Table 2 Steps in the implementation of the model
Step 1 | Step 2 | |
|---|---|---|
Purpose | Estimate ρ (relative probability of attending screening among cases versus non-cases) | Estimate the true prevalence and the detected fraction |
Geographical resolution | Each MSF project | Each screening session (results then totalled over each project) |
Model inputs | Project-specific diagnostic accuracy parameters | Diagnostic accuracy parameters |
N = 10 000, S1 = Uniform [1–50] and S2 = Uniform [1–50] (hypothetical values) | Observed N, c, S1,obs and S2,obs for the screening session | |
Observed βc (ratio of observed prevalence at coverage c to observed prevalence at coverage = 100%) for four coverage strata (5-24%, 25-44%, 45-64% and 65-84%) | ρ values estimated in Step 1 for each MSF project, sampled from their deviance distribution | |
Observed c values sampled from within each coverage stratum and for each project | Various candidate sets of S1 and S2 (true prevalent cases) | |
Various candidate ρ values | ||
Model predicted outputs | βc for the same coverage strata (5-24%, 25-44%, 45-64% and 65-84%) | Number of observed cases (S1,pred and S2,pred) |
Number of true positive cases among those observed (S1,TP,pred and S2,TP,pred) | ||
Number of iterations | 10 000 for each project and for each candidate ρ value | 10 000 for each screening session and for each candidate set of S1 and S2 |
Fitting procedure | Predictions fitted against observed βc for the same coverage strata. | Predictions fitted against actual observed cases in screening session (S1,obs and S2,obs). |
Observed βc estimated based on a statistical model of field data. | S1 and S2 candidate sets resulting in best-fitting S1,pred and S2,pred adopted as maximum likelihood estimates of true prevalence. Joint likelihood distribution informs confidence intervals. | |
Candidate ρ value resulting in best-fitting βc adopted as point estimate of ρ. Confidence interval based on squared deviance distribution. |