Enabling Change in Health Care
by Martha Hostetter, Sarah Klein, and Douglas McCarthy
Innovation centers such as the Penn Medicine Center for Health Care Innovation — an arm of the University of Pennsylvania Health System — are using rapid, low-cost experiments to explore new approaches to improving care.
Health care reforms and market trends are stimulating local health systems to seek better and more cost-efficient ways of meeting their patients’ needs. This case study is part of ongoing research by The Commonwealth Fund to track how health systems are transforming care delivery, particularly to meet the needs of high-needs, high-cost patients and vulnerable populations.
The University of Pennsylvania Health System founded the Center for Health Care Innovation in 2012 to test new models of care and build evidence of their effectiveness. The center is also designed to help Penn Medicine — a $4.9 billion system based in Philadelphia — prepare for payment models that reward clinicians for the value of the care they deliver.
Penn Medicine’s working premise is that innovation relies not on inspiration but on having a ready infrastructure to develop, test, and implement new strategies for delivering health care. The health system also sees innovation as a discipline that can be learned.
Budget, Focus, and Staffing
The center’s 21 staff members include physicians, a former Silicon Valley executive, innovation managers, a software engineer, a communications manager, a statistician, research coordinator, and “experience designers” who focus on end users. Team members often embed themselves in clinical settings so they can partner with clinicians.
Several Penn clinicians and researchers also dedicate part of their time to help lead the center’s projects, as described below. In many respects, the center functions as an R&D arm for Penn Medicine — putting research at the university’s medical and nursing schools, business school, law school, and social science, design, and engineering departments to work to improve care delivery at the health system and beyond.
Penn Medicine provides most of the center’s annual budget, which distinguishes it from innovation centers that aim to generate revenue through grants or commercialization and are expected to eventually be self-sustaining. “A mark of an enterprise that expects to be around in the future is how much it is willing to invest in R&D for its own operations,” says David Asch, M.D., professor of medicine and the center’s executive director. The center develops its projects through:
- institutional priorities and executive requests;
- crowdsourcing through innovation tournaments, staff suggestions, and a competitive grant program;
- population health needs, including managing hypertension, increasing screening rates for colon cancer, and reducing tobacco use; and
- partnerships with payers, startups, and Penn research faculty.
In any given year, the center’s core team pursues about 25 projects, many of which revolve around:
- using technology and principles of behavioral economics to engage and monitor patients;
- expanding access to timely and convenient care;
- gleaning insights about patients through their “digital footprints,” or online activity; and
- developing streamlined care models that promote better outcomes and lower costs.
How the Center Pursues Innovation
The center’s approach to innovation is similar to the Plan-Do-Study-Act cycles of traditional quality improvement, but there are key differences. For one, quality improvement tends to focus on troubleshooting existing processes, while for Penn, innovation often entails trying new approaches — sometimes borrowed from other industries. Proponents of quality improvement also usually favor rigor, while Penn researchers use whatever tools are at hand to gather just enough evidence to build momentum for change. Their work rests on four key components.
Gain Insight Through Contextual Inquiry
Center staff and the clinicians they partner with immerse themselves in an aspect of care delivery, aiming to understand the clinical, operational, and other factors that shape it and the needs of numerous stakeholders. This may involve shadowing staff members, following the steps in patients’ journeys, visiting patients’ homes and other community settings, and mapping work processes. These efforts reveal information that may not emerge from traditional interviews or surveys, and allow innovation managers and experience designers to propose new ways to achieve desired results.
Example: Improving Ebola Screening
Problem: Though intake staff at the Hospital of the University of Pennsylvania were following the Ebola screening protocol recommended by the Centers for Disease Control and Prevention, they were not getting useful answers.
Approach: The center’s team performed a “design sprint” — a quick, two-week process — to determine why Ebola screenings weren’t generating needed information. By observing the process and interviewing patients and staff, they discovered that patients felt as if they were being profiled due to their appearance or accents and/or were worried about being denied treatment based on their responses.
Change: Screeners now begin by reassuring patients that they ask everyone the same questions, and then instead of asking “Have you been outside of the U.S.?” as they had been doing to little effect, they asked “When was the last time you went outside the U.S.?” This moved the conversation beyond yes/no responses and enabled faster identification of potential risks.
Results: The revised screening process increased the number of patients who were willing to share their travel history with staff from none in the control group to 55 percent in the test group.
