One question, two answers: do the two most commonly used methods of sampling describe the length of the prospective wait for admission to hospital?

Abstract

To demonstrate that the incomplete waits captured at the time and by means of a census may not have the same distribution as the completed waits captured as a result of admission during a specified calendar period, cataracts added to waiting lists in England, Scotland, Wales and Northern Ireland between 1 July 2003 and 31 December 2005, and admitted no more than two years after the date of their enrolment, were identified using Hospital Episode Statistics, Scottish Morbidity Records, the Admitted Patient Care database and the Hospital Inpatient System. Census- and event-based sampling produced two sets of records for each closed population in which the eventual outcome, and rules about suspension and deferral, were the same. The census-based sample (30 September 2005) yielded a smaller percentage of waits with a length of one or more days, and the same was true for every other length up to the maximum of the event-based sample, e.g. up to 728 days (1 July–30 September 2005) and up to 723 days (1 October–31 December 2005) in Hospital Episode Statistics. The two methods of sampling did not yield the same distribution: one might capture the wait in prospect for those newly added to the list, but the other does not.