Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Special acknowledgments to Emily Peterson, PhD, Senior Product Researcher at WillowTree
From predictive diagnostics to administrative automation, the potential use cases for artificial intelligence in emergency healthcare are vast — but how do we ensure that what we build aligns with actual patient needs? This question was the driving force behind our recent AI Use Case Workshop — which focused on improving the emergency healthcare experience through the lens of the patient's journey — and the two-week research sprint that followed.
Welcome to our recap of our latest ‘2 Weeks To Better' rapid prototyping process, where we'll share how our cross-functional team went from AI healthcare concepts to working code. While our standard timeline for this type of experiment is a two-week sprint, the significance of the problem at hand warranted additional time for our experts to fully develop their solutions, so this time we dedicated two weeks to each respective phase of the product development process.
I'm Kristen Duke, Senior Product Researcher, and together with fellow senior researcher Emily Peterson and a team of business development experts and technical architects, we've identified several promising use cases for AI in emergency healthcare.
In this field notes article, we'll explore:
Join us as we examine the critical challenges in emergency healthcare, reveal patient perspectives on AI in medical settings, and present innovative solutions to address unmet healthcare needs.
Our process started with an AI Use Case Workshop, where a diverse group of TELUS Digital professionals — including developers, designers, strategists, and engineers — gathered for a daylong session to tackle a specific challenge: outlining a patient journey for an emergency healthcare experience and identifying opportunities where applied data and AI models could provide valuable solutions.
Rather than tackling the entire healthcare system, this targeted workshop helped identify key moments where both patients and hospitals face challenges, creating opportunities for innovative solutions that benefit both parties. The outcome was a range of promising AI applications focused on the primary pain points within the emergency healthcare experience.
We came to the workshop with some preliminary research and customer journey mapping already completed. The team then voted on the most challenging moments in this journey, and two pain points clearly emerged as the most promising to address:
The potential benefits and return on investment (ROI) for hospitals and healthcare organizations implementing these AI-driven solutions are substantial. Pre-visit triaging powered by AI could significantly reduce emergency department overcrowding, decrease wait times, and ensure that patients receive the appropriate level of care more quickly and efficiently.
Similarly, post-visit digital discharge solutions offer equally compelling benefits. By leveraging AI to create personalized, easy-to-understand discharge instructions and follow-up care plans, hospitals can decrease readmission rates — a key metric tied to hospital performance and insurance reimbursement — while improving overall patient outcomes — another key metric tied to higher satisfaction ratings.
While well-thought-out, these ideas were based mainly on the team’s assumptions regarding the patient journey. Before moving forward with solutions, we needed to validate these opportunities with quantitative and qualitative data. Enter our Research team.
Through two short weeks of structured consumer research using a tried-and-true research methodology, we uncovered where AI could make the greatest impact in emergency healthcare and where it might face resistance, considering patient sentiment toward using AI in healthcare…more on that later.
These insights are critical for organizations investing in AI-driven healthcare solutions: without a deep understanding of patient expectations and behaviors, even the most technically sophisticated solutions risk missing the mark.
Our research sought to answer three fundamental questions:
To answer these questions, we first wanted to fact-check our assumptions about the patient care journey through an interview with someone who had recently experienced an emergency healthcare visit: our Data Advisory Director, Andrew Deatherage, had recently undergone an unexpected surgery (on Christmas Eve!).
We interviewed Andrew about his experience — from deciding to go to the hospital to follow-up appointments weeks later. We documented this journey, its highs and lows, and outlined some additional opportunities for innovation beyond those that emerged in our Use Case Workshop.
This activity and our concurrent desk research on the existing emergency healthcare landscape prompted a strategic narrowing in our approach. In Andrew’s experience, there were many more pain points related to the discharge experience compared to the triage phase. Moreover, we identified a significant gap in available solutions addressing these discharge-related challenges.
Given these insights, we decided to focus our subsequent research efforts on the discharge experience.
