>> This is Adam Wilcox and today I'm talking about Issues of Clinical Decision Support as an introduction to the next section that we'll talk in much more detail about that.  We're now at this current challenge in healthcare and how can informatics really affect it, but now I see the current challenge in two ways.  I see our challenge is that we have fractionated care and we have tyranny of the urgent.  In fractionated care, what I mean is that there are many patients who are being treated by many clinicians and this is a more recent phenomenon due in large part to the increased specialization that we've had in medicine.  Decades ago, you had your one doctor and your one doctor managed pretty much all of the healthcare needs that you had for the most part.  And there could be some specific specialization that's needed but even then your one physician would be the one to refer to those other clinicians and do a lot of the coordination.  Now, so these have shown that patients over 65 see on average 8 physicians in a year and it's not clear that those--that there's a lot of coordination among these physicians and at each point as the patient moves to a different physician there's an opportunity for information to be communicated at that transition and--but it's a challenge and so because of this fractionated care we need to improve the communication we have that transitions and we need to improve the identification of who's going to coordinate that care.  There have been some efforts in medical models such as the patient-centered medical home around care coordination and other research around care coordination that have helped in these areas but it still is a big need.  A second issue is the--what I'd call the tyranny of the urgent.  Our healthcare system has really been developed around an acute--around acute problems.  You go to the doctor because you have a healthcare problem.  You're usually not going to the doctor because you're trying to prevent those acute care problems and a lot of times it's just not understood and our system for efficiency is most designed around, you know, solving problems as they are identified.  Because of an aging population that is living longer in part due to improvements in healthcare, we're moving from this acute to a chronic population and the result is that we have different needs from our healthcare system than we've had before, whereas before our greatest needs were around managing these acute care systems and, you know, we built up whole systems, emergency departments, ICUs, other specialized groups to address the acute problems.  We don't have as good systems around, how do we just manage the chronic issue for the patient especially when they are not in the doctor's office.  So these are kind of two of the biggest challenges and two of the--and the needs that go along with them, the fractionated care and the tyranny of the urgent.  And in informatics there is an opportunity for solutions that can improve--that can affect these two challenges.  Now, I recognized that many of you have already heard much about meaningful use.  In fact, probably all of you had been a major part of this class understanding meaningful use and how it relates and it's a very significant initiative in terms of healthcare informatics and health information technology.  So, I'm not going to go over this information as you know it, but I would like to focus--but I would like to focus on what the deeper components of meaningful use are and this from a presentation that I heard David Blumenthal give where he said--where he is describing the overall goals of the meaningful use initiative.  First, we have the increased adoption of electronic health records then the coordination of information across the different health records in terms of health information exchange.  But the real goal is this third item, the clinical decision support.  You know, just having electronic health records doesn't solve all of our problems, that doesn't really facilitate the transformation in healthcare and just having all of the information that can be helpful but that doesn't lead to the transformation that is necessary as part of healthcare reform.  What the real opportunity is, is how do you take that information that's gathered from the different sources and the information that's within the EHR and use that electronics' platform to support decision making, to improve the actual decision making at the point of care that it can be safer and that it can be more coordinated and more patient centered and that's the opportunity.  So, going through this next unit where the focus is on clinical decision support, recognize that the other issues around electronic health records implementation and health information exchange are really there just to support the real transformation that can occur when you have clinical decision support systems.  Now, clinical decision support systems--it's interesting because we want them to make a difference and--but we don't always know that they have made a difference.  Hunt in 1998 did a study of a lot of different projects that have been using clinical decision support and evaluated these projects that were in the published literature whether or not they actually made an improvement, whether or not they improved practice and the results were that most of the time they did.  But 34 percent of the time--one-third of the time they actually didn't make an improvement.  So, you know, it's better than nothing.  