TRANSCRIPT
[00:00:00] Jordan Messler: Hi everyone. Welcome and thank you for joining us to this informative session on preparing your health system for the new CMS severe hypoglycemia and hyperglycemia measures, the eCQMs. With a focus this evening on understanding glucometrics.
We’re excited to have our keynote speaker with us, Dr. Curtiss Cook, Professor of Medicine at the Mayo Clinic, Alix School of Medicine. Dr. Cook has been with the Mayo Clinic in Arizona since 2003 and holds the title of Chair of the Division of Endocrinology. His clinical research interests have been in the area of quality improvement and inpatient diabetes management, and he is a leader in the field of glucometrics.
[00:00:51] Jordan Messler: Before I hand it over to Dr. Cook, I wanted to take a couple of minutes to recap some of the details and the timelines of the new CMS measures related to glycemia. If you’ve been aware of these measures, you know they’re coming soon, starting in January. CMS, with the introduction of these two measures has made it clear that severe hypoglycemia related to insulin should be a never event, and that hospitals really need to prioritize early treatment and optimal management of severe hyperglycemia.
[00:01:23] Jordan Messler: There’s been too much variation evident in the work, that Dr. Cook and others have done that’s persisted. There’s been a call to action to improve glycemic management for 20 plus years, and yet that variation opportunity to improve continues to exist. I wanted to highlight some of the details of, the two measures and Dr. Cook will discuss further about them. These two measures are part of the hospital inpatient quality reporting program. They’re eCQMs eClinical Quality Measures. There’s about, just under a dozen of them, and two of them starting in January will be around glycemia. These two measures again, are severe hypoglycemia, which will be measured as percent patient stays of blood sugars, less than 40 within 24 hours of the administration of insulin or an antihyperglycemic occasion. And the second eCQM around glycemia severe hyperglycemia, which will be measured as the percent hospital days with one or more blood sugar greater than 300. Excluding that first 24-hour period. Again, the intention here with severe hyperglycemia is that we often delay treatment, and we know that treatment and treating timely and that unnecessary prolonged hyperglycemia, can impact the patient’s ability to recover.
[00:02:39] Jordan Messler: A reminder of the timeline, if you’re not familiar, they were introduced last year by CMS, just over a year ago. During this time, hospitals have been working with their EHR providers to implement the reporting packages. Starting in January begins that report, that data collection period.
[00:02:58] Jordan Messler: In February of 2024, hospital quality teams, executive teams will review the data, decide which eCQMs to report. Again, there’s about 11 of them. They’ll be less by this time, two of them will be around glycemia and hospitals can choose to report zero, one or two. Obviously, we would hope that hospitals report both of these. These are balancing measures.
[00:03:17] Jordan Messler: Then they’ll be publicly available in October 2024. To highlight on the top right there, even, you know, again, these are elective measures, even if a hospital chooses not to report them, there will be internal visibility on these measures. And then the financial penalties, again, these are elective measures to report, but you have to report at least four of these eCQMs, and if you don’t report them, there are financial penalties associated with them, which will begin in fiscal year 2025.
[00:03:45] Jordan Messler: Good to know and understand the history of a lot of these measures from CMS that they often start in this pattern. They’re elective initially if they go well, that they often become mandatory measures and that may happen here. And you’re gonna hear more from Dr. Cook on not only these measures, but other ones that may be following in the hands of these measures.
[00:04:05] Jordan Messler: The important piece of these types of measures are, again, there has been a lot of variation. There is a challenge of not having the right me metrics and the right benchmarks around glycemia. So important here to raise that awareness, begin to develop those benchmarks and hopefully drive hospitals to implement the solutions that, that are needed. To better understand the landscape of glucometrics and help us better understand how the CMS measures fit into that landscape, there’s really no better person nationally to talk about this than Dr. Cook our day two keynote presenter. If you have any questions for Dr. Cook, please add them to the chat throughout the presentation. Any questions that we’re unable to get to live, we’ll make sure that a member of our team follows up with you after the conference.
[00:04:48] Jordan Messler: I’ll now hand the microphone over to Dr. Cook
[00:04:55] Curtiss Cook: Well, thanks for the, introduction and invitation talk at this. let me have minute, activate my pointer here if I can.
[00:05:07] Curtiss Cook: So, yeah, I appreciate the opportunity to come talk about this topic. I’ve been probably involved in the area of some form of glucometrics for probably the last 20 years, and I’m gonna kind of give you my perspective. I know at this meeting in previous meetings, you’ve had some really good glucometrics experts talking to you.
[00:05:26] Curtiss Cook: And so, So I’m gonna kind of give you my perspective and having worked in this for the last 20 years, where I think we are now where we need to go. So I’ll talk about definitions, rationale, some benchmarking examples, how we apply these benchmarks in particularly future directions.
