Written by Clay Smith
An artificial neural network used temporal and weather data to accurately predict trauma volume and severity.
Don’t miss tomorrow’s weekend feature on A.I. in the E.D.
Why does this matter?
As we covered yesterday, machine learning and artificial intelligence can not only predict patient outcomes, this study shows it can predict trauma center volume based on a few readily available variables: day, time, and weather. Again, this is just the beginning.
No wise cracks about artificial intelligence among trauma surgeons…
This was an artificial neural network (ANN) that considered time, day of the week, daily high temperature, and active precipitation compared to trauma admission data. The ANN was trained with 70% of the 10,612 trauma admissions and 1,096 days, then validated with 15%, and tested in 15%. The ANN accurately predicted the number of traumas, penetrating traumas, OR cases, and average injury severity score with a high degree of correlation (correlation coefficient, r = 0.8940; numbers close to 1 indicate a very strong correlation). This was a single center in Nashville, and the same ANN may not work in a different setting unless retrained for the unique setting. But this is pretty slick: temperature, date, +/- rain = trauma forecast.
Artificial Intelligence Can Predict Daily Trauma Volume and Average Acuity. J Trauma Acute Care Surg. 2018 Apr 19. doi: 10.1097/TA.0000000000001947. [Epub ahead of print]
Peer reviewed by Thomas Davis