A PCMH Scheduling Model That Optimizes Primary Care Team Continuity
Presenter: Dr. James Benneyan, Northeastern University
Recorded on: May 15, 2014
Or anything right now.
> Yeah, sorry, just making sure everyone was muted, and we can get started here. I think we'll get started right now, if that's okay. Welcome, by the way, for those of you joining us. This is Nick Edwardson and Jim Benneyan going back and forth.
The boys here working out a few last minute kinks, but just in the nick of time, we appear to be ready to go. Of course, if you're here today, you are probably aware that you are attending a Center for Health Organization Transformation, or CHOT, Deep-Dive Webinar Series. Today's webinar's going to be hosted by Dr. Jim Benneyan of Northeastern, who will be talking about a PCMH scheduling model that optimizes primary team continuity.
And a brief reminder before I hand over the floor to Dr. Benneyan, today's webinar is being recorded. We will take the recording today immediately after the show, and host it to CHOT's members-only section where this can be viewed asynchronously. Also as a reminder, we have a number of functions available to you today during the presentation.
Dr. Benneyan's talked about, probably only going the shorter end on 30 minutes, but that would leave us more time for questions afterwards. Those questions can either be posed via the live call-in option, which you guys are all dialed into now, or it can also be made available through the chat function.
Which I will send out a chat right now to show you exactly how that all works. It should pop up soon here, excuse me. I just sent a text out to everyone saying send questions here. If throughout the presentation you'd like to send in a question for Dr. Benneyan to answer, I will make sure and will get him all of those questions when he gets to an opening.
With that, a couple other quick reminders, a few housekeeping chores. This is the first of ten Deep-Dive Webinar presentations that we will be making this summer. It's gonna be a very intense series but we, of course, welcome and encourage your attendance at all of them. If you have a webinar that's not of interest, but you have a colleague within your organization that comes to mind, please feel free to send along the invitation to them.
Because we're ultimately trying to expand our reach here at CHOT into our 15 IAB member organizations. One further reminder is, most of you might received an email from me yesterday evening, reminding you about the upcoming CHOT Fall Meeting, which will be in Wassam, Massachusetts on October 16th and 17th.
And the hotel information can be found on our website, With that, I'm gonna hand the floor over to Dr. Benneyan, Jim.
> Thank you so much, Nick, and thanks to my team at Northeastern for helping troubleshoot the technology just then. So, as I've mentioned, Northeastern University, of course, is one of the currently four sites, universities involved in our National Science Foundation Center.
To apply systems engineering and improvement methods, health services and so on to healthcare. So I really have two objectives to talk about this project both to disseminate a project that was partially funded by our CHOT Center that all of our members essentially co-own with us. But the other is we have nice pilot results.
So we've done kind of a phase one, and as part of the mission of all the universities involved, we're really seeking secondary test beds and spread opportunities. As we go through this, and let me put up the agenda and the outline. But as we go through it, I'm and all of us, I think, are tremendously interested in connecting into other health systems other than our initial test beds with so many of our projects.
So we'd love your help connecting and seeking other colleagues where we could do very similar work or try to replicate this. So with that, I'd like to give sort of just a brief overview of the general problem we've been working on. And where we are so far in terms of developing really two tools to, we believe, create better scheduling templates for team-based care.
So, it's really broader than primary care. But that's where we've been focused, but it could be anywhere, where you have sort of this team-based care approach evolving. Give a summary of that, then summarize where we are in terms of the results of a pilot case study, which we've completed.
And again, basically, hopefully this'll take 20 minutes, and then we could have some conversation about additional applications, and how we might go about that. So that's what I'm hoping to accomplish. Just a quick acknowledgement, this is really a joint project. We have another grant from, so this is partially funded by our CHOT Center, our National Science Foundation Center, Texas A&M, Penn State, Georgia Tech, and Northeastern.
And partly funded by some grant we have through CMS to do similar type of work and really try to drive stuff into practice. So we're gonna kind of look at this at the surface level, but if anyone is interested in diving deeper, reach out. We have working paper on our website.
We're trying to get tools up, and you can see our contact information there. So again, I'll hammer on this, we're really hoping to replicate test, that's where we are in the life cycle. So the general context, I think everybody's aware, patient centered medical home, or if you work inside the VA healthcare system, they call it patient activated therapy.
