MONAHRQ's Preventable Costs and Hospitalization Mapping Tool (continued)

Webinar Transcript

The following is a transcript of a technical assistance conference entitled MONAHRQ's Preventable Costs and Hospitalization Mapping Tool held on July 21, 2010.

John Bott: Thank you. Now we will move to talk more specifically about one of the modules within MONAHRQ, the Preventable Hospitalization Costs and Mapping tool. One reason for the creation of such an instrument is there is quite an opportunity for improvement, as we read almost daily in health care. In regard specifically to the Prevention Quality Indicators from AHRQ, in running those we see an annual national cost of slightly over 30 billion a year, and those are 2006 figures to take into account, so potentially it could be greater in today's dollars. This also represents issues of prevention, primary care, and sufficient access in care coordination. If those were in place, that could potentially translate to avoiding some of these hospitalizations that we're talking about measuring in this module. 

To really be successful in analyzing and understanding where the opportunities are, we need greater information, and that's what we're trying to convey with such a tool to provide some drill-down capacity as in what types of care have the most room for improvement. We'll talk about that further and see a live demo of some of those activities. 

A couple of these points may be slightly redundant with what Anne said, but as a good teacher once said, you need to say things five times for optimal education, so some of these may be slightly redundant. As Anne noted, the AHRQ Quality Indicators are based on electronic hospital administrative discharge datasets that draw on data elements that are already contained in that administrative claim. 

Basically, there are two compartments of the Quality Indicators. One is measuring the hospital's performance directly, and one is using this hospital data to understand what is occurring in the community, and, in this case, the State or the county level. This discussion about preventable hospitalization events is on the latter. It's using that inpatient activity to understand the portal for what's occurring in the community. These are what we call oftentimes "area-level measures," whereas a number of the other AHRQ QIs are hospital-level measures. 

A little bit of history on this prevention hospitalization component. In 2007, AHRQ launched this Preventable Hospitalization Cost Mapping tool with area-level Prevention Quality Indicators, and it provided the hospital's cost with the calculation of these measures. 

Then last year, seeing that MONAHRQ was coming down the pipeline, the decision was made that that tool would be integrated into MONAHRQ. As noted by Anne, MONAHRQ is the coming together of a number of tools, so this tool was rolled into MONAHRQ in 2009 and 2010. Just to reiterate, this particular tool is no longer available as standalone software on AHRQ's Web site. 

Essentially, this module of MONAHRQ has two capacities; it creates maps and it creates tables. The maps show the rate of hospital admissions by quintile for selected health conditions that we'll look at here in a minute. Then, it also produces the hospital's cost for those types of hospitalizations by county of residence for the person and statewide as well. 

Just a couple of notes on these cost calculations, this aspect of MONAHRQ. The cost is the hospital's cost, so it's based on cost-to-charge ratios to reflect that given hospital's cost and this is using data that the hospital submits to Medicare at the end of the year as part of their accounting function. But, of course, I take it there are payers out there to some degree on the phone or the Medicare commercial, so what one actually pays could, of course, be different than the hospital's cost, so when you're figuring what your costs are, you have to have that enter into your equation. Note that, of course, the costs are based on the period of time that that data is used, and like most things in life, there's inflation and cost raises over time, so if you are trying to understand it in term of today's dollars, one would need to consider that. 

These cost-to-charge ratios are a calculation that's performed on the inpatient component only of the claim. Of course, there's a professional component of the claim as well to understand the full hospitalization cost that's not present in these figures that we'll look at and that's in part of this mapping tool product.  

So, what are the measures inside this mapping tool? It lists all of the AHRQ Prevention Quality Indicators [PQIs]. Here we show the PQIs by a couple of clinical areas, some chronic lung conditions, four diabetic measures. Several heart conditions are in the family of the Prevention Quality Indicators. Then there are five other Prevention Quality Indicators that don't fit nicely into one of those clinical areas, such as UTIs and admissions related to dehydration. 

In addition to the AHRQ Prevention Quality Indicators within this mapping tool module, there are some other area-level measures as well. You will see here there's a number of AHRQs Inpatient Quality Indicators that are area-level measures: CABGs, PTCAs, hysterectomy, laminectomy. Then there's also a number of the AHRQ area-level Patient Safety Indicators, and there are about seven of them there that you can see. Altogether, we have in the upper 20s, if I'm counting correctly, of area-level measures that are currently part of this particular module. 