Example: Streamlining Treatment for Oncology Patients
Problem: Oncology services at the Hospital of the University of Pennsylvania were often so full that no beds were available for patients scheduled for inpatient chemotherapy, forcing staff to reschedule them. Staff members pointed to delays in the discharge process as the likely source of the problem.
Approach: Inquiry revealed that staff members often booked three hospital days for oncology patients, to ensure that a bed would be available if needed on the second day of treatment, when patients tend to experience the worst side effects of chemotherapy. That meant beds appeared to be unavailable even if each patient did not end up staying all three days. The innovation team also found that staff members were using the first hospital day to perform lab work that could be completed in outpatient settings.
Change: The team suggested a new approach focused on a key metric: treating patients as quickly and efficiently as possible. The problem shifted from “how do we get patients discharged earlier?” to “how can we minimize delays in treatment?” Staff members now do lab work before admission, about 40 percent of patients begin their course of chemotherapy in an outpatient setting, and staff preorder drugs patients will need while hospitalized.
Results: The average length of stay fell by 12 hours, greatly shortening treatment delays for all patients. This approach also reduced the costs of care and enabled patients to spend less time in the hospital, which can be debilitating, and more time at home with their families.
Explore Divergent Approaches
The innovation team uses rapid, low-cost experiments to explore different approaches to solving a problem and to uncover the advantages and disadvantages of each. This discourages clinicians from prematurely choosing a particular direction in their eagerness to respond to an urgent need, and also allows for mid-course corrections.
Example: Tracking Postpartum Hypertension
Problem: Hypertension was the most common reason new mothers were readmitted to the Hospital of the University of Pennsylvania within seven days of discharge. However, clinicians were having trouble obtaining blood pressure readings after the women left the hospital. The existing approach — encouraging new moms to participate in a bimonthly blood pressure clinic or phone calls — was ineffective, because the women did not answer calls or return messages and the clinic was poorly attended. Women also typically developed hypertension within a week of giving birth, before their next appointment.
Approach: The innovation team realized that new mothers are preoccupied and have unpredictable schedules, making telephone contact difficult. The team also noticed that women often texted while waiting in the clinic — a sign that they embraced this mode of communication. Working with an obstetrician and a family medicine physician, they decided to use text messaging to collect blood pressure readings and educate the women about hypertension.
Change: During the first innovation cycle, clinicians discharged seven new mothers with electronic blood pressure monitors, and sent daily text reminders asking them to check and report their blood pressure. Most women responded. Follow-up interviews led to three further insights: the women liked communication with their providers that felt personalized; they found it difficult to predict when they would be free to check their blood pressure; and they did not clearly understand why the checks were important. The team used this information to test various approaches among small groups of patients, to see if they could boost engagement even more.
Results: Of 32 patients enrolled across all iterations of the pilot, 84 percent reported blood pressure readings at least once during their first week postpartum — up from just 15 percent. And 69 percent reported blood pressure readings on five of seven days. Enrolled patients had no readmissions, compared with 5 percent of patients monitored before the intervention. Goal messaging — the final variation — yielded somewhat greater engagement, but overall engagement remained high throughout the trial.
Team members recognized that to expand the use of this solution, they would have to automate it. In a more recent pilot, they used a scripted set of questions to respond to patients who sent blood pressure readings, mimicking an automated system. Engagement remained high and had the same impact on readmissions. The team is now using a randomized controlled trial to test this approach on a larger scale. Aside from benefiting individual patients, the innovation promises to give clinicians a much more detailed picture of postpartum hypertension, which is not well understood.
Use Low-Cost Experiments to Quickly Validate New Approaches
The innovation team uses low-cost experiments and workarounds to try out proposed solutions and build evidence of their effectiveness. The goal is to avoid entrenched thinking, regulatory impediments, technological barriers, and other hurdles while building a case for change.
Example: Offering Same-Day Appointments for Orthopedic Patients
Problem: Penn Medicine’s orthopedic physicians had a reputation for being “hard to see,” so the department’s practice manager wanted to offer same-day appointments. However, the physicians repeatedly rejected the idea, concerned that it would lead to overwhelming demand and inappropriate appointments.
Approach: The practice manager learned from the innovation center about a “fake back-end” — a tool used by software developers to try out an idea without building a whole new structure to support it. A classic example is IBM’s “testing” of users’ experience with voice recognition software before programmers had even developed it. They did so by hiding a fast and accurate typist behind a wall, convincing users that a computer was automatically registering the words they spoke into a microphone, and monitoring their reaction. 
