OBJECTIVES: The main purpose of study was to evaluate the utilization of implantable cardioverter defibrillator (ICD) with or without cardiac resynchronization pacing (CRT-D) in Lombardy, the most populated Italian region with universal healthcare coverage for about 10 million inhabitants, from 2000 to 2010. The second aim was to assess healthcare resources utilization after a first ICD implantation, highlighting differences between implant indication (primary and secondary) or ICD type (single-chamber, dual-chamber and CRT-D). METHODS: Data were extracted from: i)the data warehouse DENALI that organizes healthcare administrative databases concerning all subjects covered by Lombardy Health System; ii)ICD National registry data. Linking DENALI and registry data, we developed and validated (with Cohen’s kappa coefficient) an algorithm to distinguish hospitalizations for ICD in first implant and replacement in order to estimate the annual rates of first implant (per million person-years) and replacement (per hundred person-years). We selected a cohort of patients who underwent a first ICD implantation and we followed them until 12/31/2010, recording vital status and healthcare resources consumed: hospitalizations, drugs and outpatient claims. RESULTS: We identified 25,358 hospitalizations for ICD implantation: 17,864 for a first ICD and 8,034 for a replacement. 86% of hospitalizations found a match in the National registry and the Cohen’s kappa coefficient showed values over 0.8 from 2005. The annual rates were 232.5 (95%CI:228.5-236.4) and 10.5 (95%CI:10.3-10.7) for first ICD and replacement respectively. We selected a cohort of 12,525 patients who underwent a first ICD: 55% were implanted in primary prevention and the ICD types were: 35% single-chamber, 29% dual-chamber and 35% CRT-D. Among these groups, we detected differences in survival, ICD duration and healthcare resources utilization. CONCLUSIONS: The combined use of information from national ICD registry and healthcare administrative databases could overcome the limitation of both data sources and it could improve the monitoring of ICD therapy.