There are also three composites, which are becoming more popular for people to understand performance in a succinct way, and these are a couple overheads to express what the composites are. One is all 14 of the Prevention Quality Indicators. Another is the acute care composite, and you'll see what they are here, dehydration, bacterial pneumonia, UTI. Then there's a third composite that are of chronic conditions such as diabetes, heart failure. In our demo we will look a bit at performance by composite and performance by individual indicator. 

Just prior to moving to the live demonstration, I have one note on closed captioning. For those of you who would like to continue to view the closed captioning, when the demonstration is initiated, your captioning panel will drop from your view and the toolbar will appear in the lower right-hand corner of your screen. If you wish to reopen your captioning panel, click the icon furthest to the right on the toolbar. There are two arrow buttons in the upper right corner of the captioning panel. First click on that right button, then the left one. This will increase the amount of text in the panel.

Now we will move to the live demonstration. This is the initial home page you would see within MONAHRQ, and the module we're talking about at this time is here in the lower left-hand corner, maps by county showing potentially avoidable hospitalizations. This is synthetic data for the State of Maryland and the State is interested in looking at the types of costs and variation by county for potential savings. So perhaps one would look at the overall map of the composite of the Prevention Quality Indicators. 

Here is the composite of all 14 Prevention Quality Indicators. Let's say, for whatever reason, we're interested in Anne Arundel County. They are in the fourth of the fifth quintiles, so there could be a logical opportunity to look for potential improvements there, since there's room for improvement. Then you have the opportunity in the upper right-hand corner here to look at a summary table. To walk you through just a couple of the columns in the summary table, I think it's probably fairly intuitive for folks looking at Excel files since it is probably more friendly than your typical Excel file that I've seen. In Anne Arundel County, we see there is an observed rate, and this is per 100,000. We see what the risk-adjusted rate is, appreciating demographics of people in that given county, and we look at cost savings. It expresses it by cost savings 10, 20, 30, 40, 50 percent, if we saw a reduction in the utilization of these potentially avoidable hospitalizations. We see in Anne Arundel County $18 million, if you're looking at the 20 percent cost savings, just to provide an example. 

Well, unfortunately, we'll have to exit the live demo. I apologize for the time it's taking for those maps to come up; it usually just takes a fraction of a second, but apparently we're seeing some slowness in the server response time.

These are just a couple of screen shots of that module. Before, in the live demo, you saw the listing of the measures organized in a pretty logical way. Here they are by certain clinical conditions, again, a map of all of the counties in Maryland and performance in a given measure. This one happens to be a Patient Safety Indicator, accidental puncture or laceration, again, divided into quintiles. 

One thing here I was going to note on the live demo is, when you see a map like this, you can very quickly get a sense of the degree of variation in this given potentially avoidable hospitalization measure. You can see the lowest quintile has a rate of 10 per 100,000, the highest quintile being a rate in the 90s, so we have ninefold variation across the State. One may want to quickly move through the maps as an example to see where do we see a lot of variation if we might be interested in a project across the State to have some cross "pollinization" across the counties to learn from one another. 

This is another table, and we essentially covered an example of these tables in the live demo that shows observed rates. In particular, this is a measure that is not risk adjusted, so you'll see that column blank, and the opportunity for savings from 10 to 50 percent if hospital admissions of this type were reduced. 

A State may be interested in, "Well, we can understand that's our State's performance and the county's performance within our State, but is there an opportunity to understand ourselves against another picture?" Through HCUPnet, and this is the initial foray into HCUPnet, the link that you see at the bottom, a State can look at their performance against the Nation. A State can also look at their performance against their particular region that they happen to sit in or where they are at. Just to note, to be sensitive to the fact that HCUPnet will provide performance across a number of years. So if you put in, say, 2007 data, you'll, of course, probably want to look at that against, say, the national 2007 data as well.

Some examples of potential uses—and at AHRQ we keep a tally of case studies. So they do have some very specific examples of what you see on the screen for most of them. Targeting priorities, there is an example in the State of Connecticut, the Office of Health Care Access, and their hospitals are using—and I'll talk about the PQIs or the PQI mapping tool—the hospitals are using such measures. In community health centers they're using such measures in local health departments to design community outreach services, particularly for those for the care and management of chronic illnesses such as diabetes and asthma. 

In regard to the next potential use, I have calls into a couple of States, and I haven't heard back from them, so I wasn't sure about that particular example for the second bullet. For the third bullet of identifying best practices, in the State of New York, their Department of Health, we did a case study last year and found that several new interventions for the geographic areas were identified by the AHRQ Prevention Quality Indicators. They are now in operation in the State of New York as a result of their recent legislative reform. For the first time, Medicaid in that State is paying certified diabetes educators and certified asthma educators to work in clinics and physicians' offices to provide self-management education. 

