Machine Learning Models for Predicting Appendicitis in Emergency Patients
Author Information
Author(s): Schipper Anoeska, Belgers Peter, O’Connor Rory, Jie Kim Ellis, Dooijes Robin, Bosma Joeran Sander, Kurstjens Steef, Kusters Ron, van Ginneken Bram, Rutten Matthieu
Primary Institution: Radboud University Medical Center
Hypothesis
Can machine learning models improve the prediction of appendicitis in patients presenting with acute abdominal pain in the emergency department?
Conclusion
Machine learning models can accurately predict appendicitis in patients with acute abdominal pain, outperforming traditional scoring systems and some emergency department physicians.
Supporting Evidence
- The machine learning models achieved AUROCs of 0.919 and 0.923, outperforming the Alvarado score.
- ED physicians' performance was also compared, showing variability based on experience.
- Integration of machine learning models could assist in early diagnosis and reduce unnecessary surgeries.
Takeaway
Doctors can use computer programs to help figure out if someone has appendicitis when they have stomach pain, and these programs can be better than just asking questions.
Methodology
Data from 336 patients with acute abdominal pain were analyzed using two machine learning models to predict appendicitis based on clinical data.
Potential Biases
Potential bias due to reliance on physician documentation and variability in clinical practice.
Limitations
The models depend on the accuracy of medical history and physical examination notes, and were developed in a single-center context.
Participant Demographics
Patients included were from a Dutch teaching hospital, with no exclusions based on age, pregnancy status, or comorbidities.
Statistical Information
P-Value
0.033
Confidence Interval
0.919 ± 0.023 for HIVE model; 0.923 ± 0.020 for HIVE-LAB model
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website