Skip to main content
Artificial Intelligence in Healthcare

Understanding how artificial intelligence and the power of deep learning are used in healthcare

In the past 6 months, it feels like the whole world has gone virtual. With office building and schools closed, in-person meetings and conferences moved to digital, online platforms, and leaders across all industries adjusting to a new “normal” for business operations, embracing digital technologies isn’t just wise – it’s a necessity to survive and thrive in today’s environment. During the COVID-19 pandemic, artificial intelligence in particular is being leveraged for everything from vaccine development to contact tracing. Now more than ever, artificial intelligence plays an integral role in healthcare transformation, driving measurable impact for clinical quality outcomes and economics across the healthcare ecosystem. This article explore how artificial intelligence is being used in healthcare.

What is artificial intelligence?

The term “artificial intelligence” was first coined in the 1950s, and it has been used to refer to multiple types of applications, ranging from the simple to the complex, where each application of artificial intelligence technology offers differing trade-offs in performance, transparency, speed and training data requirements. At its core, artificial intelligence “makes it possible for machines to learn from experience, adjust to new inputs and perform human-like tasks.” Most applications of artificial intelligence today leverage deep learning and natural language processing (NLP), which allow computers to be trained to accomplish specific tasks by processing large amounts of data and recognizing patterns in the data.

Machine learning capabilities such as natural language processing are the “most mature and widely adopted parts of [artificial intelligence] today,” with NLP powering many applications and products that have positively impacted businesses and individuals on a significant level.

Artificial intelligence in healthcare has quickly become a powerful tool for clinicians across the healthcare ecosystem for care gap detection to adverse event prediction to augmentation of clinical decision-making—becoming a transformational force and paving the way for more efficient care delivery and creating space for innovation and growth.

There’s no question that artificial intelligence tools are shifting how patients and providers search for and use data. Advancements in big data analytics in healthcare stemming from interoperability and artificial intelligence technologies are rapidly evolving, and health plans, providers and other organizations across the industry are heralding the potential of artificial intelligence in healthcare to improve the user experience and overcome clinical and financial challenges.

Artificial intelligence (AI) and machine learning (ML): The definitions and differences

Artificial intelligence is when a machine can perform functions similar to human cognitive functions, such as learning and problem solving. AI is often trained with repetitive problems and situations in order to truly “learn” the desired behavior/outcome that is trying to be performed.

Machine learning is a subset of AI. Machine learning is a series of algorithms that analyze data, learn from it and use those learnings to make informed decisions based on the data available. Machine learning often automates work – but a key difference is that it can’t show learned behaviors beyond data analysis; it cannot make a decision out of various data inputs or perform other cognitive-like functions in the way that AI can.

How artificial intelligence is used in healthcare

Applications of natural language processing in healthcare, powered by machine learning, can be used to evaluate clinical information found within patient medical records with greater efficiency and completeness over the traditional human clinical review approach, delivering greater care and bottom-line savings in the process. And with 75% of clinical data found in unstructured form and more than 100 million clinical record reviews performed each year, natural language processing in healthcare can accelerate and automate this process to deliver rapid insight into clinical quality, risk score accuracy, disease outcomes and cost.

Artificial intelligence technology has also proven itself useful in other areas of healthcare — for instance, in assisting physicians with patient care by providing up-to-date medical information from journals, textbooks and clinical practices to inform proper patient care.

How machine learning is used in healthcare

Machine learning has various use cases across the healthcare landscape. Providers can use it to gather more information about their patients without having to dig through various portals or applications – instead, machine learning can compile various forms of data and make sense of those inputs in one fast approach. The analytic capabilities of machine learning make it a valuable tool in drug studies, population analyses, and other forms of research. Machine learning can also be used in monitoring disease outbreaks, as was critical during the early days of the COVID-19 pandemic.

Myths about artificial intelligence

Before any industry—especially one that so directly impacts the life of individuals—jumps full throttle into the world of artificial intelligence and machine learning, careful consideration must be given and the benefits and consequences appropriately weighed. Many within and outside of the healthcare industry have debated at length how artificial intelligence can change healthcare, and with that debate has come no shortage of misunderstandings and misperceptions.

Myth #1 | Artificial intelligence functions just like the human brain

One of the biggest misperceptions today regarding use of artificial intelligence in healthcare is that the technology is similar to or functions like an actual human brain. While certainly capable of identifying information it has been engineered to recognize (e.g., image recognition performed by artificial intelligence technology is often more accurate than what a human can achieve), in general, artificial intelligence is very skilled at solving one task at a time—one that it has been specifically trained to complete—but if any of the circumstances surrounding the task shift, the technology will fail. The artificial intelligence system that has been trained to identify gaps in care not indicated in the patient’s record is not the same application that can identify patients at risk of experiencing a fall. Artificial intelligence systems are incredibly specialized and are far from modeling full human intelligence.

Myth #2 | Artificial intelligence technology is capable of learning on its own

Another myth that permeates the conversation around artificial intelligence in healthcare and raises concerns is that intelligent, artificial intelligence-powered machines can learn on their own—a misperception that stems from the idea that a finished machine learning product possesses the ability to independently learn. While artificial intelligence technology has both the capability to learn—either through online learning or batch training—and correct itself based on information provided to and feedback derived from the system to improve accuracy, all of this requires a significant amount of human decision-making and brainpower. Humans use data from clinical activities to train and continually update machine learning software to support learning in artificial intelligence-powered machines—pushing the integration of new knowledge and data into the next learning cycle.

