Searching for a needle in a haystack: Use of ICD-9-CM codes in drug-induced liver injury

Objectives: The aim of our study was to compare three search strategies using a computerized administrative database to identify cases of idiosyncratic drug-induced liver injury (DILI) due to amoxicillin/clavulanic acid, phenytoin, valproic acid, and isoniazid. Methods: In search 1, electronic medical records from patients seen between 1994 and 2004 with an ICD-9-CM code of acute liver injury were identified and cross-searched for the specific drug names in the dictation text. In search 2, all patients with an ICD-9-CM code of drug poisoning/overdose due to one of the four study drugs were identified. In search 3, patients with a poisoning code as well as an acute liver injury code were identified. Results: Review of the records from the 7,395 search 1 patients yielded 51 DILI cases (0.7%). In contrast, the 566 search 2 patients yielded only three DILI cases (0.5%). Finally, search 3 provided the greatest specificity but a low rate of detection with only two patients (3.9%) having DILI due to one of the four drugs. Conclusion: Acute liver injury ICD-9-CM codes combined with a text search of the dictated medical record yielded the greatest number of DILI cases but was less specific than crossing acute liver injury and poisoning codes. Use of ICD-9-CM codes to identify rare adverse events like DILI remains problematic and highlights the need for prospective surveillance networks.