Identifying and treating hypertension is an elusive goal that exposes millions of people in the country to the risk of heart attack and stroke. According to the Centers of Disease Control and Prevention, nearly one in three U.S. adults—about 68 million people—have high blood pressure, which is undiagnosed in 7 to 8 percent of the population.
An interdisciplinary team of clinicians, informaticists, and quality experts at NorthShore University HealthSystem, Evanston, Ill., has taken a major step in controlling this disease in its outpatient population, by harnessing its electronic health record (EHR) to screen for and evaluate individuals with previously undetected hypertension in its multi-specialty medical group. The system, which went live in January 2011, has helped the integrated health system make significant progress in eliminating undiagnosed hypertension among its patients receiving care within its primary care network.
The Undiagnosed Hypertension Project grew out of NorthShore’s Quality Mini-Fellowship, part of the health system’s Practice-Based Research Network, whose goal is to link practicing physicians with research and quality improvement that produce outcomes that can be incorporated into daily practice.
The idea to identify people with undiagnosed hypertension came from Michael Rakotz, M.D., the lead physician at the NorthShore Medical Group Primary Care office in Highland Park, Ill., who was in the first group of four family physicians to participate in the fellowship program. Halfway through the six-month program, each participant was given the opportunity to choose a project. This was in 2010, and NorthShore’s enterprise-wide data warehouse had just come on line and was available for clinicians to start using it, Rakotz recalls.
“I had first crack at asking a question,” he says; and he chose to ask this one: “How many undiagnosed patients were in our system, based on the criteria of having hypertension that was not documented in the electronic medical record?” Rakotz says he has always been fascinated by the fact that hypertension is easy to diagnose, yet it is often missed in clinical practice. He notes that traditional manual blood pressure readings taken in doctors’ offices have long been recognized as problematic for a variety of reasons, including so-called “white coat syndrome” and poor technique for taking in-office pressures. The bottom line is that manual office blood pressures correlates poorly with cardiovascular disease, especially when compared to 24-hour ambulatory blood monitoring, the gold standard around the world for accurately determining if a patient has hypertension, he says.
The initial query as to who had undiagnosed hypertension was limited to ambulatory settings of primary care offices and specialist offices across the NorthShore enterprise. Rakotz worked with the clinical informatics team, led by Ari Robicsek, M.D., vice president of clinical and quality informatics, and Chad Konchak, director of clinical analytics. Their first step was to examine records back to 2006 to identify all of the patients who met Joint National Committee (JNC) 7 criteria for hypertension of multiple blood pressure readings of greater than 140/90. They found over 6,000 patients who had three elevated blood pressure readings and had not been diagnosed. (When they followed the patients forward to the present, they found that most of them were eventually diagnosed, suggesting that NorthShore physicians did a good job of diagnosing hypertensive patients in the long run, Rakotz says.)
Michael Rakotz, M.D.; Steve Smith, Chief Information Officer; Ari Robicsek, M.D., Vice President, Clinical & Quality Informatics; Chad Konchak, MBA, Director, Clinical Analytics; Jonathan Silverstein, M.D., Vice President, Clinical Research Informatics Photo by Jon Hillenbrand, NorthShore University HealthSystem
According to Robicsek, the team was able to identify the cohort of patients in specialist offices; after pulling data on the patients, they observed that primary care offices accounted for the majority of the patients with undiagnosed hypertension. He says this illustrated the difficulty of “connecting the dots. Electronic health records, unless you ask for it, don’t show you the whole sequence of a patient’s blood pressure results. All you have in front of you are today’s blood pressure readings,” he says.
A Quest for Actionable Information
Rakotz and his informatics colleagues set their next task: to employ actionable information that can be used to inform physicians that their patients might be at risk of hypertension. Blood pressures taken in the traditional way with a blood pressure cuff tend to be inaccurate and an overestimate; and they realized that just because a patient had three high readings did not necessarily mean they were truly hypertensive.
Rakotz notes that much of the research on blood pressure so far has been done in a research setting, not an office setting. “We’re doing it wrong; we’re using inaccurate blood pressures that correlate poorly with predicting disease, and we’re using them every day,” he says.
The team created five algorithms that would go into NorthShore’s enterprise-wide database to identify patients at risk for undiagnosed hypertension. One was the traditional criterion of three elevated blood pressures on separate occasions. Others included the mean blood pressures of the three most recent encounters; and a single blood pressure of greater than 180/100. All of the algorithms were fairly close in predicting hypertension, Rakotz says. “We found that using all five algorithms at the same time helped us identify most of the patients, with the least amount falling through the cracks,” he says.
