Greg Pavlik, Senior Vice President and Chief Technology Officer, Oracle Cloud platform.
There has been a lot of hype surrounding the use of artificial intelligence (AI) in various industries – so much hype, in fact, that it can be difficult to separate truth from fiction. But healthcare is one area where the importance of this technology cannot be overstated.
AI and machine learning have already had a huge impact not only in scientific research, but also in epidemiology, diagnostics and, most importantly, in the mostly manual work that keeps labs, doctors’ surgeries, clinics and hospitals running.
Locate outbreaks in advance
For example, the emergence of better AI algorithms, which analyze increasing amounts of publicly available data, is already helping epidemiologists monitor the outbreak and progression of infectious diseases. Researchers use Internet searches on common symptoms related to geographic data to predict the outbreak of flu and other diseases in different areas.
The advantage is time. People know they are sick before they call a doctor. Many try to self-diagnose over the Internet before seeking professional help. If epidemiologists see a spike in keyword searches like “sore throat” or “difficulty swallowing” coming from IP addresses in a particular zip code, they can use machine learning models to predict the flu outbreak in that area with high probability.
Government health agencies also use public data and population maps to track crowds on site and analyze that data to predict the risk of future outbreaks. For example, health authorities in Europe, Israel, China and elsewhere use anonymized data on mobile phone traffic density to collect users’ location and then run machine learning models to predict how many people visit certain locations on a given day. If the total gets too high for pandemic conditions, venues could limit attendance, shorten visiting hours, or even close.
Using this process, the agency can monitor the potential spread of disease. If a dozen exposed people in Museum A later move on to Restaurant C, the government could warn the restaurant that the chance of an outbreak there has increased.
AI can also help people with long-term or lifelong conditions to function better. For example, machine learning models can analyze data from cochlear implant sensors to give subjects feedback on how they sound so they can better communicate with the hearing world. AI technology can also help doctors plan treatments that fit a patient’s specific needs.
Finding the needle in the haystack
AI-based computer vision, unlike the human eye, can scan and process thousands of images in a relatively short time to detect patterns. This technology could be a huge advantage in medical diagnostics, where overworked radiologists can’t pick up every nuance of one scan after looking at hundreds of others. In such applications, AI helps human experts by prioritizing images most likely to indicate a problem.
Speed up the paperwork
While helping to identify and treat disease is vital, AI is also invaluable in what some call the mundane reality of office work. AI-based speech recognition systems let doctors and researchers dictate their notes and fill out forms verbally, freeing up time otherwise spent behind a keyboard for more important patient care.
By automating form filling, AI can speed up important processes and spot errors before they get pricey. Companies with domain expertise in medical coding help healthcare organizations check for errors early in the workflow.
As noted, AI can make human experts more productive by reading scans faster and automating data entry. By taking such busy work off their plates, AI allows healthcare professionals to spend more time actually interacting with patients. Most health care providers will say that such face-to-face contact is the most important diagnostic tool at their disposal.
For example, if a code for a condition that typically costs $5,000 to $7,000 to treat gets an estimate of $40,000, that’s an immediate red flag to check for fraud or just plain old human error. Based on historical data, the system knows whether an estimate falls within normal parameters. This is not only good for insurance companies, everyone benefits if claims approvals take 20 minutes instead of 20 days and have fewer errors.
For healthcare executives, the potential benefits of AI are clear: faster, more accurate diagnoses and lower error rates in claims processing. But they also need to understand that none of this powerful technology replaces the human experience.
Business leaders should also be aware that bias can sneak into AI algorithms based on historical attitudes and data sets, and put in place guardrails to overcome that problem. For example, there has been historical bias about how different segments of the population are diagnosed and/or treated for medical conditions. Preventive measures include, yes, the use of software specifically designed to watch for such biases, but it is also the duty of healthcare providers to include data scientists, managers and clinicians from different backgrounds and perspectives, according to experts at Harvard TH Chan School of Public Health.
We are at the point in the traditional hype cycle where some are beginning to question the effectiveness of this technology. I’d argue the opposite – that AI’s ability to improve healthcare will soon live up to the hype.
Forbes Technology Council is an invite-only community for world-class CIOs, CTOs, and technology executives. Am I eligible?