How To Get The Most Out Of Your Existing OR Capacity Webinar transcript
This webinar is actually one of– it’s the first one of a five-part series that we will be conducting between now and October the 4th. And so when we send the recording of this out, we will also send a link that you see at the bottom of this page to you, in case you’re interested in any of the other webinars or in the series itself. So the structure of this series is that today we will do a high-level overview of multiple topics, how to take away block time, how to release and request block time, how to engage surgeons and administrators in performance metrics for the. OR, and how to allocate time. But we will go into much more detail on each of those topics in the subsequent webinars.
So let’s get started. There are three things we’re going to cover today. The first is dive deep into the major reasons why we think OR efficiency sometimes can be less than optimal and the five big reasons why ORs leave time and money on the table. The second is how each of those five problems can be solved using predictive analytics, mathematics, combined with transforming the processes, the core processes around block allocation, block release and request, as well as how to right-size block. And then finally, this is a big, big problem– I’m preaching to the choir here, I know. And it is only going to get worse with all the macro trends that we see, with increases in volume and need for surgical cases, as well as what reimbursement levels are doing with Medicare.
But before we dive into it, just to give you a sense of where all this is coming from– so we, as a company, work with multiple large health care institutions, academic medical centers, like the University of Colorado. Health, New York Presbyterian. We recently brought onboard Duke. University and Oregon Health. They will be live by the end of the year. But across the spectrum of these 15 or so large systems that are on Epic or Cerner or. MEDICTECH, some of them community-based and some of them academic, we’ve learned a lot. And so what you’re seeing in this webinar series is really a synthesis of the learnings from having built products and deployed them across multiple OR suites, totaling almost 500 now. And these learnings are being captured in this webinar series.
So what is the basic problem we’re trying to solve? And I’ll start with an anecdote from a children’s hospital that I personally visited a few months ago. At that hospital,. I asked them, hey, if I brought my child in today to get elective surgery done, when could you fit them in? And they said, well, you know, I have block time every Wednesday, but. I’m fully blocked out for the next six Wednesdays. And so, yeah, if your kid could wait, we could fit them in six to seven Wednesdays from now. And the question I asked them is, well, doctor, I understand. Seven weeks is a long time.
If I walk through your ORs today and tomorrow and the day after and next Monday, are you telling me that everyone that has reserved capacity through block time is actually using their time the way you expected them to? And he just laughed, and he said, no, absolutely not. It’s a Venn diagram. Some people are, some people aren’t. There is always capacity on the table, which to me was funny because here is a system that was fully blocked out, 98% blocked out. And yet, every day, there was capacity on the table. So that’s like saying, I just spent six hours in line to buy a $2,000. Super Bowl ticket. And I showed up at the stadium, and a third of the stadium’s empty. And especially when you think about the economics of the OR– the OR is the economic backbone of every hospital that we work with. And if you think about some rough numbers, every percentage point of block utilization or just utilization, room utilization, is worth a lot. So even for a hospital, let’s say, 20 ORs big, if you could improve utilization about 3% by increasing patient access, by increasing surgeon access to the OR, that’s worth a lot of money. Plus, it’s worth a lot of patient access sooner.
So my child could actually have been treated sooner than seven weeks. So the dichotomy we’re talking about through this webinar series is OR time is precious and never available, and yet, reserved time is being left on the table each day. Here are the top level five reasons we’ll talk about and touch on briefly, each of them, today, and then dive deeper into in the subsequent webinars. Fundamentally, what’s happening we believe, in looking across the set of customers we work with, as well as prospects we talked to, we’re using 20th century tools and processes to solve problems like block allocation, block release and request, and right-sizing blocks. And the five major problems we see are, first, the metrics we often use, like block utilization, to try and compare surgeon performance and right-size blocks don’t work. And we’ll dive deep into why they are neither surgeon-centric nor accurate, and there are far better ways of looking at the performance in the OR that then you can discuss with individual block owners when you try and right-size blocks.
The second is much like block utilization. As an average is a blunt instrument, another blunt instrument that we see used a lot is this idea that, I will, in Epic or Cerner, or any of my EHRs, create this thing called an auto release. Seven days, 14 days, three days before the fact your block gets released, and other folks get access to it. We can do a lot better than these auto release mechanisms to create open time and create capacity in the OR. So we’ll talk about that. The third common thread we see is that it’s not the lack of the availability of reports or dashboards or just data. The fact is that sometimes there is just too much of it. There is not enough belief in the metrics that are being used and disseminated, the way surgeons are engaged, administrators are engaged, in understanding the data. As well as the simplicity with which current performance reports are being proliferated and sent and believed leaves a lot to be desired. And if that happens, and there is low engagement in the metrics, it’s really hard to make decisions based on metrics. So we’ve seen that pretty– a pretty big theme across the board.