We also recognized the need to expand our research scope and gather insights from a broader, more diverse patient population to ensure a comprehensive understanding. This approach allowed us to validate our initial findings and uncover nuanced perspectives that could inform more detailed solutions.
Given our two-week timeline, we needed an efficient method to expand our qualitative findings into quantitative insights. We launched a quick survey to gain confidence in understanding the patient journey and identify opportunities where AI could help solve unmet needs. We recruited and surveyed 50 individuals who had undergone an emergency healthcare experience in the past 6 months and had some knowledge of AI. This approach ensured that their perspectives were fresh, relevant, and informed in that they had some understanding of how to work with AI solutions.
Note: a larger sample size would be ideal in a standard research engagement, but our intentionally constrained timeline and budget necessitated a more focused approach. We determined that a sample of 50 respondents would allow us to identify key themes and patterns in emergency healthcare experiences and to achieve a balance between depth of insight and breadth of demographic representation, encompassing diverse backgrounds, age groups, racial identities, and genders. Although not exhaustive, this approach enabled us to capture an adequate cross-section of patient experiences and perspectives, providing a solid foundation for our analysis and subsequent recommendations.
The survey we launched included questions exploring the patients’ satisfaction with their discharge experience, post-discharge tasks, and unmet needs within all of the above. Our research revealed several major frustrations in emergency care that echoed those we heard in our initial interview with Andrew:
These findings emphasize the need to digitize the discharge process and the recovery period that follows. They also indicate a need to provide reputable, doctor-backed resources for patients to reference post-visit.
Our survey also included a Jobs-To-Be-Done (JTBD) question set. We chose the JTBD research methodology specifically because of its ability to narrow down many hypotheses, focusing on unmet needs. Essentially, it allowed patients to rank a list of needs or “jobs” by both importance and satisfaction, producing an opportunity score.
(Importance - Satisfaction) + Importance = Opportunity Score
We presented participants with a diverse range of 23 potential needs (or, 'jobs to be done'). These ranged from straightforward tasks like "Receive help scheduling follow-up appointments" to more complex, AI-integrated services such as "Have access to a virtual (AI) health assistant that can help me interpret/decipher lab results and scans from my visit."
This research methodology allowed us to identify and prioritize patient needs, distinguishing those that are most important from those that are currently underserved. An underserved need is one where patients feel existing tools, services, or support systems fail to adequately address it.
Surprisingly, our analysis revealed that the majority of these needs are underserved in the current healthcare landscape. Specifically, 15 of the 23 needs we evaluated emerged as promising opportunities for innovation and improvement. This high proportion of underserved needs underscores the significant potential for enhancing patient experiences through targeted solutions, particularly those leveraging AI technologies.
It's worth noting the atypical nature of these JTBD analysis results. In standard JTBD studies, we typically observe a more balanced distribution of needs across overserved, adequately served, and underserved categories. Overserved needs, where satisfaction exceeds importance (indicated by the gray line surpassing the black one in our visualizations), are common in most analyses.
However, our findings revealed a striking absence of overserved needs and an unusually high concentration of unmet needs (and, therefore, potential opportunities). Of the 23 needs identified, 15 scored above our innovation threshold of 5.0, indicating prime opportunities for development. An additional 4 needs scored just below this threshold, presenting potential secondary opportunities. Only 4 out of the 23 needs scored below 4.5.
This atypical result underscores the depth of unmet needs in the emergency healthcare discharge process, indicating a significant opportunity for innovative solutions to improve patient experiences and outcomes.
Of the 23 jobs given to patients to rank, 15 ended up with a "good" opportunity score (above 5), with two jobs scoring above 6. This indicates substantial room for improvement and potential AI integration. Below are the jobs with the eight highest opportunity scores.
Our research showed that patients already leverage technology during and after emergency healthcare visits. A significant 82% of respondents reported using either the internet or an AI solution to seek additional information about their condition or treatment during their visit beyond what doctors, nurses, and other healthcare professionals provided. This usage increased to 92% after discharge, indicating a growing reliance on digital resources for post-visit information. This also means that the vast majority of patients are looking for medical information and possibly recommendations from the internet at large, and not necessarily from doctor-backed resources.
