An Online Survey of Healthcare Professionals to Understand the Use of Clinical Prediction Models in Practice
This study surveyed 157 healthcare professionals to reveal that while clinical prediction models are generally valued and widely used, their effective implementation is hindered by workflow integration issues, regulatory uncertainty, and inconsistent adoption, necessitating a coordinated effort to ensure models are transparent, user-friendly, and rigorously validated for real-world practice.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the human body as a massive, bustling city where millions of tiny events happen every second. Doctors are the city planners trying to keep everything running smoothly, but sometimes they need a little help predicting traffic jams, power outages, or storms before they happen. This is where Clinical Prediction Models (CPMs) come in. Think of these models not as magic crystal balls, but as highly sophisticated weather apps for the human body. They take a bunch of different clues—like a person's age, their family history, or their current symptoms—and crunch the numbers to guess the likelihood of a future health event, such as a heart attack or a stroke. While these digital tools are supposed to be the ultimate sidekicks for doctors, making decisions faster and safer, there's a mystery: why aren't they being used everywhere? Some doctors swear by them, while others barely know they exist. Understanding this gap is crucial because if we can't get these helpful tools into the hands of the people who need them, we might miss out on saving lives or preventing illness.
This paper is like a giant, global "show of hands" survey sent out to healthcare professionals to find out exactly what's going on with these prediction models. The researchers, a team from the University of Liverpool, wanted to know: Do doctors and nurses actually use these tools? Do they like them? And if they don't use them, what's stopping them? They asked 157 healthcare workers from around the world (though most were from the UK) to fill out an online questionnaire.
Here is what they found: Most of the people surveyed (82%) knew these models existed, and a solid majority (71%) actually used them in their daily work. When asked how useful they were, the professionals gave them a high score, with a median rating of 8 out of 10. It seems like when these tools work, they work well. The most popular models were the ones that predict age-related risks, like the CHA2DS2-VASc score (which checks stroke risk) or QRISK (which looks at heart attack risk). Interestingly, even though the survey had a lot of doctors who specialize in pregnancy and women's health (Obstetrics and Gynaecology), the specific tools designed for them were surprisingly underused compared to the general ones.
But the story gets more interesting when we look at how these tools are used. The doctors and nurses have very strong opinions on what makes a tool "good." They want their prediction models to be as easy to grab as a snack from a vending machine. The favorite ways to access these tools are through websites (47%) and mobile apps (33%). Nobody wants to be flipping through a paper manual; they want to tap a screen. When it comes to seeing the results, they prefer clear, visual formats like a simple percentage (e.g., "15% chance") or a "heat map" (a color-coded grid that shows risk levels at a glance).
However, there are some serious roadblocks. The biggest hurdles aren't that the tools are bad math; it's that they are often hard to fit into the doctor's busy day. The survey revealed that if a tool isn't built right into the computer system the hospital already uses (like the electronic patient records), it's likely to be ignored. Other barriers include not knowing which tool is the right one for a specific patient, worrying about whether the tool is legally considered a "medical device," and a general lack of consistent training. One doctor noted that if a tool isn't taught in medical school or recommended in official guidelines, it's hard to get the whole team to use it.
The researchers suggest that for these tools to truly take off, they need to be transparent, user-friendly, and deeply integrated into the software doctors already use. They also point out that as Artificial Intelligence (AI) starts building these models, it becomes even more important to know how the model made its guess, rather than just trusting a "black box." The paper concludes that while healthcare professionals value these models, we need to stop just building them and start making sure they actually fit into real-world practice, backed by clear rules and good education. Until then, these digital helpers might remain in the toolbox, gathering a little dust while the doctors try to figure out how to use them best.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.