Last updated on 06/03/2024
What do you think of when someone mentions Artificial Intelligence, or AI, in healthcare? Many think about how AI has improved the speed and quality of diagnosis in medical imaging. Some think about the history of AI in differential diagnosis of diseases.
The truth is AI touches all aspects of medicine, from discovery of drugs to their prescription. Some applications, especially those in drug discovery and imaging, are relatively well known. They’re flashy and photogenic.
However, I find AI often shines brightest in the dullest spaces. The boring. The mundane. Let’s look at some oft overlooked ways AI improves our healthcare.
Scheduling Care Providers
There’s an enormous amount of scheduling that happens in the medical field. Most of us are familiar with making appointments, but this is just one variety of scheduling.
Medical providers also need to schedule their care providers. It’s every bit as complicated as scheduling a retail floor. Some people don’t get along. Everyone wants to take vacations, especially around holidays. Sometimes your employees have emergencies.
That’s the baseline, but additionally, many medical professionals are specialized. A hospital needs a certain number of anesthesiologists at any time. They also need cardiologists, and the two aren’t fungible. Keeping the right coverage of specialties makes the scheduling problem truly challenging. So challenging, in fact, that it is a common benchmark in AI research.
This can be further complicated by the need for in-home care, a domain I’ve previously discussed. In-home care is a great way to let people stay in their homes, where they’re more comfortable and happy, longer. However, it complicates the above staffing problems by adding in travel time. As nice as in-home care is, every moment spent driving is one not spent tending to someone’s needs.
All of the above are logistics problems. While humans can and do solve these scheduling problems, AI is perfectly capable here. In fact, it’s one of the most common AI applications we see. Capturing all of the things that the human scheduler ‘just knows’ is challenging. Once you do, the process can be automated.
Keeping Your Records Sorted
As is somewhat typical of the rural Midwest, I have an uncle with an abundance of first names. He’s always gone by Jim-Bob, and I’m sure he’ll continue to do so until the day he dies. Uncle Jim-Bob always has a difficult time in the doctor’s office though. Not because he’s bullheaded and refuses to follow doctor’s orders, but because of his name.
He never knows if he’s going to be James, Robert, Jim, Bob, or Jim-Bob in the patient records. In some older systems, looking up his medical records and identifying him can be a real nightmare. This is all because names are, quite naturally, used to identify people, and he has so many.
It’s not just names either. Are Main St and Main Street the same, or Abbey Road and Abbey Rd? If two humans enter data, there’s a good chance they’ll record it three different ways. AI can help by predicting when two names likely point to the same entity. This is called record deduplication.
Preventing Harmful Drug Interactions
Keeping medical records in order is useful for a couple of different reasons. Doctors need accurate medical history to properly diagnose current issues. Not being hassled by an insurance provider is also near the top of the list. One of the most important reasons is to prevent harmful drug interactions. Some medicines don’t work as well in concert, and some will outright kill you when taken together.
Software can be used to detect potentially harmful interactions between multiple drugs. Once you have accurate records, detecting drug interactions is a database lookup away. AI can assist where records are inaccurate or incomplete. AI can predict what medical issues a patient is likely to have given their demographics. These predictions can supplement incomplete medical records. Then, a pharmacist or other care provider can follow up with the patient.
Reducing Delay on Lab Work
Every year when I have my physical, they do a blood draw and lab work. This tends to delay the actual reports from my physical by a day or so. I consider the one day delay in lab works up to be pretty speedy. Waiting on labs is a huge part of medical procedures. Any reduction in the time it takes to get the labs done reduces the time to intervention. Reducing the time to intervention improves outcomes.
Time to complete labs can be improved by improving the scheduling of the lab. This uses the same AI techniques described above. However, it can also be improved by reducing the time it takes to perform a particular study in the lab. Many labs are automatically performed by machines. Then humans interpret the results. These machines can take a while to run their tests. AI can often reduce the time it takes to run those tests. If a machine has programmable moving parts, AI optimization can help reduce how long those motions take. Reducing the time the machine takes to do a test improves lab throughput.
Taking Dictation
There’s a shocking amount of dictation that occurs in the healthcare field. Medical professionals need to keep records of their interactions. Some do this by typing, but others prefer oral dictation.
In the past, medical transcription would be done by an entire pool of professionals dedicated to reducing dictation to text. Today, most dictation is taken by AI using natural language processing (NLP). Some voice-to-text systems are one size fits all. Others are specialized to their domain. Specialization improves the accuracy of capturing the spoken word as text.
Diagnosis
AI systems have diagnosed illnesses for decades now. More modern systems assist with medical imaging. Historically, AI has used a list of symptoms to aid in differential diagnosis. While doctors often perform diagnosis themselves, they had to be taught. It turns out you can encode these decision procedures for computers as well.
Summary
AI can help in medical settings in a variety of ways. We mostly hear about the flashy applications, namely diagnosis. However, it’s useful in the day-to-day work needed to provide medical care at scale. From scheduling shifts to taking dictation, AI takes the grunt work out of providing care so that experts can focus on using their expertise. $CALL_TO_ACTION