By now, you have probably heard of AI and tools like ChatGPT. But you still may be wondering how these tools apply to your clinical practice.
Let me share an experience I recently had that completely changed how I think about AI (artificial intelligence) in health care.
In one of the MBA classes I am taking at NYU, we write code to help analyze large datasets. With the AI assistant turned on in the platform we use to write code, the assistant started suggesting code before I even started writing it, based on a prior comment. My comment had a spelling error, but this was no barrier; the AI assistant generated completely accurate code.
Figure 1: See the visual representation of Department of Health restaurant inspections in the Manhattan borough of New York City mapped from an inspections database. This figure shows the Google Colab platform using Python with the built-in optional AI assistant enabled. At the bottom of the frame, you can see the code automatically generated by the AI assistant after the comment is interpreted as a prompt.
This made me think: What if we could use AI to analyze data we collect about patient volumes, traffic, and the use of health care services to improve the staffing of our hospitals?
We have all felt the staffing crunch over the past few years, especially since the COVID pandemic began. The American Hospital Association reports a drop in health care workers overall compared to 2019, with an increase in staffing costs of 16 percent over the same time period.
It is clear that there will never be enough doctors, nurses, techs, or coders to keep up with demand. Yet many of us have gaps in our day or have spent empty time on call.
Can we leverage technology to help us to deploy staff better? Rapidly evolving technologies like artificial intelligence allow us to automate data collection and analysis, especially over large and complex datasets. Here are a few examples of useful ways we can lighten our staffing burdens:
1. Quantify and analyze patient flow over time through different departments, such as the ED, floors, radiology, and the operating room. This information can be monitored in real-time to help determine where patients are now and predict where they will be in the future. This information can help us get the right amount of staff where they are needed.
2. Plot the distance of patients from the hospital and what services they use at what times. Knowing where our patients are located can help us understand which services can be delivered in person or via telehealth. Assessing information can also help us plan the location and timing of our physical services and staff with greater precision.
3. Dynamic call staffing. Imagine if we could determine what on-call services we needed based on current and predicted patient volumes. Instead of stabbing in the dark, we could accurately predict and staff our needs and also decide which services need to be provided in person rather than via telehealth.
As a surgeon with a busy bariatric and general surgery clinic, I knew none of this data. We had minimal ability to predict patient volumes and use of services and accommodate them with appropriate staff in the clinic, the operating room, and especially on call.
I do want to throw in a word of caution here about AI. AI results are only as good as the inputs and training we provide for these tools. We sometimes lack clear mechanisms to validate the accuracy or reproducibility of results. As we move forward in deploying both human and artificial intelligence, we need to keep safety and liability top of mind in health care settings.
Using AI and related tech tools, we may be able to better use our limited resources. Just like the AI assistant helping me write better code in my data class, AI assistance can help our hospitals deliver more personalized and effective health care to our communities. When we continue to stumble in the dark, we are wasting our resources, which can lead to shuttered services, hospitals, and, ultimately, communities left without access to health care.
Maria Iliakova is a bariatric and general surgeon.
