Insight

The availability of patient information is exploding, which in turn is fueling the creative minds of IT leaders that collect, index and store healthcare information. In response, business leaders are collaborating with IT architects, application designers and data scientists to develop their organizations’ analytical skills to deliver breakthroughs in disease management and other healthcare-related processes. The trend began, of course, with the electronic health record (EHR). Now, information is being made more robust by data from personal health devices (such as Apple Watches and Fitbits) and other remote monitoring devices. 

Add insight from genomics (which in terms of sheer volume rivals Twitter, YouTube, and astronomical data^¹), and you have an environment flush with oceans of data, which can often paralyze an organization. With AI, it’s quite the opposite as volumes of unstructured data sets keep flowing across a wide number of systems. Interest in AI, to correlate information, recognize patterns, and generate actionable insights, are all trends that are explored in a recent comprehensive, in-depth study of the market (which will grow to $6 billion by 2021²).

Organizations such as 23andMe craft ethnic profiles from genetic material they receive from curious consumers. Many consent to have their information used for research, aiding scientists in their understanding of what makes us sick. Scientists have long suspected that genes play a role in attributes such as intelligence and susceptibility to diseases like cancer, Alzheimer’s, and autism. But, the more complex these interactions become, the less we understand them. 

This is where AI’s deep learning capability enters the picture (the process by which an expert system ingests data then uses its extensive knowledge drawn from analyzing other data insights to further refine and interpret its knowledge). 

Using emerging AI techniques, scientists are finally able to process the three billion base pairs of DNA that comprise an individual’s genome along with the genomes of others. One of the major facilitators of this process is the Precision Medicine Initiative. Launched in 2016, the initiative engaged a million people to volunteer their health data to help enable a more precise medical treatment plan. Software from Deep Genomics has been deployed to help scientists predict the effects of a particular mutation based on its analysis of hundreds of thousands of examples of other mutations.³

As machine learning refines predictions from DNA databases, in vitro specialist Nathan Treff is working on ways to get ahead of birth defects, using a combination of computer models and DNA tests. His company, Genomic Prediction, was founded on the idea of predicting which IVF embryos in a laboratory dish will be most likely to develop type 1 diabetes or some other complex disease. DNA testing in IVF clinics to spot rare diseases has been in place for years, but Treff’s preimplantation tests look more deeply at an embryo’s genome to create more informative statistical forecasts about the person it would become.⁴

AI enables breakthrough operational efficiencies

Discovering new insight that wasn’t possible in a pre-digital age represents a breakthrough achievement. But, such insight is diluted if it’s not acted upon to deliver an equally impressive breakthrough result. Magellan Health is one provider that is applying disease insights from AI to operational inefficiency by applying expert systems to its large datasets. As a result, Magellan is rapidly becoming a more disciplined data-driven organization, which it reports is delivering higher quality patient care at lower cost.⁵

Expert systems have improved operations at Magellan Health by bolstering clinical quality and its associated outcomes. AI systems have also increased the consistency and productivity of care givers. Systems are hosted on a HIPAA-compliant, highly secured Google Cloud platform (a technique that any medical practice can use to reap its benefits). Another example of how AI helps operational efficiency is Watson Oncology (from IBM), which sorts through thousands of cases and articles, helping the physician to decide the most effective treatment for a specific type of tumor.⁶

AI helps reverse the high cost of readmissions

Lowering the high cost of readmissions is particularly relevant since Medicare and many other payors are no longer covering their costs, as they believe they are preventable. Magellan’s AI platform identifies patients nearing the end of their stay and recommends keeping them a bit longer to reduce the high cost of readmission. The healthcare provider also uses AI techniques to track discharged patients in order to garner additional insight into relapse prevention. 

The cost of readmissions can be particularly high in cases of patients suffering from chronic conditions. At Intermountain Healthcare, AI is helping teenagers with Type 1 diabetes transition to self-care as they grow into adulthood, using a tool called Cognitive Cloud (from IBM), which delivers actionable insights for those suffering from conditions that cannot yet be cured.⁷

AI predicts likelihood of disease

Operational efficiencies that improve healthcare delivery, while lowering readmissions, represent a welcome breakthrough, but even bigger benefits can be realized by disease prevention. Israel’s second largest health maintenance organization, Maccabi, is doing just that with its ColonFlag AI platform, which predicts 100 different conditions that can lead to colon cancer, stomach cancer, diabetes, sepsis, and organ dysfunction—all built using data available from medical records. 

