1 Comments
Do you have thoughts on how common data elements (CDEs) may be used in NIH-supported clinical research? If so, please share them in response to this recently released Request for Information.
CDEs allow for easier exchange of data across different research areas because they are standardized, consistent, interoperable, and defined. They represent one way that NIH implements FAIR (Findable, Accessible, Interoperable, and Reusable) data principles. In addition, as we mentioned in this post from 2021, CDEs help us implement one aim of the NIH Data Science Strategic plan, which is having data resources that maximize the value and rigor of NIH research investment.
Identifying the right CDEs can be difficult. This is because data collection methods and models vary between disciplines. Systems may need to be adapted to collect new information. Common vocabulary and definitions must be agreed upon. And researchers need time to find the appropriate CDEs for use in their project.
NIH-supported clinical research represents an area where using CDEs may have promise though. Currently, there are no established CDEs for use across all NIH-funded clinical research activities. But if appropriate CDEs are adopted, then our supported clinical research data could be integrated with other healthcare information obtained from electronic health records, coverage claims, and patient-reported outcomes, thereby allowing for new analyses to be performed. With this in mind, we seek your input on the following:
- Recommended CDEs for NIH-funded clinical research/trials, including a set of minimal core CDEs
- Tools and technologies that could enhance the use of NIH CDEs
- Policies facilitating and/or incentivizing broader CDE usage in research and in data sharing and management
Perhaps you have thoughts on what demographic characteristics CDEs should collect for participants in all NIH funded clinical research? Maybe how to establish and determine the minimum core CDEs necessary? What about sharing how you selected the right CDEs for your clinical research area that others may find helpful, especially if they help harmonize data? Where are the potential roadblocks to adopting CDEs, and how might they impact data sharing from clinical research?
We would greatly appreciate your feedback by April 20, 2024. Comments may be submitted online via the RFI or by emailing a PDF response to [email protected].
For behavioral data, the following would be helpful: sex, age, race, ethnicity, diagnosis (e.g., no disorder, name of disorder[s], and how that was established) region where data was collected, inclusion/exclusion criteria, SES or an SES proxy. We have found that asking directly about income deters participation. Therefore, an evidence-based proxy (e.g., Mother’s education level) needs to be allowed. Relevant to disorder status, how the status was established is important. Many studies of normal processes just assume that their participants were in fact normal, when they may or may not have been and some disorders are high enough prevalence in the population that “normal” samples are likely to include them. Conversely, some behaviorally-defined disorders are being ascribed to participants based on self report, which may or may not be accurate. If data is re-used, we should know the level of rigor behind the diagnostic status.