Background:
New South Wales has a large number of hospitals performing oesophagectomies and pancreatectomies for cancer with curative intent for a small number of cases. A programme to foster expert multidisciplinary care for people with oesophageal and pancreatic cancers has recently been implemented. This will result in fewer hospitals performing oesophagectomies and pancreatectomies, and increase the volumes and improve outcomes in the remaining hospitals.
Aim:
The aim of this project was to provide data to inform the programme and to plan for the provision of data to monitor the effect of the programme.
Methods:
Linked population-based hospital data were used to identify people with an incident case of cancer up to the 2013 calendar year. The percentage of people receiving surgery with curative intent was estimated. Linked hospital and death registration data enabled measurement of service delivery and mortality outcomes.
Results:
The number of people diagnosed in health districts per year ranged from 25 to 110 for oesophageal and cardia cancers and from 35 to 180 for pancreatic, ampullary and periampullary cancers. The percentage of people diagnosed who received oesophagectomies or pancreatectomies was 17% across the state for both oesophagogastric and pancreatic, ampullary and periampullary cancers, with variations between health districts in the percentage of people who underwent a procedure. Around 210 pancreatectomies and 130 oesophagectomies were performed for cancer in 2013 with 30-day mortality of 3.1% and 3.5% respectively for the 2009-2013 period.
Conclusions:
Measuring the percentage of people diagnosed who receive surgical resection and post-operative mortality is important for monitoring the effect of the programme. However, small sample sizes and a low event rate make estimates of resection percentages and 30-day mortality unstable and vulnerable to random variation. Novel statistical techniques are required to provide meaningful information on hospital and health district performance. Use of Bayesian methods will provide more reliable estimates of performance.