During the study period, 8581 patients had an ICD implanted by 1959 physicians. Of these physicians, 1840 (93.9%) were male. shows the number of patients in each specialty category and the baseline characteristics of patients by physician specialty. The distribution of ICD implants by physician specialty is consistent with the distribution of physician specialty from the American College of Cardiology National Cardiovascular Data Registry-ICD registry (electrophysiologists, 79%; non-electrophysiology cardiologists, 17%; cardiac surgeons, 3%; other, 0.25%).10
Notably, patients who had their device implanted by a non-EP cardiologist were significantly more likely to have CHF than patients who had their ICD implanted by a thoracic surgeon. Patients who had their ICD implanted by a thoracic surgeon were more likely to have cerebrovascular disease than patients who had their ICD implanted by physicians of other specialties. The regional distribution of ICD patients varied by physician specialty. For example, 46% of all ICD implants performed by thoracic surgeons occurred in the South region as compared with 39% of ICD implants by electrophysiologists. In addition, the volume of ICDs with cardiac resynchronization therapy was significantly lower for thoracic surgeons than for electrophysiologists or non-EP cardiologists.
Baseline Characteristics of Patients by Physician Specialty
shows unadjusted ICD complications and mortality by year of ICD implantation. The number of ICD implants increased from 1644 in 2002 to 2374 in the first three quarters of 2005. The rates of any complication following ICD implantation declined from 18.8% in 2002 to 14.2% in 2005 (P < .001). A decline in mechanical complications was observed (P < .001). Two other important observations stand out. First, the majority of complications following ICD implantation occur during the index hospital stay or up to 1 day following discharge. Second, mechanical complications are the most common index complication with 7.2% in 2002 and 3.8% in 2005.
Unadjusted ICD Complications and Mortality by Year of ICD Implantation
The unadjusted rate of index ICD complication differed by gender (10.3% in men and 12.1% in women, P = .03). No significant differences in the rates of any ICD complication were observed by georgraphic region (15.4% in the West, 15.5% in the South, 17.3% in the Midwest, 16.2% in the Northeast), or by race (15.5% in blacks, 14.2% in other, and 16.3% in whites). The rate of any ICD complication was significantly higher if the ICD was implanted following an admission from the emergency room (17.2%) versus other inpatient (16.5%) or outpatient (12.3%) setting (P = .002).
The distribution of ICD complications and mortality by physician specialty is shown in . The rate of any complication following ICD implantation was significantly higher for thoracic surgeons than for physicians of any other specialty.
Unadjusted ICD Complications and Mortality by Physician Specialty
From 2002 to 2005, we observed a decline in 1-year mortality (P < .001). This decline is illustrated in with Kaplan-Meier mortality curves stratified by year of ICD implantation.
displays the results of a multivariable model of 1-year mortality following ICD implantation. After controlling for other confounders, factors associated with an increased risk of 1-year mortality are older age, history of MI, CHF, chronic lung disease, dementia, diabetes with and without complications, metastatic cancer, peripheral vascular disease, renal disease, and admission from the ER. Hypertension, an outpatient implantation setting, and more recent years of ICD implantation were associated with a lower risk of 1-year mortality. None of the physician specialties was associated with increased mortality.
Multivariable Model for 1-Year Mortality
displays the results of a multivariable model of time to any ICD complication. Independent factors associated with an increased hazard of any complication include chronic lung disease, dementia, renal disease, implantation of the device by a thoracic surgeon and implantation with removal/replacement. History of CHF, an outpatient implantation setting and more recent years of ICD implantation were associated with a lower risk of complications. In multivariable models, men were at a significantly lower risk of index pneumothorax (HR, 0.50; 95% CI, 0.34–0.75; P < .001) than women. As measured in the 5% sample, physician volume was not significantly associated with the risk of complications even after testing for thresholds of < 20 implants per year, < 15 implants per year, and < 10 implants per year in sensitivity analyses.
Multivariable Model for Time to Any ICD Complication