Assisted Bedside Manner:
Assessing the future of voice assistants in clinical settings
Client
Institutional Investor
Role
Project Lead
Team
Solo
Duration
4 weeks
Problem
What would the adoption rate of an ambient intelligence product for clinical settings be in the next 2-5 years?
My client wanted me to learn more about an unreleased ambient intelligence product for clinical settings to help them with their investment decisions for their clients’ portfolios. They were especially interested in how physicians and their patients would view this new technology, and to see whether the solutions this technology offers is really what users need and want.
BACKGROUND
While the introduction of Electronic Health Records (EHR) hospitals have enjoyed the easier process of submitting insurance claims, physicians have been spending more time looking at their computer screens to type patient notes than connecting with their patients during visits, which has led to decreased patient satisfaction.
This new ambient clinical intelligence product promised to take the burden off of physicians’ shoulders to help them reconnect with their patients. This product would act not only as a voice assistant to pull up patient records but also transcribe doctor-patient conversations, interpret gestures, and prepare visit notes for the approval of the physician by the end of the appointment.
HOW AMBIENT CLINICAL INTELLIGENCE WORKS
Exam rooms are fitted with a device and microphones that will capture the natural conversation between a patient and doctor
Machine vision technology helps the AI translate gestures into correct medical terminology and enters it into EHR
Doctors can pull up past visit notes when a patient can’t remember when a certain procedure was done by activating the virtual interface
By the time the patient visit is done, ACI has transcribed all relevant notes, performed a quality check, and has prepared the document for approval
Showing a story-board of how this technology would assist doctors gave my client further insight about their investment decisions.
MY ROLE IN THIS PROJECT
My task as the researcher was to assess the current attitudes toward similar existing technologies such as speech-to-text dictation apps and find any comparative health technology adoption cases that might shed light onto near-term adoption behavior.
RESEARCH PROCESS
For this project my client asked me to conduct quantitative user surveys with patients and physicians to understand the needs and pain points of these users when it came to medical appointments.
However, my preliminary domain research revealed that most physicians are affiliated with larger hospital networks and the product would have a hardware component that would need to be installed in the examination room. As a result, I proposed to my client that I also conduct interviews with decision-makers as well as physicians.
In a hospital while doctors and nurses can give feedback as to what equipment to buy, the decision ultimately lies with the C-suite.
I convinced my client, yet because these additional interviews were not included in the original project scope, I had to source the stakeholders on my own. I used my university alumni networks to speak with ten chief medical officers and lawyers to investigate the decision process for adopting new IT equipment and learn about their needs and pain points.
PROJECT CHALLENGES
My client did not understand how the technology worked since it was in closed beta testing and my client had no prior experience with the technology.
The usual survey panel vendor my company worked with went out of business so I had to search for a new vendor that could provide the mix of physicians I needed for this project and still have it be within my client’s project budget.
Due to compliance reasons I was not allowed to speak to anyone at this product owner company or former employees of this company who were employed there in the past five years.
RESULTS
My recommendation to the client was that unless the product could show solid ROI, we would not observe a fast adoption rate by hospitals or private practices given the costs involved. My conclusion was based on the following observations I have made throughout the project:
Doctors are divided on the usefulness of current speech-to-text technologies and are skeptical of any future ambient intelligence systems. Their attitudes appear to be shaped by their training and field of specialty.
Privacy is a concern for patients, with older patients appearing less comfortable with the use of ambient intelligence than younger patients.
Hospital administrators’ primary concern is cost savings and efficiency. Hospitals have invested a lot in their Electronic Health Record (EHR) systems. They want to figure out how they can use EHR data to cut down on waste and increase efficiency.
“We have a lot of older patients and they don’t like being recorded. They are really wary of it. For our hospital, we wouldn’t want to adopt such technology. I think it really depends on your patient population.”
— Hospital Staff Member in CA Health System
“Is it going to seamlessly integrate into my hospital’s existing workflow? I have never heard of a software that gets it right the first time. We adopted EHR seven years ago, and are still trying to configure the workflow.”
— CMO at a university affiliated hospital in CO
IMPACT
During my client hand-off meeting, the client disagreed with my assessment, because they interviewed some beta-tester physicians who had positive impressions of the product. I disagreed with my client for the following reasons:
Funding Bias
These users got access to this expensive product without having to pay for it, and therefore, their impressions might be biased towards the product.
Domain Bias
The interviewed physicians work in orthopedics, which happens to have more structured patient visits compared to other fields of medicine, while the physicians I surveyed and interviewed were practitioners of various types of medicine.
Business Bias
The physicians my client interviewed were affiliated with a private practice that marketed themselves as early-adopters of the latest technology, yet looking at the adoption curves and the timeline of adoption rates I was investigating, this group of physicians do not represent the average adoption behavior.
CONCLUSION
After discussing our findings I was not fully able to convince my client about my research results. However, by the end of the meeting they decided that it might be good to have another iteration on this research in six months after the product was released. Despite our disagreements on the results of our research, I was able to make my client be more cautious about their investment decision and delay it for two quarters.
REFLECTIONS ON THE PROJECT
This project taught me how to push back for the client even if I was pushing back against my client’s priorly held beliefs.
Talking to different stakeholders taught me about how the pain points felt about current products can still carry over to unreleased future products slowing adoption rate. This project was also the project that got me interested in UX design.
I learned about legacy systems used in the healthcare landscape and how that might affect the addition of any new future software to the already existing workflows.