[00:05:50] Curtiss Cook: So, we’ll start off by just talking about the categories of glucometrics, how I propose and categorize them. Then talk about rationale for measuring inpatient glucose control and describe the current state of national benchmarking efforts. Dr. Messer kind of gave you an introduction on one aspect of that already. I’ll provide examples of what I would call applied glucometrics. How do we use the data and talk about future state, You know, how do we go just beyond, just measuring stuff, you know, how can we, how can we look at this data, you know, in terms of forecasting, glucose control, spatial analytic techniques and, and how we measure insulin use in the hospital.
[00:06:22] Curtiss Cook: So what kind of started this whole, this whole field, you know, unfortunately I’ve been, around really a long time. And I kind of remember the time when, as an endocrinologist we didn’t really care much about what happened in the hospital as far as glucose control. We sort of got consults every now and then.
[00:06:39] Curtiss Cook: For glucoses, over three or 400 people asked us to come by for help. But it really wasn’t much attention paid to this topic. And endocrinology by legacy really is an ambulatory practice as an outpatient specialty. So we went to the hospital every now and then to consultants of cases. I think if I were to pinpoint a time when things changed, it’s this Vandenburg paper, which I’m sure probably most of you are familiar with this paper that came out in 2001, that, looked at the impact of intensive insulin therapy critically ill patients.
[00:07:10] Curtiss Cook: Yeah, there was a lot of controversies surrounding the paper, but I think up until then, up until this paper, there really wasn’t much attention being focused on inpatient glucose control. And this, this paper, I think, kind of set off a whole new feel. After this, It seems to me that everybody started, started focusing their attention on what’s going on in the hospital.
[00:07:30] Curtiss Cook: So what is the definition of glucometrics? This is the definition that, Goldberg, came up with, systematic analysis of data on blood glucose levels of inpatients. I kinda look at these as kind of different categories. It’s not just a single glucometrics, there’s multiple ways we can look at glucose control in a hospital, for example, glycemic exposure.
[00:07:50] Curtiss Cook: These are measured central tendencies such as medians, means patient day weighted means there’s a hemoglobin A1C, which the Joint Commission does recommend having updated A1C within 90 days of admission. This kind of term is a reflection of long term or chronic glycemic controls exposed as an outpatient. Then you have kind of what I call measures of efficacy of control targets, things like, getting glucose time and frequencies, proportion measurements, proportion, days. When it comes to some of the work that I like to do? I like to look at proportion measurements cause that’s how we assess, That’s how you’ve been, what I’ve been using to assess provider behavior. Then you’ve got measures of variability, standard deviations, coefficients, variations there of other papers out there that show that, variability, glucose variability is independently associated with mortality, and outcomes in the hospital control over time, that statistical process control, so-called control charts. Then we have outcomes. You know, hypoglycemia is obviously one of the things that’s the forefront right now, how often, how severe. Readmissions are another outcome. You know, how many patients are being readmitted within 30 days of discharge. Infection, surgical site infections. That’s kind of big among our, among our surgical colleagues. I think the measure you decide to look at really kind of depends on the question you wanted to answer. So, if we’re looking at institutional level data, things like patient day weighted mean, or, or being glucose as be appropriate, but if you wanna look at, provider behavior, I’ll show a couple examples of that.
[00:09:22] Curtiss Cook: I think looking at proportion measurements is more useful. So, why do we care? Why do we, why do we wanna measure inpatient glucose control? Well, if we decided, and I think there’s plenty of data out there that proved, that has shown this. If we decide or accept the fact that inpatient hyperglycemia is bad and inpatient hypoglycemia is bad too, then we would have a way to measure, track it and assess the impact of intervention or control measure.
[00:09:50] Curtiss Cook: The assumption here, of course, is that measuring glucometrics could somehow lead to, to increase safety or better outcomes. And then, but measuring glucometrics without applying glucometrics, no acting on the data in my views. A little, little, little use, little value, you know? But this is the part that’s gonna require a lot of institutional resources.
[00:10:13] Curtiss Cook: The big question though, which I think is unanswered: Do institutions who monitor glucometrics have better outcomes than those who do not? I don’t think there’s an answer to that question. I’ve not seen any data. I’ve not seen a randomized control trial that said, you know, hospital A who measures glucometrics has better outcomes than hospital B.
[00:10:31] Curtiss Cook: That does not measure glucometrics. I think that’s an important question. Cause I think, you know, if you’re good at convince hospitals to do this. Then you are gonna have to show some information to convince hospitals that, that the effort they’re gonna invest in this is really gonna be important.
[00:10:51] Curtiss Cook: It’s gonna have some, some better outcome than if they didn’t measure. Of course, as Dr. Messer’s pointed out, all this might be moot at this point because the need to measure now may be driven by regulatory expectations. He touched on those, just, just a second ago. So what are the goals of measurement?
[00:11:12] Curtiss Cook: What are the goals of glucose measurement in the hospital? Well, we wanna define a current state with respect to some desired statement people would call benchmarking. We want be able to monitor, if for when a process goes out, control, for example, statistical process control chart is one way to do that.
[00:11:28] Curtiss Cook: I’ll show you an example of that later on. We wanna be able to identify a broken process. Identify areas for improvement. We also wanna be able to follow up on interventions designed to improve hyperglycemia management.
