Digitally benchmarking public attitudes to secondary use of health data, with NLP data extraction from general practices as a case study

Significant benefits can flow from re-using people's health data for research, including gaining new insights that can be used to improve preventive healthcare, diagnosis and treatment. Such data is sensitive, however, and individuals may have concerns about its secondary use even when data has been de-identified. Secondary use without appropriate consultation and justification has sometimes created public scandal that damages trust. We need better ways to understand people's views about different types of secondary use of health data so that the public interest can be better served and trust protected.

1. This project has two related aims:
2. The first is to establish a robust electronic method to gather evidence of public attitudes regarding secondary uses of health data. The team will scope and evaluate alternative electronic means of gathering and benchmarking public attitudes toward specific secondary uses of health data. Alternatives will be evaluated for cost-effectiveness, issues of inclusivity, representation and equity, and their ability to provide useful insight into public attitudes toward secondary uses of data. The most promising method will be tested in practice in relation to the second aim.
3. The second aim is to gain insight into public attitudes to data extraction from medical records for primary care research using different natural language processing (NLP) tools. Tools will be tested according to:
   1. location
   2. robustness of de-identification
   3. types of health data.

This will provide insight into public acceptability of working with industry-leading medical annotation companies that require medical text to be sent to on-line services.