Waiting time information services: An evaluation of how well clearance time statistics can forecast a patient's wait
Introduction
For many years, a waiting list has been the principal administrative tool for managing access to elective surgery within publicly funded health care systems. Although it is simply a register of patients waiting for a date for surgery, a waiting list fulfills several important roles. In any system working close to capacity, and in which demand varies stochastically, a backlog of work will form (Street & Duckett, 1996). A waiting list is one mechanism by which this backlog can be managed, with patients generally being scheduled for surgery only a short time ahead of the planned date. It also allows access to be prioritised on the basis of clinical need (Mullen, 1994). Yet, controversy surrounds waiting lists because some people wait excessively (AIHW, 2000), and ensuring patients have timely access to elective surgery continues to challenge doctors, managers and policy makers.
Many types of initiative have been implemented to reduce the length of waiting times (Yates, 1987; Frankel & West, 1993). There have been one-off programs aimed at alleviating bottlenecks. At the other extreme, there have been radical changes in the organisation and funding of health services, with the aim of overcoming systemic problems. Many initiatives have concentrated on the supply of elective surgery, focussing on increasing throughput or efficiency (Frankel & West, 1993). But increased activity has often led to increased levels of demand (Street & Duckett, 1996), and it is now generally recognised that focussing attention only on this part of the interface between primary and secondary care is insufficient. Increasingly, initiatives and policies are including strategies to improve waiting list administration and demand management (Naylor, 1991; Hanning, 1996; Hadorn & Holmes, 1997; NSW Health, 1998).
One type of demand management initiative being adopted in several countries is the publication of waiting time information, often on websites. The general aim of these services is to assist general practitioners (GPs) and patients when considering referral to secondary care. With this information, it is believed that patients can avoid waiting longer than necessary, and that imbalances in the distribution of referrals will be reduced.
There is currently little evidence about the effectiveness of these services, and such policies are not without risk. Some services publish only inpatient waiting times, yet West (1993) describes how such an initiative produced a bottleneck in one specialty because one surgeon sought to minimise his inpatient waiting times, while another sought to minimise his outpatient waiting times. Worthington (1987) also demonstrated that publishing waiting times could lead to people with equal need having unequal access to surgery.
In addition, the information services raise various statistical issues. Data quality could suffer if, in response to unfavourable statistics, surgeons delay notifying hospitals of an intention to admit (Kent, 1999). Another issue concerns the accuracy with which patients can make inferences about their likely waiting time. A review of the statistics presented by six services in Australia, Canada and the UK (Cromwell, Griffiths, & Kreis, 2002) highlighted two issues in particular. First, inappropriate levels of aggregation could result in statistics being biased by, or unresponsive to, changes in behaviour. Second, no service stated when different waiting time statistics might reflect real differences in performance among the surgical units.
A later study examined the forecast accuracy of average waiting time statistics similar to those used by the six services (Cromwell & Griffiths, 2002a). The accuracy of all tested statistics could be poor, especially once the waiting time of patients joining the list was, on average, over 6 months. This was partly due to the high level of variation in waiting times among patients that joined the list around the same time, but it also related to the type of statistic used. Consequently, a study was undertaken to examine whether a different type of waiting list statistic, the clearance time function, could perform better.
Section snippets
The clearance time function
Most average waiting time statistics are either based on the data from admitted patients (throughput data) or the data from patients still on a waiting list (census data). These statistics produce different types of information, and each suffers from various drawbacks (Don, Goldacre, & Lee, 1987). For example, throughput data do not include the waiting times of patients removed from the list without admission. Census data do not measure a patient's complete wait, and are affected by
Data collection and evaluation design
Waiting list data were collected from a teaching hospital in Sydney, Australia, and described elective surgical activity between 1 July 1995 and 30 June 1998. Anonymised data were extracted on all patients admitted or removed from a waiting list during this interval, and on the patients still waiting on 30 June 1998. This provided information on all patients added to the list over the 3 years. The data covered 46 surgeons in 10 specialties who had operated for the whole period (see Cromwell &
Results
Table 2 summarises the performance of the two sets of clearance time functions across the 46 surgeons in terms of the root mean square error (RMSE). The functions within each set often performed at similar levels when forecasting the waits of patients. The range of RMSE values among the functions tended to be small, although it could exceed 30 days. The largest differences in performance typically occurred among the functions without the removal rate adjustment; the inclusion of the adjustment
Discussion
The primary focus of this study was the accuracy with which different types of waiting time statistic can forecast the wait of patients at the time they join a waiting list. This issue has assumed importance with the growth in web-based information services that encourage patients and GPs to consider waiting time statistics when making a referral decision. Yet, little is currently known about the performance of either commonly used waiting time statistics or the clearance time statistic.
Acknowledgements
This study was undertaken while the author studied at the University of Wollongong, Australia. The views expressed in this article are the author's alone and are not attributable to the Commission for Healthcare Audit and Inspection, UK. The University of Wollongong Ethics Committee gave ethics approval for this study, and thanks go to Prof. David Griffiths for his statistical advice.
References (31)
Maximum waiting-time guarantee—an attempt to reduce waiting lists in Sweden
Health Policy
(1996)- et al.
Are waiting lists inevitable?
Health Policy
(1996) - AIHW (2000). Waiting times for elective surgery 1995–96 and 1996–97. Canberra: Australian Institute of Health and...
- et al.
Statistical methods in medical research
(2002) First steps in analysing NHS waiting timesAvoiding the ‘stationary and closed population’ fallacy
Statistics in Medicine
(2000)Statistical methods for SPC and TQM
(1994)- et al.
General practitioner perceptions of surgical waiting times
Journal of Quality in Clinical Practice
(1996) Waiting listsSome problems of definition and a relative measure of waiting time
Hospital and Health Services Review
(1980)- et al.
Waiting time information servicesHow well do different statistics forecast a patient's wait?
Australian Health Review
(2002) - et al.
Waiting time information servicesWhat are the implications of waiting list behaviour for their design?
Australian Health Review
(2002)
Surgery dot.comThe quality of information disseminated by web-based waiting time information services
Medical Journal of Australia
Waiting lists and health policy
Waiting list statistics 3Comparison of two measures of waiting times
BMJ
Rationing and rationality in the National Health ServiceThe persistence of waiting lists
Fundamentals of queuing theory
Cited by (10)
Cost-sensitive classification with respect to waiting cost
2010, Knowledge-Based SystemsEfficiency and welfare implications of managed public sector hospital waiting lists
2008, European Journal of Operational ResearchOperations research applications in hospital operations: Part II
2016, IIE Transactions on Healthcare Systems EngineeringAn improved artificial bee colony algorithm for minimal time cost reduction
2014, International Journal of Machine Learning and CyberneticsArtificial bee colony algorithm to minimal time cost reduction
2013, Journal of Computational Information Systems