It's better than 50 percent but still a lot of times clinical decision support systems are pretty expensive to implement and having that 34 percent failure in terms of improving care that's pretty significant.  Later, Kawamoto and all--and others did a study around these same--they did a deeper study around the same studies that Hunt was evaluating, and they found that there were four factors and I'm listing them here in decreasing order of importance, the four factors that were critical to success at those systems.  Number one there was an automatic provision of decision support as part of the clinician workflow.  So workflow became a very important concept of making a good decision support so it was actually effective.  Decision support at the proper time and location of decision making was just a specific component of workflow was found to be important on its own.  Recommendations rather than just assessments as the decision support output was also seen as important.  And fourth, it was helpful if it was actually computer-based, if the computer was being used.  But that was of these four the least most important.  So, a sticky note solution where someone would manually put stickies in the chart, sticky notes or Post-Its in the chart around different concepts would--could--if it fit within the workflow and could be addressed at the time and location of decision making and have recommendations rather than assessments, could be more effective than a computer-based decision support system that didn't do those other three items.  So, then they studied whether or not these certain components were important for improvement in practice.  So all systems as we have seen there was a 66 percent success rate and a 34 percent failure rate, but when we include systems--computer-based systems with workflow and timing and recommendations suddenly the success rate rises up to 94 percent and only a 6 percent failure rate.  That was pretty important in showing what really was the impact of understanding these issues.  So when we think about clinical decision support, we need to understand that these components are very important.  Now, in terms of how to integrate with the workflow and provide appropriate--the timing of the information and how it should be delivered as recommendations, I wanna focus on kind of five modes of decision support and these five modes of decision support should be considered like a toolbox that you--as you are working in this field should consider as options.  You know, maybe the first one is a hammer then a saw then a level or a screw driver and a wrench, you know, different tools would be useful for different tasks.  And I wanna go through these examples because my main goal is for you to see that at each point there is a different way to deliver decision support that--and the different way to deliver it could be more effective depending on what the situation actually is.  So, the first one is task correction meaning that a clinician is engaged in a specific task and there is something about that task that can be improved.  A good example is drug-drug interaction alerts where you're actively correcting the tasks they're currently involved in.
^M00:10:03
>> Something that can be effective in this type of scenario is a pop-up where you interrupt the workflow of the clinician because their workflow is around the task that needs to be corrected.  Now, just because pop-ups work here it doesn't mean that they would work well in other situations.  Let me give you an example.  It can be very helpful to have a pop-up that corrects your prescribed--your medication prescribing process while you're engaged in that medication prescribing.  It sometimes is not as helpful if you get a pop-up alert about a patient who is now being seen in the emergency department while you're treating another patient and writing the prescription.  And unfortunately there have been a lot of issues with the clinical decision support systems where they used pop-ups may less effectively and led to alert fatigue and other challenges.  So, pop-up should be seen as one of the tools.  It's not the only way to deliver clinical decision support.  It's important that I address that specifically because of the concept of alert fatigue and a lot of failures in the application of task correction maybe to the wrong place.  But for actually affecting an existing task, pop-up alerts can be very helpful.  Contrast that with a second type of decision support which we've--the checklist reminders.  So, I'm showing here what was developed--what we developed in the demand of healthcare it was called the patient worksheet.  This provided a summary.  It was actually an electronically generated piece of paper.  We printed it out and the clinicians would use it during their care in the ambulatory setting or as their meeting with the patient and came in a pretty good concise view of information that they needed to review, but what was important about that wasn't just that we had some results specific to the patient problems but we also had reminders, by the way, this patient who has diabetes they have not had a hemoglobin A1c measure in the past six months.  They're overdue for that and this is reminding the clinician while you're reviewing what other--whatever the patient has come in for that you can consider that they also may have some condition, you know, there may be some opportunity for influence.  Now, this is effectively done not as a pop-up but rather as a checklist or a list of things that need to be considered as you're reviewing the patient.
