OJAI, CALIF.--With the right application of data analytics a CMIO can achieve much, perhaps even capitalize on a career-making opportunity, said David Kaelber, MD, CMIO for the MetroHealth System in Cleveland, during a session on data analytics at the AMDIS Physician-Computer Connection Symposium.
During the session, Kaelber, who also is a practicing internist and pediatrician, presented three case studies, one of which dealt with financial issues his organization faced several years ago. It was the kind of situation that was “a make-or-break opportunity for a CMIO,” he said. “In my case it was a make.”
In 2011, MetroHealth was having trouble reaching its financial targets because expected patient volumes were down in both primary and specialty care. In wondering what he and his informatics team could do to help increase patient volume, Kaelber realized that since MetroHealth was an integrated healthcare delivery network it “was our own biggest provider, so I just had to question how good we were doing with our referral completion rate.”
Improving that rate wouldn’t help much in the case of primary care, “but could really help in specialty care,” he added. “And the data is just sitting there if you are referring to yourself.”
Kaelber and his team looked at how many consult and procedural orders had been written on a particular day and how many of those orders had been completed by the following day, with the assumption that most of those orders would be completed.
“I thought my patients usually did what I wanted them to do,” he observed. “But when you looked at the data in my system it was only 48 percent.” Meaning that when patients were given referrals to see a specialist or have some type of test or exam performed they did it less than half the time.
So MetroHealth came up with a system—still in use today—where a report is run on a daily basis showing orders that hadn’t been completed, which is then distributed to the appropriate schedulers so they can contact patients. The result? The percentage of orders completed increased from 48 percent to 61 percent in one year and substantially increased revenues. “A big impact on the bottom line,” he said.
Another example he shared was from a “big data” perspective. He and his team wanted to see what Explorys—a cloud-based service for big data analysis of healthcare—could do when it came to post-market drug surveillance of an anti-tumor necrosis factor medication.
Explorys has aggregated data from about 40 million patients, so according to Kaelber it makes a great tool for something like this because the drug is rarely prescribed and has rare side effects. “So you need millions of patients to know what’s really going on here,” he said.
Using that huge database Kaelber and his team were able to answer questions relating to the side effects of the medication compared to similar drugs.
“We have a tool . . . and the techniques and methodologies where we can use the data, combined with the tool, to do things that without this sort of infrastructure it would be impossible,” he pointed out.
A third example demonstrated the potential of data analytics in dealing with the clinical issue of pediatric hypertension.
In his first year out of residency Kaelber was the medical director of a pediatric weight management program in which he determined that not only was he being referred children with weight problems, but children who were suffering from high blood pressure as well.
In conversations with these children’s parents, not only were they generally unaware that children could have high blood pressure, but wondered why he was the first doctor to tell them this, considering the record showed they had been exhibiting high blood pressure for years.
Kaelber wondered whether this was just an anecdotal problem or a major clinical issue in the area in which he specialized. He decided to see whether there were informatics tools that could address the situation.
One issue from an analytics standpoint was the different measures used in determining high blood pressure for adults compared to children. For adults, for example, high blood pressure is reflected in a single number, while for a child it is determined by whether his or her blood pressure is the same or higher than a certain percentage of children (usually defined as 95 percent) who are the same sex, age and height.
By running those numbers on a computer and extracting the data and converting numbers to percentages, Kaelber could figure