The fourth has to do with how we fundamentally allocate block. So we think block allocation is akin to creating carpool lanes, where you reserve capacity for certain– whether it’s by surgeon, surgeon group, or service line. And once you do that, it’s almost impossible to take it away. But when you allocate capacity, you’re making an assumption that the person or group or service line you’re allocating capacity to is going to be able to use it well. And if you’ve ever been in traffic and been in a non-carpool lane, watching how the carpool lane owners are using their capacity, it’s frustrating. And so we’ll talk about a much more sophisticated way to allocate time.
And then finally, from a day of optimization perspective, one of the ways in which the chaos on a daily basis can be avoided, or at least toned down, is if we had a better sense of how long a case was actually going to take. And means and medians just don’t work. Across the board– and. I’ll keep saying this– when we talk about averages like block utilization, like seven-day auto releases, like mean case length, these are meaningless because here’s how averages work. If my head is in the freezer, and my feet are in the oven, on average, my body temperature is OK, but I’m dead. And so most averages are blunt instruments. They are a baseball bat when you actually need a surgeon’s knife, so to speak, to make good decisions. So let’s get into each one of these.
When we say block utilization is not an actionable metric to right-size blocks, what do we mean? Every institution we work with and every institution that I have spoken to, and I’ve probably spoken to 60 or 70 perioperative committees, block has been allocated, and an. OR committee occasionally meets to look at performance. And the metric we all love to use is block utilization. [INAUDIBLE] and let’s look at the pattern of usage of their grid. This happens to be an Epic grid. This actually happens to be real data. And if I look at surgeon A and surgeon B, what’s common about them is both of them have a block utilization of 75%.
Now, surgeon A happens to be someone that does more predictable length cases, so, for example, a sports orthopedic surgeon or a cataract surgeon. And so when they are given 10 blocks a quarter, they manage to fit all their elective cases in in seven and 1/2 and pack them really tightly. And so their block utilization is roughly 75%. Now, if you’re surgeon B, and you’re a cardiovascular surgeon that’s doing quadruple bypasses, or I’m a neurosurgeon that’s opening brains up, my surgical case lengths can be highly variable. So I’ll come back to the start in a second.
But the point we’re trying to make is surgery is not like making Toyota cars. And so surgery, as we all know, has a wide variation in the length of the same case. So if I’m a neurosurgeon, for example, my cases could last anywhere between five and seven hours for pretty much the same patient type, with the same comorbidities. And before I walk into the OR, it’s very hard– Einstein couldn’t predict precisely how long it’s going to take. So now, imagine I’m surgeon B, and the patient I operated on on January the 3rd took seven hours to do a case. The exact same patient type on February the 1st took five and 1/2 hours. So post-fact, I used all 10 of my days to fit my elective cases in. And roughly, I used six out of eight hours on each day. What is my block utilization? 75%.
So the basic reason why block utilization is a one-dimensional metric that’s really hard to act on is that the number 1 minus block utilization is a fairly meaningless number. So if my block utilization is 75%, are you really telling me you can take 25% time away? No, as you can see in this example. Now, if block utilization is what makes me look like a hero, and I’m surgeon A, I can beat the system all the time by pretty much releasing two blocks well in advance of whatever your policy is. And then my block utilization becomes 75 over 80. I look like a 92% utilizer.
There is a better way. And the better way is to focus on, what is the real problem we’re trying to solve? The real problem we’re trying to solve is, who is leaving enough time on the table repeatedly that we can actually fit a case in? To quote a senior surgeon who we work with, the smallest unit, or the smallest quantum of time that matters in an. OR isn’t a minute. It’s the smallest length of case you can fit in. And just to illustrate that through this Epic grid, on January the 3rd, it’s a terrible, terrible thing that I started 30 minutes late as a surgeon, and the patient experience was horrible. But from a patient access perspective, even if I had started exactly at 7 o’clock and left 30 minutes at the end of the day, it’s unclear that anyone could really have done much with it. Or on January the 24th, if I predicted my case would take seven hours, but it took six and 1/2– I was efficient– block utilization is going to penalize me for being efficient.
So what we’ve done is basically said, let’s, in fact, use machine learning, or just pattern recognition, if you will, to look at historical usage of time and really figure out what are the usable chunks of time that you can actually do something with to increase patient access. So block owner by block owner, we actually focus on three things. One, am I repeatedly leaving large chunks of time on the table? What is large? It’s larger than the minimum amount of time you need to do a case. So instead of these little blue triangles, which are small grains of sand that surely we should focus on, but that’s not where the access is.
Access is in these large red diamonds, so whether it’s the middle of the day or the beginning of the day or the end of the day, if I’m leaving large enough chunks of time. The second is if I am not being a good steward of my time and not releasing my time in auto-releases kicking in. Because even if you have seven or 14-day auto-release, if I wanted to do a robot case, I need more than a couple of weeks to get the patient ready, and many other instances, not just robot cases. The point is early visibility into open time in the future is better. And so if I’m not releasing my time manually when I know, I’m not going to use it. That’s something we can certainly look at and should look at as part of collectable and reusable time. And the third has to do with how much time you let me release.
