Clinical Trial Matching
How might we leverage machine learning to improve the efficiency and accurate of clinical trial matching processes?
BACKGROUND
IBM was developing a clinical trial matching product with partner hospitals. The product design team was tasked with discovering opportunities where the product could transform and optimize the clinical trial workflow, then envision a product experience and help devise a product roadmap.
the team
The clinical trials and genomics design pod consisted of
A design and delivery lead (myself)
2 user researchers
1 UX architect
1 front end developer, and
supported by SMEs, IBM fellows, and the product team, who oversee the entire oncology effort
our Approach
Our daily mantra was “to solve the human problem” to keep us grounded on humans. The solution itself needed to account for organizations, players, regulations, processes, data flow constraints, and data needs for both users and cognitive engines. In addition to the experience and process design, we are also the discovery team for opportunities where our products can improve efficiency and accuracy in a clinical trial setting.
PROCESS
2-week design sprints using IBM Design Thinking framework, the team conducted activities such as collaborative workshops and rapid prototyping to iterate on the product and process design.
Each design sprint consisted of user interviews, design iterations, rapid prototyping, and playback.
Outcome
Workflow
A key outcome for discovery is a generalized clinical trial workflow from trial inception to recruitment. This workflow enables the team to discover product opportunities and growth strategy.
Product Launch and speculative roadmap
As per product requirement, MVP was launched in partner hospitals in August of 2016.
The team also provided proof of concept designs to inform future iterations. Each iteration is designed to work with speculated Watson development, mapped to the most valuable workflow moments where our product can bring the most improvement of efficiency and accuracy to the clinical trial matching process.
MVP/General Release
Only used by trial nurses on site
Ingest one protocol at a time
Document coding isn’t real-time
Fully functional product
Usful for multiple touchpoints in the clinical trial process, from site selection, to enrollment optimization across several trials, to patient matching
Ingest multiple protocols at a time
Fully optimized product
Real-time and fully automatic protocol ingestion
Can aid with site selection and enrollment optimization with synopsis alone
Real-time, bedside patient matching