Table 1.
Prior distributions
| Parameter | Prior | Rationale | ||
|---|---|---|---|---|
| p | Attack rate | Beta(2,4) | During previous influenza pandemics in the populations with no pre-existing immunity, the attack rate has been estimated to be up to 50% [9]. Another analysis [10], using computer simulation, estimated the cumulative attack rate during the 2009 pandemic as 31–38% for European countries. A serosurveillance study showed that 33% of children in England were infected in the first season of the A(H1N1)pdm09 epidemic, while the attack rate in other age groups was lower [11]. | |
| Mode: 0·25 | ||||
| Mean: 0·33 | ||||
| 90% PI: 0·86–0·66 | ||||
| s.d.: 0·17 | ||||
| s | Severity | Beta(1·33,34) | Most authors define severity as the proportion of hospitalized in all symptomatic infections, while we use the proportion in all infections. The probability of hospitalization for symptomatic infection was estimated [12] to be 0·12–0·26% during the 2009 pandemic, with probability of developing symptoms ∼50%. | |
| Mode: 0·01 | ||||
| Mean: 0·04 | ||||
| 90% PI: 0·01–0·1 | ||||
| s.d.: 0·047 | ||||
| g | IC/hospitalization ratio | Beta(5·3,40) | Estimated [12] as 16%. | |
| Mode: 0·1 | ||||
| Mean: 0·11 | ||||
| 90% PI: 0·05–0·2 | ||||
| s.d.: 0·047 | ||||
| αM | Mild case ascertainment probability | Beta(1·33,34) | According to previous studies in similar populations only a small proportion of cases was ascertained: 0·7–2% [13]; 10% [11, 14]. | |
| Mode: 0·01 | ||||
| Mean: 0·04 | ||||
| 90% PI: 0·01–0·1 | ||||
| s.d.: 0·047 | ||||
| αH | Hospitalized non-IC case ascertainment probability in the first season | Beta(9·6,39) | Assumed to be high. No hospitalized cases were ascertained during the second season. | |
| Mode: 0·75 | ||||
| Mean: 0·71 | ||||
| 90% PI: 0·5–1 | ||||
| s.d.: 0·119 | ||||
| In the second season | = 0 | |||
| αI | IC case ascertainment probability | = 1 | Due to the high attention to IC cases we take αI = 1. |
PI, Probability interval; s.d., standard deviation; IC, intensive care.
Prior distributions were defined for the six model parameters. The mode and spread of each distribution reflect our prior knowledge about the model probabilities. Regarding the prior distribution, a 90% PI means a probability interval within which lies 90% of the distribution.