So if I used to be a prolific surgeon, who you gave two blocks a week– I have a 100 blocks a year. And I release, say, 10 of them, you know, life happens. Yes, it’s not predictable. That’s fine. But if I repeatedly release a block a week, 50 blocks a year, you probably gave me too much time to begin with. So whether you’re surgeon A or surgeon B, these three forms of reusable time are the same, large chunks of time left on the table, abandoned time, or too much release. And then if you actually look at your data, and obviously we have tools for this, but the point is if you’re looking for these patterns, what you can create is what we call a collectable timetable, where by block owner, by day of week, by location in your ORs, you can find out, of the time that I gave this block owner, how much of it is collectable?
Many, many good decisions you can make based on this. Number one, if you have a lot of collectable time at a location on a given day of the week, you may be running too many ORs. We all have a nursing issue. We all have an anesthesia shortage issue. Well, part of it is, are we using the staffing we currently have in ways that make sense? Second, if you’re trying to grow, and you’re trying to hire new faculty or surgeons, and you can only hire folks who can do cases, say, on a Wednesday, or on a particular day of the week, when you look at this analysis, you get the ability to understand the lowest hanging fruit, in terms of which block owners can I take a block Wednesday– Wednesday block away from without hurting their practice, and yet, freeing up time for others?
And then, perhaps the most potent reason to do this is when you have that difficult conversation with a block owner, and you walk in, and you say to Dr. Dyer that on Mondays in a particular location, they have x amount of collectable time, the data you can show them is no longer talking about the small grains of sand, the little blue triangles. Now, you can provide very fact-based drill-downs that say, OK, you have two blocks a day every Monday, so roughly, 27 in this– or 26 or 27 this quarter. Of that, we are going to give you credit for 21 because that’s the [INAUDIBLE],, plus all of the little blue triangles. Let’s just focus on the repeated large red diamonds. And remember what they were.
If you’re not using your entire block, that’s a travesty. If you’re releasing too much of your time, that just shows that if you’re doing this repeatedly, we gave you too much to begin with. And then if you are leaving more than three and 1/2 hours, in the case of this institution, on the table, someone else could have fit a case in. And then for each of these buckets, if you will, drilling all the way down to the Epic or Cerner or MEDITECH or Paragon timestamps to provide that kind of visibility. So the punch line here is that if you’re trying to right-size blocks and increase block utilization, focusing on block utilization is actually not the right thing.
Focusing on the collectable time that you can take and repurpose is the way you can bridge the holes in the Swiss cheese. And so just to show you some results from institutions we work with, many large well-run institutions, each of whom, we found at least 15% to 20% of collectable time. And when you have this conversation with highly data-driven, highly scientific folks, surgeons, the OR committee, this is a much easier conversation to have. And even if a portion of this access or this revenue is captured through decision support tools, like collectable time, you’ll be able to free up a lot of time without– and do it in a way that’s fair, surgeon-centric. And whether I’m surgeon A or surgeon B, you can defend it.
So we’ve done this study across multiple institutions we’ve worked with. And, yes, it’s important to lower first case delays. Yes, it’s important to get our turnover time right. Yes, in many cases, we have case cancelations. But when you actually look at the pie chart across these institutions of how much time has been left on the table, a majority of the time that’s been left in the table is because life happens. When we allocate block, we make an assumption that the owner of the block will be able to fill exactly the block we gave them with cases on those days.
Well, clinics happen. Conferences happen. Vacations happen. Sometimes there is seasonality. And so more than half the time that’s actually left unused from a block perspective is left unused because cases weren’t scheduled, not from a room perspective necessarily. The OR finds ways to get add-ons or get people into the OR. But from a planned, predicted nature of how you’re allocating block, it’s worth looking at collectable time and saying, how can I right-size block? And by the way, what you do with collectable time is up to you. You can close some ORs. You can allocate capacity to existing surgeons or service lines or groups that are obviously growing and need more time. If you’re growing as an institution and hiring new faculty or surgeons, you can give time to them.
Or, we’ll come back to this, you can put it into an open pool for first come, first served use. So that’s the first piece of the five principles, if you will. There will be another webinar will go much deeper into this, in terms of how the tool can be used, what can actually be done, and take much more follow-up on this notion of collectable time, instead of block utilization. But let’s jump to the second concept. Remember we said that the way block time is released and requested leaves a lot to be desired. It’s not to say some folks don’t schedule directly on the EHR and don’t have visibility into potential open time in the future. The problem is that block time is so sacred, and that no block should be left behind, it’s far better if we are much more fine-grained and look at the data.
