Artificial intelligence (AI) has seen rapid innovations over the last few years. If used properly, it can empower care providers to make better diagnoses, provide better patient care, and improve access to healthcare. It can also transform healthcare privacy and security. Healthcare AI use cases might include clinical decision and research support, voice-to-text transcription, and cybersecurity.
On a recent webinar, Iliana Peters, HIPAA legal expert and shareholder at Polsinelli law firm, joined FairWarning’s Vice President of Product Management & Engineering, Chris Arnold, to discuss the ways that AI are improving healthcare – and the steps all healthcare professionals should take to ensure a responsible and beneficial use of the technology. Here are some key takeaways from the webinar “Harnessing the Power of AI in Healthcare: Ethical and Legal Considerations.”
What is “AI”?
First, it’s important to understand some key terminology:
- Artificial intelligence, or augmented intelligence, is a branch of computer science concerned with simulating “intelligent” behavior in computers.
- Machine learning refers to a branch of AI where computer systems are fed large amounts of data, which they then use to learn how to carry out a specific task. In healthcare, according to Iliana, “AI is about how we enable machines to help us provide better healthcare to patients.”
- Natural language processing (NLP) is another branch of AI that helps computers understand, interpret, and manipulate human language, translating it into the code they naturally “speak.” It’s often coupled with machine learning to help the computer understand various inputs.
- Algorithms are generally a set of instructions that tells a computer how to perform a task. There are many types of algorithms; machine learning algorithms specifically are designed to rewrite themselves as they work, picking up new sources of information that inform their output.
At its core, Iliana emphasized, AI should be used to improve patient outcomes and professional satisfaction of physicians, researchers, privacy and security professionals, and other staff members.
“According to the AMA, AI is really about understanding how to promote the development of thoughtfully designed, high-quality, clinically validated healthcare,” she said. The purpose? To implement best practices that…
- Are transparent
- Conform to leading standards for reproducibility
- Address bias
- Avoid introducing or exacerbating healthcare disparities
- Safeguard patients’ and other individuals’ privacy interests
- Preserve the security of personal information
So how are healthcare organizations using AI and machine learning to empower their staff, enhance care, and protect patient data? Here are five healthcare AI use cases.
Healthcare AI Use Case #1: Research and Clinical Decision Support
AI for research and clinical decision support (CDS) aims to improve clinical practice and research, said Iliana. Essentially, the computer acts as a knowledge base that can generate patient-specific assessments or recommendations, which are then presented to clinicians or researchers for consideration.
“[The AI] can provide recommendations and support at the time and location where the physician is making the decision, which helps [the physician] provide the best service possible,” Iliana said. “All machine learning does is augment that already available clinical decision support.”
A great example of this type of use in AI is an algorithm that automatically extracts risk factors from unstructured notes in clinical records. In a pilot program, IBM’s Watson supercomputer was given an algorithm that sent it looking across clinical records, note sections, and other data sources to learn key markers and predictors for congestive heart failure. After analyzing 21 million patient records in six weeks, the code achieved an 85 percent accuracy rate identifying patients at risk of developing congestive heart failure within one year.
Healthcare AI Use Case #2: Medical Imaging
Similar to CDS tools, medical imaging AI empowers physicians with additional information that can help them more quickly and accurately diagnose particular conditions. An algorithm could be written, for example, to review a million different chest films, both with and without malignancies, and learning to spot malignancies. The goal is to enhance the care provider’s decision-making.
“It’s very important to remember, though, that the AI is not making the decision,” Iliana said. “The point is to provide all that information to a physician who makes the decision based upon the initial first run of a comparison of this particular chest film to hundreds of thousands of other chest films, and also the additional data that’s available to that physician because it’s available to the clinical decision tool that has AI built into it.”
Healthcare AI Use Case #3: Voice-to-Text Transcription
Many people are familiar with voice-to-text, most commonly through home assistants like Alexa. In healthcare, this technology can be used to help healthcare professionals transfer chart information into an EHR. This, in turn, helps ensure that the right information is available to be searched and used to provide additional CDS or research support.
Many of these tools are also designed to be interpreted by human beings. A transcription service, for example, may start with AI that can transcribe multiple speakers and pick up on complicated conversations. Then, a human transcriptionist would review the information to ensure it was transcribed accurately.
“The whole point here is to make sure that we have the most accurate data fed into the EMR system, so the machine learning application has the best data to make suggestions to physicians or researchers, so they have the best information available to them,” said Iliana.
Healthcare AI Use Case #4: Fraud Detection
Much like they can be taught to spot malignancies on chest films, machine learning algorithms can learn to look for fraudulent claims. Feeding a variety of claims data – fraudulent, not fraudulent, investigated, etc. – helps the machine proactively identify potentially fraudulent claims before they’re submitted.
“You could affirmatively run these audits on your claims as they’re being processed before they’re submitted, such that you can reduce levels of potential fraud before the government identifies any of these issues,” Ilana remarked.
Healthcare AI Use Case #5: Cybersecurity
Trying to keep up with cyber-issues is a never-ending battle, Iliana said. “What we do know is that most criminals are pretty lazy. They’re going to go for the easiest target, and the idea is to make yourself not the easy target. This is where AI can be helpful.”
Your IT and cybersecurity specialists simply don’t have the time to wade through every potential vulnerability or attack to your system. A virtual analyst, or algorithm, however, can help you identify potential issues and rank them for your physical analyst. This, in turn, helps her determine which advanced threats need to be addressed most immediately. These might include ransomware, identity fraud, or drug diversion.
The healthcare AI market is predicted to reach $6.6 billion by 2021, according to Frost & Sullivan – a compound aggregate growth rate of 40 percent. And 43 percent of health IT leaders plan to increase AI spending this year. It’s no wonder, as there are myriad benefits to using AI in healthcare – including improved outcomes, increased efficiency, and more private and secure patient data. Take the time to fully vet any AI-enabled partners, and explore what applications could be right for you. From clinical decision support to privacy and security, AI can be an excellent tool to aid care providers in their ultimate goal: better patient care and patient trust.