But a central tenant of where we're going is team-based care, and in family medicine and in primary care, it's bigger nowhere else there. For a variety of reasons, home-based care leads to greater continuity, leads to greater relationships, leads to all sorts of things that I'll talk about in a minute.
The challenge of creating scheduling templates is like a Rubik's cube or a Sudoku puzzle. They're all sorts of constraints on everybody's time, other obligations, and how do you create a reasonable time table, a template that is optimal or reasonably good? And what does even optimal or reasonably good mean?
We'll talk about that. But the challenges in doing this is the complexity, and everybody's other things, whether they have teaching or research responsibility. Particularly in family medicine, if you choose to go into family medicine, you probably have a focus on families, and may have one of your own, and may be starting one, and may not be full time.
So how do we create schedules where we have good coverage across days of weeks, sessions of the day, morning and afternoon, if they're evening and weekend hours? How do we create good templates that better serve the patients? And then if we have residents in the mix, that's a whole another layer of complexity.
So there's literature in the healthcare literature. This is our causality thought exercise moving from from left to right. If we can improve the coverage of a team, so if three or four of us are just primary care provider, an RN, a medical assistant, and a resident, and collectively, we're the team managing the panel.
A patient would like to be seen when they're available, at their convenience, and if they can't, they have to do one of two things, wait Wait to see their provider or somebody on their team of choice, or come in and have a visit with a more familiar individual.
Both of those are sort of bad things. So if we can create better templates that provide better team coverage, that'll lead to greater continuity defined in a very specific way. But percent of times I see somebody I've seen before is one operational definition and or reduce delays to come in.
Those two things, there is experience and literature, lead to good stuff happening in in-patient health. They lead to greater satisfaction, it's more patient-centered approach. They actually tend to drop in-patient utilization, unnecessary visits to the ED, for example. Because I can't see the person I'd like to see. So this is our driver diagram and on the right, how would better scheduling impact what IHI and CMS would call the triple aim, better care and better care experience, better overall health.
Here, better health is more trusting relationships, greater compliance to annual or periodic business whatever their latest device is and greater compliance to all the wellness. The well child visits, the screening for colorectal, mammography, pap smears, etc. And in the long run, less utilization of other avenues to get care or lower utilization of the system because of better population health would drop down cost.
So that's kind of the CMS view of this type of project. So here's the concept in the top of this slide, conceptualized four care teams. Green, blue, red, and yellow and they each have a number of individuals who are working or able to provide clinic time some part of their FTE some part of their week.
And it's gonna differ for everybody. And if we breakdown the blue team, you can kind of see the breakdown there. But the real challenge, oops, I have a blank slide here, is how do we create something like the right-hand side of this slide, rather than the left-hand side.
So on the left-hand side, if you look at certain sessions, certain days of the week. Tuesday morning we've got a lot of people covering from the hypothetical red team. One person covering from the orange team. But if my provider and the people I've seen before the RN I know is on the green team and Tuesdays' my day I can do this I either need to take time off from work wait until another day where I can see a provider that's familiar with me.
Or I really need to come in because of that nagging feeling I have whatever is the cause for the attempt to schedule. So I'm going to see somebody I'm unfamiliar with. Right? So if we can create better template such as envisioned on the right, where any day of week, any session somebody from every team is covering.
Now, whether they have availability or not, this is kind of a zen thing. Just the fact that there's coverage, increases the likelihood that there'll be availability. So whenever I can peel away from work or the rest of my life, and come in for a visit, I can see somebody that I'm familiar with.
So, that's the general concept, and it's an easy idea, it's complex to reason out how to create a better template. So this is what got us on all of this. And so I should acknowledge courses, I'm a faculty, so I'm architecting and involved in work. But we've had a number of students, PHD's, Masters, and undergrads working of this problem.
So we have basically two tools. One is a spreadsheet tool that just allows you to play what if. If this is our current template, so the mess on the left is an example of a template, over the top are individuals 1 through 14, and the rows are days of week and sessions, morning and afternoon.
And, where am I, am I in the family medicine clinic, am I pulling administrative responsibility, am I doing some didactic lecturing, etc., etc. But quickly we can analyze that template, which you can kind of just maybe look at visually and see you don't have good coverage by team in terms of uniformity.
