PPRL Boosts Patient Matching Across Clinical Research Networks

Privacy-preserving record linkage (PPRL) technology helped enhance patient matching within the National Patient-Centered Clinical Research Network (PCORnet), according to a study published in JAMIA.

PCORnet network partners harmonize EHR and claims data to the PCORnet Common Data Model (CDM) specifications and use DataMart resources to respond to queries.

Researchers used a PPRL solution from Datavant to quantify patient overlap across ∼170M patient records and report a de-duplicated analysis of the PCORnet population. PPRL technology produces unique sets of de-identified tokens to match patients.

Researchers found a high degree of variation in the overlap between DataMarts, with the highest percentage occurring between geographically close organizations or those with an organizational or health plan relationship.

Overall, the study authors found the greatest success in matching records with tokens based on first and last name, gender, date of birth, and current zip code.

After using PPRL, the researchers saw increases in the prevalence of clinical characteristics such as asthma, depression, and cancer.

“Given that most of the characteristics in our summary table are describing somewhat chronic conditions, we would likely have seen an even bigger increase if we examined acute events or conditions that might occur out of system, and therefore be missing from an EHR-based DataMart,” the study authors wrote.

“This is evident by the fact that we saw the largest absolute increases in prevalence with depression and the presence of a result from an LDL laboratory test,” they said.

Mental health conditions are often treated outside of traditional health systems, and diagnoses associated with those visits might be more likely to appear in an external claim within a health plan research network (HPRN).  

While outside facilities may share laboratory test results for care coordination, the receiving facility may not be able to use those data for research. Therefore, record linkage can be particularly useful for studies that need outcomes or baseline clinical characteristics that are more likely to fall into this “out of system” category.

Researchers noted that the study is limited as it did not conduct a validation of the linkage algorithm itself by comparing the matched charts at different health systems to confirm they were the same patient.

“In addition, while we were able to show that supplementing EHR data with administrative claims from health plans resulted in an information gain related to prevalent conditions, governance challenges limited the scope of data that could be used in our analysis (number of variables and timeframe),” they wrote. “A more robust characterization of clinical features would illustrate the full breadth and depth of this linkage.”