Thirty-day postoperative mortality after colorectal cancer surgery in England
These eAtlases contain information on the 30-day postoperative mortality rates of all English NHS Trusts and Cancer Networks undertaking major surgery for colorectal cancer. Full details of the methods used to generate these figures are available in the paper: Morris EJA, Taylor EF, Thomas JD, Quirke P, Finan PJ, Coleman MP, Rachet B & Forman D. Thirty-day postoperative mortality after colorectal cancer surgery in England. Gut 2011 doi:10.1136/gut.2010.232181.
The first eAtlas contains funnel plots that are based on crude analyses and others adjusted for the risk factors of age, sex, year of diagnosis, tumour site, socioeconomic deprivation, co-morbidity, stage of disease at diagnosis and operation type. Funnel plots are available for the time periods of 1998 to 2002, 2003 to 2006 and 1998 to 2006.
The methology described in this paper has also been applied to data covering the period 2007-2008 for eight Trusts with significantly better or worse postoperative mortality rates than expected during 2003-2006. Although subject to further checking, preliminary results show that the performance of all of the Trusts with worse than expected 30-day postoperative mortality rates in 2003-2006 has improved in 2007-2008 and that none of these remains an outlier. Results for these eight Trusts are displayed in the second eAtlas. Full results for this period will be published once completed.
This study is the first to provide a comprehensive, national perspective on the 30-day post-operative mortality associated with colorectal cancer surgery across England. An NCIN data briefing summarises some of the key findings.
Overall, 6.7% of the study population died within 30 days of surgery amounting to 10,704 deaths. There was significant variation across the population with post-operative mortality greater in the elderly, men, the socio-economically deprived, those with advanced stage disease at diagnosis or with additional co-morbidities and amongst those operated upon as an emergency.
Significant variation, independent of case-mix, was also observed between hospital Trusts and Cancer Networks. This variation is quantified in this eAtlas in the form of funnel plots. Each dot in the funnel plot represents an NHS Trust or Network. The solid horizontal line represents the national post-operative mortality rate for England over the study period whilst the coloured lines are the 95% and 99.8% control limits. Trusts that lie above the upper control limits have significantly worse 30-day postoperative mortality rates than expected and those below the lower control limits significantly better. Trusts lying outside the upper 95% control limits can be considered to be in a ‘warning’ zone whilst those outside the upper 99.8% control limits are in an ‘alarm’ zone. Such differences may warrant further investigation. For further information on how to interpret funnel plots see Speigelhalter DJ. Funnel plots for comparing institutional performance. Statistics in Medicine 2005; 24: 1185-1202
The study is based on the routine health data available within the National Cancer Data Repository. In this study this is essentially linked cancer registry and Hospital Episode Statistics data. These data are submitted by NHS providers to inform the commissioning of services. This study is reliant on the quality and accuracy of these data. If there are errors in the data submitted by Trusts then the results of this study may be incorrect.
The study has shown that many factors influence 30-day post-operative mortality. These may relate to the patient (for example, stage of disease or level of co-morbidity) or the institution offering care (such as the specialisation of the operating team, the quality of post-operative care or the availability of beds in high dependency and intensive care units). Whilst the study has identified providers with outlying 30-day post-operative mortality it is not possible to determine from the data available what aspects of care or, indeed, if the quality of care within these units is deficient. The outlying status could be explained by problems in data quality, chance or case-mix factors not quantified in this study.
Efforts should be made to understand why some institutions appear to have an outlying status. With this information it should be possible to learn from those achieving good outcomes, by seeking the underlying causes, adding to and spreading the adoption of best practice guidelines and, ultimately, reducing post-operative mortality following colorectal cancer surgery.