Probabilistic Reasoning Technology : Ada’s AI-powered Health Companion
Ada’s AI-powered health companion helps individuals understand and manage their health. In practice, Ada works much like a GP in your pocket, asking relevant, personalized questions, enabling people to self-assess their symptoms and identify possible causes and appropriate next steps. Ada is supported by a sophisticated AI engine and curated medical knowledge base, which also augments doctors by providing earlier health information and decision support.
Ada’s deep medical foundation and our approach is really what distinguishes us from similar applications. Our AI platform has been built over six years of research and development, enabling us to refine our reasoning technology and build up a very robust medical knowledge base covering many thousands of real medical cases, conditions, symptoms and findings. Doctors have also played a key role in Ada’s development and continue to contribute to Ada’s learning.
Our bespoke AI uses a probabilistic reasoning technology, reviewing multiple pieces of data to output probabilities. This means that Ada can consider a variety of diseases to make a health assessment, providing the most likely explanations for a user’s symptoms. We continue to update our medical knowledge base with the latest information, and we’ve integrated machine learning techniques that enable Ada to learn from every interaction. Ada doesn’t exclusively rely on information from one type of source.
Over 2M assessments have been completed on Ada, and as Ada’s user base continues to grow and incorporate different data sources (i.e. medical history, sensors, lab tests, etc.), Ada is continuing to become more intelligent, putting us ahead in the market.
Ada’s Foundation: Grounded in Medical Precision & Human Insight
Ada’s technology was initially designed to support medical specialists with diagnosis, but early testing and subsequent healthcare data quickly made it clear that there was a real need for patient support before individuals even stepped into the doctor’s office. We chose to start with medical experts, because we wanted to make sure that Ada’s reasoning and knowledge was intuitive and accurate, and then continued to build on that, expanding Ada’s reasoning system to GP’s, and finally translating our core content, knowledge base and technology into a tool that would also be accessible to patients, and culminating in the consumer-facing personal companion app that you see today.
Data played an important role in helping to inform the design of our health companion app. We conducted an extensive market research study on health care attitudes and the perception of e-health tools, along with surveying doctors, family and friends. What we learned was that there was a real demand for easily accessible, credible health information and something more personalized than a broad web search.
So when developing our health companion app, we took this into account in the design. We really wanted Ada to be more personal and work like a human doctor with no time pressure- friendly, conversational and underpinned by medical precision. We are very excited about Ada’s future and are continuing to evolve our product to incorporate the feedback from the people who use Ada on a daily basis. Their feedback remains an integral part of how we write content, design solutions and the inspiration for how we’ll continue to evolve Ada’s capabilities in the future.
Diagnosis Support and Population Health Benefits
Already, we have learned much from collaborating with research institutes. First, we have gathered medical knowledge and insights about individual conditions which has been incorporated into our service, even before the medical research is published. Second, we have also been able to verify just how useful Ada is in terms of furthering the capabilities of doctors for diagnosis, particularly for rare diseases.
We’ve done retroactive studies with specialized research institutions, and have found that in many cases, if the patient had used Ada at the onset of initial symptoms, the patient could’ve saved years of suffering and been directed to the appropriate specialist for treatment. There’s a lot more potential for Ada in this area, and as Ada continues to build its user base, the possibilities of its population health benefits will grow substantially – including the potential to identify and track health trends and symptoms, patterns of disease and flag potential health risks early on.
The Future of Healthcare: Increased Digitization and Patient Empowerment
In the future, Ada will continue to provide increasingly more sophisticated and personalized health assessments and will also become more of an ongoing health companion, allowing individuals to have more responsibility and control over their data and their health, a lot of it even from home.
Consider this: a woman with one or more long term conditions might use Ada to monitor symptoms on a daily basis and to continuously track a range of data points such as heart rate, blood pressure, body temperature and activity levels. In addition she might choose to enable regular automated updates of her Ada profile with lab results and other clinical findings following interactions with her doctors. By gathering data over time, Ada effectively ‘gets to know’ the patient, a bit like a regular GP does, and in doing so is able to provide increasingly relevant and personalized insights and suggestions for how individuals can manage their symptoms and long-term health more effectively.
Eden Duthie is the Head of Data Science at Ada Health in Berlin and is originally from Melbourne, Australia. He has had an extensive career applying machine learning to disrupt a wide variety of industries such as vehicle manufacturing, drug discovery, weather forecasting, and sports betting. He was also the original developer of the Kaggle competition platform. Eden leads the data team at Ada where they apply machine learning to primary care patient-focused medicine, developing conversational health assessment services. He believes that healthcare is in the initial stages of world-changing advances based on the availability and flow of data.