Home >> ALL ISSUES >> 2014 Issues >> Big gain theory—data warehousing pays off

Big gain theory—data warehousing pays off

image_pdfCreate PDF

Anne Paxton

November 2014—It’s been said that “data” is the plural of “anecdote,” and debate has swirled around whether that is true. Something about data makes most of us feel that it is qualitatively different: more manipulable, more reliable, more helpful in drawing useful conclusions. But is there a new stage that might be considered the plural of data? These days, the health care industry is excited about a concept that promises to catapult the value of laboratory information into a new era. It’s the Enterprise Data Warehouse.

Dr. C. Terrence Dolan of Regional Medical Laboratories, which has been in the vanguard of the data warehouse trend. “We’ve got a major store of very clean data and have already been able to discover things that were not known before in health care,” he says.

Dr. C. Terrence Dolan of Regional Medical Laboratories, which has been in the vanguard of the data warehouse trend. “We’ve got a major store of very clean data and have already been able to discover things that were not known before in health care,” he says.

In the years ahead, get ready to hear the term “big data analytics” often. “The data warehouse is probably the most powerful management system I’ve ever seen,” says C. Terrence Dolan, MD, president and CEO of Regional Medical Laboratories in Tulsa, Okla. “It takes transactional data—that is, the electronic medical record data—plus hospital information, financial data, and more, and brings it all into a common database optimized for massive analysis.”

He tells health care groups: “If I’m in the health system and if I don’t have an effective Enterprise Data Warehouse in the next five years, I’m in serious trouble.”

RML is one of the largest regional labs in the country, performing 9 million tests a year, and has been in the vanguard of the data warehouse trend. With its Enterprise Data Warehouse, containing an array of data points on 2.3 million patients going back as far as 15 years, RML has been able to achieve cost-saving process improvement and more sophisticated guidance for clinical care.

“We happened to be earlier than most in developing a data warehouse in health care,” Dr. Dolan says, pointing to Intermountain Healthcare and Geisinger Health System as other leaders in this area. “But the gap will be closed soon, because there is so much money pouring into data warehousing and so much work going into it.”

Massive databases, relying heavily on laboratory data, provide a new variety of evidence-based medicine that has vast possibilities waiting to be tapped. Current applications are already shaping key business decisions, directly aiding clinicians, and steering patient treatment. And laboratories, manufacturers, and researchers envision a multitude of potential applications.

What makes massive amounts of data so different from more modest amounts? “Some people will say, ‘I already have a data repository, and it’s the EMR.’ But data warehousing is totally different. Its content goes far beyond the EMR, and the way content is structured within the database allows us to do these new analyses,” Dr. Dolan explains.

Lawrence R. Johnson, MD, RML’s director of hematology, coagulation, flow cytometry, and urinalysis, says it can be challenging to convey the concept of a data warehouse. “Usually when we explain the data warehouse to people, my impression is that they really can’t wrap their minds around it. They don’t quite know what it is.”

One useful comparison is what happens when, after being tested in a controlled population of patients, a drug is approved and goes on the market. Suddenly being used by millions of people, the pharmaceutical may turn out to have significant adverse side effects in a certain subset of patients—effects that weren’t apparent in a trial that had only thousands of subjects. The increase in the quantity of data has now changed the questions that can be asked about the drug’s effects.

That kind of discovery is called “finding things out the hard way,” through adverse consequences, Dr. Johnson says. But RML hopes that the kind of massive database the laboratory has developed could catch or predict such adverse outcomes before they happen.

Academia developed data warehousing years ago, says Dr. Dolan. But the retail trade is the sector that has mastered it. A classic example of how data warehousing can be used is a supermarket’s positioning of items for sale. “With data mining, they determined they should put diapers next to beer, because the dad was sent to pick up the diapers and he gets a six-pack at the same time. The analysis is done by looking at millions and even billions of transactions.”

Similar analytics will transform health care, Dr. Dolan believes. “Information technology has become far more cost-effective. You can buy hardware much cheaper than ever before, and now, as all this data is being captured, it’s cheap and easy to store it. The real trick is what you do with the data.”

Data warehousing is different from data mining, Dr. Dolan points out. “Data mining is a very sophisticated analytic tool where they use mathematic formulas to look for associations no one knew existed. With the warehouse, we are really doing online analytical processing, where we know the questions we want to ask.” But he believes the warehouse, with its 15 years of data, will increasingly move toward data mining. “We’ve got a major store of very clean data and have already been able to discover things that were not known before in health care.”

In microbiology, the data warehouse has fostered process improvement, leading to shorter TATs. “Historically, we would read cultures every morning, and we didn’t necessarily adjust for time of incubation.” Urine cultures are read now in precisely 18 hours. “We have determined using the data warehouse the optimal time,” says Dr. Dolan, here with medical technologist Carol Powers.