[00:11:46] Curtiss Cook: But the term benchmark implies that there’s some industry standard against which a comparison can be made. It implies that there’s some threshold above which is bad, and below which is, is our target, which is good. And I just don’t think we’re, I don’t think we’re quite there yet. About what that benchmark should be.
[00:12:06] Curtiss Cook: There’s a lot of requirements for establishing a benchmark. For example, what is your standard type of measure? How you gonna calculate that, measure the numerator and denominator? What population are you gonna be looking at? How are you gonna report the data? What about the source of glucose sample? You know, blood glucose is not the same as a point of care glucose.
[00:12:24] Curtiss Cook: You really can’t be cooling that data, into a single denominator because we, we just, they’re different measurements. What instrumentation’s being used? If we are gonna look at point of care, point of care data, point of care, glucose data, we are using the same instruments across every institution. I’m not certain that’s that’s the case currently. I’m not sure if every hospital’s using the same instruments.
[00:12:44] Curtiss Cook: And we still gotta establish that benchmark. Some benchmarks that, you know, that, CMS have proposed, which we’re just reviewed. I’ll go of the more detail as we go along here in the talk. Here’s a good example of, of a national, I think, why we need a national standard.
[00:13:05] Curtiss Cook: So this is a result of a survey we published back in 2008 on 269 hospitals. And, this data’s not been replicated. It’s quite old at this point. I think some sort of follow up should be done. But we simply asked a question, you know, what, what are your, what are your target ranges and what’s your, what are you trying to achieve in terms of glucose?
[00:13:29] Curtiss Cook: We looked at both responses from for ICU as well as non ICU. I think you could see from this graph that responses were all over the place. In other words, these hospitals all had had different sorts of target ranges to aim for, basically a lack of a standard. Again, this is from 2008, things that may be different now than they were back then.
[00:13:50] Curtiss Cook: You know, data’s had more time to filter through the system, so we might see less variability than if the study were conducted now, that’s we were looking at back in 2008. Really? No, consensus in terms of what to aim for.
[00:14:07] Curtiss Cook: So what else do we need to do before we reach a benchmark? Well, what is most meaningful to hospitals that are stakeholders? You know, I know that, I know that CMS and other institutions are, are coming up with what they think is necessary, but is that really meaningful to hospitals? Is that what they need to know?
[00:14:24] Curtiss Cook: Is that what their stakeholders need to know? Is that what their hospitalists and surgeons need to know? I’m not so certain that hospitalists and surgeons wanna know more, know so much about what is the glucose level is, what is a glucose level of, So what is their risk rate of surgical site infections related to a glucose?
[00:14:39] Curtiss Cook: Who’s gonna look at the data? Is it some national organization looking at the data. Right now, there’s probably several different organizations looking at data, not a single organization. and what are the, what are the metrics again, we have to get back? What are the actual metrics we’re going to be looking at?
[00:14:56] Curtiss Cook: Cause the metrics that we, we decide upon, are gonna have to be within the capabilities of a hospital’s informatics and quality assurance functional areas. You know, does a critical access hospital in, you know, in, in the high country of Arizona, for example, which is where I am, where I’m in Arizona is a critical access hospital in the high country, the mountains of Arizona, you know, on, in one of the, IHS facilities, could they do this stuff?
[00:15:28] Curtiss Cook: Could they do these analysis? You know, can our VA here do these analysis? So, so whatever we decide upon has to be, has to be dual people, the institutions have to have the ability, to pull and pull and extract and, and slice and dice the data. And what are the consequences of not measuring to to a facility?
[00:15:50] Curtiss Cook: You know that right now the measures that are proposed are not mandatory. They’re, they’re really, among a list of things that a hospital can look at. What’s the consequences of not meeting a benchmark? You know, are you gonna ding a critical access hospital that’s already struggling financially because they’re not meeting a glucose inpatient, glucose benchmark?
[00:16:09] Curtiss Cook: We have to think about that kind of stuff, And we all know that achieving benchmarks is tough. Anybody who takes care of patients in the hospital with diabetes or hyperglycemia know that it’s really, really hard to achieve, a stable glucose control. Steroids are on the steroids are off the patient’s fasting, they’re not fasting, maybe on tube feeding than oral feeding. They’re septic and not septic. You know, there’s just all kinds of, of things going on in the hospital that affect the ability to really stabilize blood sugars in many cases.
[00:16:45] Curtiss Cook: So what are, what are actually, what are hospitals actually looking at? We just talked about some of the differences in targets, the variability in targets that hospitals reported on. Well, the state survey also asked the question on what are the primary metrics you’re interested in doing? And you can see here that that about, about 35% of hospitals, at least the time of the survey in 2008, didn’t really have any metric clue and of the ones that they were concentrating on were hypoglycemia. Hyperglycemia really is the top two, you know, followed by a bunch of other things, which are maybe not so much glucometrics metrics, but more outcome type measures. Such as infection rates, length of stays, ventilator socio pneumonia, things that are probably associated with the glucose control, but not necessarily measure glucose control itself.