In regard to the next bullet, public reporting at various geographic levels, people may be familiar with the Commonwealth Fund report that they put out annually. Essentially, what they do is they take 37 indicators, and a number of the AHRQ Prevention Quality Indicators are included, and they report on several different rubrics. The two rubrics that the Prevention Quality Indicators are used in are in regard to equity and efficiency, and they make composite statements then for where a State is placed using the AHRQ Prevention Quality Indicators. 

The last example here is in relation to more broadly establishing budgets. The AHRQ tool reinforced what General Motors was seeing in their internal reviews in what areas they might focus on in relation to their employees and retirees. They were considering phasing out some of their efforts that were related to diabetes and asthma. Based on the data that they were seeing and the expenditures that they were paying due to use of the software, the data reinforced and highlighted for GM that they, in fact, do need to stay the course—that was one of their quotes—and stay invested in prevention activities or well-care activities related to diabetes and asthma. 

There are actually a couple other uses in combing through the case studies, after I developed these slides, that I saw. The measures are also used as a basis to identify areas for quality improvement for stakeholders to work on. GM worked with a number of coalitions in their area to express the fact to coalitions in their area that we all have such common interests to work on, and it served as a galvanizing factor to work across a number of stakeholders. Also, the measures have been used to cut the data to look at these rates of hospitalizations by race and ethnicity. An example of that was the State of Connecticut; again, the Office of Health Care Access looked at the measures in that capacity. 

Looking at future work, where it related to this particular module, a quality improvement toolkit is currently in the works to assist areas, States, and counties in reducing potentially avoidable hospitalizations as defined by the AHRQ Prevention Quality Indicators. The anticipated date that will be released is early next year, early 2011. Second, there've been some requests to look at the results by not only county level but by ZIP Code level, and we're doing some analysis to see if that would be feasible with this tool.

Last—and Anne may have noted this in her presentation—this particular module is adding the remaining AHRQ area-level Quality Indicators. So with that I'll turn it over to Margie.

Margie Shofer: Thanks, John. We have entered another Q&A session, so as you guys know very well by now, there are two ways you can submit questions. You can type your inquiry into the "Q&A" box located on the right-hand tool bar of your screen, or you can click the "Raise hand" button, and actually, I believe most of you have a phone icon next to your name, so I really encourage you to click the "Raise hand" button and we can unmute your line and you can ask a live question.

John, I do have a question for you. "How do you account for variation in small numbers and counties from year to year?"

John Bott: One thing a user definitely needs to be in tune with is the fact that these are relatively infrequently occurring events, so it's really at the user's discretion at what point do they want to suppress cell size. Anne, are there any cut points where the cell size is suppressed within MONAHRQ where MONAHRQ just will not deliver results?

Anne Elixhauser: The cell size suppression is actually something that's controlled by the host user, so you can get a cell size of whatever you want. Some States actually have legislation that tells them or requires them to report no data on less than 10 cases or less than 3 cases, whatever it may be. There is an option in MONAHRQ as you're doing your Web building to input that. 

Just in terms of how to handle that analytically, one of the things that we've been talking about within MONAHRQ and that is on what we refer to as our "future enhancements list" is allowing users to input multiple years of data to analyze that data simultaneously. Not to look at print but to aggregate the data across 2 or 3 years, and that is something that is not going to be in MONAHRQ Version 2.0, but we're hoping it is something that will be included within 3.0.

John Bott: We just realized we actually answered a couple of online questions inadvertently.

Margie Shofer: I have another question for you. "When you say you will be adding other measures in time, what are the area-level measures you are referring to? "

John Bott: That's the Pediatric Quality Indicator area-level measures, and those contain such things as diabetic-related admissions, and UTIs are a couple of the examples, so I believe that's five additional area-level measures.

Margie Shofer: I have another question. "If you're trying to understand what was the full cost of the hospital for a given hospitalization, do you have a ballpark figure as to the additional cost for the professional component of the admission?"

John Bott: A senior economist here within our center, Bernard Friedman, has done some work in this area with data that's a few years old. It's not published, but from his work, and again this is internal analysis, we see that for the given Prevention Quality Indicators we're talking about is, on average, the professional fees tend to be between in addition of 12 and 18 percent. Again, that's for these types of particular admissions. If they were types of other more complex admissions such as a transplant or a CABG, that percentage may be higher, so it was specifically for looking at the Prevention Quality Indicators; that would be a ballpark figure. 

