So, now we'll tackle the second chunk of knowledge. What's in the scientific literature? What we care about the literature is that the fact that it's authoritative. So, talking to people in your institution is great because that's where you are. But if you want to make a claim that people should follow but inside this system, you want to have an appeal to more than local authority. So, the literature is that wider authority. Over the last 30 years, the area of evidence-based medicine has emerged. There's this pyramid of authority that's based not on and who says it but how the evidence was generated. So, at the bottom, we have expert opinion. So, the whole last session we had on eliciting human expertise would be called the bottom most level. Having said that, there are practice guidelines, and there are imperatives, and there are policies. They may be based on evidence of this high level evidence, but the guideline itself generally is bottom-line expert opinion. So, that's the kind of the bottom of this little stack if you will. Next up is evidence you get for patients, thank goodness. They could be a case series. Here's my experience or here's our experience with patients like this. Or I can do a case-control study where I say " Okay, patients who developed insomnia and I compare patients who did not develop insomnia. " I look backwards and see how the patients with insomnia defer or deferred from the patients who did not develop insomnia. The next level up is called cohort studies where I do the opposite. I look at people at the beginning of the course and follow them through, then I see what is the life history of somebody with insomnia? And what is the life history of somebody without insomnia? Now I can say, maybe either what's dangerous about insomnia, how bad is it, what to expect from insomnia, and maybe even now you want to know how to treat it, don't you? So, now when we get to how to treat, then we get into randomized control trials, and there an intervention is randomly assigned, and then you compare the cohorts who were assigned drug A or drug B, let's say treatment A or treatment B, or the case of institutions decision support A and decision support B. Then you follow them into their outcomes and see what's happened. Now you know the joke, what's worse than finding a worm in your apple and the answer is finding half a worm in your apple because that means you've eaten the other half. So, the flip side is what's better than one randomized controlled trial? Well, two randomized control trials. They can be synthesized either informally as critical appraisals or more formally into a systematic review which somebody has looked for all the evidence in the literature. So, the process of using literature follows a process. So, I'm going to teach you the knowledge how to use evidence-based medicine. So, step one is define your question. So, let's say how should I manage type two diabetes after diet has failed? Well, you know when I say you should, you are thinking about the you should the system two type of decision-making. When in evidence-based medicine they have a simpler framework called picot. Who's the patient? Which is like the H of you should, the intervention. What do I want to evaluate? This how should I treat? How should I manage or what am I doing? Then you want to compare the intervention that you want to learn about to something that you already know about. So, this is basically the list of actions in the you should model. Always the outcomes just like you should, and T is the timescale again just like you should. Well, the uncertainties are, what you're going to get from the literature? The trade-offs are actually really mentioned explicitly. The perspective is generally the patient's perspective. So, once I define the question, now I have to figure out would I find an answer? No, if I'm asking a medical question I should go to the medical literature. So, that's PubMed. But there are other literatures to look at. There are databases of the nursing literature, database of European literature, one on the South American literature, one on psychology, and psychiatry. So, depending on what your question is, there might be better places even than PubMed to go look. But PubMed is pretty good. Now if you want to be even more thorough than just looking for articles using what I'm about to show you, if you have a good article and you want to see what happened since then, that's called cited reference search, and the freely available one is Google Scholar. There are other proprietary ones such as Scopus. But librarians like the proprietary ones because the searches there are reproducible. If I run Google Scholar search today and then again in a week, I'm going to get a different list. Whereas, if I use a proprietary method, I could say next week, I can say, "Give me the search I would have gotten a week ago." So, somebody else wants to recreate your search they can do it, and reproducibility is important. The next step is to actually do the search and that's translating the question you have into a query. That notion of the difference between a question and a query will be an organizing principle in the fourth course that we have in our specialization. Once you get you search back here, you want to look all the title of abstracts and figure out are you happy with what you got? Do you have enough to answer the