Takeda Using Chatbot To Improve Patient-Centricity And Trial Design
By Deborah Borfitz
March 29, 2021 | Takeda Pharmaceuticals has developed a digital chatbot assistant to provide integrated and continuous support for patients with chronic diseases at risk of serious event occurrences and frequent emergency room visits, according to Yaozhu Juliette Chan, director and global evidence & outcomes leader for the company, during a presentation at the recent Summit for Clinical Ops Executives (SCOPE). The learnings are expected to help clinical trials become more patient-centric and less labor-intensive to conduct.
Innovative solutions—including digital health, artificial intelligence (AI), and real-world evidence—have the potential to transform both patient care and clinical development, Chan says. They are integral to a future healthcare world where patients are empowered at the point of care, making it easier for them to take care of their own health via health trackers, wearables, and sensors.
The technologies provide real-time, longitudinal information they can communicate with stakeholders, including physicians, and that will help keep patients informed and motivated, she says. These patient-centric solutions could help patients and physicians work together to “detect abnormalities with actions to immediately follow,” and potentially also lessen the burden on family caregivers.
For clinical development purposes, initial screenings could be done from home and perhaps virtually using these innovative options, says Chan. Patients best know their own needs, which should be “driving decision-making” about trial design and accommodations.
The solutions include predictive analytics using real-world data, such as electronic heath records, to inform protocol design, she says. Mobile phone applications are also available to help patients manage information, track their status, and receive timely support.
Wearable, sensors, and connected devices can additionally enable remote and continuous monitoring, as well as passive listening that can potentially inform digital endpoints and markers, adds Chan. Finally, social media and web platforms can be utilized to garner patient insights about their journey, experiences, and the perceived impacts on their life that can help inform trial design.
A Bot Is Born
As a case example, Chan shares how Takeda developed a chatbot for a chronic disease population with multiple unmet needs. The condition currently has no appropriate treatments, and some clinicians are slow to make the diagnosis.
To better understand the patient experience, and help individuals track disease changes, Takeda developed a digital bot patient assistant toolkit, she continues. The smartphone app is designed to help patients with integrated support.
In the first phase of the chatbot’s development, the focus was on the prototype, Chan says. This stage involved empathizing with patients (i.e., who is living with the disease and how is it affecting their life), defining the main concerns that needed to be addressed, considering the technologies that could be designed to meet those needs, settling on the “doable ideas,” and then testing with feedback from stakeholders.
The second version of the chatbot was tested with hundreds of patients, says Chan. Future enhancements to the patient assistant toolkit is intended to help users better understand their prognosis and provide improved support.
The mobile application is integrated with a digital chat board and dashboard, together with some of the sensor and wearable options, Chan says.
Among features of the patient assistant toolkit is a patient self-monitoring tool for symptoms, episodes, triggers, healthcare resource utilization, and impact on their life (e.g., sick leave from work or school). There is also a tool to surface evidence, such as patients’ symptom patterns over the past weeks or months to support communication with their healthcare provider and human resources at their workplace.
A resource tool provides links to patient communities and information specific to medications, nutrition, mental health, and managing appointments and flu shots, says Chan. The toolkit also includes a remote monitoring tool where the wearable is tracking vital health parameters such as temperature, heart rate and blood pressure, potentially serving as a screening tool for clinical studies.
A dashboard feature provides summary information on the user’s most regularly used and reported parameters, Chan says, such as triggers, symptoms, medications used, and encounters, as well as some information coming from the wearable.
Key features of the digital bot patient assistant tool include profile creation (basic demographic and clinic information), diagnosis history (e.g., age of diagnosis, initial symptoms), disease history (e.g., comorbidities, medications) and disease triggers (e.g., what prompts symptoms, observed patterns).
During each self-defined check-in—which might be daily, weekly, biweekly, monthly, or another customized frequency—patients report on their disease symptoms and, if they have none, “will not be bothered with any further questions,” says Chan. But if they are experiencing symptoms, they are asked about the frequency and severity, medications taken, and any related medical visits.
Quality-of-life questions also surface physical and mental health issues that get tracked on a regular basis, along with educational information on suitable topics, she adds. Patients can get a recent view of their disease experience from a dashboard, such as the number and length of recent episodes, recent physician encounters, and a recap of symptoms and medications used.
Information coming from wearables also gets reflected in the dashboard summary interface screen, Chan continues. Dashboard data is intended to help patients self-monitor as well as enable them to speak “quantitatively” when engaging with physicians and other stakeholders.
Ongoing development for the “practical-use version” of the chatbot will deploy AI to enhance the patient experience, says Chan. For example, the bot is learning to recognize the intent of users based on their free text responses to questions to formulate future interaction patterns with individual patients.
The digital bot and patient assistant toolkit could be applied to multiple diseases where patients need integrated and continuous support, says Chan. For clinical development purposes, the technology could help make trials more patient-centric as well as less labor-intensive from a data collection standpoint. The digital evolution is translating to specific diseases and compounds, and “may start at the very beginning of the whole development cycle.”
As Chan points out in a follow-up panel discussion, understanding patient behavior to improve future trial design and personalize patient engagement strategies will require tapping the treasure trove of data generated outside of clinical trials. This includes passive listening using wearables “to help us hear patient voices via biomarkers” for a more comprehensive picture of individuals’ health status.