What I’m showing you here is a lead time analysis of booking patterns across multiple institutions we work with. The data is all anonymized. And I’ve taken four service lines, general surgery, neurosurgery,. OB-GYN, and urology. And what these graphs are are the percentage of cases booked in advance of the day of surgery and how far in advance. So on the x-axis, you have number of days. On the y-axis, you have percent of cases. So each of these colors represents a different service line. Now, whatever your data is, it doesn’t have to be exact, but there is some distribution. These vertical lines are the medians, which means there are service lines– in this case, it’s the gray OB-GYN– the blue is, in this case, general surgery– where 50% or more– these are the median lines– 50% or more cases were booked 40 days in advance, 24 days in advance, 21 days in advance, and 17 days in advance.
So when we come up with this artificial construct that, let’s do an auto release, say, seven days or 14 days in advance, you really have to ask the question, where did that come from, right? Much like block utilization, you have to ask the question, where did the target 78% come from? And part of this is because it seems like a good and a nice thing to do. The problem is if you have this kind of booking pattern, you could do a lot better. For example, if I’m Dr.. Agarwal, I’m an OB-GYN, and I have blocked time 25 days from now because my practice tends to book a majority of my cases more than 40 days in advance, why couldn’t my office get a little reminder saying, hey, Dr. Agarwal had signed 25 days from now. By now he should have put a case or two in his block time. He hasn’t. Is he on vacation? Does he have a clinic conflict? Please consider releasing that time.
So when you think about having tools that allowed proactive release and then allow you to pick up time as a block owner or a non-block owner when you need it, so imagine being able to free up time and then giving your clinic’s access to, say, a mobile and a web tool. A mobile tool could look like this. It’s like OpenTable for open time. That said, someone needs time in the OR. Why couldn’t we expose all available inventory of open time in the OR to them? When the clinic sees what time is available, they get to request a chunk of time in one of the locations that time is open in, pick the slot that they can do the case in, and then while they’re requesting that time, put in any special requests– I need the robot room, I need an [INAUDIBLE],, I need a scope– and notify everybody in the. OR that they need the time.
What if they could also put their name on a waitlist or an Amazon wish list if no time was available in the inventory of time that we’re making available through this tool– so, alert me if time opens up. Similarly, if I’m on vacation, and I realize that I forgot to release my time, and I tell my clinic scheduler, or myself, I can go in on this mobile tool and release my time. The point is when we do block allocations and block time out in reserve capacity, we need a bit of a– we need an ability to take the pressure off when folks are not going to be able to use their time well, predict who isn’t going to use their time well, find ways to get them to release it, and then make that capacity available.
There’s a lot of benefit to block allocation. But block allocation taken to an extreme will result in carpool lanes, where there is no pressure valve to release time that people are not going to be able to use. So this is just an example of one way in which you could make that time open and create this capability of a common lane in this set of carpool lanes that multiple cars could drive in, as the occasion presents itself. So whatever your current process is of block release and request, chances are that OR schedulers and clinic schedulers have a bit of a tough time fitting in cases when they need access to the OR for their surgeons. Because chances are, if you are like any of the institutions we work with, there are multiple phone calls and text messages and Post it notes and people walking into the OR scheduler’s office, asking for time. If you use tools that help you create open time and expose them in an egalitarian fair way, all of that goes away because now you have a very streamlined lean process for creating and exposing open time.
And that example of walking into the children’s hospital and saying, you know what– what. I wish that surgeon would have said is, I don’t have block time except on Wednesdays, but hold on for a second. Let me go to my mobile device. Let me look for open time. Oh, you know what, there’s a three-hour chunk of time open next Friday. I can fit your kid in. And if everybody could do that, you’d get more cases in in business hours and pack the ORs. Now, the flip side of it is creating open time. So remember that booking pattern slide I showed you? We’ve condensed it to the left side of this page. Because based on this booking pattern analysis, imagine my office or a clinic scheduler in my office getting occasional reminders, saying, hey, Dr. Jones, Dr. Agarwal, or. Dr. Smith, has time in the OR.
And this is out of sync with his or her natural behavior and booking patterns, meaning by now, they should have put a case in. And this is not based on an artificial seven-day, 14-day, three-day deadline. It’s based on my prior history. And so it’s like a dentist’s reminder coming in and saying, hey, you’re going to show up? If not, this is precious inventory that I can give to someone else. Not that you’re going to lose your block, you’re just going to lose the one instance that you’re willingly giving up because you didn’t put any cases in, and/or you know you’re not going to use it well.
This, in combination with the other feature, this ability to put your name on a wait list, right, and say, look, let me know if time opens up, this is like a marketplace in the sky. This is you saying, I want to go to a Michelin two-star restaurant. No time’s available. I put my name on a wish list. And then this tool,. OpenTable, goes out and ferrets out an open table for me and says, who has reservations on the day I want to go? It’s not John or. Jane being a nag and running around and following me and saying, Dr. Agarwal, release your time. It’s a machine, 24 by 7, going and looking for time and then finding an instance or a slot that they could open up. So this is what we mean by– we have access to tools now with mobile and web, thanks to Steve Jobs and thanks to data science, that allow us to get far more precise about how we create capacity and match supply and demand because supply and demand is stochastic in surgery. It’s not deterministic.