^M00:12:52
[ Pause ]
^M00:13:00
>> Another type of decision support tool is one that facilitates prioritization among the existing patients.  So, here we have a tracking system within an emergency department that was felt also within around healthcare at a specific hospital where we have a very dense information display listing individual patients and the information about those individual patients in a way that one could look at it and see where the next action should be performed.  So, the decision that you're influencing at this point is kind of which patient I should see next and that's really a prioritization task.  If you wanted to do prioritization at the patient level that's often done with a summary view with the checklist, you know, and checklists so that you can see all of the information that needs to be done to get to be prioritized among those.  Across multiple patients' prioritization is often facilitated by a dense information display showing those multiple patients and so it looks a lot like this tracking board.  There is a lot of value to this type of display.  Number one is that the clinicians can assimilate a lot of information very quickly.  Second, the value of this is that the clinicians by reviewing this can really--they can choose or they can work it out amongst their team what area they need to focus on.  You know, if it's a standard display across multiple roles then the multiple roles can also see what needs to be done and you don't have to alert one person and say, you know, someone needs to take care of this if it's not their job you can have the multiple clinicians at multiple disciplines looking at this information and each can respond appropriately and it does facilitate interdisciplinary care and working together.  Contrast that with how some prioritization mechanisms abuse just alerting systems or--I mean not alerting but pop-ups where they had a system that had to be delivered to someone and so the consideration was well, you know, ultimately the clinician is the most responsible for this and so much delivered the pop-up alert in an interrupted form to the physician.  That has been shown to also generate a lot of alert fatigues.  So one of the benefits of this dense information displays like this tracking board is that you can kind of work it around--work it out among a team, who's going to take care of what.  According to not just the individual team dynamics but maybe even adapted to the situation among the team.  This is another multi-patient view but it's different than the emergency department tracking board.  This is actually designed for a care manager who is coordinating care for patients and writing follow-ups.  So, the care manager who's working with the patients with chronic conditions and if this patients have not been seen and not been monitored for their chronic conditions [inaudible] she would contact them, follow up with them such as, you know, you haven't been into get a test for your hemoglobin A1c if they had diabetes or I just want to follow up on your depression, how are your medications going, how was that working, do we need to bring you in to adjust the medications or something like that.  So, that was kind of the clinical role of this care manager.  What was effective in facilitating the care manager was a list of the things that need to be followed up.  So, different than a tracking board which shows a panel of patients all together they can be prioritized among them.  This was an encountered tickler list that showed who needed to be followed up for this period of time.  And so this was a very effective tool.  The care managers, they would print this off and then, you know, we kind of--we had the care manager fax us an example of the tickler that she was using and then kind of show how that was used.  So, I like the little scribbled notes down below 'cause it indicates how well this form was being used.  This is the concept of follow-up reminders.  It's done in one specific way but there can be other types of follow-up reminders especially for patients with chronic conditions where they need to be managed outside of the care system or outside of their--the care encounter.  So, you know, you're--for the first time or not for the first time but in this instance you're gathering up information and saying let me tell you who needs to be coordinated with and here you're not considering in order to bring them in.  And that's an effective form of decision support that fits within the workflow of certain clinical roles.  And then another form of decision support is an update when patients actually change status.  One of the simplest ways that they change status is when we have results available.  So, again this is around chronic care management or ambulatory care management it can be also used within the hospital and we did have some status updates in the emergency department tracking board that I showed you earlier where--when certain results were back and laboratory results or radiology results were back there was an indication that this patient has now changed status rather than waiting for results their status is now results are available.  In this case we use a notification component or like an inbox component of the EHR for the outpatient system when they had lab results available.  So, the daughter would now know that the lab results, they asked the patient to get and send the patient off to have obtained were completed maybe a day or 2 later and now there is no indication saying oh these are completed and now the doctor can act on that and follow up appropriately.  This is different than the follow-up reminders that I showed for the care manager, because a lot of this follow-up reminders are more timed based on what is an appropriate amount of time for a dressing and ensuring compliance or appropriate follow up for a patient whereas this one the note--it's really a notification.  There is something about the patient's status that has been changed and so it's driven by that information rather than just time.
