Data management and statistical analyses
Data were extracted from and onto a secure, encrypted SQL server controlled by NHS Improvement. Data were analysed within this secure environment using standard statistical software: Microsoft Excel (Microsoft Corp, Redmond, WA, USA), Stata (Version 13, StataCorp LLC, College Station, TX, USA) and Alteryx (Alteryx Inc, Irvine, CA, USA).
ITSA was used to investigate changes in practice in response to each intervention for each recommendation. ITSA allows temporal trends in data to be examined prior to an intervention, any immediate changes at the point of the intervention and longer term changes in trend post-intervention to be examined.6 The itsacommand in Stata was used to conduct the analysis.7The procedure used ordinary least squares to estimate the model coefficients with Newey–West standard errors used to produce confidence intervals for the estimates.
Data were categorised into the 30-day periods outlined above for the analysis. For all three datasets, periods at the start and end of the time series with fewer than 200 admitted patients per trust were excluded. Periods with smaller numbers of patients at the start and end of the time series arose since the trusts visited first had a very long post-intervention period and the trusts visited last had a very long pre-intervention period. For the day-case surgery for TURBT dataset this involved removing 6 (3 at start and end) of 94 time periods. For treatment for the use of ureteric stents dataset this involved removing 11 (7 at start and 4 at the end) of 94 time periods. For the waiting times for TURP dataset this involved removing 22 (12 at the start and 10 at the end) of 82 time periods.
Autocorrelation was investigated using the Cumby–Huizinga general test
for autocorrelation and the test implemented using the actestcommand in Stata.8, 9 Autocorrelation was only detected for the TURBT dataset and was corrected for by a lag of one. For all other datasets and outcomes studied the lag period was set at zero.