Waiting-time targets in healthcare markets. How long are we waiting?

Presenter: Luigi Siciliani, University of York

Abstract

Waiting times are a major health policy concern in many OECD countries. Mean waiting times for non-emergency care are above three months in several countries and maximum waiting times can stretch into years. Increasingly, information on waiting times is made available to patients, who would like to take informed decisions when choosing where to seek treatment, and researchers, who would like test the effect of policy interventions on waiting times. Information on waiting times is also used by policy makers as a target or as a performance indicator at provider's level (hospitals, general practioners). Typically providers with longer waiting times are penalised or monitored more strictly.
There are two commonly used measures of waiting times. The first measure is the waiting time of patients treated in a given year. This takes all of the patients treated throughout the year, and measures the difference between the time the patient was added to the waiting list and the time the patient receives the treatment (the completed waiting time). The second common measure is the waiting time of the patients on the list at a census date: it is a cross-sectional measure which takes the list of patients at a point in time (census date) and measures the difference between that time and when the patient was added to the waiting list (for most patients this is an incomplete waiting duration since they will still be waiting after the census date).
In this study we first investigate from a theoretical perspective the link between the distribution of the waiting time of the patients on the list with the distribution of the waiting time of the patients treated. We show that in steady-state there is a one-to-one mapping of the two distributions, so that given one of the distribution the other one can be derived. This has the important implication that policymakers can use the up-to-date waiting time of the patients on the list not only for monitoring purposes (or performance assessment), but also to infer or predict the waiting time of patients treated before it is available. Moreover, we compare common statistics like the average and median waiting times, and the proportion of patients waiting over x months under the two distributions, which are commonly used to set waiting-times targets.
Second, we apply the theory using data from the English National Health Service. Using data at hospital speciality level, we find that the average waiting time of the patients treated is higher than the average waiting time of the patients on the list across all the specialties considered. Moreover, the proportion of patients treated waiting more than x months is also higher than the corresponding proportion of patients waiting on the list. Therefore, waiting-time targets based on the proportion of patients on the list provide an under-estimate of the actual time waited by the patients treated.