody with an ankle sprain needs to get an x-ray was derived at one site. It was then validated at three or four sites. And then an impact trial was performed to see if it reduced unnecessary x-rays being ordered. >> Clinicians receive quite a lot of training and have a pretty good sense for what they need to do for their patients when they present with certain problems.
When should they rely on clinical prediction rules? >> First and foremost, clinicians should rely on a clinical prediction rule that's well validated. So part of the trick of a prediction rule is that the derivation and validations may be published separately. And we would like to know that a rule has been validated several times so that we know it's definitely accurate and something we can use. And we find that physicians tend to like prediction rules in areas where they feel they don't want to order a test or they feel the test is overordered, but they don't have the probability to help them think that through.
So on the example of an ankle sprain, the clinician may think to themselves, "Look. I don't think this is a fracture. But how do I actually prove that?" And the Ottawa Ankle Rule says when you apply the rule, which is is there point tenderness at the base of the foot, and can the patient walk four steps, if there's no point tenderness and the patient can walk four steps, there's extremely low probability that this person has a fracture. Therefore, you don't need to get the x-ray. So it gives the clinician the confidence to do what they kind of really want to do.
And that's where we find prediction rules come in handy. The other example would be the strep pharyngitis rule which has four or five points depending on the scoring system. If the patient has a very low score, you would not prescribe antibiotics in that case. So it helps the physician making these kind of split second frontline decisions at the time of care. >> Dr. McGinn, there are several well-known clinical prediction rules that are used in clinical practice. These include the Wells' Criteria for both pulmonary embolism and venous thrombosis, the CHADS2 score to determine the probability of requiring anticoagulation for patients with atrial fibrillation, the Pneumonia Severity Index which determines the need to admit patients to the hospital who have pneumonia, and the Heckerling Model that helps determine if antibiotics should be used when treating pneumonia.
Based on this list, it's apparent that clinical prediction rules can be quite useful in clinical practice, but where can physicians go to find a reliable listing of clinical prediction rules? >> There are many sites you can go to that will have the prediction rules listed and even some calculators attached to them, but what you're really looking for is the level of evidence grading system for each one of those. And there's no real good consistent webpage that does that right now. >> What grading system do you recommend?
>> Well the grading system that we developed and is published in the chapter, and it basically scores level one being a highest, level four being the lowest. A prediction rule that was simply derived on say a retrospective database, it would be a level four prediction rule which means it should not be used in clinical practice. A prediction tool that has been validated prospectively in several sites and shown to be accurate would be a level two and is ready to be used broadly in clinical practice. A prediction rule that's been validated and shown to have an impact analysis performed would be a level one.
So the Ottawa Ankle Rule and the Wells' Criteria reach a level one validity criteria and can be used broadly with a good knowledge of what the impact of applying that rule can be. >> Clinical prediction rules can be helpful for clinicians to increase their confidence in a diagnosis and minimize the use of expensive testing. Unfortunately, there is no good single source for finding these clinical prediction rules. Before using them, physicians should ensure that they have reviewed the literature of a clinical prediction rule's derivation, validation, and impact.
A scoring system to assess the utility of a clinical prediction rule, as well as a more detailed explanation for how to use clinical prediction rules, is found in the Users' Guide to the Medical Literature. This is Ed Livingston, Deputy Editor of JAMA. Thank you for listening.
