J Clin Epidemiol Vol. 52, No. 6, pp. 499–502, 1999 Copyright © 1999 Elsevier Science Inc. All rights reserved. 0895-4356/99/$–see front matter PII S0895-4356(99)00004-9 Positive Predictive Value of ICD–9th Codes for Upper Gastrointestinal Bleeding and Perforation in the Sistema Informativo Sanitario Regionale Database Chiara Cattaruzzi, 1 Maria Grazia Troncon, 2 Luisa Agostinis, 3 and Luis A. García Rodríguez 4, * 1 Azienda Ospedaliera S. M. Misericordia, Udine, Italy; 2 Direzione Regionale della Sanità, Trieste, Italy; 3 INSIEL S.p.A., Udine, Italy; and 4 Centro Español de Investigación Farmacoepidemiológica, Madrid, Spain ABSTRACT. We identified patients whose records in the Sistema Informativo Sanitario Regionale database in the Italian region of Friuli–Venezia Giulia showed a code of upper gastrointestinal bleeding (UGIB) and perforation according to codes of the International Classification of Diseases (ICD)-9th revision. The validity of site- and lesion-specific codes (531 to 534) and nonspecific codes (5780, 5781, and 5789) was ascertained through manual review of hospital clinical records. The initial group was made of 1779 potential cases of UGIB identified with one of these codes recorded. First, the positive predictive values (PPV) were calculated in a random sample. As a result of the observed high PPV of 531 and 532 codes, additional hospital charts were solely requested for all remaining potential cases with 533, 534, and 578 ICD-9 codes. The overall PPV reached a high of 97% for 531 and 532 site–specific codes, 84% for 534 site–specific codes, and 80% for 533 lesion–specific codes, and a low of 59% for nonspecific codes. These data suggest a considerable research potential for this new computerized health care database in Southern Europe. J CLIN EPIDEMIOL 52;6:499–502, 1999. © 1999 Elsevier Science Inc. KEY WORDS. Medical record linkage, ICD-9 codes, positive predictive value, population-based study, auto- mated databases INTRODUCTION The feasibility of a pharmacoepidemiologic study relying on computer-based case ascertainment largely depends on the validity and reliability of data recorded in the automated databases [1,2]. Few studies have assessed the validity of au- tomated health care databases comparing the information from recorded discharge diagnosis data with the one in orig- inal medical charts [3,4]. We identified cases of upper gas- trointestinal bleeding (UGIB) and perforation in a recent case-control study carried on to estimate the risk of hospi- talization for UGIB associated with intake of nonsteroidal antiinflammatory drugs (NSAIDs) and other drugs [5]. The aim of the current analysis was to examine the positive pre- dictive value (PPV) of International Classification of Dis- eases (ICD)-9th revision codes used in this case-control study of UGIB and recorded in the Sistema Informativo Sanitario Regionale (SISR), a new health care automated database in the northeastern Italian region of Friuli-Vene- zia Giulia (FVG). METHODS The region of FVG maintains complete computerized records of hospitalizations since 1985 and outpatient drug prescriptions since 1991 for the 1.2 million inhabitants of the region. The Hospital Services Database collects data on all admissions to public and private hospitals of FVG. This database contains information on personal demographics of enrollees, admitting and discharge date, vital status, one ad- mitting diagnosis, and a maximum of four discharge diag- noses (ICD-9th revision) [6] and four surgical procedures. The Outpatient Prescription Database keeps a record for each prescription dispensed to FVG residents and covered by the National Health Service. The information includes the drug name, data or prescription, and strength and num- ber of refills. The Patient Identification Database maintains a record for each resident in FVG. It includes a unique per- sonal identifier (present in all computerized files) that allows a link the various databases, name, gender, date of birth, address, general practitioner (GP), and special medical conditions such as diabetes, cancer, and glaucoma. This enrollment file is updated daily for address and GP changes, births, deaths, and the arrival and departure of residents from the region. The source population of the original study included per- manent residents aged 25–89 years between January 1991 *Address correspondence to: Dr. Luis García Rodríguez, Centro Español de Investigación Farmacoepidemiológica, Almirante 28, 2°, Madrid 28004, Spain. Accepted for publication on 12 January 1999.