In microbiology, the data warehouse has fostered process improvement, leading to shorter TATs. “Historically, we would read cultures every morning, and we didn’t necessarily adjust for time of incubation.” Urine cultures are read now in precisely 18 hours. “We have determined using the data warehouse the optimal time,” says Dr. Dolan, here with medical technologist Carol Powers.

To protect the confidentiality of the 2.3 million patients included in the database, RML has smaller “data marts” that are subcategories of the warehouse where confidential patient information has been removed. “Sales, the blood bank, human resources—you have access to the data you need for your field. Then desktop software can pull the data into tables and graphs for automated reporting.”

For the business aspects of RML, the data warehouse is “extremely powerful and central to everything we do,” Dr. Dolan says. “We drive the whole company through the warehouse by doing massive analysis of our cost accounting, sales, human resources, financials, etc.”

The financial side of dealing with the government is an example. “Hospitals are penalized if they re-admit within 30 days, and starting next spring they will be penalized if they dismiss someone before two midnights. We’re using the warehouse to analyze all of that to determine why it happens and develop methods to correct it.”

On the medical side, Dr. Dolan cites public health as one area where the data warehouse is helpful. “Our health departments know we have the capability of monitoring emergency departments as well as doctors’ offices. We know their lab ordering patterns and can determine if a pattern changes, so if we’re starting to see a lot of people who have a viral respiratory infection like influenza, we can zero in to the source where an outbreak could have occurred.”

The data warehouse has also facilitated automation of anatomic pathology. “For quality improvement, we’ve put two-dimensional barcodes on blocks of tissue and the slides, then we wrote software to stop processing if there’s a mismatch,” Dr. Dolan says. “Historically, this has been a significant problem in pathology, but we’ve virtually eliminated it with the technology we’ve developed through the warehouse.”

Throughout the laboratory, spectacular improvements have been seen in elapsed time to complete tasks. “None of us ever thought we could improve throughput like that, but we’ve improved quality, we’ve become far more efficient, and we can do far more tests per square foot of lab.”

RML has found the data warehouse the only truly effective means of refining reference ranges. “Since we have all these patients and data for many years, we can truly identify patients who are ‘normal.’ To develop a reference range, we never use less than 100,000 people, and many times we have 200,000. That gives us the ability to eliminate a lot of false-positives.”

Before data warehouses, laboratories lacked adequate tools to adjust reference ranges for their own populations, Dr. Dolan points out. “The accrediting groups all say you need to adjust your reference ranges to actual populations, but in reality, the analytic tools have not been there.”

To analyze ALT testing, the data warehouse was able to look at 317,000 laboratory patient results. By refining the reference ranges, RML reduced the number found “abnormal” from 10 percent to 8.2 percent. “That’s almost 6,000 patients we reported as abnormal who were in fact normal under a more sophisticated reference range. A thousand dollars per workup of a patient is nothing; if you have a potential liver disease, you’ll be doing scans and all kinds of studies and may spend $10,000. But even at $1,000 a patient, that’s $6 million you’ve avoided.”

The data warehouse also has demonstrated increased productivity at RML over the past 11 years. “Most people measure productivity by year and by month; ‘by week’ is unusual, and ‘by hour’ is unheard of. But we look at productivity by all those parameters. Measured by billed units per full-time equivalent, we’ve averaged about a three percent increase in productivity per year for 11 years, a total of 33 percent improvement.” Some of that is due to growth and economies of scale, he says. “But a lot of it, we believe, is due to improving processes.”

Another use of the data warehouse: analyzing workflow to change staffing levels. “We track the average productivity per FTE at 12 weeks, four weeks, and one week, and can predict productivity based on those averages.” By examining the difference between expected productivity and actual productivity through average FTE per billed unit, RML could see it had underestimated staffing in the early morning and overestimated it for afternoon and evening. “Since that time, we’ve made changes and we’re now more optimal on staffing, so people have a more steady workflow, with minimum peaks and valleys.”

In microbiology, too, the data warehouse has fostered process improvement. “Historically, we would read cultures every morning, and we didn’t necessarily adjust for time of incubation. Now, as urine cultures come in, we set them up and read them precisely in 18 hours. We have determined using the data warehouse the optimal time. We can do that reliably because the microbiology department is working 18 hours a day, 365 days a year, so we report them at 18 hours and do an addendum at 24 and 48 hours, if necessary, using a follow-up read to make sure we did not miss an organism. That’s reduced our turnaround time significantly.”

Among the population studies RML has done, one has focused on the 30 percent of the population with chronic diseases who account for 70 to 80 percent of total health care dollars spent. A large percentage of that group are those with diabetes, who are monitored primarily by laboratory results.

“When we looked at these patients, they weren’t getting the consistency of testing as recommended by national groups for good outcomes. So we said okay, this is our criterion: Have the tests done as recommended,” says Dr. Dolan. The plan is for RML to email patients quarterly that it’s time for their hemoglobin A1c, lipids, albumin, etc., “and really monitor these patients appropriately.”

CAP TODAY
X