Context Implantable cardioverter-defibrillators (ICDs) have diffused rapidly into clinical practice with little evaluation of their real-world effectiveness.

Objectives To determine the effect of the adoption of ICD on patient safety, particularly with respect to physician volume and early outcomes.

Design Retrospective cohort of all ICD implantations in New York state from 1997 to 2006, with follow-up at 90 days and 1 year.

Setting New York state non-federal hospital discharges in which an ICD was implanted during the admission. Patients were followed forward for 1 year for subsequent admissions.

Patients New York state residents undergoing ICD implantation.

Main outcome measures Effects of annual and career ICD implantation volume on 90-day complication, readmission, reprogramming, mortality and revision of the ICD within 1 year.

Results This cohort (N=38 992) represents a period of rapid adoption and implementation of this new technology, with frequency more than tripling between 1997 and 2006. We identified 6439 (16.5%) post-implantation complications and 1093 (2.8%) deaths within 90 days of implantation. The majority (73.4%) of physicians implanted one or fewer ICDs per year, and 11.0% of all implantations were performed by these very-low-volume operators. Patients treated by very-low-volume operators were more likely to die (RR=1.8, 95% CI 1.3 to 2.4) or experience cardiac complications (RR=4.7, 95% CI 3.3 to 6.8) even after the adjustment for case mix compared to operators who frequently performed ICD implantation.

Conclusions These findings suggest a need for safe and effective implementation strategies for new medical technologies, which minimize patient risk due to rapid diffusion among inexperienced providers and assure that the intended benefit can be maximised rapidly.