It's unquestionable that accurately diagnosing a patient's symptoms is a critical part of medicine. The effects of a misdiagnosis can be fatal. Dr. David Mathison, a pediatric emergency physician, regional medical director at PM Pediatrics, and assistant professor of pediatrics and emergency medicine at George Washington University, well understands the risks in diagnostic errors.
Two years ago, Mathison turned to Isabel, a web-based diagnosis checklist system, for assistance. Diagnostic artificial intelligencesolutions like Isabel "help focusthe primary provider by avoiding anchoring bias, preventing over-testing, and over-consulting, and potentially getting a patient to the correct diagnosis and treatment plan faster," Mathison says.
Isabel Healthcare, the company behind the diagnostic tool, was founded by Jason and Charlotte Maude in 1999 after their daughter Isabel was nearly fatally misdiagnosed by a local hospital. Isabel provides a diagnosis checklist that serves as a standalone tool or can be integrated with electronic medical records (EMR). It analyzes structured and unstructured data to create a list of possible diagnoses with critical "don't miss diagnoses" flagged.
Thanks in part to changes under the health reform law pushing for improved patient outcomes and reduced costs, companies are pouring investments into artificial intelligence-based tools. Isabel is one example of a growing set of tools that analyze information about patients in EMRs and other data points to identify patterns and deliver insights faster than a human.
Artificial intelligence consists of many capabilities, but a key area expected to have a large impact in healthcare is machine learning in which computers process data in huge volumes, notes Mani Kumar, MD and principal at the consulting firm West Monroe Partners. "If you think of a physician's role as diagnosis, prognosis, and therapy for a patient, you can begin to see how AI will affect the different clinical specialties," he says. "Pretty soon, a computer may be able to elicit a patient history, perform a physical exam, and diagnose a patient with appendicitis."
AI may also impact pathology, radiology, and dermatology. These specialties involve recognizing patterns in microscopic specimens, X-rays, CT scans, and skin lesions. "Computer algorithms for image recognition have recently demonstrated error rates lower than that of trained physicians and it's only a matter of time before this part of the job gets outsourced to machines," Kumar maintains.
Additionally, mobile technology and facial recognition is being used to improve medication adherence. "Pharmaceutical companies are combining Big Data with AI to identify new drug compounds and new targets for existing FDA-approved drugs," Kumar adds. "AI will have an impact in almost all areas of healthcare, it's just a matter of where we'll see evidence of it first."
IBM is one such company that is heavily investing in technology to make healthcare more data driven. IBM Watson Health has access to 50 million anonymized U.S. medical records which it gained through the acquisition of firms likeExplorys, a cloud-based dataanalyticscompany that is a spinoff of Cleveland Clinic. Explorys has built one of the largestclinicaldata sets in the world, IBM says in a press release.
The anonymity of the medical records is a key factor, says IBM Watson Health Vice President Bill Evans. "HIPAA compliance isn't an issue because the data is randomized, but it gives us a huge data set for Watson to analyze and learn from," Evans maintains.
IBM Watson Health's approach to healthcare is divided into four pillars: wellness, care, productivity (e.g., population health), and research and discovery. The company has already formed partnerships with organizations like Boston Children's Hospital to identify rare diseases. Watson will analyze massive volumes of scientific literature and clinical databases on the Watson Health Cloud to match genetic mutations to diseases and potentially uncover insights that could help clinicians identify treatment options.
IBM is also working with the American Heart Association and Welltok to build workplace health technology for combatting cardiovascular disease. The solution, which is delivered on Welltok's health optimization platform, combines the cognitive computing power of IBM Watson and the heart association's science-based metrics and health assessments, to assess the employer's workplace health environment, as well as employee health based on AHA metrics. The tool provides employers with data-driven suggestions on how they can best support their employees' health, such as how best to design and deliver health benefits and health promotion programs.
One of the values of artificial intelligence is that it allows physicians and other healthcare professionals to focus on more complicated areas, notes Michael Schmidt, founder and CEO of Nutonian, which was founded in 2011 out of the Cornell Artificial Intelligence Lab. "AI is automating the tasks that humans don't want to do or can't do more effectively and efficiently," Schmidt says. "This gives people the freedom to dedicate more time to more difficult tasks and get results sooner."
Nutonian delivers a machine intelligence solution that helps researchers find patterns in data and iterate quickly to enhance the results. The company serves a number of industries, including life sciences. With the help of Nutonian's technology, for instance, a research group was able to develop a way to predict appendicitis in children while limiting their exposure to radiation.
AI can also potentially replace expensive machines or systems, maintains Tony Keating, CEO of ResApp Health, which developed an app that can diagnose respiratory diseases when users cough into a smartphone. A ResApp can be "a significant step forward from the tools doctors have available today," Keating claims. "Consider locations such as in the developing world, where the doctor does not have access to x-rays or other tests, or a telehealth situation where even a stethoscope is unavailable."
But despite the advances that are being made in AI, humans are unlikely to be replaced. "The last thing we want is HAL (a sadistic computer featured in Stanley Kubrick's classic film 2001: A Space Odyssey)," deadpans Mitch Lawrence, president of Next ITHealthcare, which creates virtual assistants. "There's always going to be a need for some human oversight, just to make sure any technology is operating the way it's supposed to."
Mathison agrees that while artificial intelligence has the potential to be very helpful, it's not a replacement for the medical decisions made by humans. "There are certainly some disciplines where more focused decision making could be made by computers," Mathison says. "Mid-level practitioners, for example, are used in many fields to take 'routine' decision-making out of the hands of specialists and experienced physicians." AI can play a similar role as part of a management team.
The area where AI falls short is in connecting with a patient in a relationship built on trust and human understanding. The doctor-patient relationship is "intimate," Mathison notes. And AI "won't replace the more personal interactions that are necessary, such as thediscussion of different treatment options when there isn't one clear answer, the role of surgery versus pharmacologic management, and the ability to gain the confidence of a 2-year-old boy so he lets you look in his ears when he is irritable with fever."