Successful health care innovation follows the pattern of successful science; it requires laboratories where experimentation is encouraged and can proceed safely so that change seems less fraught. Three approaches below can help support this essential kind of experiment: Avoiding having to seek consensus up the chain of the command; enable exceptions to the "rules" for small, quick experiments, and "free the data" by creating platforms that stand between clinicians and cumbersome legacy electronic health record systems.
Every week, medical journals bring us news of astounding scientific discoveries: CRISPR gene editing, or CAR-T cell therapy for cancer. And yet just as frequently we hear, "Why can't health care be more innovative?" The resolution of this paradox lies in recognizing that when people lament health care's lack of innovation, they're referring to how we deliver services to patients. That distinction makes the paradox even starker: "So, you're telling me that you can reprogram T cells to find and kill cancer cells, but it took four months to get my mother an appointment with a neurologist; she spent two hours in the waiting room; and then she got an exorbitant bill that read, 'This is not a bill'?"
Improving patient scheduling, service, or billing should not be as hard as harnessing clustered regularly interspaced short palindromic repeats to edit nucleotide sequences. And yet it seems to be. Why is that, and what can we do about it?
A big difference between innovating in the molecular sciences and in care delivery is that molecules don't fight back. Bacteria may exchange plasmids, but they don't practice guile. Nor do they take comfort in doing things the way they've always been done. The resistance to change in health care is sometimes simple intransigence but mostly it is the natural byproduct of thoughtful professionals trying to avoid mistakes in a setting that is expensive, regulated, and high stakes. Yet, successful health care innovation follows the pattern of successful science; it requires laboratories where experimentation is encouraged and can proceed safely so that change seems less fraught. We've found that the approaches below can help support this essential kind of experiment.
Delay consensus. The CEO of an oil company has more organizational power than the CEO of a health system because the latter cannot tell the aortic surgeon how to operate any more than the aortic surgeon can tell the perfusionist how to manage the heart-lung machine. Highly-specialized expertise and narrow licensing and credentialing make health care organizations so matrixed that it seems anyone can say no, and no one can say yes.
We have been testing a program in which patients with advanced obstructive pulmonary disease alert us of exacerbations before heading to the emergency room. Using existing ride sharing services and "borrowing" the medical supply backpacks already prepared for our helicopter-based trauma team, we rapidly dispatch to patients' homes EMT-trained nurses carrying intravenous tubing, corticosteroids, bronchodilators, and diuretics. The pilot has prevented seemingly inevitable admissions at low cost. When we describe programs like this at meetings, we often hear responses like "If I suggested that at my institution, a dozen people would tell me no."
Traditionally, health system leaders presented with this concept would insist we first coordinate with a range of clinical and administrative services (medicine, nursing, pharmacy, finance, security)-certainly delaying and likely dooming the project from the start. Instead, when leadership's first exposure to the new model is in the form of promising results from initial tests, conversations focus instead on to how to work out kinks, make it part of the regular business process, and scale it up.
The program might not have worked, but those risks are easier to take at lower organizational levels where getting input, which is directionally useful, doesn't turn into requiring consensus, which is often directionless. Health care innovation requires allowing teams pursuing novel models to get started without all the permissions they will eventually need to scale what works. This isn't about recklessly going rogue, but recognizing that small experiments offer their own safety checks when they're stepwise and measured.
Enable exceptions. Guidelines recommend early post-partum visits to manage hypertension among women with pre-eclampsia. But clinic appointment show rates within our health system remained low despite efforts to engage this population, leaving hypertension as our leading cause of maternal rehospitalization. Our rules limited patient communication to only four channels: face to face, telephone, mail, or email transmitted through a patient portal requiring sign-in. Each of these channels is foreign to women of this age. None seemed likely to work, and none did.
The argument that we should try a text-based monitoring system with these women encountered, at that point in time, concerns that texting is not secure and reminders that it's not allowed. Changing the request from "Can we text patients?" to "Can we try it, for a limited time, in a limited population" made it safer by bounding it in an experiment with an automatic sunset. In health care, even seemingly small exceptions to protocol create outsized concerns about setting new precedents. The gambit here is that those concerns can often be overcome with the explicit stopping criteria experiments provide. This project's success (it more than doubled the rate of post-partum blood-pressure measurement) made it essential to continue texting. The predictions about setting new precedents came true, as texting use cases grew based on patient preference and results, but the concerns about those predictions were managed.
Free the data. The opportunities arising because health care data are increasingly digital sit alongside laments that these opportunities remain out of reach. Processes created by electronic health record (EHR) vendors and hospital information technology (IT) policies aim for scale, reliability, standardization, and security. The threat to innovation is that these processes typically lock systems down, limiting experiments that explore new ways to leverage data.
Leading health systems select clinicians and staff who can take health care to new places. Serving on the front line, and aiming to make that care better, they will always be ahead of the EHR vendors, and frustrated by standardized information systems whose upgrades solve for what most people needed in the past but not for what leading organizations need now. While clinical uses and needs should dictate design and pace, the felt experience in hospitals is often the other way around.
For us, success has required creating platforms and extensions that sit between the EHR and clinicians, allowing data manipulation and presentation in new interfaces outside of the locked down systems. It entails having a dedicated development team operating outside the IT organization and protected from the enterprise priorities of the moment to focus on opportunities of the future. Effective health care CIOs support infrastructure that makes data available for experimentation.
We had been identifying only a small fraction of medical inpatients needing behavioral health support, and even then finding them too late-resulting in high use of restraints, the need for 1:1 coverage by staff, longer hospital stays, and incomplete or delayed care. To address this challenge, the language within clinical notes was ported outside the EHR to quickly design and test algorithms for early patient identification, with automated communications to interdisciplinary behavioral health teams. With a few months of iteration and testing we were able to identify eight times as many patients with needs, deliver behavioral health consultations on day one of their stay instead of day five, decrease hours in restraints by 30%, reduce 1:1s, safety events, and patients leaving against medical advice, and cut a day from the average length of stay. Those pilot results built organizational support to refine and test further. This pace was enabled by rapid experimental iterations that would not have been possible had this work been attempted within the existing enterprise IT infrastructure.
Successful innovation requires experimentation-following many of the same pathways of the successful science that has brought us CAR-T cell therapy and CRISPR. But health care change requires we tinker with the health care system we depend on, affecting critical resources organizations understandably protect. To support the people determined to drive change quickly, we need to find ways to bend institutional norms safely.
All Rights Reserved © 2024
