The Virtual Assistant for Clinical Trials

Problem

The challenge for our client, a startup focused on digitalizing daily diaries and monthly diary entries for patients in clinical trials, was to create an enterprise-level, AI-powered, GDPR-compliant chatbot and analytics platform. The client needed a solution that could be used by patients, clinicians, and regulatory personnel in clinical studies. The specific pain points and obstacles included the need for regular engagement and form-filling for 2+ years, the tracking of private/anonymous data, and the requirement for an empathic virtual assistant to track data with a dashboard to synthesize data for doctors and sponsors.

Solution

The SmartCat team addressed the challenge by helping the client define the MVP and optimal trade-off combination to yield the best ROI. The team structured the ML work in usable components to bring value even before the full-blown AI platform was completed. They prioritized functional over non-functional requirements, resulting in a highly available platform prioritizing security and privacy in line with the latest regulatory and industry standards.

The team also ensured the platform’s stability through a suite of over 1000 functional end-to-end and unit tests, allowing for changes and iterations on changing requirements with confidence.

Regarding the AI aspect, the team used a novel approach to overcome the lack of data. They generated a dataset that simulated patient behavior based on several weeks of real user activity. The predictive data model built on top of the generated data showed surprisingly good results. Over time, as more users started using the platform, the pipeline created more value and became more tuned according to real-world data instead of the generated one.

Results

By making smart and hard trade-offs, the SmartCat team managed to put the platform into production on time for a clinical study to start. The platform provided an empathic virtual assistant to track data with a dashboard to synthesize data for doctors and sponsors. The platform was also GDPR-compliant, highly available, and prioritized security and privacy aspects.

Smart Tip

When building an AI-powered platform with limited data, consider generating a dataset that simulates user behavior to create a predictive data model.

Smart Fact

According to a study by the Journal of Medical Internet Research, patients enrolled in clinical trials using digital health tools were more engaged and had better adherence to their treatment regimens compared to traditional methods.

About the Client

The client is a startup focused on digitalizing daily diary and monthly diary entries for patients in clinical trials through a chatbot solution instead of complex forms.

Technologies Used

The project utilized IBM Watson and Java.