Myth #3 | Artificial intelligence is primed to take healthcare jobs

There is some concern from healthcare professionals, such as nurses, that artificial intelligence-powered machines will replace them in the foreseeable future. Machines, however, lack characteristics like empathy, creativity, judgement and critical thinking. In aspects of patient care where human compassion is essential, artificial intelligence-powered machines or solutions won’t be a replacement because they lack emotional intelligence, which is critical to providing holistic, patient-centered care. Certainly, the use of artificial intelligence in healthcare will evolve to minimize the need for humans to take on repetitive tasks and some basic tasks, such as appointment scheduling, but the healthcare industry is moving forward with the focus on patient-centric care—something that cannot be accomplished without the decision making skill of humans.

Myth #4 | More data equals AI success

For artificial intelligence technology to learn, it  needs data. But it needs data of the highest integrity. If your artificial intelligence technology relies on inaccurate or irrelevant data, then decisions made based on that technology will be inaccurate or irrelevant. The key is not to overwhelm your artificial intelligence technology with copious amounts of data. Rather, you should aim to train your artificial intelligence or machine learning technologies on high-quality data that has been verified and put through rigorous integrity testing on a regular basis. A high-quality data lake solution can makes it easier to train and deploy more accurate machine learning and artificial intelligence, which thrive on large, diverse datasets and serve as a powerful foundation to support the training of new algorithms for these technologies.

Artificial intelligence in healthcare: Where we are and where we’re going

Opinions about how artificial intelligence can and should be used in healthcare are as varied as the applications of the technology itself, but it is impossible to deny apparent benefits of artificial intelligence within the healthcare industry. According to a report by the center of Stroke and Vascular Neurology (SVN), artificial intelligence can use algorithms to “learn” features from a large volume of healthcare data, and then use the obtained insights to guide clinical decision making. This learned behavior can streamline the clinical record review process—a significant advantage that organizations choosing to implement artificial intelligence can benefit from right now.

Such applications of artificial intelligence in healthcare are great news for patients, physicians and even hospitals, as any technology designed to improve patient care is certainly worth exploring.

Three examples of AI in action

  1. AI is changing patient behavior for the better. Think of all the apps available on our smartphones, smart watches, and wearables to track steps, sleep, and other daily activities. All of these systems track certain activities to improve them. They notify us when it’s time to stand, when we should exercise, when we need more sleep, and more. In mass, AI is improving the health of many simply because they opted to make a simple change to improve daily activity,
  2. AI is helping providers identify acute conditions. A prime example of this is the use of AI for early cancer detection. For breast cancer, artificial intelligence can review and translation mammograms to increase the speed and accuracy in which providers analyze the results – and take action when necessary.
  3. AI is driving decision making at the point of care. Consider the use cases mentioned above: natural language processing can deliver fast insights on various aspects of the patient condition, and artificial intelligence can provide up-to-date medical information from various sources in one single view. Both of these lower the time it takes for physicians to access the clinically-rich patient details they need in order to deliver the best possible care.

Artificial intelligence solutions hold promise for closing care gaps & advancing patient-centered care

Although significant misunderstandings remain about what artificial intelligence is and how organizations across the healthcare ecosystem can reap its benefits, there are plenty of examples of how artificial intelligence can be applied successfully.

In the healthcare market today, the payer space presents the lowest barrier opportunity for artificial intelligence applications to meaningfully improve operational efficiencies and economic performance at a massive scale. Approaches in machine learning and neural networks can further bolster gap detection by permitting more variables to be accounted for, expanding beyond diagnosis codes to procedure codes, drug codes and much more. Thousands of codes can easily be used across any healthcare policy model existing today. This information can also be used to assess overall patient risk.

From identifying the patients and providers most at risk for gaps to providing financial estimates to locating patients within a membership who are over a specified risk threshold, artificial intelligence is perfectly aligned to help move the needle on providing patient-centered care. Solutions leveraging artificial intelligence, when combined with deep learning algorithms, are helping clinicians provide more—and better—care. In 2018, Apple launched Health Records, a personal health record feature that aims to combine patients’ existing healthcare-related data found on their personal Apple devices with data from electronic health records from hospitals they have visited.

Two other initiatives launched last year—MyHealthEData, a government-led initiative to give people greater control over their medical data, and Medicare’s Blue Button 2.0, a secure way for Medicare patients to access and share their personal health data—paved the way for innovative applications of artificial intelligence in healthcare.

The future of artificial intelligence in healthcare is now

Applications of artificial intelligence like machine learning have already been incorporated into our society. Today, it is a part of everyday life, and we often interact with artificial intelligence-powered technologies in ways even we don’t realize. Scheduling a doctor’s appointment, for example, is now often an automated process that allows one to make an appointment. And as health services have increasingly shifted to telehealth encounters, healthcare organizations would benefit from integrating aspects of artificial intelligence into their services to enable seamless patient encounters that drive greater clinical quality outcomes  for their members.

Once an idea that could possibly be applicable in the future, artificial intelligence in healthcare is here now, and like any other breakthrough technology, it isn’t likely to slow down. Organizations that choose to eschew artificial intelligence technology from their business strategy are putting themselves at a competitive disadvantage.

As more healthcare organizations adopt artificial intelligence tools, the amount of data powering the industry will grow exponentially—to the benefit of the industry as a whole but more specifically to patients. Now is the time to strategically incorporate artificial intelligence into workflows and collaborate with business and technology leaders to drive transformative and relevant innovation powered by applications of artificial intelligence.

To make sure your organization is ahead of the artificial intelligence curve, contact our team to learn more about our AI, data sharing, and interoperability capabilities.

Inovalon and design® and Inovalon® are trademarks of Inovalon, Inc. 

About the author

Hemanth Doma, Associate Vice President, Product