Table 3.
Adherence to the five phases of the development of a classifier.
| Studies | Phase (1) | Phase (2) | Phase (3) | Phase (4) | Phase (5) | Phase (6) |
|---|---|---|---|---|---|---|
| Problem Selection and Definition | Data Collection/Curating Datasets | ML Development | Evaluation of the ML Model | Assessment of Impact | Deployment and Monitoring | |
| A decision algorithm to promote outpatient antimicrobial stewardship for uncomplicated urinary tract infection [43]. | Yes | Yes, information about the dataset and training/test split is provided. However, no justification was provided for the train/test split. | Logistic regression was chosen based on a comparison with decision tree and random forest models; however, no justification was provided on why these three models were chosen. | Yes | No information | No information |
| Hybrid Method Incorporating a Rule-Based Approach and Deep Learning for Prescription Error Prediction [44]. | Yes | Yes, information about the dataset is provided. However, no information was provided for the train/ test split. | No justification was provided for choosing the ML model; however, the authors reported that the rules they used with their model were based on consultation with a clinician. | Yes | No information | No information |
| Evaluation of a machine learning capability for a clinical decision support system to enhance antimicrobial stewardship programs [45]. | Yes | Yes, information about the dataset is provided. However, no information was provided for the train/ test split. | No justification was provided for choosing the ML model. | Yes | No information | No information |
| Personal clinical history predicts antibiotic resistance to urinary tract infections [46]. | Yes | Yes, information about the dataset and training/ test split is provided. However, no justification was provided for the train/test split. | No justification was provided for choosing the ML model. | Yes | No information | No information |
| Using machine learning to guide targeted and locally tailored empiric antibiotic prescribing in a children’s hospital in Cambodia [47]. | Yes | Yes, information about the dataset and training/test split is provided. However, no justification was provided for the train/test split. | No justification was provided for choosing the ML model. | Yes | No information | No information |