Article Text
PDF
Abstract
Background Analysis of the ‘weekend effect’ and discussions about ‘seven-day working’ in the NHS are witnessing a resurgence. Funnel plots have been used as the graphical tool of choice to compare institutional performance, but have not been used to compare performance of days of admission so far.
Aims
To compare the performance of different days of admission using risk-adjusted standardised mortality ratios in a funnel plot.
To use logistic regression analysis to verify the validity of funnel plot interpretation as Simpson’s paradox may confound the comparative analysis among subsets of the same group of population
Methods We analysed all admissions from 2008–14 using PIM-2r for risk-adjustment using PICU mortality as the primary outcome measure. Funnel plots for risk-adjusted SMRs with 95 and 99.7% control limits were drawn for elective, emergency and all admissions combined. Weekend admissions (versus weekday admission as reference), emergency admission (versus elective as the reference), out-of hours admission (22:01 to 06:59) and an illness severity marker (using adjusted PIM-2r with logit of nature of admission component removed) were used as explanatory variables in the logistic regression model.
Results 1857 (20%) of the 9133 admissions to PICU during the study period occurred during the weekend. 82% of those were emergency admissions compared to only 59% during the weekdays (p < 0.001). Median probability of death using adjusted PIM-2r for weekend admissions was higher than for weekday admissions. [3.7 (1.4–7.2) versus 2.1 (0.7–5.4)%; p < 0.001]. Funnel plots (Figure 1) and outputs from logistic regression (Table 1) are shown below.
Results of the logistic regression analysis
Conclusion We did not find any evidence of weekend effect in PICU. We have shown that funnel plots may be used for comparison of different days of admission. However, caveats associated with interpretation of even adequately risk-adjusted analysis should always be considered.