Figure 2.
Ninety-five percent prediction intervals of symptomatic case fatality ratio (sCFR) estimators based on 10,000 recruited cases. Pyramidal sCFR estimators are obtained by multiplying estimates of pi |j, the probabilities of progressing to severity level i for cases in level j, which are based on 4 severity levels: symptomatic cases (S), medically attended cases (M), hospitalized cases (H), and dead cases (D). The 95% prediction intervals (T-shaped bars) are the intervals surrounding the true sCFRs (bullet points) in which future sCFR estimates will fall, with a probability of 95%, if the assumptions about progression probabilities are correct. Narrow prediction intervals indicate good precision. A) Severe (1918-like) influenza pandemic scenario; B) intermediate (1957-like) severity scenario; C) mild (2009-like) severity scenario. Maximal precision is ensured for each estimator through optimal resource allocation between surveillance levels. Recruitment costs are assumed to be equal at all surveillance levels of all estimation strategies.