Predicting Emergency Room Visits and Trauma Cases Using Machine Learning
Author Information
Author(s): Luis Filipe Nakayama, Lucas Zago Ribeiro, Caio Vinicius Saito Regatieri
Primary Institution: Escola Paulista de Medicina, Universidade Federal de São Paulo
Hypothesis
Can an ophthalmologist without coding experience create a machine-learning model to predict emergency room visits and trauma cases?
Conclusion
The study successfully developed a machine-learning model that closely predicted actual emergency room visits and trauma cases.
Supporting Evidence
- The model predicted a mean emergency patient volume of 216.27/day in p90.
- In January 2020, actual visits were 13.5% less than the p50 prediction.
- The model was developed by ophthalmologists with no prior coding experience.
Takeaway
This study shows that even doctors who don't know how to code can use machine learning to guess how many patients will come to the emergency room.
Methodology
The study used a dataset of emergency visits from 2014 to 2019 to train a machine-learning model using Amazon Forecast.
Potential Biases
Potential bias due to the limited dataset and the specific time period chosen for the study.
Limitations
The study was conducted before the COVID-19 pandemic, which may have affected patient volume predictions, and there were issues with incomplete ICD codes in medical records.
Participant Demographics
The mean age of participants was 40.55 years, with a total of 356,611 visits and 7,027 trauma cases.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website