[00:17:36] Curtiss Cook: That’s why I say that if you’re thinking about, about metrics you wanna keep in mind, what are the outcomes hospitals are interested in, in looking at. Here’s another benchmarking study that we did, a few years ago. This is in 2012.
[00:17:54] Curtiss Cook: We extracted data from 635 hospitals. This is all point of care glucose data, and you can see the, the sample sizes here. This is ICU, on he top graph, bottom graph is non ICU. Again we collected, we collected POC data. These, all these hospitals were on the same, device. I believe it was the ACU check three inform at that time.
[00:18:19] Curtiss Cook: We looked at percent patient days with hyperglycemia, looking at different cutoffs or hyperglycemia in the ICU, non ICU, you could see that probably about a third of these hospitals had, blood glucoses that were above 180. Of course, at that time, 140 to 180 was felt to be the target range for, for blood glucoses in the hospital.
[00:18:43] Curtiss Cook: So with this told us is that hyperglycemia was really prominent in US hospitals. I think some of the later data I’ve seen published from other institution. Kind of kinda look like this is about the same frequency as it has been even 10 years ago. What about hypoglycemia? So this is again, the same study, same hot number of hospitals.
[00:19:03] Curtiss Cook: Everything’s the same, same technology used to measure POC, glucose levels. This type looking at percent of patient days with hypoglycemia, again looking at different cutoffs. And you could see that, about 6% of, hospitals had hypoglycemia frequencies that were about 6%. And then of course the frequency of severe hypoglycemic was far, far lower than that I think may have some implications when you look at the CMS requirements.
[00:19:31] Curtiss Cook: So this told us that, you know, that hypoglycemic was actually uncommon in US hospitals. And even though it is uncommon, it is certainly, as I showed you in that previous graph, one of the biggest concerns that hospitals may have because, you know, hypoglycemia can result in acute complication, you know, seizure or coma particularly of severe.
[00:19:51] Curtiss Cook: So, but again, hyperglycemia, very common. Hypoglycemia not as common. This is the same study this looked at. This looked at mean glucoses, according to different hospital characteristics, according to hospital size, hospital types, so rural, urban, and academic and region of the US and we looked at the mean glucose levels and so when we adjusted each of these for the other two, we found there was significant variation, glucose levels across hospital size. The bigger hospitals having lower glucose control, hospital type with academic and urban, you know, bigger hospitals, academic hospitals having better glucose could show say than these rural community hospitals.
[00:20:38] Curtiss Cook: And then some significant variation. Across regions as well. And, yeah, I would’ve, I had a kind of a dream back then that we could develop this, compendium, you know, based on hospital size type and region. That a hospital can go reference to see how they fell within that standard.
[00:20:59] Curtiss Cook: But after this study, the program kind of kind of folded and we haven’t really done anymore after this point. So anybody interested in doing something like this, let me know. So that’s some of the work we’ve done looking at national data in terms of benchmark, in terms of where we are with glucose control right now.
[00:21:17] Curtiss Cook: Let’s talk about moving towards some of the more national things that have, that have come forward, in recent years. This, this is not exactly a glucometric benchmark, but this is the hospital acquired benchmarks. This is put in place by the deficit reduction hatch in 2005. As a means to cut, you know, cost to identify high cost, high volume conditions that resulted in higher DRG payments that could have been prevented.
[00:21:47] Curtiss Cook: And, and this went live in 2008 as part of the inpatient prospective payment system that Dr. Messer kind of mentioned a few minutes ago. And in 10 categories of conditions, one of which was related to glucose control. Again, not a glucometric benchmark, but they were interested in knowing which of these conditions actually occurred in the hospital.
[00:22:10] Curtiss Cook: Not that patients were admitted with these already, but conditions that occurred within the hospital. These are called to be preventative conditions ones for potentially which hospitals could be, could be dinged if they, if they occur.
[00:22:26] Curtiss Cook: So let’s talk into a couple of, couple about the national, benchmarking measures that, that are being put into place. Again, this is, Dr. Messer touched on this briefly in the beginning. CMS has developed two of these eCQMs, which stands for electronic clinical quality measures. They developed, they’ve developed two, one for hypoglycemia, one for hyperglycemia, but right now still voluntary reporting.
[00:22:57] Curtiss Cook: Yeah, you look at, when you look at their numerators denominators, I, I think it’s quite complicated. This is the one for hypoglycemia. So the population, their population are looking at is adults. They’ve had at least one antidiabetic drug administered well in the hospital, which really, quite frankly, is gonna be insulin for the most part, also includes patients in the ED and observation status and, instances where harm was suffered.
[00:23:25] Curtiss Cook: And if you look at their numerators glucose less than 40 and a diabetic drug administer within 24 hours and no confirmatory blood glucose was done within five minutes. For the result of greater than 80, only the first event counted as one event per hospital encounter. The denominator hospital encounter.
[00:23:47] Curtiss Cook: So when I look at this, you know, I look at this, I think to myself, I don’t, I don’t know what this means, quite frankly. I, I mean, I don’t how, I don’t know how hospitals gonna calculate this is where require a lot of data abstraction, a lot of qualifiers being applied to the data calculations being done.
