College of American Pathologists
Printable Version

  Putting glucose data to hospitalwide use


CAP Today




July 2008
Feature Story

Anne Paxton

Implementing a tight glycemic control protocol in the hospital should be straightforward: Monitor blood glucose levels, assess how well they’re meeting target ranges, use the information to improve, and reap the benefits in shorter lengths of stay and lower mortality and costs.

But even when hospital staff are eager to comply with the protocol, the challenge of getting the right data together can hamper hospitals’ ability to benefit from tight glycemic control, or TGC.

Michael Blechner, MD, found this out in 2006, three days into his job as director of pathology informatics at the University of Kentucky Medical Center. “I was pulled into a meeting with a nurse who needed help from the laboratory because she was responsible for monitoring glucose levels and had been doing it all by hand,” he said, speaking at the Lab InfoTech Summit in April. “Every month she was going into the electronic medical record system and pulling a random sample of ICU patients, and manually putting that data in spreadsheets.”

Dr. Blechner and his team were able to bring order to the data collection process via an LIS-based system that monitors the success of the university’s TGC initiatives. “Our solution essentially was to extract data from Sunquest using a query tool, put it in a database in Microsoft SQL 2005, and use SQL tools to provide people with embedded reports over the hospital intranet.”

His response is one example of how laboratories are trying to cope with, and take advantage of, the vast amount of potentially useful data that tight glycemic control programs are generating.

That more than 1,300 laboratories are using Medical Automation Systems’ RALS-Plus information management system for point-of-care testing confirms that hospitals are keen to tap into comparative data on patient glucose results—both their own data over time and the data of other hospitals with TGC protocols.

Philadelphia, for example, is a “well-RALSed region,” says Bette Seamonds, PhD, DABCC, director of point-of-care testing services at Mercy Health Laboratory, part of Philadelphia’s Mercy Health System.

But Dr. Seamonds confirms that integration with the LIS is still an aspiration at many area hospitals rather than an achievement. “More have interfaced to the LIS in the last couple of years, but some still haven’t. The reasons could be either they were not given the funding, or it could be they have a scripted interface—as opposed to an HL7 interface—that takes a long time to dump data, can cause roadblocks, and may potentially crash the system.” Mercy Health Laboratory was hooked up to RALS in 1999 but had no RALS interface until a couple of years later. “We had to wait until we got our Sunquest LIS, because it would have crashed our previous LIS,” Dr. Seamonds says.

The interface also lagged behind because of funding, but now it is functioning and making point-of-care results available to physicians. “Our next venture will be getting funding to have the urine analyzers, which we’ve had linked to RALS for a year now, also interfaced with the LIS.”

Still, seamless integration of the RALS and LIS is an elusive goal. “We have an IT department that’s corporate and they do things we don’t even know about,” Dr. Seamonds says. On one recent weekend, the primary systems crashed, and some of the medical record num­bers were not being recognized. “Thirty percent of the data was not crossing the LIS because of that. A phy­sician alerted us to it because he was looking for patient results and there were none for three days. But, fortunately, these kinds of problems crop up very rarely between RALS and the LIS.”

Adding to the difficulty is that the glycemic protocols differ at Mercy’s three hospitals. “We’re doing reasonably well in one place and not quite so well in another, so we’re trying to get them standardized toward the seemingly better protocol.”

Dr. Seamonds strongly advocates hospitalwide TGC, not just in the intensive care unit. “The reason is you can get somebody in great control in the CCU and lose them in the general hospital environment, and they go right back to where they were. So I think it’s crucial that this become a hospitalwide effort in the long term.”

But a protocol goes nowhere if it doesn’t have a physician champion, she points out. “If you have an intensivist assigned to the CCU who has a passion for tight glycemic control and is really driven to make sure patients are being treated according to the protocol, and then the patient leaves and comes under the care of some other physician who is not in tune with the protocols, you have different people writing orders—and then you run into political issues.”

A comprehensive team approach, with buy-in from every sector, and especially from nursing, is critical, she says. “Some of the nurses are up to speed; some are not. I’ve trained several nurses to use the tight glycemic control software and they simply don’t bother using it.”

Her training of the intensivist in the medical ICU, on the other hand, turned out somewhat better. “He loves it but says he just doesn’t have time to go in and tap the data, so when he needs something on an individual patient, or overall performance such as comparing the last three months of this year with the last three months of last year, he calls me and I go in and pull the data out.”