Change: The orthopaedics practice manager noted on the department’s website that physicians were offering same-day appointments, and listed his number as the point of contact. That allowed him to circumvent the call center and quickly test the idea at minimal cost. He persuaded physicians to try it by noting they could end the experiment at any time.
Results: Without any other advertising, requests for same-day appointments came pouring in. The experiment revealed patient demand for better access. In fact, the practice manager had to take down his cell phone number after three days because he could no longer handle the calls.
Although many requests came from new patients needing joint replacements, most did not demand to see surgeons first: they accepted referrals to sports medicine or physiatry when indicated. And most patients wanted a convenient appointment rather an immediate one: a time in the near future, at a location near their homes.
These results convinced the department to adopt a scheduling system that improves patient access, including same-day appointments for those who request them. Since launching the system in April 2013, the department has seen average wait times fall from 13 days to 8 days, and surgeons have attracted new patients. Yet the department has not added providers, making better use of appointment times instead. Other departments are considering adopting the approach.
Changing the Culture
The innovation center aims to instill a culture of innovation across the health system. One way it does so is through innovation tournaments, a way of crowdsourcing ideas for improving health care delivery from among Penn Medicine’s employees. The first tournament, held in 2012, garnered over 1,000 submissions from more than 5,000 employees about ways to engage patients in their care. The innovation team also trains staff to pursue their own projects and gives them the resources to do so, including grants. Through Penn Medicine’s Performance Improvement in Action program, Roy Rosin, chief innovation officer, has trained nearly 1,000 Penn Medicine employees to explore complex problems, rapidly test ideas, and validate assumptions. The program complements traditional training in quality improvement by encouraging staff to move beyond efforts to fine-tune existing approaches to find new ones.
The center measures success in its efforts to instill a culture of innovation partly by how effective it is at helping frontline providers understand and apply new approaches it has developed. “When employees stop Roy or David or the team and say ‘I was at your lecture, I heard the ideas, and I’m using that same approach on my floor’ — I think that’s our biggest success,” says Kevin Mahoney, the health system’s executive vice president and chief administrative officer. “In an organization of this size, you can’t hire enough people in the innovation center. So you have to teach, you have to change the culture.”
The center has also created tools to enable staff to develop their own solutions, including Agent, which allows clinicians to select the information they want to receive on patients, when they want it, and though what channel. Clinicians can customize the alerts — for instance, setting one up to notify them via text or email when orders for antibiotics expire — and provide feedback on which alerts are most useful.
Another tool, Clinstream, gives providers access to real-time clinical data, a server, and a “sandbox” they can use to develop their own solutions. For example, they might use clinical data to test a theory or create a prototype app. Clinstream empowers the growing cadres of frontline clinicians who are interested in medical informatics and able to code.
With this infrastructure in place, clinicians can develop solutions to problems quickly at low cost. For example, the medical director of the ICU used Agent to address a safety issue for patients on ventilators. He had been working to educate and remind staff about the importance of taking certain precautions when removing endotracheal tubes from difficult airways, but had not achieved compliance in use of these protocols. Using Agent, the medical director now receives alerts whenever a patient is intubated without a staff member setting the required order set for extubation, so that he can then intervene in to educate residents in safe extubation procedures before they occur. Compliance quickly went from zero to 100 percent.
Spurring Innovation Beyond Penn Medicine
Some Penn Medicine clinicians are promoting new approaches that reach beyond the health system’s borders. A good example is IMPaCT — for “individualized management for patient-centered targets” — developed by internist and pediatrician Shreya Kangovi. In this approach, trained laypeople, known as community health workers, help some of the poorest and sickest patients navigate the health system and find resources and supports, such as housing assistance or a workout partner.
A randomized controlled trial showed that IMPaCT improved participants’ health outcomes. However, Kangovi feared that the program would go nowhere after her grant funding ran out.
She approached the innovation team for help in developing a plan to sustain the program. The team focused on IMPaCT’s key strength compared with other such efforts: its rigorous approach to training the community health workers. The training ensures a consistent approach, and has reduced turnover among the workers. The innovation team also developed a three-pronged strategy for encouraging wider use of IMPaCT. That approach included providing a free toolkit for implementing the program, an online platform for managing workflow and data, and training and technical assistance.
Penn Medicine charges other health systems for use of the online platform and for technical assistance, helping to cover the costs of the program. More than 400 organizations have used the tools since February 2014. Kangovi and her team have also received a federal grant to expand and test the model across different settings, including the Veterans Health Administration and a community health center.
