TABLE 5.

Data regarding literature-based prior distributions for the probability of inapparent CHIKV infection

Prior distribution category	Reference(s) summarized by the prior	Prior beta distribution hyperparameters (α, β)a	Approx median (%)h	Precisionb,h	95% HDCIh (%)c	Mean (%)h	P(x) > 28.0 (%)d,h	P(x) > 49.2 (%)e,h
Overall	10, 14, 20–27, 30, 31, 85–92	1.49, 3.69	25.6	30.18	0.3, 63.3	28.8	45.9	15.0
Epidemics within the 3–28% range	21, 30, 31, 85–92	2.97, 16.72	13.9	161.46	2.0, 30.5	15.1	6.9	0.1
Epidemics caused by the Asian lineage	10, 14, 20, 24–26, 86	4.95, 5.35	47.9	45.29	19.7, 76.6	48.1	90.7	46.8
Epidemics caused by the ECSA lineage	21–23, 27, 30, 85, 87–92	1.68, 6.39	18.2	54.98	0.3, 46.9	20.8	26.7	3.8
Epidemics caused by the IOL strain of the ECSA lineage	21, 22, 27, 30, 85, 87–92	1.69, 6.50	18.0	56.15	0.3, 46.4	20.6	26.1	3.6
Epidemics caused by non-IOL strains of the ECSA lineagef	23
Epidemics with A. aegypti as the primary vector	10, 14, 20, 21, 23, 26, 90, 92	2.17, 3.92	33.9	30.90	3.9, 69.6	35.6	62.2	23.3
Epidemics with A. albopictus as the primary vector	22, 30, 85, 87–89, 91	1.87, 7.43	17.8	64.09	0.6, 44.3	20.1	24.1	2.6
Epidemics in the Americas	10, 14, 20, 23, 26	13.86, 16.94	44.9	128.47	27.9, 62.3	45.0	97.6	31.7
Epidemics outside the Americas	21, 22, 24, 25, 27, 30, 31, 85–92	1.35, 4.29	20.4	36.49	0.0, 55.9	23.9	35.0	8.9
Epidemics in Africa	27, 30, 85, 87, 91, 92	2.51, 10.37	17.8	88.43	1.9, 40.2	19.5	20.2	1.1
Epidemics in Asia	21, 22, 24, 25, 31, 86, 89, 90	1.00, 2.21	26.2	19.60	0.0, 74.3	31.2	48.4	22.4
Epidemic studies with a convenience sample	21, 22, 26, 30, 31, 85, 86	1.21, 3.75	20.4	32.31	0.0, 58.6	24.4	36.1	10.5
Epidemic studies with a nonconvenience sample	10, 14, 20, 23–25, 27, 87–92	1.65, 3.80	27.5	30.60	0.8, 64.3	30.2	49.3	16.3
Updated, overallg	10, 14, 20–27, 30, 31, 85–92; this study	1.50, 3.67	25.9	29.95	0.3, 63.6	29.0	46.5	15.3
Updated, epidemics caused by the Asian lineageg	10, 14, 20, 24–26, 86; this study	5.20, 5.61	48.0	47.31	20.4, 76.0	48.1	91.3	46.9

^aHyperparameters are the parameters that specify the particular probability distribution to be used as a prior distribution. Beta hyperparameters were estimated by numerically optimizing the likelihood after weighting each identified study’s contribution by the number of CHIKV-infected individuals.

^bPrecision is the reciprocal of variance. In Bayesian analyses, precision is a preferred way of quantifying the variability in a distribution. Higher values indicate a less variable distribution.

^cValues for 95% HDCIs (highest-density credible intervals) are usually presented for the medians of posterior distributions and not prior distributions. However, we list them here to quantify the variability around the medians for the probability of inapparent CHIKV infections given the state of the literature.

^dProbability of observing a proportion of inapparent CHIKV infections, under this prior, that exceeds the upper limit of the expected range, 28.0%.

^eProbability of observing a proportion of inapparent CHIKV infection, under this prior, that exceeds the observed proportion in this study, 49.2%.

^fHyperparameters for a prior distribution could not be estimated via maximum likelihood or method-of-moments estimators from the single non-IOL ECSA study identified in the systematic search, so no parameters could be calculated.

^gThese data were derived by treating the present study as another study contributing to the understanding of the proportion of inapparent CHIKV infections. Results from the present study were incorporated along with those of the identified studies into a prior distribution, as before. Because our index cluster study concerns an epidemic caused by the Asian lineage, only the overall and Asian lineage priors are updated. Data in this column could be used as prior distributions for future Bayesian studies, public health planning, or mathematical modeling studies focusing on epidemics of CHIKV-specific lineages.

^hStatistics were calculated from the prior beta distribution with the given hyperparameters in the respective row.