^M00:20:01
>> So, these are five modes of clinical decision support.  The task correction, the prioritization, the checklist reminders, the status updates, and the follow-up reminders that can be effective in fitting within the clinical workflow, providing decision support at the time and location of decision making, providing recommendations rather than just assessments and being computer based that can increase the likelihood of success.  And you can see from these that they're all acting somewhat differently and they can all--they can all effectively use actually improved care.  These last two slides I wanna talk about an overall concept of clinical decision support as it relates to quality improvement and the evolution of the quality improvement process.  Since much of the variation in healthcare that we're seeing is around quality and a related concept of efficiency.  It's important that we understand how quality is really delivered and--or how quality--the concept of improving quality really interacts with the healthcare system.  First of all, it's important to understand that quality improvement is a process and it's an evolutionary process.  Devin [phonetic] discussed this early with his model of plan and do study act understanding that there is different stages of quality improvement.  There is a planning stage, there is a doing, there is studying what you did and then extending the action more broadly.  In Six Sigma, which is a quality improvement mechanism which is quite much more broadly than just a healthcare environment, they list these stages in terms of defining measure, analyze, improve, and control.  And this is based on some research that I did with John Chelico where we studied the interaction of the--of clinical decision support systems across this evolutionary continuum of quality improvement.  So, first you start with patients--you start with a clinician who's--or a researcher who is trying to figure out really what's going on.  They're trying to define the problem and that really is the more inception level of quality improvement.  In the task, there is, you know, just answering questions.  You need to be able to support ad hoc queries often done with the clinical data warehouse rather than a pop-up alert for example.  And then once that's kind of corrected or understood well, once you've really defined the problem and addressed that you move on to a different stage where the users often will ask.  Okay.  Now, I see what are the challenges, can you give me this report or can you run this query for me every week?  And at that point, what you're trying to do is something different than just defining the problem.  You're actually trying to observe the trends and you've moved from defining and actually measuring where you're measuring the effect of different things that you might do by how they influence the trends.  So, at this point rather than just having supporting an ad hoc query, you kind of wanna create a report that delivers multiple queries over time so you can see those trends.  Eventually, just seeing those trends may not be--may not solve the full problem and you may need to dig deeper about why there's plateaus or why there is variation within those trends and at this point users are often looking to dig a little more deeper into the data online analytical processing tools or other analytic tools can be used to discover this.  All of these are kind of precursors to actually delivering clinical decision support at the point of care which happens during the improved stage.  And in terms of this evolution process through the Six Sigma quality improvement stages, at this point the tool really is not--it's no longer an ad hoc query, it's no longer some automated report or online analytical processing tool.  It's some information display where the information can be assimilated and this is where dashboards or prioritization screenings or, you know, patient summaries can be very useful.  At the fifth stage, once the workflow has been defined and once it is clear what actions are taken at the point of care, you can make it easier to do those actions by incorporating the support for those actions within the application within the EHR and that's more the classic decision support which is done--which has been--is done by alerting or some notification.  One of the challenges though is that most systems or most applications of clinical decision support skipped that improved stage and they--once the problem is identified they jump right to control where they're trying to implement it within the EHR.  This makes sense because a lot of time the EHR is really the only tool that you have and the EHR is defined by a specific user and so someone has to be notified because that's, you know, the most standard tool that we had for delivery of information is some type of notification often done through a pop-up.  But because, you know, a pop-up is really, you know, affecting the workflow in a very precise way.  And so that's really the control stage and until you really know who should be receiving the pop-up and how they should be acting it's different to be effective at implementing a pop-up and one of the challenges is the EHR they were first developed.  They did not include dense information displays like dashboards or registries or summary views and so this improved stage was in large part skipped.  That seems to be changing somewhat, there is hope in this area and a lot of the tools or a lot of the existing electronic health records are starting to include more dense information displays that can be used more for improvement and more for inner disciplinary care and for teams to organize around the workflow and tools that support the adaptation of the workflow rather than forcing the specific workflow action through a pop-up.  This kind of shows the same information though little more details when we were researching this we--instead of just looking at, you know, what the task and what the use was and we actually--and stage in the quality improvement lifecycle we were mapping it out to a bunch of different components but this is basically showing the same concept that delivery decision support needs to fit within an overall quality improvement life cycle and jumping down to one mode of decision support when what the users are really needing is maybe identifying independencies or observing trends, things that are earlier in the--at an earlier stage of the quality improvement lifecycle is important to recognize so they can be applied more effectively.  So, this is the lecture to introduce the concepts of decision support.  I was doing it in the context of how, you know, what the opportunity is in terms of the variation of medicine in the context of what historically has been, the role of computers in medicine, and how it had gotten to this point where a clinical decision support is very important.  The importance of clinical decision support for the whole meaningful use objectives, and then finally the tools that can be--the tools that are part of decision support toolbox not only how these tools can be used but at what stage of quality improvement they should--really should be applied.  Thank you.