From: Quantification of parasite clearance in Plasmodium knowlesi infections
Characteristics | WWARN PCE (Method 1) | Bayesian ‘bhrcr’ package (Method 2) |
---|---|---|
Advantages | Simple and straightforward method that is user friendly Computationally fast | All patient data (including patients with ≤ 3 blood samples) are included in the analysis Has a robust sampling method that uses a changepoint model to determine the lag and tail phases accurately |
Disadvantages | Requires frequent sampling for best results (e.g., recommends blood samples taken every 6-h) Omits patients from the second stage (calculation of population mean) if the parasite clearance rate cannot be calculated in stage 1 (due to fewer samples), potentially leading to selection bias Requires at least three consecutive and distinct positive blood sample measures above the microscopic detection limit per patient to successfully estimate parasite clearance rates | Complex method of calculation which requires in-depth statistical knowledge for execution, including understanding the hierarchical structure and prior distributions for each parameter included in the model Computational time for sampling from posterior distributions (e.g., about 10–36 h, depending on the number of samples and processing system used) Integral convergence issues may occur if there is too much variation between patients’ parasite clearance profiles Patients with frequent parasitaemia measurements contribute more information to the model parameter estimation leading to potential bias |