Proceedings of the 2013 annual meeting of the Netherlands Epidemiology Society
Volume 1 Issue S1 Abstract 6
Klinkenberg, Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, Utrecht
Nishiura, School of Public Health, The University of Hong Kong, Pokfulam, Hong Kong
A correct description of time events for infectious diseases in individual hosts is important to develop control options. Key time events are the incubation period (from infection to symptom onset) and the generation time (from infection of primary to infection of
secondary case). The problem is that these are seldom observed because the time of infection is generally unknown, and that they are likely correlated. What is observed however, is the serial interval, i.e. the time between symptom onset of a primary and secondary case. The aim of this study was to estimate the joint distribution of incubation period and generation time with two datasets of serial intervals of measles in households, and a separate incubation period dataset.
We have developed a method to estimate the joint distribution of incubation period and generation time, with Approximate Bayesian Computation. The method takes explicitly into account the possibility that both cases were infected from outside the household. Identifiability of parameters was addressed with simulated datasets.
Both mean incubation period and generation time of measles were around 11-12 days, and they were positively correlated (r = 0.41 and 0.84 for both datasets). With simulated data
the model produced reliable estimates of correlation coefficient and most other parameters, but only if a bivariate gamma distribution was used, and not a lognormal distribution.
We have shown the possibility to estimate the joint distribution of incubation period and generation time from serial interval data. For measles, these appear to be positively correlated.
Published: 06 Jun, 2013