Premium
Detection of adverse drug reactions: evaluation of an automatic data processing applied in oncology performed in the French Diagnosis Related Groups database
Author(s) -
Quillet Alexandre,
Colin Olivier,
Bourgeois Nicolas,
Favrelière Sylvie,
Ferru Aurélie,
Boinot Laurence,
LafayChebassier Claire,
PeraultPochat MarieChristine
Publication year - 2018
Publication title -
fundamental and clinical pharmacology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.655
H-Index - 73
eISSN - 1472-8206
pISSN - 0767-3981
DOI - 10.1111/fcp.12333
Subject(s) - pharmacovigilance , medicine , database , false positive paradox , medical diagnosis , adverse drug reaction , drug reaction , adverse effect , medical record , drug , pharmacology , pathology , machine learning , computer science
The aim of this study was to assess an automated detection method of serious adverse reactions induced by oral targeted therapy ( OTT ) in patients with cancer, performed in the French Diagnosis Related Groups ( DRG ) database. Patients with cancer of the Poitiers hospital who started an OTT between 2014 and 2015 were included. This study focused on adverse drug reaction which required inpatient hospitalization ( ADR h ). All diagnoses coded in the DRG database for hospital stays that occurred within 3 months after OTT initiation were collected (potential ADR h ). Filters (exclusion criteria) were automatically applied on potential ADR h to exclude diagnoses that were not adverse drug reactions (false positives). A pharmacovigilance review was carried out to identify ADR h in the medical records (reported ADR h ). The sensitivity and specificity of the detection method were estimated for each filter combinations by comparison between potential and reported ADR h . This study included 129 patients. The medical records review led to identify 19 ADR h (all coded in the DRG database) in 14 patients. To maintain a 100% sensitivity of the method detection, the best specificity obtained was 58.3% (95% IC : [55.2–61.4]).The use of restrictive filters (‘drug’ in the diagnostic label, specific diagnosis code for adverse cancer drug reaction) resulted in a 97.8% specificity (95% IC : [96.6–98.5]) with a 38.2% sensitivity (95% IC : [23.9–55.0]). Our method has detected the third of ADR h with an excellent specificity. Complementary experimentations in pharmacovigilance centers are necessary to evaluate the interest of this tool in routine in addition to spontaneous reporting.