[00:24:06] Curtiss Cook: I don’t know what they mean by harm. They don’t define harm. I think they also indicate in their, outlet, in their requirements could either a blood glucose or point of care glucose. Again, those are apples and oranges. You really shouldn’t, to my view, shouldn’t be mixing that data together into a single, into a single numerator.
[00:24:25] Curtiss Cook: I don’t know what they mean by harm. That’s not defined. Why do they only count a single event? I gotta tell you, as a clinician, when I round on a hospital, a patient with diabetes in the hospital, I, I look at how many events there were. Are they having events every day? That’s why what I used to make clinical decisions.
[00:24:42] Curtiss Cook: So I’m not sure why they’re focused on a single event. I’d be more curious to know what a total number of events, Cause I can guarantee almost every hospital is gonna have at least one. I’d be curious to know more about how many events are happening at a hospital. Why does one hospital have, you know, twice as many events as say as is hospital B does.
[00:25:00] Curtiss Cook: That’s what I’d be more interested in knowing. Why 40? I mean, as I showed you on the previous slide, at least based on our data from 10 years ago, the, the frequency of glucose is less than 40 is pretty rare. I’m not sure how much data they’re gonna find, with that kind of a low cutoff. I don’t know why, why they’re using such a low cutoff.
[00:25:22] Curtiss Cook: It’s gonna, I think, require to meet these requirements is gonna need some pretty sophisticated programming. I can tell you the work that we’ve done with the work we’ve shown is it’s required some fairly sophisticated programming just to, just to get some averages, hyperglycemia, this is their benchmarking for severe hyperglycemia, 18 or older either had a, have a diagnosis of diabetes or had at least one anti-diabetic drug, or at least one glucose, five more than 200. Again, three different criteria right there. You can defining, extracting your population that could be fairly complicated. The numerator all encounters with a glucose event within the 10 days of the encounter, minus the first to last 24 hours a day with at least one glucose of more than 300 or a day.
[00:26:16] Curtiss Cook: Glucose was not measured and it was preceded by two consecutive days where at least one glucose failure during each of the two days of one than 200. If you find that confusing, I’m. Because I think it’s gonna be really hard for, for institutions to figure this kind of programming out. And here the denominator is gonna be the number of hospital days here again, what is, what do they mean by harm?
[00:26:40] Curtiss Cook: I don’t know what they, they don’t define harm. They don’t really measure overall glucose control. in the, if you look at these guidelines, carefully, it acknowledges right up front. That there’s no accepted cutoff for severe hyperglycemia. It actually cites some of the work we did at Mayo, just as a support for the fact that people use different cutoffs for different definitions of hyperglycemia.
[00:27:04] Curtiss Cook: And again, I worried about the program sophistication needed to pool this kind of data.
[00:27:11] Curtiss Cook: Here’s another major benchmarking organization, National Healthcare Safety Network. I had a lot of trouble, finding data on this. In fact, I went to our colleague, Dr. Greg Mayard over at UC Davis, because he’s been very active in this area. He’s a beta site for testing these things that, hey, what, what is the update on this stuff?
[00:27:35] Curtiss Cook: And he couldn’t point me to any, any published or online references. But the NH, the National Healthcare Safety Network, is also delving into glucometrics. And so this first line here, it uses the same, the same definitions as eCQS for hypoglycemia, but it also includes some different cutoffs, which I think is good because I think it’s nice to stratify the data between, you know, mild, moderate and severe.
[00:28:05] Curtiss Cook: But it also includes other, measures for recurrent hypoglycemia, not just a single event, but repeated events they call a recurrent hypoglycemic day as an inpatient day with document hypoglycemia that’s preceded by another inpatient day within 24 hours where there was a hypoglycemic event. So here they’re attempting to.
[00:28:27] Curtiss Cook: Not just a single value is, not just a yes or no, but, but how many? And to me, I think that’d be more interesting comparison between institutions, just whether you had one or not. And they’re also looking at the, the median time between the index measurement and the first glucose of less than 70. So I think this is an important measure.
[00:28:47] Curtiss Cook: So just to digress a bit. You know, our hospital, I think probably most hospitals are going to have some type of nurse driven protocol for the recognition and treatment of hypoglycemia, and part of that protocol is once the event is recognized, probably to get a repeat value, then to treat in 15 minutes later, to get a second value and to repeat that cycle to the glucose is above 70 on two occasions.
[00:29:16] Curtiss Cook: That’s how our protocol is here. I didn’t think I’d have time to show this data, but we just published a paper that looked at compliance with that and nurses are struggling. They really are struggling to stay within that 15, 15 minute timeframe. You know, it turns out that the timeframe between the index event and, and the glucose above 70 is probably more like 20 to 25 minutes, and so I’m not sure whether just looking at less than 40.
[00:29:45] Curtiss Cook: When somebody’s less than 40, they’re probably gonna be unconscious to need IV, dextrose. So I’m not sure whether they’re looking, looking at 40 I, I’d prefer they just look at everybody with hypoglycemia less than 70. But these are the NHSN proposed metrics.