What that means is Einstein could do your block allocation for you and your block schedule for you. The day he would do it, it would be wrong because life will happen. Everybody takes two weeks of vacation. Everybody has clinic conflicts. Everybody today knows for the next six months which conferences they’re going to go to. Well, what do you do? Do you take that time and put it in a pot and make it available to everyone else? Or do you believe that hope is a great strategy, and I’ll sit on i? So where this has been successful, there is a web version of that tool, but the tool is less important.
The principle of opening time up and making it available to everybody is a big deal. So at University of Colorado Health, for example, when they did it across 42 ORs, in two years 2,500 blocks have been requested, over 3,000 released. Now obviously, when a request comes in for a case or a block, your OR schedulers get to decide whether to approve or deny that request, because remember that OR might asked for– or a specific piece of equipment that I need may or may not be available. But if you notice the numbers, almost 80% to 90% of time that has been requested has actually been approved in each of these cases. In a world before tools like this, there would be, you know, gentlemen’s agreements, or a whisper in the alley saying.
I can’t use my time, can someone else use it. I call that doing a garage sale. Why do garage sales when you have eBay? And eBay is a far more effective way of getting rid of stuff I can’t use, especially if it’s as precious as block time, the $100,000 block. And so it’s not just one institution. It’s not just academic medical centers. It’s not just EPIC. It’s not just Cerner. Large, small, academic, community, all kinds of institutions have benefited a lot from tools like these. In fact, if you’re a community center or an institution where you have splitters, people who can do cases at your institution or at a competitor, frankly, where would they go first where you gave them access to the OR sooner?
So if you showed people, like OpenTable, the availability of inventory, that’s a big deal. Some of these other numbers are pretty impressive for us anyway, not just the financial impact of being able to pack your ORs better. That’s huge. University of Colorado Health actually has gone on record and said they’ve improved their block utilization 4 percentage points. That’s roughly half a million dollars per OR per year based on this tool. What’s equally impressive is this release lead time. So imagine, in a world even where auto release was working and you gave people 7 day lead time, now if you gave people almost a month’s lead time or more into future capacity, my kid could have been seen sooner.
The ability to promise open time to new surgeons and not necessarily give them block and say, look, I don’t have a block, but I’ll always give you open time. Build a book of business, show me you can use Wednesdays well, and then let’s talk at OR committee and we’ll give you a block. So many, many benefits of creating open lanes on the highway for people to use and prove that they can use them well. So that’s the second principle. So the first principle, if you remember, was right sizing blocks based on collectible time instead of block application, the second principle being opening up time– because even with the best block allocation and the best block schedule in the world, four to five weeks in a year every block owner will have a conflict, and that’s a lot of precious block time that won’t be used well.
The third has to do with, how data driven are we? Are we an institution where there is a single source of truth everyone believes? Are we an institution where our data is current? I call– frankly, not just in health care but in almost every industry, reporting can be used in one of two ways. It can be used as yet another weighing scale which I get on and it says you’re overweight again– I call that admiring the problem– as opposed to, no weighing scale on the planet that I’ve ever gotten on has helped me lose weight. But there are dashboards and there are reports and there’s paper and multiple different places people pull reports from. The question is, what do we do with it? Is it easy to access? Do our surgeons and our administrators see the same number then believe them? Do we understand the definition of block utilization? When we use two rooms, how does that count?
So this idea of credible, transparent data in which we can all engage, then we have a shot at potentially making decisions based on them that actually matter. So back in high school and in college I would get a homework grade every week, and there were two types of professors I had– one that said, you got a 30 on 100. Good luck. And then there were the ones who said, you got a 30 on 100 again. Go read chapters 4, 5, and 6. So that’s a little bit more helpful than saying, you know what, you suck at math, again. So imagine a world in which all of us get access across the institution to data being pushed to us that’s relevant, credible, and transparent.
And so imagine if every surgeon, or if the surgeons didn’t want to participate, their offices and administrators, got a simple text message. You know, I don’t know if you guys are signed up for this, but I’m signed up for a tool where when the market closes every day I get a text message where my portfolio performance is pushed to me, simple text message. I open it, it says, you know,. Trump did this, China did this, market went up, market went down, this is what happened to my various stocks or bonds. Right? Very similarly, what if I got a simple text message saying, hey, Dr. Agarwal, the week just ended. It’s Monday. Last week, here was your utilization, your volume, your delays, your room in block utilization, your case length accuracy, whatever are your five key metrics are. And then imagine I could click on a little link that launched a mobile browser, so you don’t have to download an app or anything. But imagine if even 90 seconds or less I could pretty much slice and dice my performance over the last week.