Margie Shofer:  Well that is it for the questions for right now. I'm going to turn this over to Brooks.

Brooks Daverman: All right. Hi, everybody. My name is Brooks Daverman, and I work for the Tennessee Division of Health Planning. I'm presenting on our use of AHRQ's Preventable Hospitalization Mapping tool, which we used in our State Health Care Report Card, and I'll tell you all about our State Health Care Report Card. The first version came out in March 2009. We're working on version 2 now. It should be coming soon. The first version looked at diabetes and hypertension, and the new version is going to also include asthma. Basically, what this report is county-level maps, and I'll show you more about that. 

A little bit about why we did this report. We had a group of stakeholders get together to talk about quality of health care in Tennessee, a pretty distinguished group. One of the outcomes was somebody proposed that we would just look at an easy first step of what can we do given what we have right now? Everybody agreed to look at chronic disease and the treatment of that first because that is a major issue in the State of Tennessee. The idea was to pull together all the data that we have available now, whether it's in the State or in other places, any measures that we have available now, and try to report on the quality of care at the county level. 

We chose the county level because we wanted it to be local, but we didn't want to get into provider-level reporting on quality of care yet, just because, first of all, the attribution is difficult and sometimes expensive, and also because there's always a lot of controversy at the provider level. So we wanted to look at the county level and see if we could get a little more granularity on some of the information that we had at the State level.

We also saw this report as kind of a test, maybe a first step toward Tennessee getting an all-payer claims database. Actually, in June of 2009, the legislature passed a law to make Tennessee an all-payer claims database State. What that means is we've always been collecting hospital discharge data, but now we're also starting to collect data from insurance companies, all of their claims, which shows us not just hospitals but also providers in a doctor's office or claims related to pharmacy. We're one of 10 to 12 States now that are collecting that data, and we started having that data flow in June. This report, the Health Care Report Card, was a first step, using just hospital discharge data and some other data that I'll talk about on the next slide.

There are three types of data in the report. First, there's BRFSS survey data, then there is HEDIS measures, and basically, we had major insurance companies voluntarily submit HEDIS measures, numerators and denominators for each county, and we kind of aggregated them. We also have HEDIS measures on everybody in Tennessee's Medicaid, which is also called "TennCare," so we aggregated that in there too. The third type of data was the AHRQ mapping tool using the hospital discharge data, so I'll talk about that last one the most. 

First, here is the map that we included based on BRFSS data on whether or not you've been told by a doctor that you have diabetes, so this is based on survey data. We couldn't get it down to the county level because it's a survey, there's not that many people who are being called and answering that question. We do have the regional level at least, and there is variation. Of course, before you note the variation, you've got to note that a lot of people in Tennessee have been told by a doctor that they have diabetes. It's a major problem in our State. 

I'll go on to talking about the HEDIS data. These are some of the screening measures from HEDIS that we looked at, HbA1c testing rates and then LDL-C screening rates for both diabetes and cardiovascular disease. I'll just show you one example of the kind of maps that we generated with the HEDIS data. This is the HbA1c testing rate, and you can see that there is variation in the level of quality. This is a measure of quality, whether or not somebody with diabetes got their HbA1c test when they went in, or whether they got it at all, I guess. 

Another thing to see is that Tennessee is a State where when you look at the county level you really do get good information because we've got 95 counties, and they're all about the same size, and it's really nice. It works really well with the AHRQ mapping tool. We did a bunch of other maps with HEDIS. These are the other HEDIS data that that we included.  

We also had some statewide HEDIS data that we couldn't get down to a county level because we had four major insurance companies that participated that cover a lot of people in the State, a high percentage of people in the commercial market, and we had them not send the HEDIS results for every person. We had them just send the county-level numerator and denominator for each of the measures that we could add together, so it's not like a scientific survey, but it is getting a lot of people when you put the major commercial insurance companies together with Tennessee's TennCare, which is Medicaid. The condition control measures in HEDIS are based on a survey of charts, so it doesn't have the same level of granularity, so, again, we couldn't do the county level; we could only do a statewide number. 

Now I'm showing you the first of the maps that we generated using AHRQ's Preventable Hospitalization Mapping tool. This was short-term complications of diabetes and uncontrolled diabetes put together. You can see that there is variation across the State. We colored these by quintile, with a risk-adjusted 100,000 population. In West Tennessee, there are a couple other groups of the darkest blue that's sort of the worst, and I'll just jump ahead to the next one. 

