Automated data collection in cancer care: State of play among registries in the United Kingdom and Europe.

BACKGROUND: Automated clinical coding can use statistical or artificial intelligence-based technology to transform unstructured clinical data into clinical codes. These processes have the potential to enhance the quality and accuracy of data collections, save resources and accelerate research. OBJECTIVE: To evaluate the use of automated clinical coding in the United Kingdom (UK) and European cancer registries. METHOD: An online electronic survey was formulated to evaluate the use and user opinion of automation within cancer registries. The survey was distributed to members of the United Kingdom and Ireland Association of Cancer Registry and the European cancer registries. Data analysis was performed using Microsoft Excel 2015® version 15.13.3 in order to summarise the results. RESULTS: Twenty-three of the 117 cancer registries responded to the distributed survey; 15 (12.8%) cancer registries used automation within their registry, mainly in the form of natural language processing or machine learning. Most of the sampled registries (73.3%) used these technologies to automate data collection from pathology reports; 87% of respondents reported automation as efficient; and 26.1% reported improved data quality; 12 (52.1%) of cancer registries still manually checked all the automations; and 17 (74%) respondents believed that the algorithms for difficult tasks require further development. CONCLUSION: Various computer-based algorithms have been used for automated clinical coding in the UK and European cancer registries in the past few decades; however, to date there are no published data to validate its use. Further research and development of these technologies is needed to ensure external validity and maximise the potential use within other cancer registries globally.Implications for health information management practice:It is clear that while automation can be advantageous in areas of clinical coding, the role of the "human" (HIMs and clinical coders) in coding and classifying registry data, and in overseeing the transition, will be required for some time yet.