To the Editor: Clinical indications for cardiac rhythm management devices (CRMDs) have expanded dramatically. The impact of this expansion upon rates of CRMD infection, a complication of device therapy that often requires system explantation (1), has not been clearly defined.We hypothesized that the current growth in implantable cardioverter defibrillator (ICD) utilization would impact rates of CRMD infection, both by expanding and changing the demographics of the population at risk. We analyzed a nationally representative database to compare trends in rates of new CRMD implants with rates of hospitalization for CRMD infection. We also sought to define the rates of in-hospital mortality associated with such events.We collected files from the National Hospital Discharge Survey (NHDS) from 1996 to 2003. Cases with a primary discharge diagnosis of pacemaker (PM) or ICD infection (International Classification of Diseases-9th Revision-Clinical Modification [ICD-9-CM] code 996.61) were identified. In addition, patients with device explantation (ICD-9-CM codes 37.77, 37.79, 37.89, or 37.99) and a primary discharge diagnosis of sepsis (ICD-9-CM code 038 or 785.59), bacteremia (ICD-9-CM code 790.7), endocarditis (ICD-9-CM codes 421.0, 421.9, or 424.90), cellulitis (ICD-9-CM code 682.9), or fever (ICD-9-CM code 780.6) were defined as having CRMD infection. New CRMD implantations were identified by ICD-9-CM procedural codes 377.0 to 377.6.In order to determine the clinical predictors for in-hospital mortality among patients with CRMD infection, a control group with previously implanted PM or ICD who did not have CRMD infection was identified. This much larger group included patients with PM or ICD in situ (ICD-9-CM codes V45.01 or V45.02, respectively) who were discharged with diagnoses other than CRMD infection. Demographic factors were recorded. Other characteristics were identified, including presence of diabetes (ICD-9-CM codes 250.00 to 250.02 or 250.70 to 250.72), renal failure (ICD-9-CM codes 585, 593.9, or V56.0), and hospital size.Univariate analysis was performed using 1-way analysis of variance for continuous variables and the chi-square test for categoric variables. Multivariate analysis using a binary logistic regression test was undertaken to determine the independent predictors of in-hospital mortality. Cases were weighed using the "weight" variable for derivation of national estimates according to