A study on the use of administrative data to create a colorectal cancer database

© 2017 Caroline Jane MacCallumColorectal cancer (CRC) is a common and life-threatening disease which contributes a significant burden on global health systems. Clinical colorectal cancer registries (CCCRs) are potentially powerful tools in the study of cancer and have the ability to positively impact on the value, efficiency, effectiveness and resource allocation in the management of CRC globally. However, they are resource intensive, region specific, and compromised by reporting and recruiting bias. Hospital administrative data, which are routinely collected, are an obvious source for a more efficient and automated collection pathway of many data fields in CCCRs. They also provide comparable data across many hospitals; these data can be linked and enable comparative studies among hospitals and regions. The first hypothesis of this study was that the research output produced by CCCRs does not provide sufficient value to justify the extensive resources required to maintain these registries. The second hypothesis was that administrative data can be used to autopopulate a Victorian CRC database, and thereby be used to measure and compare clinical outcomes in CRC among health systems. In order to investigate the first hypothesis, a systematic review of CCCRs was performed which aimed to both identify and characterise the existing CCCRs, and also to measure the comparative research impact of each. 18 CCCRs were identified with sample sizes between 104 to 1,400,000 cases. Data fields, published aims and outcomes were similar between registries. The most frequently published outcomes related to anastomotic leak following colorectal surgery. The American National Cancer Database formed the basis of the highest number of publications (66), the British Northern Region Audit had the highest median article citation number (28.5), the British National Bowel Cancer Audit had the highest median impact factor (4.72), and the American National Cancer Database had the highest median Altmetric score (4.5). Overall, we found that there is a significant body of colorectal cancer outcomes research generated from the CCCRs. Given the enormous resources, the overall research output and impact of CCCRs is low in proportion to the size of the data sets. However, these registries hold key oncological and surgical outcomes data; focussing on data linkage and automated data collection will enable international comparisons. The aims of the second part of this study included identifying a cohort of new colorectal cancer patients from administrative data, measuring the accuracy of that administrative data, and deriving coding algorithms to improve the accuracy of diagnoses, procedures and outcomes sourced from administrative data. The study was based on Royal Melbourne Hospital (RMH) administrative data, medical charts and BioGrid Australia data. We found that it is possible to use administrative data to identify new colorectal cancer patients who underwent surgical resection using specific coding algorithms. Further, administrative data has an accuracy of between 80-100% for most data fields, and this accuracy can be improved using coding algorithms. We plan to use these coding algorithms across many hospitals, making the possibility of a Victorian CRC registry containing all surgical CRC patients a real possibility. Such a registry would enable outcomes research across the whole state