Several case studies conducted by the IMI-GetReal project planned to use registry data to inform their data analysis. For example, by using individual participant data (IPD) from a patient registry in a network meta-analysis to establish the relative effectiveness of medicines in situations where no head-to-head RCT data is available. For some case studies, the registries seemed very willing to engage with the case study teams and provide access to registry data. However, for other case studies, registry data access was declined, despite compliance with the registry holders’ procedures for access.
To apply for data access, each case study team had to comply with the general procedures required by each registry. This generally included filling out an application form and submitting a research proposal. These documents were subsequently reviewed by the scientific advice committee, general advisory committee, or both. Afterwards a final decision was made.
Each proposal needed to be tailored to the specific registry holder. For example, case study teams were encouraged by registry advisors to involve academic groups, which conventionally perform all statistical analyses for the registry, as a means of increasing the chance of access approval. In some cases, this meant that clearance was required from GetReal (as well as funding) to involve other academic groups affiliated with the registry to participate.
Issues with access to registries
The following list outlines some of the common issues that were experienced when applying for access to registry data. We recommend that these should be considered before approaching registry holders for access to data:
- Timelines too short: timelines were sometimes unrealistic and difficult to maintain because of the time needed for the application procedure, difficulties in engaging the registry steering committees, expected timelines for approval used by steering committees, and cleaned registry data being unavailable within the case study time frame.
- Difficulty submitting a data access application: in some instances it was difficult to submit an application for data access to the steering committee. For example, sometimes little information was known about when the registry steering committee would consider the application.
- Access affected by registry type: accessibility to disease-specific registries was particularly difficult compared with product registries.
- Access affected by the disease area of the registry: access to registry data on chronic diseases may be easier than for oncologic indications. For example, the case study team of the IMI-GetReal rheumatoid arthritis case study were actively approached by the registry to inform them of the availability of a new update of the data.
- Clean data: once IPD from a registry is made available, cleaning the data is often a lengthy process.
- Ongoing academic research: in some instances, data access was declined because of a conflict of interest, for example, when academic work (such as a PhD thesis) was performed by parties involved in the registry. In other instances, data access was granted under the provision that analysis conducted would not interfere with the academic work being done by registry affiliates.
- Legal aspects: guaranteeing patient consent may be cumbersome and current European legislation on privacy of health data are also restrictive.
- Collaboration: working together with researchers directly involved with the approached registry can be constructive in gaining access to registry data. Alternatively, it may be possible to outsource the analysis to the research team directly associated with the registry.
- Purchasing registry data: some registries provide the possibility to purchase their data. Although this may be an option, it can be very costly and require substantial funding.
- Network: having a connection with one or more individuals involved in a registry is crucial to gain data access, since they can help initiate the dialogue on applications for data access.
Issues with the registry data provided
When access to registry data was granted, most often summary or aggregate data were provided, rather than access to IPD. This could present several obstacles, depending on planned analyses, since aggregate data does not conventionally include detailed information on covariates. As a result, it becomes difficult for research teams to examine the effects of covariates on observed outcomes. This may be a strong limitation when patient characteristics and treatment patterns, for example, have a considerable effect on the outcome measure (for example, overall survival).
Amr Makady and Rachel Kalf, Zorginstituut Nederlands (ZIN)