Our AI initiative has focused on four key areas:
Our features include the Daily summary, a brief report highlighting all the changes that were made to a patient’s prescriptions in the last 24 hours. By keeping medication data accurate and up-to-date, it ensures that clinicians have the most current information at their fingertips, enhancing the quality of care.
The other feature is our AI Assistant, which is embedded within the patient’s chart. It simulates the experience of consulting a colleague or searching through a patient’s prescription history. Clinicians can ask questions such as whether a specific medication has already been prescribed or if the patient has any known allergies. The AI assistant leverages the patient’s medication record to provide timely and relevant answers, supporting clinical decisions, and offering valuable insights. The functionality of this feature continues to expand as the Better team works to enhance both efficiency and the overall experience for clinicians and patients.
In addition, we also developed multiple internal tools that utilise Retrieval-Augmented Generation (RAG). These tools are designed to help us organise and access our data more efficiently. By combining natural language processing and hybrid search tools, we can interact with our data in a more intuitive and user-friendly manner. This means that our team can quickly find the information they need, ask questions in plain language, and receive accurate and relevant responses, all thanks to the advanced capabilities of RAG. It turns out that the fine-tuned hybrid search, which combines semantic and keyword search, has the biggest utility, as it is very fast and helps navigate otherwise hard-to-find data spread over multiple documents.
In the realm of clinical decision support, we have been exploring the potential of building a system that is 100% reliable. Our approach involves using AI as a supportive tool rather than a decision-maker. The AI provides blueprint sentences, which streamline the process by offering a structured starting point. The critical task of verifying and finalising the output remains with the human experts. This "human-in-the-loop" approach ensures that the final decisions are both accurate and trustworthy, as the burden of verification lies with the human professionals, not the AI.
We have also developed a tool specifically designed to assist us in crafting texts. This tool goes beyond simple spell-checking; it ensures that the texts we produce are in the appropriate tone for their intended audience. Whether we are drafting patient communications, internal reports, or marketing materials, this tool helps us maintain a consistent and suitable tone, enhancing the clarity and effectiveness of our written communications.
This topic was presented at our community event in April. For insight into other interesting topics we discussed, you can download the report below.