Artificial intelligence (AI) is an inevitable force of change in health care. Nearly every day there are advancements asserting that AI and machine learning (AI/ML) will help us better diagnose and predict disease, uncover new drug therapies and deliver the precise patient information physicians need at the point of care. Although these ambitions may be possible, thorny challenges stand in the way of a truly AI-enabled health care system. We hope to shed light on some of these challenges, where AI is already having an impact and what patients can expect today — and in the years to come.
What AI can do
For all the hype, it’s important to remember AI isn’t magic (or robots coming to replace your doctor) — it’s just math. Terms like “machine learning” and “deep learning” are simply ways of explaining statistics-based computer algorithms. These algorithms need a lot of data to identify patterns and become powerful prediction tools. For instance, you’d need a lot of patient data to train an algorithm to recognize the markers of sepsis, and then use it to predict sepsis in future patients. Although more health care data than ever are available via electronic health records, remote monitoring tools and genomic tests, the data still tend to be siloed, messy and proprietary. Patients ultimately benefit if researchers can draw from the data of patients worldwide instead of data isolated to a single institution — and AI can help encourage collaboration for scalable impact. As Google Ventures partner and physician-scientist Dr. Vineeta Agarwala noted, “AI is uniquely capable of forcing data silos to break down and forcing large institutions, which have historically had legitimate fears about data-sharing, to finally see a source of self-interest in it.”
As an early stage health care technology investor, we see many startups leveraging AI/ML. From 2011 to 2017, investors poured $2.7 billion into AI/ML digital health startups. While media attention has largely centered on AI/ML’s ability to revolutionize the delivery of clinical care, the technologies are even more quickly transforming the business of health care. Companies are creating more efficient processes, from the drug development pipeline in pharmaceuticals to clinical operations, scheduling and documentation in the hospital. While exciting, we think some of the more clinically-focused use cases — like diagnosis and robotic treatment — will take longer to build and scale due to the risk of integrating new technology into direct patient care.
Why you should care
What might this all look like for you? Though the algorithms may be hidden, AI has likely already touched some part of your care. Many consumer-facing products integrate AI to offer tailored recommendations. For instance, symptom-checker apps like Babylon and Ada are powered by machine learning algorithms, while Woebot offers people emotional support via an AI-powered chatbot. Other solutions help providers manage their patients. For instance, Omada Health uses machine learning algorithms to help health coaches predict which of their pre-diabetic patients could use additional support sticking to their nutrition and fitness routines. And many hospitals use predictive algorithms — like those offered by AgileMD — to monitor patients and flag emerging clinical issues.
We live in a world with an unprecedented and growing mass of health care data collected at-home and on-the-go, from our phones, wearables, sensors and voice tools. To capitalize on this treasure trove of information, we’ll need AI/ML to comb through the data, find patterns and derive insights that can be used to improve health. As investors and digital health enthusiasts, we’re eager to support companies applying AI/ML to create a more efficient, accessible and intelligent health care system.
