Adaptation and Validation of a Bidirectional Matching Algorithm to Enrich Hospital Data With Information From the French National Death Registry: A Case Study in Patients With Lung or Prostate Cancer (MO-REAL)

OBJECTIVES: To validate the performance of the adaptation and integration in a RWE Platform of the matchID algorithm to automatically link data from hospitals EMRs to the National French Death Registry (INSEE).

METHODS: We linked EMRs from 3 hospitals (Tarbes, Mont de Marsan, Troyes) with INSEE database via the quasi-deterministic matchID bidirectional algorithm, exploiting personal identifiable information (name, surname, birthdate, sex)

RESULTS: Between January 2022 and December 2023, we identified 544 patients with a diagnosis of prostate (26,5%) or lung (73,5%) cancer in EMRs and a recorded death (discharge status) from the hospital information system (69% at metastatic stage). For lung cancer, the mean age was 69 years old at time of death among which 29% were female patients. While for prostate cancer, mean age was 81 years at time of death. Using all personal information available, nearly 100% of patients with a death record in the hospital system were identified in the INSEE database (range 94,4% prostate cancer to 100% for lung cancer in Tarbes). Few differences were observed by cancer type (98,61% matching for lung cancer; 100% for prostate cancer), with a homogenous performance of the mapping algorithm between the 3 hospitals [range % - 100%]. Removing date and place of death did not alter the sensitivity of the algorithm. In a confirmatory analysis on all deaths recorded in the 3 hospitals in 2022 and 2023, 98% of all death events (5,183 out of 5,289) were successfully matched to the National French Death Registry.

CONCLUSIONS: The study validates the quasi-deterministic matching algorithm between hospital and mortality death registry in France. Since a large proportion of deaths occur outside the hospital, the study addresses endemic underreporting of hospital mortality in long-term studies and validates the accuracy of the public mortality database.