Estimating the length of waits: a description of the period lifetable method and comparison with census and event based methods

Abstract

To discover whether the period lifetable provides more valid estimates of length of wait in prospect than are obtained using the lengths either of (current) waits captured at the time of the mid-period census or of the (prior) waits of those extracted over a specified period. We determined whether there was a surplus (or a deficiency) of extractions within the cross-classification of cohort and waiting time category which straddled each census. We used census-, event- and lifetable-based methods to produce three period-specific estimates of the percentage of waits of 0–2 months, and we determined whether length of wait grew shorter (or longer) from one period to the next. We used Lambda B to indicate the extent to which we were able to predict the direction of change in length of wait once we knew the direction of change in size of list. We found a direct correlation between change in length of wait and change in size of list, as expected under the stock-flow model, when length of wait was estimated using the lifetable for the period (LB = 58.33, 95% confidence interval [CI] = 29–88), but we obtained a null correlation when we used census-based estimates (LB = 6.45) and we obtained an inverse correlation when we used event-based estimates (LB = 57.14, 95% CI = 31–83). The period lifetable appears to provide more valid estimates of length of wait and should therefore be substituted for census- and event-based methods of estimation, wherever possible.