Artificial intelligence (AI) – long an impactful technology in other industries – is in the midst of rapid adoption across the healthcare industry. What was once seen as having great potential is now making its way to real-world implementation, driving improvements in care through enhanced clinical decision support, empowering payers, providers and healthcare organizations across the ecosystem with actionable insight.
Inovalon’s William Kinsman, senior manager of Product Innovation, recently sat down with Dr. Gordon Gao, co-director of the Operations and Information Technologies Department at UMD and renowned AI expert, to discuss the current uses of AI in the healthcare sector and the projected impact that this dynamic technology will have on the healthcare industry in the near future. Gao also touches on some of the projects he’s worked on at UMD, that focused on the evolution of AI – specifically the adoption of AI in healthcare. Dr. Gao was candid about his thoughts on the impact of adoption of AI in healthcare and how UMD is using the technology to further its research goals and objectives, as well as the future impact he thinks adoption of AI will have on the industry.
In the past year, research students from the Center for Health Information and Decisions Systems (CHIDS) at UMD conducted a study on the effect of mobile health apps on patients with diabetes. The main objective of the research was for students to gain insight into how personality differences might explain why mobile health apps help some patients with diabetes more than others. The CHIDS research team conducted a 13-week clinical trial, during which they designed and tested an app for managing diabetes; the team discovered that users who leveraged the app were more active in keeping up with treatments, diet and exercise than the users who rarely used the app. Greater use of the app was also associated with greater reductions in HbA1C — a key clinical indicator for diabetes. The study also revealed that users who used the app frequently started to experience a decline halfway through the trial, suggesting that perhaps participants got bored with the app.
The team is also focusing its efforts on big data. Specifically, big data, typically found in unstructured data sources such as social media sites and CRM ratings/comments. These sources are often very difficult to analyze; due to the breadth of this type of data, it often goes underutilized. CHIDS is researching adoption of AI in healthcare to augment these data sources and provide payers and providers with key insights on social determinants of health, behaviors, mental health, patient intervention and drug adherence.
Patient-generated behavioral data is becoming increasingly common for care management. Smart gadgets such as Fitbits, smartphones and other wearable sensors collect patients’ data, helping patients and physicians better understand habits and set realistic healthcare goals. While data garnered from these sources has been beneficial for both patients and physicians alike, the discrepancies in patient-generated data has caused misunderstanding, making it harder for them to identify opportunities of improvement in the healthcare journey. CHIDS is researching how AI technology can be leveraged to build a better sensory self-tracking to improve the accuracy of the data reported and ameliorating the identification of future areas of development for patients.
Many of the latest AI technology innovations such as facial recognition, self-driving cars, Google Earth and diagnostic imaging in healthcare are all byproducts of deep learning technology. The dynamic technology first emerged on the tech scene in 2006. Today, it is used in innovative image recognition solutions. Deep learning technology is a more focused, advanced version of machine learning, all housed under the umbrella of artificial intelligence. Typically, machine learning technology takes in data and processes it through algorithms, and then makes a prediction; deep learning use of neural networks allows it to make predictions and decisions in real time, without the delayed thinking stage that machine learning technology often must go through. Because of this capability, smart phone users can unlock their phones using face recognition, without waiting for the technology to confirm their identity. While deep learning has made major improvements in the tech world, there’s still more progress that needs to be made with image recognition.
Since the early 2000s, we’ve seen deep learning perform well in image recognition and in game environments. Applying models like these continues to be a challenge in the healthcare space; however, image models are making leaps and bounds in augmenting radiology tasks.
Language processing is a particularly clear and present application as well. The increase of adoption in computers in patient documentation at the point of care, and the increase in performance of models in this field is particularly progressing, and that we see as fantastic.
All of them will contribute and already are today. With so much happening in the industry, it is hard to say exactly where the next big breakthrough will come from. On the provider side, the electronic health record vendors such as Cerner and Epic are very aware of this technology and are keen to work with APIs to drive use cases early on within their computer-hosted products.
Big tech companies such as Amazon, Google and Microsoft are determined to make waves in this space. Last year it was reported that Amazon is partnering with JPMorgan Chase and Berkshire Hathaway to pursue a new healthcare venture named Haven. I think the creation of new organizations like Haven are creating new opportunity.
In an August 2018 issue of Nature Medicine, researchers published a report that focused on the use of AI for the diagnosis of retinal disease. The implementation of the technology has effectively addressed the challenge of reviewing complex diagnostic imaging at a pace that experts were unable to keep up with; allowing clinicians more time to assess each patient and provide a referral recommendation to a specialist. This is like the CT machine, where with its invention, we are now able to see things that we could not previously see on a patient with the naked eye. Deriving new information about a patient that was previously not available will incrementally improve each patient’s quality of health in a unique and positive way.
Overall, we are excited to see the increase in demand for innovative AI solutions. The ease of use of these tools and the increase of adoption of AI in healthcare is opening more unique opportunities quickly, further driven by the collaborative development culture surrounding it. With so many advancements happening in the industry, it is safe to say that the use of AI and emergence of big tech in the healthcare space is not going to slow down any time soon. If anything, it is just getting started.
Click here to watch Doctor Gao and his colleagues discuss the impact AI has had on the healthcare industry, common misunderstandings and speculation surrounding AI’s ability to add tangible value to the healthcare space, and how the technology has enabled real-time access to the right insights at the right time to drive improved patient outcomes and economic performance.