Calculating the need for intensive care beds.

Abstract

Prompted by high refused admission rates, we sought to model demand for our 20 bed paediatric intensive care unit. We analysed activity (admissions) and demand (admissions plus refused admissions). The recommended method for calculating the required number of intensive care beds assumes a Poisson distribution based upon the size of the local catchment population, the incidence of intensive care admission and the average length of stay. We compared it to the Monte Carlo method which would also include supra-regional referrals not otherwise accounted for but which, due to their complexity, tend to have a longer stay than average. For the new method we assigned data from randomly selected emergency admissions to the refused admissions. We then compared occupancy scenarios obtained by random sampling from the data with replacement. There was an increase in demand for intensive care over time. Therefore, in order to provide an up-to-date model, we restricted the final analysis to data from the two most recent years (2327 admissions and 324 refused admissions). The conventional method suggested 27 beds covers 95% of the year. The Monte Carlo method showed 95% compliance with 34 beds, with seasonal variation quantified as 30 beds needed in the summer and 38 in the winter. Both approaches suggest that the high refused admission rate is due to insufficient capacity. The Monte Carlo analysis is based upon the total workload (including supra-regional referrals) and predicts a greater bed requirement than the current recommended approach.