Other physicians, however, are on the system all the time, looking for trends, she says. “The software has actually given us the opportunity to pull data physicians can present to corporate to show the impact of their efforts, and maybe bargain for additional resources, because it’s become so much easier for them to access the data.”

From her point of view, there’s still too much paper. For example, nursing still records manual results because of a problem so chronic in hospitals that she believes it deserves a forum of its own: “If a glucose is repeated within five minutes at the point of care—which usually is due to a bad fingerstick—the results don’t pass through the computer. We set up that guideline so we can actually review the results to determine whether or not they’re discrepant, and if the results agree within 15 percent, we’ll allow one result to cross. If they’re discrepant, neither gets posted in the LIS.”

“But the nurses are still documenting results manually, and the concern we have is that somebody might act on something when they should have sent a specimen to the laboratory for a real value.” While only a very small percentage of results are discrepant, she says, they are often for patients where intervention is needed. “So there are potentially erroneous results floating around the nursing stations. We’re wrestling with this problem, and other institutions are too, and no one has an optimal solution.”

The next acquisition she is hoping to bring about is insulin management software and hardware. “It would be about $130,000 to fund acquisition of laptops and the software and licenses for a year, and even though I can show how much potential savings there would be to have this, there’s just no way the hospital is going to approve that kind of funding.” Dr. Seamonds is looking for outside funding to get the project underway.

“I believe we have a really good point-of-care program. But I can see how much more we could accomplish if I had the tools,” she says.

The data from RALS reports at the five-hospital Crozer Keystone Health System in Delaware County, Pa., led the system to standardize, and then tighten, its glucose target ranges for critical care patients, says Susan McAneny, MT(ASCP), laboratory manager at the system’s Dela­ware County Mem­orial Hospital in Drexel Hill.

The RALS quarterly and annual reports, which the hospitals have been receiving for three years, show results of the Roche Accu-Chek glucose meters tabulated by total number of glucose measurements for ICU, non-ICU, and all patients combined. Those results appear side by side with the same data for all users of Roche Accu-Chek and for other members of Crozer Keystone’s buying group, Novation, giving the hospitals benchmarks with which to compare their own figures.

“They had originally started with a goal range of 90 to 150, but after looking at the data and analyzing it, they felt they could lower it,” McAneny says.

RALS breaks out the data by hospital profile as well, according to number of beds, type of hospital, and geographic region. “It’s the single best way they can validate the data they’ve collected internally by giving them a good view of how they compare with everyone else,” she says. The data have been particularly helpful to Crozer Hospital, with its level two trauma center.

“When I get the report, I forward it to the director of clinical utilization and outcome, and she presents the data to the multidisciplinary critical care team,” McAneny says. Most recently, based on the report, “the data showed that, by doing hourly glucose measurements, they were able to get to their target mean glucose a lot faster. It went from an average of 18 hours down to eight hours.”

One thing McAneny would like to see changed: “Right now, the reports include the total number of glucoses and the average of all glucoses. What we’d really like to see is how many patients that equates to, since one patient might have 20 glucoses done.” Medical Automation Systems and its affiliate company Medical Decision Network are in the process of developing other benchmarking products and services, among them percentage of blood glucose results in the clinic­ally relevant range; com­pari­son with other participating hospitals according to number of beds, region, type, and number of blood glucose measurements; time to target; and percentage compli­ance with the guidelines of the American College of Endocrinology.

In setting up the University of Kentucky HealthCare TGC program, “we spent an inordinate amount of time developing standardized order sets, protocols, policies, and algorithms with associated educational programs,” said Dr. Blechner in his Lab InfoTech presentation. But the sticking point turned out to be the metrics for evaluation. “It seems kind of obvious and simple, but the lack of integrated information systems that allow tracking metrics is a challenge. The improvement teams are faced with the task of devising regular reports to summarize and trend parameters describing glycemic control or hypoglycemia rates, but they lack standardized methods to do it.”

To get maximum usefulness from the data, the pathology department needs information on patients’ clinical parameters as well as costs associated with their admission. As part of its most recent restructuring of the database, the department obtained much of this data from the University HealthSystem Consortium and integrated it into its system. “This data has in it the actual medical record numbers for patients discharged from our hospital, as well as visit identifiers. We put that data in our database and cross-referenced it with the POC glucose results we have,” Dr. Blechner told CAP TODAY.