[00:30:04] Curtiss Cook: So let’s talk about what I call applied glucometrics. You know, what do you, what does a data mean? What do you do with it? And one of my interests really has been over the years provider behavior with respect to, uh, glucometric data, you know, within the hospital. You know, our, our, provider behavior really has to do with insulin therapy and are we applying the right insulin regimen in the hospital to the appropriate glucose level.
[00:30:33] Curtiss Cook: And so this is one of the very first papers we published back in 2007, from our hospital here in Mayo Clinic in Arizona. We just simply asked a question, Okay. How does, how is insulin therapy being applied with worsening levels of hyperglycemia? So we, we have here on the, on the bottom here, the different turnstiles of mean glucose, uh, and these are the, these are the averages in each toile and the black bars represent.
[00:31:02] Curtiss Cook: Basal bolus insulin therapy, which is still the optimal form of, treatment in the hospital. And the gray bars are, are, are bolus only the so-called correction or sliding scale. And then we ask the question, Okay, how does this vary? How does this therapy vary with respect to worsening sever of hyperglycemia?
[00:31:21] Curtiss Cook: So, so the one hand, we saw some very good news here, is that, is that this is a severity of hyperglycemia worsened. Basal bolus insulin therapy used increased, which, okay, that was encouraging. But then we looked at the other top half of this graph and we saw that about, about 40, about 50% of people were still on bolus insulin therapy, despite, despite having severe hyperglycemia.
[00:31:47] Curtiss Cook: And we saw that as a negative. We turned, we used the term, we applied the term clinical inertia to that, to that situation in the hospital. So showing some evidence of clinical inertia. When it comes to inpatient care, the clinical inertia has often been applied to, to outpatient management of diabetes.
[00:32:04] Curtiss Cook: But I think we are the first ones to apply this concept to the inpatient setting. So we clearly have some work to do here or had some work to do here. Here’s a nice study we do with our surgeons. Our surgeons are all over hyperglycemia. They don’t like it. They don’t like the infections, they don’t like the wound existence.
[00:32:23] Curtiss Cook: So a few years. We entered a collaborative study with them because they were interested in knowing and better applying insulin therapy to their postoperative inpatients. So we, we did a baseline study, similar to the one I just showed you, except here we use measurements, not means, but measurements.
[00:32:46] Curtiss Cook: What percentage of percent of measurements were than 180? Again, I like measurement data. When I’m looking at provider behavior, I like to look at measurement data. When I’m routing in a hospital, I’m looking at a diabetes patient. I’m not sitting there calculating patient day weighted means or means from the previous day.
[00:33:02] Curtiss Cook: I’m looking to see how many values were high and how many values were low the previous couple days. That’s what I make decisions based on, not based on means. So in this case, we’re looking at the use of insulin therapy with respect to increasing frequencies of hyperglycemic values are more than 180. And again, this is the baseline study and you can see here the black is basal bolus, the white is bolus, and the gray is no insulin.
[00:33:28] Curtiss Cook: And you can see that across increasing frequencies of hyperglycemia. There was absolutely no change in provider behavior. Insulin therapy did not change at all. So what we did, the way we responded to this, this is how I think the glucometric data ought to be used, the way we responded this is we came up with a quality improvement project.
[00:33:48] Curtiss Cook: Where we educated our surgical residents on proper use of insulin therapy. And given an algorithm, I went to start applying basal bolus insulin therapy. And which you can see here is a follow-up analysis. Shows still not optimal, but look at the big improvement use of basal bolus therapy with a rising, frequency of hyperglycemic values.
[00:34:11] Curtiss Cook: And when we looked at the hypoglycemia data with this, there was no increased risk of hypoglycemia frequency did not increase. Here’s another study that kind of shows the difference between the value of having an APP or NPPA involved with day-to-day management diabetes compared to, not having somebody involved.
[00:34:35] Curtiss Cook: So what you have on this, on this figure here on the left side is use of basal bolus insulin therapy, according to, again, toiles of, of frequency, of hyperglycemic measurements, and pretty good response. You know, providers. We’re responding, but this is what it looked like when we actually had an endocrinology NPP actually involved with care of the patient.
[00:34:58] Curtiss Cook: You can see almost, almost a hundred percent use of, of basals insulin therapy with the highest levels of hyperglycemia. Kind of makes an argument that you may want to have some specialists, specifically dedicated, to a glucose control service in the hospital.
[00:35:15] Curtiss Cook: So the last 10 minutes, let’s talk about Just moving, I like to say moving beyond glucometrics Okay. you know, a lot of our data is, is just a lot of the glucometric data. Everything we talked about really is retrospective. You know, it’s looking back and, when you, when you’re looking at retro, when you’re looking to data retrospectively, it may be too late.