To build a case for sustaining the program at Penn Medicine, the team used results from Kangovi’s randomized controlled trial to develop a financial projection. The analysis accounted for opportunities to avoid Medicare’s penalties for excessive hospital readmissions and gain revenue from payers that reward cost-effective care. The projection prompted health system leaders to establish the Penn Center for Community Health Workers, which supports more than 1,500 patients a year.
Preparing for Value-Based Reimbursement
Clearly, the innovation team cannot provide this level of support to everyone across the health system working to improve care, partly because Penn Medicine still earns most of its revenue from fee-for-service contracts, which provide few incentives to try different approaches to care. Still, health system leaders recognize that they will soon face more financial risk as payers shift to contracts that tie payment to improving health outcomes and providing more efficient care. “The innovation center has been so successful because it’s trying new approaches on a small scale, helping us get ready for the future,” says Mahoney, Penn Medicine’s executive vice president and chief administrative officer.
The approach may show that some new ideas may not be worth pursuing without evidence of their impacts on outcomes and costs. He cites one project supported by an innovation grant that enabled clinicians to provide genetic counseling to cancer patients via secure video conferencing. The new service was unveiled with “great fanfare,” he recalls. However, although virtual counseling saved patients a trip, it required the same amount of physician time and physical space. And participants asked to build a tele-oncology unit with four soundproof rooms, even though four exam rooms were already available for in-person counseling. “I scratched my head and said, ‘I’m not sure what we’ve accomplished,’” he notes, especially as the program draws no new reimbursement from payers.
Still, because a pilot of the approach found it improved clinical outcomes as well as patient access and satisfaction, Penn Medicine is working with Independence Blue Cross, its largest payer, to evaluate whether the results justify the additional expense.
Although the Center for Health Care Innovation has been in existence for just three years, its record thus far offers lessons for other organizations seeking to transform health care delivery.
Real-world testing of innovations is critical for producing the evidence needed to overcome resistance to change. “You can’t just think in the abstract,” Rosin says. “You need to be able to translate ideas to pictures and sketches and think in code. People have to see ideas progress at least far enough so they can touch them, feel them, and get a sense for what’s possible.”
The ability to accept and move on from failure is also important. Rosin’s mantra, learned from his days in Silicon Valley, is to “fail fast, fail cheap, and fail often” — to try out new ideas and quickly discard those that don’t appear to work. “When you do a rapid experiment and it doesn’t work out, you really didn’t fail, you invalidated a hypothesis,” he says. “If it took years and millions of dollars and didn’t work, then you failed.” The key is to use iterative rounds of experiments and do just enough to determine the effectiveness — or lack thereof — of new approaches.
“It’s important to … love small dollars and small wins that are indicative of directions to go, and especially indicative of directions not to go,” says Asch, the center’s executive director.
To implement new approaches to health care delivery, diversify your staff. At Penn Medicine, the center’s experience designers play a critical role is capturing and synthesizing the views of patients, providers, and health system executives, enabling the innovation team to develop approaches to care that meet a broad spectrum of needs. “Three years ago, I didn’t know that there were such things as experience designers. Now I don’t understand how you can run a health system without them,” observes center director Asch. Staff members who can navigate contracts with payers, privacy regulations, and other hurdles are also critical to effective innovation.
Research a problem before reaching for a solution. Would-be innovators often skim over a problem in a rush to find a solution. But developing an effective approach entails exploring the problem from the vantage of patients, clinicians, payers, and other stakeholders. “You really have to get at the professional and bureaucratic practices that cause things not to happen correctly,” says Ralph Muller, University of Pennsylvania Health System CEO.
New approaches also often fail because they are brought to scale prematurely. Taking the time to fine-tune them and gather evidence of their effectiveness is important.
Health care innovations can improve the bottom line, but requiring an immediate return on investment may stymie new approaches. Several innovation center projects show that making care more convenient and accessible attracts both new patients and a mix of private and publicly insured patients — even when that is not their explicit goal.
However, demanding rapid payoffs from innovations may lead to inadequate investment, particularly in those that can improve patients’ long-term health. Aiming for small wins that can add up over time may offset some of the risk of implementing new approaches to care.
 See A. Savoia, Pretotype It: Make Sure You Are Building the Right It Before You Build It Right, Amazon Digital Services, March 2012.