You can play around with it and say well, I wonder if I swap out this and that, I wonder if I swap out this and that. What if Benny and I change places? Does this get any better or any worse? And you could hunt in pack or a better template which is useful so we have this what if tool.
But if you wanted to try all possible combinations of who works when, where, subject, all sorts of constraints on any given Wednesday morning in the sports clinic we have to have at least one person covering. And, so you've gotta satisfy all that. It turns into two things, a very long exercise and a real headache.
So that other thing that we did is took an optimization approach. So I wanna show you the concept of a linear program. It's actually what's called an integer program. But I'm not gonna dive deep. So, we have this mathematical algorithm where the objective function is to create better templates.
And we just mathematically said we've tried a couple of things here. We said, let's maximize the number of sessions is there morning and afternoon, five days a week, there's ten sessions. Let's maximize the number of those that have at least one person from all four teams covering. You could define this in other ways but we want to maximize coverage and we had to have an operational definition so is it objective function that's expressed mathematically and then what can we perturb, what can we do parrelized changes on?
We can change the providers assignments, right. So we can decide who's working where. So you basically see this type of equation and so if I'm a provider that's the left hand side across all possible responsibilities on Thursday morning. The sum of everything I'm assigned to, places to be, has to be equal or less than 1.
You can't tell me to go teach class and also be in clinic, right. So that's the idea of the sum over all possible responsibilities of provider i during session m. You can't put me in two places at once. And then there are all of these other constraints. We have to satisfy under contract or culture, everybody should work a certain number of clinic hours.
Everybody should have responsibility here, there, and everywhere. And so you can write all of those constraints. And even nobody should work more than X hours a week and if I'm .5 FTE divide that in half and so on. So you can define all of the particulars about the situation mathematically.
And if you're familiar with integer programs, you can kinda see how this and the equations on the left map to the words on the right. And if you're not that's the general idea. We can define this mathematically. So actually let me give you a little insight. So what's going on just one of these equation is the following.
> We have a certain clinic, clinic 7, which happens to be a sports clinic on Wednesday morning. And we wanna provide that for parents to bring their kids in if they play sports, type of thing, Wewanna have that. So somebody's assigned to it. And if we sum over all of the residents, if I'm assigned to it, B is equal to per resident Jim, B is equal to one.
And if there is not a sign to it, B subscript Sarah is equal to zero. So if we add up all those zeros and ones, it should equal one. Somebody's gotta work it. And so that's how basically these equations work, and then the subscript W is all the leaks where we want to have the sports clinic open.
This is what it really looks like, and it looks at two things. It either looks like, huh, not that big, some equations are quite overwhelming, depending on your experience with this type of thing, but this is the type of solution that's produced. Who does what when where. And none of the constraints are violated.
No-one works more in any function than they're supposed to. Every session has as much coverage as it's supposed to, and so on. There are a couple ways to accomplish this is finding a way to accomplish that, that maximizes the number of sessions that have at least one person from every team working.
So that's the idea. And we've taken a couple of cracks at this. And tried different objective functions. But we're getting almost the same result and pretty good results. So, we've done a pilot test on the left hand side, the bullets describe that. But a mid size family clinic in the greater Boston area, you can sort of see the size and scope of it in terms of the number providers and the numbers of first year residents and second and third year residents in the mix a lot the full timers are part time because of other responsibilities.
I'll show you the results but again what we're hoping to do next year is really replicate the system and or adopt it. And learn from that and sort of refine if needed. And so we did two things. Once we came up with a method rather than tweak the actual system and schedules.
We looked back over the last year and said, what was the session coverage like last year? That would have been 2012-2013, July to June. And then ran our model and compared how coverage would have been if we had been running our model. So this just sort of retrospective offline.
And any given week, in theory, coverage went up 20%. So if there's 20% coverage, hopefully, all the other good stuff will happen. That was proof of concept. Then what we just completed is this prospective test where around Labor Day 2013, we went live with new templates. And then with the health system had been measuring continuity and access.
So continuity is described as anybody who's had a visit in this health system in the last 36 months. Was their visit with somebody that they have seen previously. That was just our measure. So continuity has gone up. Days until third next available appointment. Well actually this is a different measure now.
Days until people are being seen has gone down. And then we're looking at inpatient admissions and use of inpatient admission, the title there I apologize utilization I believe the are in there. But utilization of the system are going down and what's quite interesting is compliance and wellness visits are going up.