South Africa - Assessing the cause of death pattern as reported during the Agincourt routine surveillance: an initial attempt to define reliable period from death to verbal autopsy interview.

The Agincourt's mode of operation has been to do an annual census update round, followed by a round of VAs for the deaths registered. Therefore, there is a range of recall periods from as short as one month to around one year which currently questions the validity of the VA mode of application. It is an important analysis to do, because if there were major differences between, for instance <6 months versus >6 months, it would be an argument not to rely on annual rounds. On the other hand, if the recall effects are not significant, it's another important piece in the overall validation puzzle - we could then clearly recommend that it is reasonable to cover deaths with VAs up to a year after they occur.

Individual level; cause of death, time of death, death symptoms and indeterminate COD.