March 2024 Pick of the Month

Vital Information

As technology and healthcare slowly align, the theorem of “garbage in, garbage out” becomes more important than ever. This principle essentially asserts that any output of either human or machine learning will be only as good as the quality of data used for training.

The idea for an omnipresent learning algorithm that can alert clinicians about the risk of unexpected deterioration based on patterns in data has been around for decades. Progress has been slowed in part by the lack of large datasets (generally defined as more than one million encounters) representing heterogenous persons with known outcomes and within a specific context. Contexts may include the battlefield, emergency department, ward nursing home, or ICU. Little has been done to create highly rigorous prediction rules in the community prehospital setting.

But in March of 2024, the work by Ramgopal et al., Correlation of vital sign centiles with in-hospital outcomes among adults encountered by emergency medical services, is helping to close these gaps. Ramgopal and colleagues analyzed over 13 million prehospital encounters to quantize vital signs based upon age and outcomes.

Any researcher who has attempted to find a simple article that answers “what is an abnormal vital sign?” in any setting for any age patient has likely ended up frustrated. The staggering gap between the dependence of emergency providers on vital signs to predict everything from driving direction to hospital code alerts, and the scarcity of published empirical data, led to the need for Ramgopal’s work. A good place to start with this research is Figure 1, which presents unique, valuable insights into the definition of abnormal for vital signs. These data are likely to move the needle forward toward developing both human- and machine-applied supervisory systems to predict unexpected, adverse, short-term outcomes, based upon data obtained in prehospital emergency care.

Jeffrey A. Kline, MD
Wayne State University School of Medicine
Editor-in-Chief