We've got five maps that we generated around diabetes and hypertension. This is long-term complications, and in West Tennessee, around Memphis, there is a concentration there. There are a couple other areas of concentration. There are a couple areas that have better than average. You can see the same thing with our lower extremity amputation rate for diabetes that we ran, and also pediatric diabetes is a little bit more mixed. You don't see that concentration right around Shelby County, which is Memphis, Tennessee, in the southwest corner, but there are some areas with a higher level. There's that same variation. Finally, this is the hypertension admissions rate, and that has a big concentration in the northern/northeastern section of Tennessee, and it's in a couple other areas as well. 

I just want to show you also that this is just a portion of one of the tables that we included in the appendix that we ran, showing the actual numbers for people, and we showed them for 2005 and 2006. We had little arrows that you could see if there was a statistically significant difference, either better or worse, up or down arrow. Across the top, we showed the benchmarks of Tennessee as a State, the U.S., but we also used that regional number of the South, which was really helpful because at least in Tennessee, everybody always asks how we're doing regarding the whole country, but they also want to know how we're doing with regard to just the South. That's an important metric for people around here. 

A question we often get with this report is, "You did the report. Did you take any next steps? Did you do any interventions based on what you found?" And we didn't. We just did this as a first test case, and we ran the numbers as far as they could be done. We're going to update them again for this year. In the future, we're really going to look to the data that's starting to come in. We've just got one month coming in now, but we're starting to collect that all-payer claims data, and we want to use that, in addition to the hospital discharge data, to try to answer some of the questions that you see when you look at a county and you say, "Well why does this county have such good numbers or such bad numbers?" 

Usually you're asking about the bad numbers. "Why does this county have such bad numbers with regard to amputations or long-term complications of diabetes?" We take that by looking at the whole State, not just the voluntary data and not just the hospital discharge but the all-payer claims data as well; maybe we will be able to answer some of those questions. We're designing the reports now that will try to answer some of those more detailed questions. So that's my presentation, and I will open the floor to any questions.

Margie Shofer: Thanks, Brooks. We are now in the third and final Q&A, and as I think you guys know really well by now, there are two ways you can ask questions. You can type it into the "Q&A" box and you can also raise your hand, and we will unmute your line and hear from you live. So, Brooks, I'm curious to know, what was the response from your report?  

Brooks Daverman: The first response everybody always has is they want to look at their own county. When you get past that, I think that with our legislature, there was some interest generated in some of these measures and kind of looking at them, and not only looking at the variation but also thinking about how Tennessee is doing with regard to the whole country and the South. 

What we found was that in some of the quality measures, we actually do a pretty good job in comparison to some of those benchmarks. Tennessee providers get it, and I think actually because they see so many patients who have some of these conditions, they know what to do when they see it. But at the same time, because we do have so many people with some of these chronic conditions, it's even more important that they get the right care. 

Those are some of the conversations we had. We were really using it as well to say we could get even more information if we were collecting all-payer claims data, so some of the reaction was that we did pass that law to get all-payer claims data starting to come into the State. Then we have some programs in the State that are trying to address some of these chronic diseases, so I think that the people working on those programs were happy to see the information, and that's some of the response that we've gotten.

Margie Shofer: Just a follow on to that, do you see your legislators doing anything in particular with some of the findings in the report? 

Brooks Daverman: Yeah. We've seen them use the report to say, "I come from this area, and we've got a real problem," and they'll point to the report, and now they've got some numbers to back them up, in addition to a sort of personal testimony. At the same time, in Tennessee, as in most of the country, we've got the tight budget situation, so there was not a lot of proposing of new programs or new expenditures this go-round with the budget. I think that that's pretty much where the conversation stayed.

Margie Shofer: Perhaps that will happen when you're flush again. Of course, we don't know when that will be. I think we are actually close to the end, so that wraps up the questions. Thank you, Brooks. I want to thank everyone on the phone for your thoughtful questions and your participation in today's Web conference, and we really hope this event was helpful to you. 

If you have questions about follow-on technical assistance opportunities, please do not hesitate to submit them to the Quality Tools e-mail address on this slide that you see here. If you have any questions or comments, including a request for slides from today's presentation, please send an e-mail to the same address. 

This outreach program will run through October 2010, so please notify us of any further information or technical assistance needs before then. 

Thank you. This concludes this Web conference, and we look forward to hearing from you. 

Page last reviewed October 2010
Internet Citation: MONAHRQ's Preventable Costs and Hospitalization Mapping Tool (continued): Webinar Transcript. October 2010. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/cpi/centers/ockt/kt/webinars/monahrqtrans/monahrqtrans2.html