[00:35:40] Curtiss Cook: By the time you analyze the data, there may be a process already out of control. It’d be nice if we could forecast, potentially forecast and predict the state of inpatient glucose control. That way we could potentially anticipate changes as opposed to waiting for something to happen discovering it later on. Now one approach to this is to apply mathematical models to existing, to, to retrospective glucose data and use that data to project forward. And this is actually a, a methodology used in industry. Things like supply chain management, diffusion of products, credit risk accounting and finance, and the mathematical model that uses something called damned trend, exponential smoothing.
[00:36:31] Curtiss Cook: And I’m not a statistician. I’m lucky to have somebody here who, who knows how to do all this. So what we did is we, we took point of care glucose data from our laboratory information system and we had a fairly large number of pieces of data. Over 55,000 pieces of patient day weighted mean data. We looked at 60 to 63 weekly observations and then took those observations, applied the damn print analysis to look forward at different different time points.
[00:37:00] Curtiss Cook: This measure called the mean absolute percent error is what we use to compare predicted against observed and it’s felt that a map of somewhere between 10 to 15% is pre reliable. There’s a good reasonable representation to underlying data. Now, before I show you the results, cause I need to show you a floor layout of our hospital because it’s important for the spatial analysis.
[00:37:23] Curtiss Cook: So currently we have two towers, an east and a west tower. Although right now we have a third tower, a hundred bed tower under construction. It’ll be the future west Tower, just become the central and east. But we have seven floors to our hospital and we have an east wing, west wing, east side, East Tower, West Tower.
[00:37:44] Curtiss Cook: And each of these wings, each of these floors we have, three pods. And each of these pods is comprised of 12 rooms. So we’re looking at forecasting data. This is what the forecasting data kind of looks like. We’ve published a couple papers on this. So this is just one quarter, quarter three of 2017 that represents the entire hospital.
[00:38:08] Curtiss Cook: These little dotted lines represent, observe, patient day weighted means, and the solid gray line represents the, damn trend or predicted patient day weighted means. So this is the retrospective data we use to. Forecast the future data. And you can see here that, that the predicted and the observe lines were right on, fell right on each other, very low map and a very, very tight prediction interval.
[00:38:36] Curtiss Cook: So then we asked the question, Okay, how low can we go? You know, how detailed can we go on these analysis? How could we go down to the floor level, the pod level? How far could we go to get meaningful data? So we next looked at the, just strictly the fourth floor, same quarter, same year. And we found really good data, very good agreement.
[00:38:58] Curtiss Cook: And then we got down to the wing level, just a wing, and can see you start losing predictive predictability here. Start seeing large interval between the predictive values and not much agreement between, the predicted then the observer and got even worse when you started looking just at the, this is actually the, a single pot.
[00:39:20] Curtiss Cook: So we were to do any forecasting would be probably at the level of the, of the higher, either the higher institution or an entire floor. So that’s a different way of applying glucometric data. The other thing is spatial analysis. You know, spatial analysis is basically reporting data on a map, and it was first utilized in 1854 by Dr. John Snow, and he was mapping collar cases in London, and he eventually pointed the source of an outbreak to a water source near a local brewery. These days we use geographic information systems. And we use, geographic locators such as zip code and latitude and longitude to overlay maps. And here’s an example, CDC publishes these counter level county level diabetes and obesity maps, you know, annually.
[00:40:11] Curtiss Cook: Now, of course we don’t have latitude and longitude in our facility. You know, we, so we had to find some other way to geolocate the glucose data, and overlay it on the hospital floor map. So in this case, we, you know, we use basically, used room numbers to identify the location of the glucose data.
[00:40:30] Curtiss Cook: And so this is basically what it looks like. So this is third floor. West wing, East Wing, fourth floor, West Wing, East Wing. These great areas represent non-clinical areas and so forth. So we divide, these are patient day weighted means divide into T tiles, and you could just see just by looking at this map that there’s some areas that are doing really well in the green.
[00:40:52] Curtiss Cook: Some areas that are not doing so well in terms of glucose control, they’re in a red, and other areas, it might be somewhere in between. So we could potentially go in and look at this day. Okay, what’s going on over here in this, in this pod over here on the west wing on the third floor. Same thing for hypoglycemia.
[00:41:10] Curtiss Cook: This is patient. Percent of patient days with hypoglycemia, again, divide into total tiles. Again, you see some areas that are really hot, a lot of hypoglycemia. Other areas are looking pretty good in the green. Some areas are in between. So this is kind of, this is kind of what we can use spatial analytic techniques, uh, for use, looking at glucometrics in the hospital.
[00:41:31] Curtiss Cook: So a quick, easy reference. The very last thing in the last three minutes is a concept of what I, what we called insulin metrics. I’ve kind of hinted to you earlier already that I don’t think glucometrics can just be considered in isolation. We gotta consider glucometrics in with, with respect to outcomes, but also how are we managing it?
[00:41:53] Curtiss Cook: So we published this term back in 2017, it, we called Insulin Metrics, which we define systematic analysis, reporting inpatient insulin therapy. So pretty much the same definition for glucometrics except substitute insulin for glucometrics. It’s pretty much patients in the hospital these days are pretty sick and they’re probably gonna need insulin therapy to control their blood sugar.