“So now we can actually do reporting and analysis. We can look at POC glucose values according to mortality, according to the projected risk of mortality, which is one of the consortium’s parameters, real billable costs, projected costs, as well as 25 codes for comorbidity and 25 procedural codes for procedures done on that patient while in the hospital before discharge.”

The university’s other reason for restructuring the database was probably more important, says Dr. Blechner, who also directs the coagulation laboratory. “We expan­d­ed the system to enable the addition of other laboratory tests that we may want to analyze, things that may be related to glucose and diabetes like HBA1c, or things that have nothing whatsoever to do with glycemic control. We set the system up so we can input results from any laboratory tests with numeric values. Once those results are in our database, we can run similar queries against them with minimal additional effort.”

Dr. Blechner describes “analytical cubes” as a way to conceptualize the analysis being done, also referred to as dimensional modeling. “The idea is to think of the center of the cube as containing the data, all of the glucose values, for example, and each face of the cube as one way to look at the data. So you can rotate the cube in three dimensions and say ‘I want to look at glucose values according to length of stay.’ Then you can rotate the cube and say ‘I want to look at it from the perspective of comorbid disease, or according to which service the patient was admitted.’”

Using online analytical processing, or OLAP, he continues, one can zero in on a subset of the data inside the cube. “You can say I want to take the intersection between this group of patients with this comorbid disease who were also admitted to any of our ICUs as opposed to floor services, and I also want to look at their length of stay, or discharge data, complications, or ICD9 codes for diagnosis. To address the issue of bad fingersticks, we can very quickly add filters saying, ‘I only want glucose values that do not have a subsequent value in the next five minutes.’ Or we can focus in on patients receiving more aggressive monitoring by adding a filter that says, ‘I want only patients who have at least eight blood glucose results during a 24-hour period.’”

Using this analytical processing technology allows a researcher or physician to do those kinds of queries almost instantaneously, he says, and these analytics can have useful clinical applications. For example, it’s been demonstrated that maintaining a glucose target range of 80 to 110 benefits patients in the ICU, and the University of Kentucky and many other institutions adopt that as their target range. But could the hospital do even better?

“Maybe if we fine-tuned and targeted a range of 80 to 100, we’d do better, but we don’t know that. And the likelihood is nobody is going to be able to show much of a difference in any single institution between setting the top of the range at 100 versus 110 in the ICU. A lower target range also increases the incidence of hypoglycemic events, with their own associated risks.”

“However, in non-critically ill patients, it’s a whole different ball game, because no one has a clear idea about what the tradeoff is.” Non-ICU patients are not maintained within as tight a glycemic range because it requires too much intensive monitoring and increases the risk of hypoglycemia. “It’s hard enough to do it in the ICU when you’re constantly monitoring someone, let alone on the floor,” Dr. Blechner points out.

So, what’s going to be the target range for patients out on the floor to maximize benefits and outcomes, decrease costs, and shorten lengths of stay while not increasing the risks associated with overtreating and hypoglycemia? “I think this kind of tool could help provide the answer,” he says.

Future directions could include applying the analytics to outpatient and home glucose monitoring, adding therapeutic data such as insulin regimen, insulin order protocol, and administration of oral hyperglycemic agents, as well as tracking followup testing for patients who experience inpatient hyperglycemia but who have not been previously diagnosed with diabetes.

His own hospital, Dr. Blech­ner admits, is not paying enough attention to the potential operational benefits of the pathology department’s data, although it’s moving rapidly in that direction. “Like most medical centers, historically we have been a typical example of what you might call ‘data rich and information poor.’” But now that the University of Kentucky is building its own clinical and research data repository, he believes laboratory data will serve as a valuable resource to answer operational questions about quality and safety.

“Our data is clinical data, and pathologists have to be on the team,” he emphasizes. “When you get into things more complicated than POC glucose values, such as microbiology and molecular, the people down in IT are not going to appreciate how to handle that laboratory data, and there is an even greater danger that it won’t be interpreted properly. We as pathologists can and should play a major role in developing systems that can generate useful information from clinical and laboratory data.”

Anne Paxton is a writer in Seattle.

Related Links Related Links