“Wouldn’t it be great if we could predict the future? If we could predict if a patient was going to do well from a certain type of treatment before the cost and the time was implemented in that treatment?” asks Judith Baumhauer, Professor of Orthopaedics at the University of Rochester Medical Center.
She tells the story of Anthony, a 56-year-old who started having knee pain over the last year, limiting his activities. His family doctor orders a knee x-ray, diagnoses him with arthritis, and implements a treatment plan: arthritis medicine, low-impact exercise, and visits to a physical therapist 2 times a week for the next 6 weeks. When Anthony returns to his family doctor after the 6 weeks, his arthritis feels no better, and his doctor says it’s time to see an orthopedic surgeon.
“Could we have figured out if Anthony was going to be a responder to that noninvasive treatment? If Anthony was going to get pain relief or improved function with his physical therapy appointments that cost him, by the way, $50 per copay per visit?” asks Baumhauer.
And now, with invasive surgery, we’ve upped the ante. But do we know if a knee replacement will help? “Anthony needs to know to make a good decision,” says Baumhauer, “and we as doctors need to know in order to provide Anthony the correct information for him to make good decisions about it.”
To get there, we need to leverage information from all past patients on whether a knee replacement is a good idea. “The piece of the puzzle we’ve been missing is the patient’s voice,” says Baumhauer. “It’s patient-reported outcomes.”
The University of Rochester collects patient-reported outcomes, which started in orthopedics, for each patient and each patient visit. Baumhauer describes how and why they do this, and how it’s changed patient care.
First, when the patient comes in for their scheduled office visit, the intake person pushes a button on the patient’s name, which pops up a QR code. The QR code is optically linked to a tablet that then displays the patient-reported outcome. Rochester uses PROMIS® (Patient-Reported Outcomes Measurement Information System) for its patient-reported outcomes, an 11-year, $100-million effort funded by the National Institutes of Health. “It’s made by very smart clinicians and psychometricians and utilizes smart testing so that within 4 to 7 questions, we have the patient pegged for a value,” Baumhauer explains.
Instead of asking disease-specific questions, the system asks about symptoms such as physical function, pain, depression, fatigue, and anxiety, because these are what patients track to see if they’re feeling better. “They’re not asking about their knee arthritis,” she says. “They’re saying, ‘Am I physically functioning better? Is my pain getting better?’”
Rochester’s orthopedics department looks into three different symptom complexes. Once that information is collected, it becomes instantaneously available in the electronic health record. By clicking the Outcomes tab in the health record, providers can view a graphic display of office visits over time and the outcomes value. “Down is bad. Up is good,” explains Baumhauer. “It’s very simple. It’s color-coded, and we can share this with the patient, and it’s improved the patient experience.”
PROMIS® has improved patient satisfaction at Rochester and allowed the organization to connect with its patients and understand what they’re saying. If patients say they’re not doing as well, Rochester can translate that information in a validated way. For Anthony, Rochester could learn how bad his symptoms are and, based on his baseline symptom severity, run algorithms and probabilities to determine if his symptoms could improve with formal physical therapy or if they are bad enough to warrant a knee replacement.
What if the algorithms show there’s nothing more for Anthony to gain after he has a knee replacement? The provider should ask how individual patients like Anthony have improved after knee surgery, compare those results to Anthony’s symptoms to form a recommendation, and discuss with Anthony whether surgery would be worth considering.
It’s also useful to examine individual data at the group level, such as improvements over time for three different surgeries at Rochester for the same problem. That outcomes data went back to the Rochester surgeons who provided their patients’ data, and, seeing that one of the surgeries — which incidentally also cost the most — was not producing as good results as the other two, the surgeons decided to discontinue it.
This type of graph could also represent providers, showing which ones are doing well with the same diagnosis and which are not; quality improvement initiatives for the provider falling behind would be a great way to use that data, according to Baumhauer. Population health is another thing to track so that health systems can allocate the right resources for the right problems.
Collecting patient-reported outcomes at Rochester began as a grassroots effort among a few orthopedic surgeons who realized there was a hole in patient assessments: the patient’s voice. “We needed for our patients to tell us how they’re doing,” says Baumhauer. “We didn’t like telling them how they’re doing. We like them telling us.”
With 17,000 orthopedic patient visits a month, Rochester’s orthopedics department needed to implement PROMIS® in an efficient way. Data collection took only 2.4 minutes per visit so that it didn’t hold up patient care. Over 3 years, that data resulted in 220,000 unique patient insights that they could leverage to answer unique questions. Other groups at Rochester liked what the orthopedics department was doing and so it spread across the health care system. Today, about 900 providers collect data in approximately 40 different divisions and departments. “We all populate the same graph, so we all learn from each other,” Baumhauer explains.
PROMIS® includes many different symptom domains; while Rochester orthopedics collected data on physical function, pain interference, and depression, cancer providers wanted anxiety and palliative care wanted fatigue — all those symptoms that are important to their patients, that they’re asking to manage.
One major challenge with this program is slow provider adoption. “Although we can collect this data very efficiently, provider viewing is not as good as I would like it,” says Baumhauer. “I’d like the providers to click on the button and show it to the patients, but sometimes that doesn’t occur.” This could be because they’re a busy clinic and haven’t made it a natural part of their workflow, or a lack of education about what to do with that information. Rochester is working on the education and workflow pieces.
Another challenge is that providers are burdened with other quality indicators that they need to address in order to get paid. “We need to get rid of those quality indicators that are, in fact, not quality and are not indicators, and put the patient’s voice back into the program here,” Baumhauer says. “If the providers do not share it with the patients, the patients question whether they should complete it. It’s a little house of cards.”
The next step for Rochester’s PROMIS® program is to make patient-reported outcomes actionable to improve patient care. They’re beginning to look at upstream screening to do that. Baumhauer shares the example of fall risk, noting that 30 to 40% of patients over the age of 65 fall each year, and half of those falls result in injury requiring medical care, costing about $50 billion per year.
“What if we could just take a 1-minute test and screen those patients to figure out if their physical function was adequate or if they fell standard deviation or two standard deviations below the norm?” she asks. Based on that information, they could then do a second screening, apply some machine learning and patient-derived information, and identify those patients who would benefit from a physical therapy program proven to help with fall risk.
“That is the holy grail of health care,” says Baumhauer. “We’re going to do preventative actions to improve the health care that we provide for patients.”
From the NEJM Catalyst event Provider-Driven Data Analytics to Improve Outcomes, held at Cedars-Sinai Medical Center, January 31, 2019.