Imagine waking up after a stroke and struggling to swallow. Food or water going into your lungs is a scary and dangerous risk called aspiration. Doctors need a way to know who is in danger before it happens.
A team looked at 18 studies designed to predict this risk for adult stroke patients. They found most of these models showed good ability to tell high-risk people apart from low-risk ones. Sixteen of them seemed ready to use in real settings based on their performance.
But here is the catch. Every single study had a high risk of bias. The designs were often weak, with small groups and limited checks. They also lacked proper external validation to prove they work everywhere. Many were retrospective, meaning they looked back at past data. This means the results might not hold up when you try them outside the lab.
These tools still need serious work before they help you. They inform future improvements, but we cannot use them to make decisions today. The evidence is too shaky to trust completely right now because of methodological flaws.