AI Precision Dosing and Pharmacokinetics Quiz

Answers and explanations appear as you take the quiz.

1. What AUC/MIC range did the 2020 vancomycin consensus guideline target for serious MRSA infections, assuming an MIC of 1 mg/L?
   • 100 to 200
   • 400 to 600
   • 800 to 1000
   • 15 to 20
2. At steady state, how is vancomycin AUC24 related to dose and clearance?
   • AUC24 equals daily dose divided by clearance
   • AUC24 equals clearance divided by daily dose
   • AUC24 equals volume multiplied by half-life
   • AUC24 equals trough multiplied by 24 only
3. Why did guidance move away from trough-only vancomycin dosing?
   • Troughs cannot be measured in hospitals
   • Troughs are too expensive
   • Troughs predict AUC poorly, and high troughs were linked to kidney injury
   • Troughs only apply to oral dosing
4. What advantage do Bayesian methods have over traditional two-level steady-state calculations?
   • They need no patient data at all
   • They ignore kidney function
   • They only work in children
   • They can use a single level, levels not at steady state and irregular dosing
5. In MAP estimation, what does the penalty term weighted by between-subject variability do?
   • It penalizes parameter values that stray far from the population estimate
   • It removes measurement error
   • It sets the MIC
   • It calculates infusion rate
6. What is shrinkage in Bayesian dosing?
   • The drug's volume of distribution decreasing over time
   • The software reducing every dose by default
   • Clearance falling with age
   • Estimates staying close to population values when samples are few or uninformative
7. Why does the guide stress accurate records of infusion and sampling times?
   • Billing requires them
   • A timing error of even an hour can shift estimated clearance
   • Software cannot run without a clock
   • They determine the MIC
8. Which statement reflects the guide's view of so-called AI dosing tools?
   • They are black boxes unrelated to pharmacology
   • They remove the need for pharmacist judgment in all patients
   • Most are built on established pharmacometric methods, with machine learning added for model selection and covariates
   • They only work without drug levels