The algorithms identified about 1,600 patients who were actively being seen in the health system and who satisfied at least one of the algorithms. The next step was to reach out to these at-risk patients and invite them for an in-office blood pressure reading. This posed another challenge: knowing the limitations of traditional in-office blood pressure readings, how could they trust the results?
With that in mind, Rakotz suggested using an ambulatory blood pressure monitor known as the Automated Office Blood Pressure (AOBP) machine. The AOBP machine takes six blood pressure readings in six minutes with only the patient in the room. It throws out the first reading and averages the next five. The machine is extremely reliable and its results correlate well with 24-hour ambulatory blood pressure monitoring, the industry gold standard, Rakotz says.
With the decision to rely on the AOBP machine, the team was ready to reach out to the patients who satisfied the algorithms to see if they were truly hypertensive. Through the Practice-Based Research Network, the team was able to recruit a large number of NorthShore primary care physicians who were willing to participate and reach out to their patients.
Of the 1,600 patients who were flagged by an algorithm, 475 came in for the AOBP test. Of those, 38 percent were verified as truly hypertensive (about 80 percent were found to have either pre-hypertension or hypertension). Rakotz says the results, verified by the AOBP machine, uncovered a large number of patients who had falsely elevated blood pressures from traditional in-office blood pressure readings. “One of the benefits of using this more accurate system was we were no longer identifying patients with hypertension who didn’t have it,” he says.
A Two-Pronged Approach
Encouraged by the results, the team considered how to turn the information into an intervention that could be used in real time on a wider basis. In discussions with the participating primary care physicians, some expressed a concern that they would have to call three patients into the office for every one patient who would end up having true hypertension. The upshot was an agreement with the primary care physicians that 50 percent would be an acceptable threshold.
The informatics team went to work optimizing the algorithms to meet that goal, by raising the manual blood pressure limit that would result in a higher predictive value of hypertension. Yet by making the algorithm more stringent, some patients would be missed at the lower end. “Our chief quality officer felt that it was important not to miss any from a population health standpoint,” Rakotz says.
This led to a two-pronged approach to a surveillance system, using two sets of algorithms. One is to notify physicians of patients in their panel who have been identified as high-risk of hypertension using the more stringent set of algorithms. The names of those patients are included a monthly “opportunity for improvement” report, recommending that the physician call in the patient for an AOBP blood pressure reading. The other is a “best practice advisory” for patients who come in for scheduled visits for any reason. These patients are flagged with the original (non-optimized) set of five algorithms, incorporated into the electronic medical record, which fires off a real-time alert that a patient is at-risk and should undergo an AOBP reading.
Rupesh Mandala; Meredith Sefa; Nilesh Patil; Raman Jathar; Joseph Alan Simmons; Mekhala Patwardhan Photo by Jon Hillenbrand, NorthShore University HealthSystem
Following the successful pilot, Kenneth Anderson, D.O., NorthShore’s chief medical quality officer, took the idea to the hospital administration for approval. The surveillance system is now up and running in all of NorthShore’s primary care offices, which are equipped with 30 AOBP machines.
Since going live with the system in January 2011, the system has been used to identify, test and diagnose more than 500 patients with previously undiagnosed hypertension, according to Rakotz. In addition, it allows NorthShore to identify newly hypertensive patients as soon as they meet the criteria, essentially shutting off the flow of new undiagnosed hypertensive patients entering the system. He also notes that 94 percent of patients who have been diagnosed using this methodology have been started on anti-hypertensive medication by their primary care physicians within 90 days of diagnosis.
The project has been a factor in NorthShore’s achievement of HIMSS stage 6 certification last year, he notes, adding that the project succeeded as a collaboration of clinical, informatics, quality, and administration areas, all with a single focus to provide the best care for the patient.
Next Steps and Broader Implications
Rakotz believes this is the “most advanced hypertension surveillance system in the world, because it is virtually impossible for you to be a patient in our system, have hypertension, and not be identified by the system, assuming you don’t have a diagnosis already in the chart.” He believes the system is highly replicable to other health systems, and that it will be an effective tool for controlling related diseases such as chronic kidney disease and congestive heart failure. The next phase of the program will be to go after patients who have been diagnosed with hypertension and who are not controlled to goal, he says.
As for the informatics team, Chad Konchak observes that “there is a ton of low-hanging fruit in healthcare; we just never had data like this to evince change.” In Robicsek’s view, the hypertension project supports “the idea that we’re able to use our electronic medical record to identify places where we could be doing more in terms of population health management.” His team has been discussing the intriguing possibility of sharing preventative health and population health management information, possibly through health information exchange, with other regional providers.