In the spring of 2020 and in the dawn of the covid-19 pandemic, the world came to a near standstill. Due to lockdown measures, the Civic Lab at Voice Tech Global had to leave our workspace at the Toronto Reference Library. As the director of Civic Lab, I was happily surprised that through the dedication of our passionate team we were able to continue our work without skipping a beat. It’s in this new remote world, I started working on an interesting project with one of our members, Patrick O’Neill, founder and developer at Toronto Chatbots.
We began working together on what became Covid Check; a rapid covid-19 assessment and scheduling chatbot and voice application. While we did not release the product, we wanted to take some time to share our research, insights and product ideas, in hopes that they might help inspire others to improve existing market solutions. To begin this project we asked ourselves the following question:
How might we leverage conversational AI to help support healthcare professionals and public members during the pandemic?
To take on this challenge, we started where we always start–with understanding our users’ needs. For this project, we wanted to define a value proposition that created value for both of our user groups; members of the public and for healthcare workers as well. To begin our discovery process, we held 1:1 interviews with a variety of different participants and documented what we learned.
The needs of healthcare practitioners
To understand the needs of healthcare workers, we interviewed an emergency room doctor, a family doctor, a telehealth nurse, and a surgeon. We also interviewed subject matter experts, including a policymaker and a product manager who were working in this domain. There are many challenges that healthcare providers are facing during this pandemic. These may be friction points or bottlenecks in service that we wanted to understand better. After our initial research, these were the themes we uncovered.
Bottlenecks The systems for telehealth triage, diagnostic testing, and vaccinations are being overwhelmed, causing bandwidth bottlenecks.
Integrated solutions New products need to integrate easily into the existing system, which often prevents them from gaining adoption.
Scaling services Human-driven call centers have not been able to scale to meet the demand for telehealth systems and diagnostic testing.
Repetitive education Providing ongoing education and conducting triage through assessment questions are core responsibilities for some healthcare practitioners. This dissemination of information and service can often feel repetitive and monotonous.
Language barriers Language barriers are a significant challenge when providing services to diverse populations. There simply aren’t enough multilingual telehealth nurses.
The ‘Worried Well’ The swell of “Worried Well”. This term describes the effects of anxiety on the public, which can cause paranoia and confusing psychosomatic symptoms.
Data governance In healthcare, it is critical to support good data governance practices. There are regulations to ensure PII is protected and any solution needs to abide by them.
The needs of the public
To help us understand the public, we spent time interviewing friends and family who had undergone testing for COVID-19 or had contracted it. We also researched social media spaces and forums better to understand the public’s general sentiments and concerns. When people are worried about whether they have COVID-19, we learned that they would often call telehealth services for advice and support. We also learned about extremely long wait times, overburned healthcare workers, and several accessibility issues.
Long hold times Unacceptable hold times on telehealth lines: During high infection rates, wait times can exceed 6 hours.
Evolving recommendations and guidelines People found the rapidly evolving recommendations to be confusing and were unsure of where to get authoritative information and guidance.
Lack of tech literacy Many people lack the technical awareness or know-how to access certain chatbots, voice assistants, mobile app and web-based solutions.
Language barriers Non-anglo residents may be left out of government-sponsored solutions if they are not available in their spoken language.
Access to technology Financial and housing insecurity or fixed incomes may prevent people from accessing solutions if they require expensive digital technologies.
Excluded Solutions are often not always accessible to people with disabilities.
After synthesizing our research, and identifying unmet user needs and opportunity areas we generated many concepts and converged on our concept–Covid Check. What we’ve built is a multimodal assistant which addresses a wide range of the highest priority needs for both of our user groups. It is an integrated, scalable and user-centred solution.
Covid Check is a voice interface and chatbot that gives members of the public instant access to diagnostic pre-assessment that can triage callers for patients who need urgent care, allowing them to bypass long telehealth wait times.
- Multimodal and accessible on a wide range of devices. Including telephony support.
- Seamless handoff to telehealth for most urgent cases.
- Using QR codes to handoff test results and integrate them into hospital systems
- Available in 44 languages to better. serve the non-English speaking population.
- Patient follow-up for easy and quick access to test results.
Hand-off to Telehealth
Callers with ‘red flag’ symptoms that put them at immediate risk are identified early in our triage and the conversation is handed off directly to a telehealth nurse.
QR code integration
A QR code is given to users after their assessment which can be leveraged as proof of recent assessment or can be scanned at a testing site or hospital to speed up processing and triage. A six-digit alphanumeric code provides a redundant system that is accessible to voice-only integrations like landline telephones.
During pandemic waves hospitals have seen bottlenecks at testing sites with long lines. Covid Check works to alleviate this by booking tests for users using time slot appointments that are spread out. This is done by leveraging location data and a scheduling backend service.
Available on a wide range of devices including; smart speakers, smartphones, web browsers and landline telephones. Chatbots and voice agents can reach those who have connectivity and access to technology, but not everyone has access. By deploying across every channel including telephony COVID Check can be made available on a 1–800 number, increasing access to some of the most vulnerable populations.
We designed Covid Check to create value for both healthcare practitioners and the members of the public and to make a difference across the entire user journey with preventative education, assessment, symptom tracking, testing and vaccination scheduling. At home, Quick Assessment gives instant access to a diagnostic pre-assessment and can hand-off to telehealth when necessary. The QR code carries a user's assessment details into the hospital, testing site, or vaccination clinic for scheduled time slots–this streamlines admissions and eases bottlenecks.
User Journey | From the home
User Journey | At the Hospital
Data Governance & System integration
We learned in our research the importance of protecting user data and ensuring governance of it, as well as designing our product in a way that integrates with existing patient record systems in a frictionless way. We came up with a relatively simple solution that I’ll describe to you.
In Dialogflow we created entities for each covid condition. When the entity is triggered during the conversation, Dialogflow will create a parameter that the system can use. These parameters are passed to the database firebase and associated with a random 6-digit alphanumeric code that we encode into an associated QR code. While the user's PII (Personally Identifiable Information) is never associated with the collected symptom data the QR code allows users to hand off their results of this test quickly, and seamlessly at the hospital or testing site. With their consent, it can then be integrated into their personal health records. This process remains independent of our database however and completely within the existing patient record system.
To accommodate this data transfer process our Firebase database is synced with a Google Sheet. We can then extract this data into a Google Document that has been preformatted with Google Scripts. This information can be easily read by a healthcare practitioner and printed out and inserted into patient records.
Demo of Covid Check
Here’s a short demo of the beta deployment of the Google Action version of Covid Check. We used the name Quick Assessment because Google has restricted the use of the word ‘Covid’ in what seems like a preventive measure against potential misinformation.
The hurdles to market for healthcare products like this are a minefield of regulations and red tape. In light of this and after some consideration, we’ve decided that instead, we’d like to offer our research and ideas to anyone working on similar efforts, in hopes that it might be helpful in the global effort to curb covid-19 and support vaccination efforts. I gave a presentation on this project to the Civic Tech Toronto community and also gave a private shareout of our work with several of the healthcare workers we had interviewed, including a product manager for Toronto's COVID-19 online assessment tool.