SimpleTrials

SimpleTrial - banner
SimpleTrial - screenshot

Inspiration

Finding relevant clinical trials is a daunting task for patients, often relying solely on doctor recommendations which may not encompass all available options. Similarly, as a busy doctor or clinical research assistant providing healthcare and facilitating clinical trials, it's challenging to adequately educate patients amidst juggling multiple trials, or get in touch with patients who could benefit from participation but are not within their immediate reach.

SimpleTrials addresses these hurdles by offering simplified, easy-to-understand modules, empowering patients to explore trial information independently. By bridging the gap between patients and trials with clear language and accessible resources, SimpleTrials enhances patient autonomy and facilitates informed decision-making in healthcare journeys. Moreover, by enabling patients to find trials, SimpleTrials contributes to accelerating clinical research, enhancing speed, and overall participant experience, thereby driving innovation in healthcare.

Special thanks to my friend Annie, who is working as a clinical research coordinator at UCSF, for discussing the project idea with me :)

What it does

SimpleTrials is a clinical trial dashboard designed to empower users without medical education background in navigating and understanding the clinical trials.

The journey begins with users inputting search parameters such as their condition of interest, type of intervention, or preferred location. The default settings prioritize ongoing trials, allowing patients to explore immediately available options.

Results are then categorized into three main sections. Firstly, the Data Summary provides users with essential insights into the trials, including study type, phase, sex requirements distribution, and trial locations, offering a visual overview to navigate the trial information. Next, the Filter and Explore trials feature enables users to interact directly with the data, allowing for trial shortlisting and column filtering based on specific keywords, with key information regarding study aims and eligibility criteria readily accessible. Finally, users can engage with Language Model Models (LLMs) agents - summarizer and comparator - to simplify complex trial information, with options to learn about individual trials, receive summaries of eligibility criteria, the point of contact for each trial, and compare two trials using their NCT IDs. These features empower patients to make informed choices and explore potential trial participation.

How we built it

During conceptualization, the primary objective was to empower patients to explore clinical trials with ease. This guided the approach to presenting data, favoring clarity over complexity. The importance of simplicity in language was recognized, prioritizing patient understanding over precision in wording. Leveraging Language Model Models (LLMs), prompts were engineered to translate intricate medical concepts into easily understandable language for patients without medical backgrounds.

In designing features, focus was on creating an intuitive user experience. Visual representations of trial data were prioritized to provide users with a quick overview, followed by specific trial details essential to patients, such as summaries, eligibility criteria, and contact information. This user-centric design ensured that users could easily access the information they needed to make informed decisions about their healthcare journey.

Sourcing resources involved utilizing clinicaltrials.gov APIs to access raw data. This allowed for further processing and analytics locally, enhancing the user experience by providing accurate and up-to-date information about clinical trials. Integration with GPT APIs facilitated the seamless incorporation of LLM prompts into the Streamlit app, enabling patients to explore trial information effortlessly.

Overall, the development approach was driven by a commitment to simplifying the clinical trial exploration process for patients, fostering informed decision-making, and improving accessibility to medical research opportunities.

Challenges we ran into

Breaking the app result page down into modules was not as straightforward as it seems. By narrowing down the focus to empowering patients as the app's current scope, it was clearer to me that LLM agents can be useful in the final explanation steps. However, it is important to give users without medical background a soft intro to the clinical trial information, hence a general data summary with visual graphs makes sense.
I discovered the hackathon and started the project relatively late (~a week prior to deadline), hence needing to spend a few days 'hacking' the solution. But I think that fits the theme of a hackathon 😉 .

Accomplishments that we're proud of

Built the app within a short period of time! Around 2-3 days of concentrated dev time.
Leveraged different APIs (from clinicaltrials.gov to OpenAI APIs) - built confident for making more complex LLM/RAG apps in the future.

What we learned

Clinical trial data - I didn't have much experience with the trial data prior to the project, so from this project I learnt about its structures and what it offers. It could be a valuable (and perhaps underutilized resource) for complex clinical trial data analysis (when combined with additional result data from research papers, from drug info sites etc.) and inspires me to try prototyping additional solutions for clinical trials!
Streamlit app dev - very easy to use and customizable!
Integrating apps with LLM APIs - how to handle secrets safely etc.

What's next for SimpleTrials

Enhanced User Experience: To further reduce barriers to understanding, SimpleTrials should allow easy search of specific terms on the page, empowering users with greater comprehension. Additionally, enabling users to save previous explorations and LLM-generated summaries will streamline their journey.

Seamless Patient-Provider Contact: We aim to facilitate direct communication between patients and trial providers by implementing a single-button contact feature. This will allow patients to easily reach out to trial managers via email or phone directly from the trial details page, fostering seamless engagement.

Deeper Analytics: Introducing custom scoring mechanisms to quantify patient burden will enable us to assess the complexity of trials more comprehensively. By evaluating criteria related to personal information and clinical lab data, we can identify potential barriers to trial participation and assist in streamlining trial designs. These analytics tools will not only benefit patients by identifying suitable trials but also aid researchers, trial designers, and healthcare professionals in optimizing trial protocols and understanding trial landscapes more effectively.