Waiting times tracker: analysis of seasonal effects

An important issue in understanding trends in waiting times – and what may be influencing changes since June – is seasonal variations. Hospital activity, for example, has long shown peaks and troughs at certain times of the year – most notably at holiday (troughs) and post-holiday (peaks) periods, especially in December/January and also at Easter and in August/September.

Similar patterns can also be seen in the way waiting times change over the year too.

For month-on-month changes in median waiting times, there is a noticeable seasonal pattern, although the timings and strength of the peaks and troughs vary between the three stages of waiting (patients still on waiting lists, patients treated in putpatients and those treated as inpatients), and between years.

Click on the graphs below to expand them

Month-on-month percentage change in median waits: Patients still on waiting lists

For median waiting times for patients still on waiting lists there are notable peaks in August and December and troughs in September/October and February.

seasonal-effects-graph1.jpg

Month-on-month percentage change in median waits: Patients treated in outpatients

The seasonal pattern for median waits for outpatients is also clear from the figure below. However, the timing of the peaks and troughs is different, with notable peaks in January, April and September and troughs in December, February and October.

seasonal-effects-graph2.jpg

Month-on-month percentage change in median waits: Patients admitted as inpatients

The seasonal pattern for inpatients is similar to that for outpatients.

month-on-month-percentage-change-median-waits-inpatients.jpg

Table: Correlation co-efficients (Pearson r1): Month-on-month percentage change in median waits between years

The seasonal trend pattern that is evident visually is also clear statistically. The table below shows the correlations between pairs of years for each stage of waiting and shows strong positive associations – ie: changes up or down from month to month in one year, move in a very similar way to changes in another year.

seasonal-effects-table1.jpg

Seasonal effects and the proportion of patients waiting more than 18 weeks

However, there is less evidence of a seasonal pattern for month-on-month changes in the proportion of patients waiting more than 18 weeks, as the more erratic pattern in the figures below show.

Month-on-month percentage change in proportion of patients waiting more than 18 weeks: Patients still on waiting lists

seasonal-effects-graph4.jpg

Month-on-month percentage change in proportion of patients waiting more than 18 weeks: Patients treated in outpatients

Month-on-month percentage change in proportion of patients waiting more than 18 weeks: Patients admitted as inpatients

seasonal-effects-graph6.jpg

Table: Correlation coefficients (Pearson r1): Month-on-month percentage change in proportion of patients waiting more than 18 weeks

Statistically, the weaker and more erratic seasonal pattern is evident too, as the table below shows, with correlations varying between years and also changing also changing between a positive and negative association.

seasonal-effects-table2.jpg

Some predictions for the future

On the basis of the seasonal pattern for median waits it is possible to offer a prediction for November's median waits based on changes in median waits from October to November in previous years. For patients still on waiting lists the median wait is likely to rise very slightly to just over 5.9 weeks. Median waits for outpatients could also rise very slightly to just over 4.3 weeks. However, median waits for inpatients could fall from 9.1 to around 8.7 weeks in November.

As subsequent waiting times data are published we will update our analysis to see if seasonal effects continue to be an important factor in changes to median waiting times, and whether they continue to play a relatively small role in explaining changes in the proportion of patients waiting over 18 weeks. We will also see if our predictions for median waits turn out to be correct.

Keep up to date

Subscribe to our email newsletters and follow @TheKingsFund on Twitter to see our latest news and content.

Comments

#286 Mary E Hoult
community volunteer

John,you have produced an impressive analysis of key statstics but what of the patient experience?to give you an example,I was told I needed to be refered Nov 2010 as a Consultant to Consultant referal, but since all these changes that means you have to go back and starting blocks agian via your GP.I received an appointment via choose & Book for Feb 2010 was seen and had further tests early march with a follow up appointment in early june 2010,3 appointments and cancellations later!! I got my test results Nov 2010.Whatever analysis is made this is not a good patient experience and could also be a patient safety issue.Nine months to receive test results for what is quite a serious condition.Perhaps further work is needed on consultant to consultant referals.

#287 Andrea
MA
KF

Agreed.

#289 Mary E Hoult
community volunteer

Sorry first date should read 2009 not 2010

#535 Kris Wright
Senior Information Manager
NHS/Scottish Government

John,
Thanks for sharing this analysis, i found it interesting as i too have investigated "seasonality" within an orthopaedic service. I found the term seasonality quite daunting and i investigate this strictly with a statistical test for seasonality. I then found out that the data that was provided was not actually seasonal though it did conform to some pattern that appeared seasonal. I would be interested to hear if you tested your data for true seasonality or whether you interpreted the pattern as seasonal from a visual perspective. The question behind your analysis is intriguing, what causes this apparent variation/seasonality? I would conjecture that the main causes in the build up of a queue is not only variation/seasonality in demand but clinicians cancelling clinics for various "summer" or "winter" reasons coupled with clinicians taking leave

#539 John Appleby
Chief Economist
The King's Fund

Kris
No, I havern't tested the seasonality of the data statistically - difficult to do as we only have a few years' worth of data. I have used the term 'seasonal' somewhat broadly simply to decribe (from eyeballing the trends plus a few basic correlations) what appears to be some regular patterns at certain times of the year. These patterns will be a combination of demand and supply factors (with, I suspect, a bias towards the latter).

Add new comment