The platform employs machine learning to analyze millions of records, cross-referencing them with other medical knowledge and similar cases. A patient-specific prediction is produced, which peers six months, two years or five years into the future to crystallize the patient’s risk factors. A unique combination of algorithms and storage formats lets Maccabi work with large datasets to return accurate cognitive insights.⁸ For example, cognitive algorithms work with conventional algorithms that have been deconstructed and rewritten to optimize performance.

Medial EarlySign, founded in 2009, uses AI to improve patient health outcomes by isolating extraordinary clinical insights hidden in existing blood tests and EHR. In a study based on data from 645,000 prediabetes individuals, the technology discovered that by isolating less than 20 percent of the prediabetes population, its algorithm pinpointed 64 percent of individuals who became diabetic within a year.⁹

The FDA readies for change as digital medicine accelerates

The FDA is creating a new unit dedicated strictly to digital health in response to a growing workload that requires faster response than the organization is equipped to handle. Once again, AI is being assessed as a solution to the task of reviewing and garnering insight from huge volumes of text and reviewing applications to bring products to market. 

Until now, it would take manufacturers years to get products ready for regulatory approval—and FDA reviewers could barely keep up. Now, as computer code is taking on more tasks (like spotting specious moles and quantifying blood flow), the workload is becoming overwhelming since software developers take months, not years, to ready a product for market. Through its digital health unit, the FDA is hiring software developers, AI experts, and cloud computing experts in order to prepare the agency to regulate a future in which health care is increasingly mediated by machines.

AI and robotics

In San Francisco, Dr. Cory Kidd (CEO and founder, Catalina Health) has built an AI-powered robot dedicated to motivating its owners into positive behavioral changes. Kidd’s invention, named Mabu, is a desktop companion with a touchpad on her belly. As a personal healthcare companion, Mabu helps patients manage chronic diseases. With her wide green eyes and yellow skin, she is often compared to one of the personified feelings in the Pixar movie Inside Out. But, as we live longer and put more stress on an already overburdened healthcare system, AI and robotics may provide a much-needed contribution. 

Says Mabu user Martha Singleton, “She’s here to mainly remind us of our medications. She’s a robot, but in a way, you feel that she actually cares when she does things that a friend would do, like calling my doctor.” As society moves towards an alarming scenario where there are not enough people to deliver healthcare services, robotics and AI could provide a viable solution to fill the gap by monitoring your health, sending encrypted data to your doctor, and reading your emotions. Robots won’t replace the healthcare worker—rather, they’ll augment the capacity for healthcare workers to handle larger numbers of patients.¹⁰

Virtual therapists help with depression

Another expert system that responds to the growing demands of healthcare providers is Woebot, a chatbot that helps treat depression by integrating advances in natural language with cognitive behavioral therapy.¹¹ The resulting virtual counselor immediately informs patients that their sessions are confidential, while offering suggestions if a situation appears to grow serious. 

People who have tested Woebot say he is able to understand a wide range of answers as he checks in with them every day, directing users through easy-to-follow steps. For example, when told about stress at work, Woebot offers ways of reframing one’s feelings to make them feel more positive. Clinical research psychologist Alison Darey, who invented Woebot (along with other Stanford psychologists and AI experts), reminds us that therapy is conversational, hence a natural way to receive emotional support. Darey also notes that people seem happy to suspend their disbelief and enjoy talking to Woebot as if it were a real therapist.

Conclusions and recommendations

Use AI to score a quick win

Don’t forget AI’s low hanging fruit, namely the use of deep learning to train sets of medical image algorithms to efficiently read x-ray studies, MRI exams, and computed tomography (CT) scans. AI for imaging is an especially promising area because computers and deep learning algorithms are getting better at recognizing patterns.

Expand your scope to include operations

Apply your investment justification process to show how AI automates and delivers breakthrough efficiencies in complex, inefficient, and frustrating administrative processes. For example, show how simple medical data saves lives and reduces costs when applied to machine learning techniques that predict an individual’s proclivity to certain diseases.

Educate your team

Executives in every healthcare sector should launch initiatives aimed at strengthening institution knowledge and competence in where and how AI is making breakthrough improvements in disease management and operations. Lean on the AI vendors through RFIs and internal presentations to get started, but make sure you have the skills to cut through vendor hype by building realistic use cases that can be addressed today.

Monitor investment trends

Heighten your understanding of AI trends in healthcare by following investment trends. For example, interest from the venture capital activity is a great way to see where industry shifts are occurring. Leading investors are BlueCross BlueShield Ventures, GE Ventures, and Khosla Ventures. Newcomers include Amazon Alexa Fund and Amgen Venture. Monitor the incubators such as Blueprint Health, which provides both funding and mentorship.¹²