question? If you are, you're going to read the articles with what's called critical appraisal and finally, you're going to answer your question. So, here's a graphic for all these. You can see in the upper right, defining your question. How shall I manage type two diabetes. The dark arrows says okay chosen, next choose a source like PubMed. The next step is to actually type in the query. A hint there that on the left is something called clinical queries in PubMed, that will actually help you find the best type of evidence for your question using that evidence hierarchy that I showed you a few slides back. So, you get the results, you evaluate them. If you're happy you go on to critical appraisal using tools like the University of Alberta toolbox. Then you get your answer. If you're not happy with your results, you have to think okay where did I make a mistake? Generally, the first thing you'll do is change your query. Maybe I can get more articles by either broadening the years, or using some different terms, or ordering it with other concepts. Or you might say, you know, wait a minute I'm looking in the wrong place. Am asking an informatics question I actually go to the IT literature and not necessarily the medical literature. So, you might want to go to inspect, let's say your I triple E or your ACM and not PubMed. Or you may say "You know what, I'm asking the wrong question." So, let's say, you're asking about asthma, and you're focused onto saturation and you realize that that's not what patients care about, what really care about is whether they sleep or not? You're going to go back and rethink your strategy. I mentioned critical appraisal. This is just showing you some tools that are available back into 1990s, the journal JAMA came up with about 30 articles over the course of several years about how to read different types of articles. Just because it's from the 1990s as being case series are any different than they were then. So, a lot of what they write then it still applies today. On the right-hand side is another set of tools for reading articles, and they give you very nice worksheets to say, look if you don't want to go all that reading, here's a one pager, here's a checklist of things to look for go and look for them, and then you can figure out whether or not you really want to believe your article or not. So, EBM is great for having articulated, what the most vetted information is, making it available to you. It's really the basis of all the clinical practice guidelines, whether the guidelines that are publicized or the ones are at the heart of a lot of decision support, or part of your practice regimes, your clinical pathways, and search at the hospitals. Pros are that they are authoritative, which I think I clarified. The cons though are a little subtle. When I read the article, I have to think about, do the patients in that study relate to these patients that I practice with? Are those patients my patients? And if they are terrific? If they're not, how do I adjust for that difference? We don't give me a lot of tools for that. A more subtle and related to that is their environment my environment? So, if that comes out of a tertiary care hospital and I'm in the community, they might have resources that I just don't have. So, is an example in the early 2000 I was involved. The American Academy of Pediatrics on diabetes guidelines for children. In a whole number of committees that these guidelines have to go through and they were at the last committee, the guideline was recommending that to manage children with diabetes, the community physician should have a nurse whose sole job is to call up daily to patients with diabetes and see how they're doing. The committee said that's nuts because the vast majority of community physicians just don't have that resource. We can't have a guideline to say that you must have a nurse. We can say it would be a good idea to have a nurse but we can't insist on it. They sent it back to the committee for them to figure out how to reformulate that. This subtle solve the issues that evidence is aimed at the average person. So, I'll tell you on average let's say, this medication will reduce your mortality by two percentage points. That's terrific, and two percentage points is a big deal. But some people might be one percentage, some people might be 10 percentages. It'd be nice to know which one you are. It could also be that on average let's say, there's a certain effect, but your patients they're not thrilled with that effect, it's not big enough for them to change for them to accept the treatment or for the physician to change their practice. So, there's wiggle room that evidence-based medicine by itself may not address. The last point is a very sad one, that the vast majority of things that you care about in practice simply are not studied. It costs money. Very often, the pharmaceutical companies have the best funds available for questions of practice. But they're not interested to help you figure out whether or not their drug is better than the old drug, because they want some to get in the market, they want to show that their drug is better than placebo, which is what the FDA demands for them to demonstrate. So, figuring out whether something is better than what you currently do is tough to study. There are a number of approaches to dealing with it, but they're really just starting.
