Innovations in Linking Human Services Data
The process of linking administrative records in human services has long relied on techniques that require a significant amount of manual oversight and input. Recent innovations in record matching have the potential to impact the field in meaningful ways, but applied examples are rare, as are clear guidelines for when and how new techniques should be applied.
What We Did
We conducted a literature review of record linkage methodologies, pulling from across disciplines. We also sought out commercial enterprises that may be implementing and documenting innovative approaches.
What We Found
- A range of classification algorithms (common in machine learning) can be applied to record linkage, with assorted potential benefits over traditional methods. However, these approaches have trade-offs in transparency and interpretability and do not necessarily scale any better than traditional approaches. There is also little research to understand which algorithms best suit which data sources.
- Collective matching techniques are largely theoretical at this point, although these techniques show promise to address several challenges in human services record linkage.
- Privacy-preserving record linkage approaches can be applied where there are restrictions on data access, but do not provide any new insights for techniques applied to linkage where full identifiers are available.
What It Means
- While options for new alternatives to traditional record linkage approaches in human services data exist, their potential for application is currently limited.
- The human services field badly needs an active community dedicated to record linkage methodology to develop best practices and provide guidance on new techniques.
Recommended CitationWiegand, E. R., & Goerge, R. M. (2019). Record linkage innovations for the human services. Washington, DC: Family Self-Sufficiency and Stability Research Consortium.