One of the most surprising takeaways was the positive sentiment toward AI in emergency healthcare. We asked patients about their attitudes toward utilizing AI in healthcare and whether they thought AI could have improved their recent emergency healthcare visit: 80% of patients believed AI could have improved their last emergency care experience.
Top reasons for positivity toward utilizing AI in healthcare included:
However, we also asked participants about reasons that may make them feel neutral or negative toward AI in Healthcare. A vast majority, 92% of respondents, reported that some level of human oversight was important to their comfort with AI in healthcare. This underscores another crucial point: patients are open to AI-enhanced care but do not want artificial intelligence to replace human expertise. Some top reasons selected for feeling neutral or negative were:
Our research synthesis revealed three promising AI innovations for emergency healthcare. What began as initial ideas in our AI Use Case Workshop evolved and expanded through comprehensive patient experience analysis and methodical research. Each innovation addresses critical unmet needs in the patient journey, supported by quantitative evidence from our study:
When we presented these findings and ideas to our technical team, AI Research Engineers Nish Tahir and Christopher Frenchi, along with Andrew Deatherage, our Data Advisory Director had many pointed questions about how this might come to life: are we using this data responsibly? What are the risks of GenAI output in the situation? What existing protections need to be considered regarding HIPAA? The team also identified existing healthcare innovators, like Nuance’s DAX, emphasizing the need to integrate our work with other initiatives to create a more comprehensive platform.
These questions stimulated deeper conversations about what these ideas could become and the range of regulations we had to consider. Ultimately, we collectively and strategically opted to prioritize the first two innovation opportunities. The team felt these solutions could work in tandem, effectively bridging the critical gap between provider-patient interactions and post-discharge recovery.
While it addressed a significant unmet need, the third opportunity — a medication manager — was deemed too high-risk given the current limitations of generative AI technology. As Nish pointed out, an AI-driven medication manager, while promising, could result in catastrophic risks where system failures or AI hallucinations might lead to life-threatening situations. This careful weighing of patient safety against technical feasibility exemplifies our thorough innovation approach and emphasizes the importance of cross-functional collaboration.
Our methodology enables us to approach innovation through a dual lens: combining real-world, pragmatic perspectives with nuanced human judgment. This holistic approach to problem-solving and innovation in healthcare tech is something that, despite rapid advancements, AI is not yet capable of replicating. The integration of domain expertise, ethical considerations, and technical acumen remains a uniquely human capability, underscoring the continued importance of human oversight and patient input in AI-driven healthcare solutions for the foreseeable future.
Now, Nish, Chris, and other members of our technical teams are hard at work bringing these ideas to life. They quickly got to work, developing solutions architecture diagrams demonstrating how a system like this could be built and maintained. Stay tuned for Part Three of our 2WTB series on AI in healthcare to see the product come to life.
For organizations investing in AI-driven healthcare solutions, patient/user research is not just beneficial — it's essential. Here's why:
Most critically, this research-first approach helps prevent the healthcare system from innovating in isolation, ensuring new technologies truly serve their intended purpose: improving patient outcomes and experiences.
Our research has clarified that the patient journey extends far beyond discharge, but the support they receive often does not. AI solutions present a powerful opportunity to bridge this gap, but only if developed with patients at the center.
When organizations prioritize real-world patient insights in AI design, they can create solutions that benefit everyone: patients may recover more successfully, providers can engage more effectively, and hospitals may see improved outcomes with reduced readmission rates and higher satisfaction ratings— all leading measures of success.
The future of AI in emergency healthcare depends on more than technological advancement or innovation for the sake of innovation — it hinges on designing solutions that truly serve people, patients and healthcare professionals alike.
Want to learn more about how consumer research can drive AI innovation in healthcare? Reach out and let’s discuss.
One email, once a month.
Elegant, performant digital products.
Personalized, automated marketing.
All powered by data & AI.
Proudly a telus digital company
