We don’t quite have it right with health care. We know that. It’s too costly, too impersonal, too inefficient, and too ineffective. It also continues to focus on managing sickness at the expense of promoting health. Given the need for change, our group is looking outside of health care for inspiration. The tech industry is making increasingly complex applications both intuitive and empowering for end users, and might provide some novel approaches to meeting the needs of patients and communities—not just providers.
Beyond the “Waterfall” Approach
Biomedical research often informs health services research. Possible solutions start with safety and efficacy trials, and then move to implementation trials. In this latter stage, each step is based on hypotheses about causality, which researchers test in a pre-defined manner, but it does not allow for modifications based on what researchers learn during implementation. For example, if you are testing the effectiveness of a new diabetes education program on patients’ knowledge of diabetes and their outcomes (as compared to regular care), and partway through you learn that participants need and want support rather than education, you can’t modify your intervention.
In the tech world, this kind of approach is called “waterfall”—each step occurs in sequence: gathering requirements, designing the solution, implementing it, verifying its implementation, and then maintaining it. But it’s recently fallen out of favor with start-ups, since you can spend a lot of time developing something that nobody wants, which is very common for early-stage products and services. In the diabetes program example above, a waterfall approach would tell you whether your pre-determined intervention was better or worse than regular care, but it wouldn’t help you land on the best possible intervention for the people you’re looking to help.
In response to this, many start-ups are using a more iterative approach that incorporates adaptive planning, earlier delivery of smaller parts, and constant testing with end users. These new approaches are called “agile”—and the best-known is the “lean startup.”
Our group applies the agile approach to helping health care professionals design models of care for people with complex needs—an area where there is a lot of uncertainty about what works and how to measure it.
For instance, we recently worked with one team on developing a community-based program to reduce readmissions of patients discharged from an in-patient psychiatric service. Prior to our first meeting, the team did a literature review to find effective interventions, and outlined the proposed service and implementation process.
Together we decided to draw on the lean startup approach to test the “minimum viable service” before building an entire program. We described eight critical components, including care coordination for community resources, support for family doctors, medication review and education, and urgent access to psychiatric care. We then followed a small number of patients through the program to see how much these components really mattered to users and whether they affected readmissions rates.
One month of data showed that three of the eight components were neither relevant nor effective. For instance, we found that the care co-ordinator had limited knowledge of community resources and several patients had no family doctor. Those who did have a doctor had little or no recent contact and no interest in reconnecting.
So we modified our components. Because engagement with the family doctor was important but challenging, for example, we developed one approach for engaged patients and another for disengaged patients. And in addition to trying to improve the program, we are working with the in-patient psychiatric service to fundamentally improve the discharge process, rather than just put a Band-Aid on it.
Our initial approach took three months to develop, but just one month of data collection revealed its flaws. We learned that while it may take a trial to determine that something does work, rapid testing of rough versions of a service can convincingly show that it doesn’t. Now we look for problems and are already thinking about alternate solutions before problems surface. For our psychiatric project, we are reconceptualizing and taking what we have learned to better plan services, including adding family physician support service to an existing platform for family doctors with complex patients.
We think this approach is well-suited to circumstances where the intervention, appropriate target group, and outcomes are unclear. It’s hard work, with few eureka moments, and so we’d like to share our tips for other teams considering the same approach:
- Assume it isn’t going to work. This will do two things: keep your emotions and motivation intact and encourage you to test multiple things at once to maintain your momentum.
- Pace yourself. Constant iteration isn’t about working like crazy until you kill yourself; it’s about creating simple tests that help you see whether things are working or whether they need to change.
- Document your journey. It’s easy to get discouraged, and it’s easy to forget the things you have already learned if they aren’t visible. We made several tweaks to the program (such as with physician communications), and recording the changes was helpful and motivating.
- Keep reaching. Most things we do in health care make only a marginal difference. If we want major insights, we need to expect more and change strategies (while keeping the same vision) until we find them.
Getting stuck is inevitable. Great teams with strong members take turns leading each other through the hard times. Getting unstuck could be a matter of revisiting what you have learned or conducting a few radical experiments. Maybe you need to all go out for ice cream. Regardless, working in an iterative and critical way is hard work and takes focus and determination.
The uncertainty we face in treating sickness will only increase as we move toward creating health by working across sectors to support people’s aspirations. But honing agile and iterative approaches in the former domain will increase our success in the latter. And we hope that by sharing our frustrations and our journey, it will help you in yours.
This post was first published as part of a blog series on Stanford Social Innovation Review – see it here.