Linking survey responses and administrative records without respondent consent to link; a discussion paper

Abstract

ABSTRACT Objectives (i) Review the application and interpretation of the Data Protection Act (DPA) 1998; clarifying whether individual consent is required for data linkage for secondary research purposes, in consideration of the policies and principles of the UK Administrative Data Research Network (ADRN). (ii) Determine ethical, logistical or ‘tactical’ factors researchers might have to take into consideration. Approach Linking survey data to administrative records offer potential advantage to both researchers and survey respondents. Informed, specific and explicit consent is typically a prerequisite for linkage. However, not all respondents consent to data-linkage resulting in a reduced and potentially biased sub-sample for analysis. In Northern Ireland consent rates for record linkage are typically about 50%. Discussion with the ICO confirms that the DPA may encourage rather than restrict research. S33 ‘research exemption’ supports secondary use of survey data subject to conditions such as that is for research purposes, it is not incompatible with the original purpose, and would not cause the data subjects substantial damage or distress. Other DPA principles remain in force; Principle 1 (fair and lawful processing) and the need to make data subjects ‘aware’ of the research: explicit consent is only one route by which this can be achieved. The Processing Sensitive Data Order (2000) protects the privacy of individuals. Research must be of substantial public interest with access only to data necessary to answer the research question. The ADRN enables access to de identified data for research purposes where identified public benefit is independently assured, maintains the privacy of individuals and ensures lawful “conditions of processing” are met. Resulting discussion Legal considerations aside, researchers may face other obstacles; the first is technical as the surveying agency may have deleted all linkable identifiers. The second is ethical as research ethics committee approval is a usual precondition. De-identification of individual subjects should be grounds for ethical approval where the research proposal includes a publication plan with appropriate methods to inform participants of research conducted and findings. Finally, the data custodians may not agree to the linkage for sound ‘tactical’ longer-term reasons, even if convinced of its legality. Conclusion Use of de identified survey data for research purposes is possible via the UK ADRN but raises other considerations for researchers and data custodians. We argue that this option should be used in limited circumstances.