This is my homework grade being sent to me with details on the chapters I should study, how I used my block time, when I started late, when I ended early, how accurate was my case length depiction, did I spend time in my own block, in a surgeon group block, in a service line block, how these numbers are being calculated. What is the definition of block utilization that we’re using? When you say you release time, did you really release time? Can I show it in. Epic or in Cerner where that time that was released all on my mobile device, palm of my hand being able to justify the numbers and provide guidance on what’s going on by day of week that is creating first case delays, that’s creating more turnover.
What is the logged values in Epic, for example, being exposed to me? Where this gets– and imagine a very similar set of tools being used by administrators to be able to have hallway conversations that are data driven. So when I walk up to the OR manager and I say, hey, John, I need more time in the OR. And they say, I need more in the OR because I worked late the last two Wednesdays, he or she being able to pull their mobile device up, use the same corpus of data and say, yeah, you know, Sanjeev, you did, but the three Wednesdays before that you left early. The point being anecdote becomes strategy when we don’t have data at our fingertips to make decisions as often as we can. Now, the earlier point I made– every other industry has early warning signals.
This idea that, don’t tell me my cancellation ratio was high three months ago, or my utilization was high three months ago– help me do a better job. And the way you help me do a better job is use mathematics to identify where certain metrics are deviating from a band. So if my turnover time is starting to get out of whack, notify me before telling me three months after the fact that it got out of whack. Or if my cancellation ratio is getting to the point where I should care about it, notify me, because reporting needs to be helpful. It needs to be my consultant, not just pejorative and saying I did a bad thing historically, so helping everybody do better, not just admiring the problem. We get into a lot more detail in the webinar around reporting and engagement and how you can use web and mobile tools in combination.
The fourth principle has to do with block allocation where, again, when we allocate capacity, we always do it on rules that make sense at the time we do it. But once we allocate capacity, it’s a rite of passage. I have never found any OR committee being able to have an easy discussion with a block owner saying, you know what,. I’m going to take time away, first of all because we use block utilization as a metric, and second, because there is, you know, always a good explanation for why life happened and my utilization is low. So just to continue with this metaphor of carpool lanes, imagine if every lane on the highway you drive on was assigned to a different make of car or a different color of car. What would happen?
And that’s the equivalent of what we do when we do block allocation. Some lanes would be half full, some lanes would be so packed, and some lanes would look at the other lanes going, how come my lane is so packed and the other one is– whether it’s hotel reservations or any expensive asset where the commodity is perishable. An OR that’s empty today is $100,000 worth of revenue that you didn’t get. And so how do you deal with a situation where, in fact– you know where carpool allocations would work really well? Is if you could precisely know how many cars would be driving on each lane every 10 minutes. Then you could allocate capacity really precisely.
Or if you were making Toyota cars with low variability, then you could say, you know, each car takes 30 minutes to make with a standard deviation of 20 seconds. I will be able to make 500. Camry’s today across four production lines. The problem is, in surgery, the two big stochastic variables are the volume of cases– and this is, again, real data, it’s never a smooth line trending upward by day of week or weekly or monthly– and second, the length of a case. So you’ve got these two stochastic measures, but we’ve allocated capacity as if we knew that the block owner was going to be able to use all the capacity well. That never happens. It’s seasonality, it’s vacations, it’s clinics, all the things we’ve talked about before. So how do you deal with this?
The way– and at the same time, this idea of, should we just move to open time completely? No, absolutely not. That’s a highly inefficient model as well. We have to balance clinic schedules, when surgeons can do cases, when they can get blocked, the fact that when I’m in the. OR as a surgeon, my ability to add cases on and my ability to have a smoother day, a planned day is better, the team I work with, the equipment and rooms that I need, all of these are really important. And so blocks are a really good thing. But assigning 100% of time or a large– the right mix of open time and block time is, I guess, what. I’m getting to. So how do you create that? Let’s show you one way to do that. And this is a methodology called bin packing. And it’s used in many other industries.
For example, UPS and FedEx use it to decide how many trucks they need to deliver packages each month. Because if you think about. UPS and FedEx’s traffic in December, it looks very different– lots of gifts, small boxes, light packages moving, versus in January where, again, we move to business traffic. So the principles of bin packing are as follows– if I take every service line and every block owner historically, I know the volume and mix based on the timestamps of cases they’ve done by service line. So let’s take an example. Imagine for orthopedics, each of these graphs is showing you, over time, how many cases of each length that that service line did historically, OK?
So just do this one more time. So the purple is the number of 1-2 hour cases. The blue is the number of 2-3 hour cases, et cetera, that this service line has done historically. And if you use this volume and mix breakdown to project what the likely number of cases they’re going to do in the next 90 days, you can be very sophisticated about a forward looking forecast. And you can say– you can accommodate growth in volume. You can say, how does Q3 typically relate to Q2? And let’s look at all the historical patterns. Here’s where some pretty precise statistical methods can be used. There’s one called. AREMA that weighs near term data points more than historical data points.
