The big buzzwords right now in healthcare information technology are artificial intelligence (AI), machine learning, and deep learning. Because there is so much noise, I would like to start off by defining what AI is not. AI is not a doctor. AI is not simply a set of rules. Rather, true AI for healthcare is a model based on clinical standards that aids clinicians and health organizations across the spectrum to deliver one thing: quality care to the patient.
Lightbeam recently published a press release announcing the integration of AI, machine learning, and deep learning functionality into our population health management platform, enhancing its predictive analytic capabilities. We have integrated a full-tech suite, providing clients who want AI-calibrated insights in their organizations with results. Lightbeam’s AI capability is learning from the data we have accrued for the entire longitudinal history of our 17 million patient population and over 1 billion data elements.
AI is delivering value. We have already seen great use cases and potential in healthcare. For example, one that Lightbeam is offering is the identification of pre-diabetes and people with undiagnosed diabetes to support CMS covered programs such as the Medicare Diabetes Prevention Program (MDPP). A future use case that Lightbeam will be offering is predicting preventable visits to the emergency department or admissions to the hospital before they occur, as well as the prediction and treatment of other chronic conditions.
AI can greatly augment the ability to identify and manage high-risk populations. Population health vendors should be using AI to identify:
- Patients who are undiagnosed with a chronic disease at this exact moment and are not currently being treated
- Patients who have an extremely high likelihood of developing a chronic condition in the near future, where both the patient and physician may be unaware of the warning signs
- The best treatment pathway that is most likely to make individual patients better in the shortest amount of time
What our data scientists have built is a true deep learning engine, not another rules-based engine. This engine is constantly learning and refining itself through the flowing of new data and information. For those interested in how it works, the AI framework begins with the foundation of clinical informatics and then adds a combination of AI/ML techniques such as CNN (convolutional neural networks) and recommender systems. Once the data is analyzed, the identification model provides a list of patients who are at risk for a certain condition and their probability of being diagnosed within a certain timeframe. The platform is very adaptive to the latest clinical guidelines, for example, it was able to incorporate the recent landmark hypertension guidelines within five hours.