[00:42:17] Curtiss Cook: So we need to measure in a basal bolus insulin therapy is the most effective regimen that we need to be able to figure out a way to measure that and how it’s being applied to therapy. We could assess for clinical inertia is, I already, pointed out. Now insulin errors I think is. You know, the Institute for Safe Medical Practices still identifies insulin as one of the top medication errors in the hospital.
[00:42:43] Curtiss Cook: So two proposed insulin metrics could be insulin errors, and again, how insulin therapy being applied and modified to a level of hyperglycemia. Here’s the example of how we look at insulin errors. This is a statistical process control chart, that looks at, 2012, 2016. And you can see there’s a big difference.
[00:43:03] Curtiss Cook: And so this is the control limit here in Upper and lower Bounds here. And you can see there’s a big difference, a lot of errors in this timeframe here. And suddenly they drop, Well, what do we do different? Well, we instituted as a barcoded medical administration process. We went from syringes, we went from vials to pens, slapped a uh patient label on every pen that was scanned by the nurse.
[00:43:26] Curtiss Cook: And with that, we saw a huge reduction in insulin errots.
[00:43:32] Curtiss Cook: And here’s an example. This is one of the metrics that we track in our system. this is use of basal bolus insulin therapy according to a portion of patients with a patient day weighted mean of greater than 180. And so what this says is that over time, about 60%, about two-thirds of patients that have blood glucose levels, above the 180, were receiving a basal bolus insulin therapy.
[00:43:58] Curtiss Cook: We don’t know. We don’t go into what the situation was, why it wasn’t higher. It stayed pretty flat. We haven’t made a whole lot of progress on this. I’d be curious to know why. That’s another way that we could measure insulin. So where do we go from here? at one point I think, I hope I, I’ve driven home and credit, kudos to organizations that are trying to do this, but I still think we have a long way to go before we reach what I call glucometric harmonization. We still have different organizations proposing different things measured in different ways in different population subtypes. We still don’t know what’s going on in different hospitals and what they are, are not doing, what the capabilities are.
[00:44:40] Curtiss Cook: We gotta know if we wanna measure point of care versus blood glucose. Point of care gives us the most data, but glucose is the most accurate data. Our, hospital is using the same instrumentation. If hospital A is using one glucometer device and hospital B is a different glucometer device, you can’t pull that data.
[00:45:00] Curtiss Cook: And what’s the source of the POC? You know, in the ICU our nurses will, uh, will draw blood from an arterial line and dab it on a, on a, on a, on a strip to measure the glucose. Well, as a arterial line glucose, point of care glucose, the same as a CAPA rate, point of care glucose. Again, how standard, how standardized that it has to be.
[00:45:21] Curtiss Cook: What is the institutional support we’re gonna get? You know, you really need a good informatics and a good quality department to. To really, get where you need to be. And in lacking that, it’s gonna be very difficult for hospitals to achieve the kind of reporting that’s gonna be desired. And again, I’m not sure hospitals gonna wanna just simply know. They’re not gonna wanna know what the glucose state is. They wanna know how does that affect my outcomes, You know, how does that improve my surgical site infection rates? How does that improve my mortality? They’re gonna want to know that stuff. So not, not, I don’t think they’re gonna want to know number.
[00:45:55] Curtiss Cook: They wanna know what does that number mean? What’s the significance of that number? And also, policies and processes are important. It’s not just the measurement, you know. Many years ago we published some of the first guidelines, uh, for instance, on, on use of insulin pumps in the hospital. We’ve got very detailed, very detail policy on recognition, treatment of hypoglycemia. We now have a policy on hypoglycemia prevention. You know, what do you do? Don’t wait for the blood sugar to drop so low. So it’s not just processes and policies are important. I think it’d be nice to know on a national level who’s got what. And, and because that’s really what drives care really is, is, is what is your policy, what is your, what are your processes?
[00:46:38] Curtiss Cook: So those are some of the thoughts I have about where to go in the future.
[00:46:48] Jordan Messler: That was fantastic. You can leave it up there. thank you so much Dr. Cook, we really appreciate you sharing your expertise, experience, insight with all of us tonight.
[00:46:58] Jordan Messler: There’s so much potential and opportunity with Glu Metrics, insulin metrics. To drive change, overcome clinical inertia, ideas to help get us to, I love that phrase, glucometric harmonization. Thanks to all of our attendees for joining this session. We’re very grateful for you choosing to spend the time with us today.
[00:47:15] Jordan Messler: Unfortunately, we don’t have time for the live q and a, uh, but if you posted questions in the chat, we will follow up directly with anyone who submitted their questions. There is still one more session left tonight at 5:15pm Eastern Time. We have some experts on the Glytec team. For our day two post show huddle.
[00:47:31] Jordan Messler: Be sure to tune in to their key takeaways from today’s sessions. So again, on behalf of the Glytec team, thank you again to Dr. Cook. Thank you so much and to all of our audience today. Enjoy the rest of your evening. Take care.
MAR-0000939 Rev 1.0