But that’s neither here nor there. We’ll get into that in the webinar on this topic– the point is, if you base your future block allocation not based on the current block allocation but on the time actually used. I could give my kids $100 a week of pocket money, but if they only use $20 all the time, I probably shouldn’t give them that. And so being able to project future case volumes by volume and mix of case and projecting how much time every service line and every block owner will use will result in the following. So this is, again, real data, where when we walked into this institution by service line, the total number of allocated blocks were 1,600, and there were only 1,600 blocks to give away. So they were fully blocked out. And so the allocated blocks were a result of all the historical decisions that remained.
When you use a model like predicting by block owner what the need for time is going to be over the next 90 days, we found that, really, all they needed to do was allocate 1,479. That’s what the forecasting algorithm said. OK? So we’ve said– let’s do a Coke Pepsi test and compare our forecast. Let’s just play this quarter out. You’ve given us the last three years worth of data, and I’m going to predict today that each service line is going to use a different amount than you’ve allocated, and here’s my prediction. In some cases you’ve given them too much, and in some cases you’ve given them too little. When we actually compared that to how much time was actually used, only 1,279 blocks were actually used.
So in every case the historical decision making that has led us to the point where we’re at, how much block has been allocated, whatever it’s been based on can be improved if we are not using sophisticated forecasting methods. So that’s step one. So right off the bat, for this institution, they could have created 121 blocks of open open time across service lines. All right? That’s level 1. Now, within a service line, if you go back to the variation and individual search and usage of the time– I’ll just go back here– if I take a graph that for each surgeon maps how much volatility there is in their usage of the OR, you can establish a base level of block allocation above which you don’t really need to give them individual block up to that point. So what that means is, within the service line there’s the ability to create what we call service line open time.
So if I’m in general surgery and my block allocation is 287 blocks– that is the right size version of my block allocation– in fact, when I then allocate that time I can create a sub division of that open time and only allocate a portion of it to the surgeons within general surgery, leaving more time open for me to manage as the general surgery service line owner. Like I said, there’s a lot behind this, and we’ll get through it. In the interest of time,. I’m going to move on. But that’s two ways to create open time, and open time is your friend in a highly volatile environment.
The fifth principle has to do with models of case length estimation where means and medians are really not that effective. And we all know the benefits of being able to be more precise in case link estimation. Now, because it’s a fundamentally stochastic process, you can get better at the estimate, but you can never get precise and exact. But if a case is running late in a given OR in a given day and you’re able to predict how long the next case is going to take, you might be able to, for example, move them into a different room. And if you do that based less on tactics on the ground, but based on an actual prediction of how long that case might take, you’d do a lot better. So let’s get into that a little bit.
There are statistical models that will take a lot more than just the medians of cases in the past. There is one called a random forest model. And basically what it does is it creates a decision tree of decision trees, and then comes back and shows you, based on the patient type, the CPT code, the patient characteristics, even the pair mix, it takes a few vectors, it takes a few data points, and predicts what the length of a particular case is going to be. When we have done this for multiple institutions, we’ve come up– and the way you measure this is you take historical data, you take a portion of that data, your test data, and you create a model based on that. And then you test it on the remainder of the data and see how well your model did relative to in-built EHR models. And in some cases, we’ve only been able to beat the estimate by 10%, in others as high as 27% to 30%.
The reason why this matters is 30 minutes to an hour’s worth of decision making capability, how long this case is going to take actually matters when you’re making a decision day of. So again, we’ll get into this in more detail as to the methodology if that’s interesting to anyone. But obviously, you know this better than any one of us does. This is good for patients, it’s good for surgeons, it’s good for staff, on the day of, it’s good for anesthesia, to know what’s going to happen during the day, and especially in a situation where one or more cases are running late. So this is another one of our topics for the webinar. I just wanted to summarize by saying some of these things you’ve seen are packaged tools. And when you package these tools, you create more open time, you have a fairer way of right sizing blocks, you have a better way of block allocation, estimating case lengths, and you engage the community of surgeons, practitioners, and administrators in the data, lots of good things happen.
We start managing precious OR resources with 21st century tools, which we use in daily life. You know, no one calls a travel agent anymore to make travel bookings. No one calls a restaurant anymore. We all go to OpenTable for a lot of things. Why do we have to call 15 people to make a single reservation in the. OR, especially for open time? And so just to share some cumulative data with you, several of these institutions that are starting to adopt these practices are seeing a kind of benefit. They’re seeing anywhere between 5 to 10 percentage room and block utilization improvement. They’re seeing collectable time come down. They’re seeing a much more satisfied surgeon population, because if I’m going to give my surgeons access to the OR, if I’m going to give my surgeons the ability to right size their blocks using a metric that makes sense as opposed to blunt instruments like block utilization, it goes over better.
