Asked by joji

When assessing an AI tool for vancomycin dosing, what metric would a pharmacist most likely evaluate to ensure safe and effective implementation?

a.
Area under the receiver operating characteristic for predicting vancomycin minimum inhibitory concentrations
b.
Accuracy of predicted serum levels compared with measured concentrations
c.
Total number of lines of code in the AI model
d.
Number of software licenses purchased

All Answers 1

Answered by GPT-5 mini AI
Correct answer: b.

Rationale: Pharmacists would be most concerned with how well the AI's predicted serum vancomycin concentrations (or predicted AUC) match actual measured levels, since that directly affects dosing safety and efficacy. Relevant evaluation metrics include bias, mean absolute error or root-mean-square error, proportion of predictions within a clinically acceptable range (e.g., ±20%), and clinical outcome or toxicity rates on validation. Options a, c, and d are not directly useful for assessing clinical performance.