The following is a guest article by Jordan Bazinsky, EVP of Operations at Cotiviti.
Artificial intelligence has taken an increasingly prominent role in the healthcare system over the last several years. Until now, however, AI’s role has been primarily focused on two domains: diagnostics and administrative functions. An example of the former is radiology, where AI supports rapid and robust detection of potential anomalies, while an example of the latter is identification of fraud, waste, and abuse in healthcare billing.
The next frontier for artificial intelligence in healthcare is direct patient care. We are rapidly entering a new era where AI plays an integral role in directly treating and monitoring patients. Here are five places within direct patient care that are most likely to benefit from AI in the next three to five years.
1. Ensuring drug adherence
The local drug store has plenty of pill container options for under $5. They work well for organizing a few pills per day. However, for medically complex cases or for patients who may experience difficulty adhering to their medication schedule, these plastic boxes are lacking. Smart pill dispensers already exist that release drugs at certain times of day and send notifications to caregivers if drugs are not taken. The next logical step in at-home drug dispensing is the AI-enabled pill case. As everyone has slightly different rhythms to their day, this device will learn an individual’s preferred time of day to take different medicines. It will wait to alert the individual (and caretakers) at that time if drugs aren’t taken. It will also make drugs unavailable past a certain time to avoid overdoses, sharing compliance information with providers. That information may then be used to titrate drug dosing over time. It will also tap into online repositories of drug contraindication information to ensure that none of the medicines conflict with each other, offering a backstop to prescribers who may be unaware of all the medicines that an individual is currently taking.
2. Avoiding falls among the frail and elderly
Falls are the leading cause of injury-related visits to emergency departments among the elderly in the United States, and the primary cause for accidental deaths in people over 65. More than 90% of hip fractures occur as a result of falls while an estimated 60% of nursing home residents fall each year. What if we could substantially reduce falls? An AI-enabled shoe could learn someone’s gait. Over time, foot pressure, stride, time of day, and relevant information could be learned to identify when any of those variables are statistical outliers of a pattern indicating increased instability. This could send a warning to the individual and caretakers of imminent risk of a fall. While it would not capture true “accidents,” this type of artificial intelligence may help reduce a subset of falls that are preceded by changed behavior.
3. Powering the next generation of prosthetic devices
One common type of advanced prostheses is the microprocessor knee (MPK) for patients with above-the-knee amputations. These sensors detect the position of the leg, weight load, and control when to bend the knee, roll onto the toes, and move the leg forward. An AI-prosthetic leg could extend the capabilities of MPK by crowdsourcing information. What types of movements have caused falls among all patients using a given device? What types of movements came immediately before or after running patterns among individuals? Looking at prosthetic hands, what subtle movement adjustments are best for picking up certain types of objects? AI could create a community of individuals that never meet, but whose daily activities support each other through a cycle of reinforcing movement patterns and learned adjustments.
4. Bringing sight enhancement to the visually impaired
Sight impairments are a particularly rich domain for AI. For example, Microsoft’s “Seeing AI” is an app designed to help people with low vision or who are blind by enhancing the world around the user with audio descriptions. But the future of AI for the blind and visually challenged is immense. For example, Researchers at Ajman University in the UAE are developing smart glasses that can use AI to read, provide navigation information, and identify faces. Foresight Augmented Reality is a company that uses Bluetooth beacons that provide precise directions for individuals to navigate indoor spaces (like GPS at a much more granular level). The WeWalk cane integrates sensors to improve safety factors for unexpected overhead obstacles, with integrated apps and vocal directions. AI is increasingly enabling the blind and visually impaired to navigate the sighted world with confidence.
5. Dispensing insulin without patient intervention
The most basic “smart pumps” for diabetics shut off the insulin pump whenever a user’s blood sugar reading crosses certain thresholds. We are now seeing systems that can deliver correction boluses of insulin when glucose values are high. The next generation of artificial intelligence will enable systems to detect missed meals and automatically deliver proper levels to prevent high blood glucose. When no manual bolusing is needed, the system becomes a “fully closed loop.” Early versions of this may not be able to factor for exercise and other exertion that alters blood sugar levels. However, over the next several years we are likely to see closed-loop insulin pumps that increasingly account for all of diabetics’ daily activities.
The future of healthcare is unpredictable, but it’s clear that artificial intelligence will continue to play a critical role—migrating from its current applications in the administrative and diagnostic realms to transform patient care delivery. AI unlocks untold numbers of future therapies, many of which sounded like science fiction only a few years ago. As the healthcare industry embraces AI more fully, it will likely enter most areas of patient care and delivery, elevating outcomes and quality of life in the process.
