Skip to main content

Table 1 Training sample size (from 2016 to 8 datasets) and validation from manually coded samples from 2019 dataset

From: Trends in reasons for emergency calls during the COVID-19 crisis in the department of Gironde, France using artificial neural network for natural language classification

Reasons Training sample size Max F1 AUC Accuracy Proportion in 2020 95% confidence interval using bias estimated by bootstrapa
Main reasons for EMS calls
 Chest pain 29,310 0.800 0.987 0.978 0.057 0.0542–0.0598
 Gastroenteritis and abdominal pain 63,446 0.710 0.958 0.946 0.082 0.0777–0.0863
 Flu-like symptoms and breathing difficulties 72,323 0.683 0.958 0.929 0.151 0.1461–0.1559
 Focal neurologic deficit, stroke 5951 0.698 0.978 0.991 0.0135 0.0118–0.0152
 Road traffic crash (RTC) 1829 0.799 0.980 0.988 0.0233 0.0214–0.0252
 Violence 3158 0.636 0.984 0.991 0.011 0.0092–0.0128
 Suicide and self-harm 5904 0.654 0.969 0.988 0.015 0.0131–0.0169
 Injury other than violence, self-harm and RTC 120,007 0.694 0.938 0.887 0.166 0.1596–0.1724
 Pregnancy and delivery problems 6222 0.804 0.990 0.994 0.013 0.0116–0.0144
 Malaise with loss of consciousness 41,468 0.492 0.935 0.958 0.035 0.0313–0.0387
 Stress and anxiety 12,198 0.479 0.877 0.956 0.046 0.042–0.05
 Other reasons 412,218 0.673 0.785 0.746 0.385 0.3754–0.3946
Alcohol intoxication 8934 0.712 0.982 0.979 0.033 0.0303–0.0357
  1. a As estimated by bootstrapping (N = 10,000) the validation sample