Patient Identity Matching

In today's rapidly advancing healthcare landscape, the effective management of patient identities is paramount. Ensuring accuracy, completeness, and coherence of patient histories not only streamlines administrative processes but also plays a pivotal role in providing quality patient care.

However, healthcare organizations often grapple with this, leading to inefficiencies, errors, and potential risks to patient safety. Deduplication, the process of identifying duplicate patients within a database, emerges as a critical solution to address this challenge. In this article, we delve into the importance and value of deduplicating patients, focusing on various methods tailored to enhance accuracy and efficiency of the process.

Complexity

Traditional methods typically rely on deduplicating with a static method, for example using First name, last name, date of birth, and administrative gender. One immediate issue with this beyond its low match rate is that many of these fields may not be available from a patient source instance.

Beyond “single-mode” methods, complications escalate. Unlike document processing, hashing is not possible as the patients we are attempting to merge have fundamentally different attributes. Computation ramps as well as the number of comparisons that must be performed goes from O(n) to O(n^2). However before delving into approach, let us evaluate attributes that are commonly available for use:

patient identification attributes

Completeness: availability of the trait

Validity: Percent of time it is correct

Distinctiveness: How unique it is to a patient

Comparability: How low is the variation of the format?

Stability: How stable is the value over the patient’s lifetime?

While the completeness value may vary from source, it is roughly consistent with what is shown here. For example, unstructured medical records list administrative gender verbatim about 25% less than what is implied here.

Eliminating low stability and low availability variables narrows the options to the following:

patient deduplication methods

We will focus on these for the rest of this blog article.

Preprocessing

Regardless of attributes used, preprocessing of patient attributes is a must. Ideally this happens before a patient entering the database, but may also be done during a merging step. Here are a few standardization options:

  • DOB: “YYYY-MM-DD”

  • Primary phone: +X XXX-XXX-XXXX (note this may require more advanced software)

  • SSN: XXX-XX-XXXX

  • First/Last Name: Forcing casing, and replacement of punctuation according to a standard procedure

These further minimize variability and reduce matching.

Multi-Modal Patient Merging

When merging patients on multiple non-intersecting methods at the same time, great care must be taken to maintain patient identities, for the following reasons:

  • Conflicting information: 2 patient entries may have 4 matching traits but 1 that is different. If the patients are to be merged, one patient will have their information overwritten for that field.

  • Chain of Events: When 3 patients are merged, in sequences of pairwise comparisons, there comes scenarios whereby order matters, given that overwriting may occur as previously stated.

  • Precision / Recall: While purely preferential in nature, attributes chosen for matching have a significant influence on the total number of patients merged, as well as the accuracy.

Tenasol makes use of the following methods for patient merging, barring client preferences:

merging patient identities

Tenasol has found that these options minimize cases of conflicting information, overcome cases of multiple or complex first names, maximize recall, and maximizes precision. Furthermore, this strategy is robust to a patient who may change non-stable attributes over the course of their lifetimes.

Conclusion

Deduplicating patient records stands as a cornerstone of effective healthcare data management, offering myriad benefits ranging from enhanced data accuracy and operational efficiency to improved patient safety and regulatory compliance. By leveraging method sets tailored to specific criteria, healthcare organizations can unlock the full potential of deduplication, driving efficiency, precision, and quality across the continuum of care. As the healthcare landscape continues to evolve, prioritizing deduplication remains essential in realizing the promise of data-driven insights and delivering patient-centered care in the digital age.

Sources

[1] Eric, H., et al. "A framework for cross-organizational patient identity management." The Sequoia Project and The Care Connective Consortium (2018).

Previous
Previous

FHIR Healthcare Data Standardization

Next
Next

AI Terminology Explained