Q&A: Planning and Design Set Up Clinical Automation for Success

Teta Alim is the managing editor of HealthTech. Teta previously worked as a digital journalist in Washington, D.C.

As healthcare organizations continue to adapt to workforce challenges and evolving patient expectations, automated solutions transforming care delivery are supporting clinical workflows.

During a HIMSS23 session in Chicago, leaders from the UCSF Center for Digital Health and Nemours Children’s Health shared lessons learned from standing up automated care programs in collaboration with Amwell that have improved patient outcomes.

“Sometimes, we go to the immediate technology solution, and we don’t spend enough time taking a step back, mapping out the journey, designing what we’re trying to accomplish and really looking at how this integrates into our current ecosystem,” Carey Officer, vice president of service delivery innovation at Nemours Children’s Center for Health Delivery Innovation, said in April.

HealthTech followed up with Dr. Anobel Odisho, an associate professor of urology and epidemiology/biostatistics at the University of California San Francisco and clinical informatics lead at the UCSF Center for Digital Health Innovation, and Dr. Patrick Barth, pediatric otolaryngologist, head and neck surgeon and medical director for telehealth specialty care at Nemours Children’s Health, to talk more about clinical automation, the design process and future goals.

Click the banner to discover how health IT solutions can help create an integrated care experience.

ODISHO: Technologically, I think a few things that have made the potential for automation better are interoperability and data integration. The main thing that has changed is the acceptability of some of these tools, from providers, administrators and patients. The COVID-19 pandemic really launched us into more widespread adoption of telehealth. Patient comfort with using remote care management platforms and secure messaging has significantly improved. There has been more progress on implementing on back-end automations.

BARTH: I couldn’t agree more. Part of me wonders if the patient or families were always willing to adopt and it was really the healthcare system and the providers who were unwilling to adopt or adapt. Now, we’re on the same page and engaged. I think our patient populations are going to continue to push us to provide innovative care.

ODISHO: At UCSF, we’ve been doing video visits for years before the pandemic, and even when they weren’t being reimbursed, we as an institution were reimbursing the physicians for their time because we felt that it was important and we wanted to start building the capability and the tool sets around it. Up until 2020, our telehealth adoption was maybe 5 to 8 percent, which is more than many other places but still very small. But having that infrastructure and muscle memory in place allowed us to go from 8 to 90 percent over a weekend when the COVID-19 public health emergency was declared.

I think something similar happened for virtual and automated patient care. In 2018, we launched the Digital Patient Experience Workgroup. This brought in operational and technology leaders from all over the health system to start envisioning what virtual and automated care could be and start experimenting. We recently brought all of the different groups working on virtual care formally into our core IT team so that instead of having little islands of innovation spread across the organization, we can build the foundation to quickly move forward as a big organization.

As far as stakeholder engagement, critically, we need our clinic staff and nurse teams engaged because they’re usually on the front lines of any patient-facing tools. They’re the ones that hear from the patients. We have to make sure patients are involved in the design and implementation. Early in the design process, we engage with our existing patient family advisory councils for general feedback and advice. We then rapidly iterate and get feedback, first from a small cohort of patients with a specific condition. Then, as we stabilize the tool, we move toward a larger cohort, do additional user testing and interviews, and then finally go into general availability. Even at this point, we’re continually taking patient feedback and revising our program. It’s critical to have broad-based stakeholder engagement and buy-in.

BARTH: At Nemours, we had a well-established telehealth program that was being underutilized. The pandemic allowed us to ramp it up to a point where it overwhelmed the system to a certain extent, so we had to come up with other modalities to engage families. We used Amwell’s automated care platform (formerly called Conversa) and automatic texts to prepare families for their telehealth visits.

READ MORE: Focus on integrated clinical automation platforms to mitigate nurse burnout.

BARTH: It’s just a different type of work. It’s all work, and somebody needs to manage that work, whether it's a clinician, a nurse or a physician assistant. They’re either getting phone calls, they’re getting messages or they’re working a dashboard. Everybody works a bit differently. Because of the tools that we’re using, we’re able to be more proactive instead of reactive. We’ve gotten positive feedback, but I think time will tell.

ODISHO: We really tried to integrate with existing provider workflows. We try to make providers more efficient where we can, but in situations where we will increase provider work, we want to make sure that's paired with improved patient care quality. If you’re increasing effort but improving quality and outcomes, you will get buy-in from providers.

We send alerts when patients are not doing well. We’re very aware of the burnout impact on physicians from managing alerts and inbound messages. We feed a lot of data back to providers, so we show them how many alerts are coming to them from these programs. But we also track what happens in the electronic health record after an alert comes through. If an alert comes through, and the provider marks it as done but doesn’t take any action, that may indicate that an alert wasn’t useful. Whereas if an alert comes through, and the provider orders a CT scan and labs and sends the patient a message, maybe that was clinically useful. So, we show that data to providers and say, “We sent this many alerts, and 35 percent of them resulted in additional action. How do you feel about that?” The providers can say there were too many alerts and that we need to adjust the threshold or that it’s about right for their practice. It’s an iterative process to make sure that we’re not having a negative impact on their workflow.

Dr. Anobel Odisho Associate Professor of Urology and Epidemiology/Biostatistics, University of California San Francisco, and Clinical Informatics Lead, UCSF Center for Digital Health Innovation

ODISHO: Instead of working in very specific disease or clinical scenarios, how do we now reach the much broader proportion of our patients with more generalizable tools? Instead of building very specific, one-off workflows, how do we build the toolkit for generalizable workflows? We are obviously very interested in the opportunity around AI. We use predictive models in our current work and have a robust, EHR-integrated AI platform at UCSF, so this is not new to us. We have an internal informatics platform that can pull EHR data, run predictive models and then take results back to the EHR for action.

Our organization has always been very forward-thinking. We will have our own HIPAA-compliant API access to OpenAI APIs. We’ll have a chance to start building some internal tooling based on large language models in the Azure platform.

BARTH: We’re moving beyond the telehealth model and looking at different hybrid versions. That way, providers don’t have to travel and can be where they need to be, and the families can be where they need to be by leveraging technology, whether it includes various types of video or audio instruments or ultrasound. We’re also looking beyond the hospital: How can we get our patients back to their homes sooner and get care at home leveraging remote patient monitoring and virtual care? As far as AI, we have different chat programs that will hopefully evolve to being less scripted and more conversational.

ODISHO: In general, it’s going to be improving provider efficiency, increasing capacity to see complex patients, improving patient quality and trying to do those things by aligning all of our various platforms into some more rationalized tools.

BARTH: We’ve learned a lot about how technology can improve the care of our child patients, but it’s also easy to slide back into how we used to do things. So, it’s that constant looking forward and pushing to leverage all of the tools that we have at our fingertips.

UP NEXT: Learn the answers to five questions about generative AI in healthcare.