My ability to grow, my ability to– in some states, like Maryland, we know that there’s an optimization of how much we want to grow versus how many ORs we want to run efficiency perspective. The tools, like collectible time, have multiple dimensions to them. They can help enclose ORs. They can help you figure out how to grow and do more cases in business hours. So the economic value that can be unlocked through using these techniques is significant. We have seen numbers like half a million to a million dollars per OR per year in environments where the institution is growing and trying to accommodate new volume and hire new faculty and surgeons. In instances where they are trying to be just more efficient and do all the cases, there isn’t enough volume and fewer ORs, your ability to shut down a few ORs. And then from a soft benefit perspective, being able to delay when you need to open new ORs, from a CFO perspective they’d love you if you don’t need capital for another two or three years if you’re growing very fast.
From a culture perspective, if we believe the numbers and the data and we’re able to drive our surgeon satisfaction up, that’s a huge deal, and our nurse and staff satisfaction up by being more planned predictive in the day of surgery. Just to summarize, the five principles were, instead of block utilization, try this idea of collectible time. It’ll go over much better. Second, get more time released sooner and create the ability to provide transparency and visibility 24 by 7 into open time. And we call that. OpenTable for open time. The third is, if it can get to a single source of truth and have a user friendly, provider friendly tool for reporting and engaging everyone in the data, we will end up in a better place, where the belief in the numbers and belief in the decisions that are being made are much higher. This idea of how we allocate block and how much to whom is a very hard problem. One way to start is to start looking at forecasts of an allocation and creation of more service line open time and open open time, and then using more sophisticated mathematical models to create visibility into what truly the estimates of case lengths are can help you manage the day better.
The last thing I will say is there are many people out there– all the tools that we use, a lot of them have reporting. Please don’t– and I know you know this– we need to differentiate between admiring the problem and reporting against numbers and what happened and why it happened with what’s going to happen and then making the right thing happen. So from an analytics perspective, because we know we all have analytics teams, are we focused on figuring out, much like Waze does, and it tells you, you know, it’s going to take you 50 minutes to go from point A to point B based on all the historical data, when you embed these predictive analytics into core processes of block allocation, block right sizing, release and request, is when magic happens.
It’s not just the math. It’s combining the math and feeding it into fundamental processes used to make decisions to drive the right actions. So here’s what we’ve talked about as a summary of lessons learned. We haven’t found any EHR that does many of the things we’ve talked about. So you do have to look beyond your EHR. You also have to look beyond what we call descriptive analytics. You have to look beyond dashboards. You have to look beyond just looking at the numbers, but getting into how we change them. Predictive is what predictive does, integrating those analytics into our core processes, working with their IT team, and being careful about what you can do in-house and what you can’t. If you can do it in-house, go for it.
But you know, for example, as a company we’ve spent $50 million building a platform and making an investment because we can defray it across 53 institutions. If you have the analytical and economic horsepower to do that, by all means, do it internally. But sometimes it’s good to look outside. You know, before Google existed, we all went to the library to look for books. This idea of, are we willing to take a look outside of our institution for people that have solved this problem, and solve this problem at scale? Everybody can solve this problem for one service line for one year, doing it by hand and getting two analysts to do it, but can you truly solve it at scale?
Let me stop here and see if there are questions that have come in. The first question was, does this require an app on the mobile phone? Great question. Thank you. No, it does not. It all runs on a mobile browser. And so anybody that has a smartphone with a mobile browser can use it. Obviously, the web based tool that you didn’t see today is also a web based tool, so you don’t need any special equipment for that. There’s a question– I’m not sure I fully understand this question, Elliana, so if you wouldn’t mind elaborating on it. It just says, break relief.
So if you would wouldn’t mind elaborating on that, happy to answer it. Do tools like this depend on the EHR? No, they don’t. They just depend on the ability to extract historical data and get a feed from the EHR. So they do not dependent with– these tools work EPIC, they work with Cerner, they work with MEDITECH they work with pretty much every EHR you have. Any other questions or thoughts? OK, so just to wrap up, we have, like I said, a series of webinars that are coming up. The next one is on. September the 13th where we’ll dive deeper into this notion of collectible time.
If you’d like to register for that, you can go on our website and register for it or we will send a wrap up of this webinar along with a link to sign up for the others. Please feel free to sign up for any ones that you like or for the whole series. And then finally, if you are interested in learning more from institutions that are using these tools, remember, I mentioned about 15 of these institutions are using these tools. We are having a conference in October. No obligation on your part except to show up and learn, if that’s what you choose to do. As a physician or an administrator or a scheduler, if you would like to come see how these institutions are integrating analytics and mobile and web into their core processes, you’re welcome to join us. We will send you more information about that as well.
HOST: OK, well thank you very much, Sanjeev, for presenting today’s webinar, and thanks to all of you for joining us. A quick reminder– I know you’re eager to get a recording of this webinar. Look for that. It should be appear in your inbox in about 24 hours. And thanks again for joining us.