Predicting missed health care visits during the COVID-19 pandemic using machine learning methods: evidence from 55,500 individuals from 28 European countries

Table 2 Performance measures

Accuracy, true positive rate and true negative rate are assessed at a cutoff of 0.135 for men and 0.17 for women. Models were trained on data from wave 1 and then applied on data from wave 2. For wave 1, all measures are means across the five cross-validation folds for each method. For wave 2, the measures for stepwise and lasso are retrieved by applying the model from wave 1 on the data from wave 2. The measures for random forest and neural networks are retrieved by applying the models from all five cross-validation folds from wave 1 on the data from wave 2, and taking the mean predicted probability across all folds. (SD) is the standard deviation of the AUC across the five cross-validation folds for wave 1, and hence cannot be computed for wave 2.

ISSN: 1472-6963