VALIDATION OF DIAGNOSES CODES FOR IDENTIFYING PATIENTS WITH HEPATITIS C VIRUS IN EHR IN A US, REAL-WORLD POPULATION
Author(s)
Ellen Stein, MS.
Principal Research Scientist, TriNetX, LLC, Cambridge, MA, USA.
Principal Research Scientist, TriNetX, LLC, Cambridge, MA, USA.
OBJECTIVES: There is a dearth of research validating diagnostic codes to identify HCV patients in the US, yet these codes are used in studies globally to identify HCV patients. There are inconsistencies in existing studies using laboratory results versus laboratory results with diagnostic codes to identify chronic versus active HCV. Given these shortcomings, this study aims to validate using diagnoses codes to identify HCV infection as compared with the gold standard of lab-confirmed infection among the general US, healthcare-seeking population.
METHODS: This study leverages TriNetX’s Linked EHR + Closed Claims network on the TriNetX Live platform, a network of de-identified EHR data within the US linked to closed claims, to identify patients with least two documented encounters at least 365 days apart since 2016, who are at least 18 years of age at second encounter. Among those patients, true positives were defined as those with a diagnosis of HCV and a positive laboratory test for HCV, while true negatives have neither. False positives were defined by a diagnosis of HCV but no positive laboratory test, and false negatives vice versa. Sensitivity, specificity, and PPV were calculated.
RESULTS: Among the 21,233,263 patients in the Linked EHR + Closed Claims network, 15,617,618 patients met the eligibility criteria, and among those 55,957 were identified as true positives, 15,268,347 true negatives, 274,726 false positives, and 18,588 false negatives. Sensitivity was 75.1%, specificity 98.2%, and PPV 16.9%.
CONCLUSIONS: Within the TriNetX Linked EHR + Closed Claims network, using diagnosis codes to identify HCV as compared with the gold standard of lab-confirmed HCV was moderately sensitive, highly specific, and had a low PPV. Low PPV likely reflects low prevalence in the overall population, which was crudely calculated as 0.4% of eligible patients. This aligns with other published literature, though PPV is lower than among other cohorts, such as cirrhotic patients.
METHODS: This study leverages TriNetX’s Linked EHR + Closed Claims network on the TriNetX Live platform, a network of de-identified EHR data within the US linked to closed claims, to identify patients with least two documented encounters at least 365 days apart since 2016, who are at least 18 years of age at second encounter. Among those patients, true positives were defined as those with a diagnosis of HCV and a positive laboratory test for HCV, while true negatives have neither. False positives were defined by a diagnosis of HCV but no positive laboratory test, and false negatives vice versa. Sensitivity, specificity, and PPV were calculated.
RESULTS: Among the 21,233,263 patients in the Linked EHR + Closed Claims network, 15,617,618 patients met the eligibility criteria, and among those 55,957 were identified as true positives, 15,268,347 true negatives, 274,726 false positives, and 18,588 false negatives. Sensitivity was 75.1%, specificity 98.2%, and PPV 16.9%.
CONCLUSIONS: Within the TriNetX Linked EHR + Closed Claims network, using diagnosis codes to identify HCV as compared with the gold standard of lab-confirmed HCV was moderately sensitive, highly specific, and had a low PPV. Low PPV likely reflects low prevalence in the overall population, which was crudely calculated as 0.4% of eligible patients. This aligns with other published literature, though PPV is lower than among other cohorts, such as cirrhotic patients.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
RWD7
Topic
Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems
Disease
SDC: Infectious Disease (non-vaccine)