Author + information
- Received April 10, 2010
- Revision received June 11, 2010
- Accepted July 6, 2010
- Published online September 28, 2010.
- James V. Freeman, MD, MPH⁎,⁎ (, )
- Yongfei Wang, MS†,
- Jeptha P. Curtis, MD†,
- Paul A. Heidenreich, MD, MS‡ and
- Mark A. Hlatky, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. James V. Freeman, Stanford University School of Medicine, 300 Pasteur Drive, Falk Building, CVRC 5406, Stanford, California 94305-5406
Objectives We sought to examine the relationship between hospital implantable cardioverter-defibrillator (ICD) implantation volume and procedural complications in a contemporary, representative population.
Background Hospitals that perform higher volumes of procedures generally have better clinical outcomes.
Methods We examined initial ICD implantations between January 2006 and December 2008 at hospitals participating in the NCDR (National Cardiovascular Data Registry) ICD Registry and evaluated the relationship between hospital annual implant volume and in-hospital adverse outcomes.
Results The rate of adverse events declined progressively with increasing procedure volume (p trend < 0.0001). This relationship remained significant (p trend < 0.0001) after adjustment for patient clinical characteristics, operator characteristics, and hospital characteristics. The volume–outcome relationship was evident for all ICD subtypes, including single-chamber (p trend = 0.004), dual-chamber (p trend < 0.0001), and biventricular ICDs (p trend = 0.02).
Conclusions Patients who have an ICD implanted at a high-volume hospital are less likely to have an adverse event associated with the procedure than patients who have an ICD implanted at a low-volume hospital.
Hospitals and physicians that perform higher volumes of procedures generally have better clinical outcomes (1–6). A significant “volume–outcome” relationship has been demonstrated repeatedly for coronary revascularization procedures (1,2,7–13), but the potential for a similar relationship with other cardiac procedures has not been as well documented. A few studies suggest that outcomes are better when pacemakers are implanted by high-volume physicians (14–18), but less is known about the relationship between patient outcomes and procedure volume for implantable cardioverter-defibrillator (ICD) procedures (19). Hlatky et al. (20) found in Medicare data that procedural mortality for ICD implantation fell significantly from 1987 to 1995 and that there was a clear “learning curve” for this procedure. Al-Khatib et al. (21) analyzed Medicare data from 1999 through 2001 and showed that mechanical complications and infections within 90 days of ICD implantations were inversely related to a hospital's procedure volume.
Because ICD implantation requires the acquisition and maintenance of high-level skills, we hypothesized that those hospitals that perform more procedures should have better outcomes. This question has not been addressed with more recent experience with ICDs or in younger, non-Medicare patients. Moreover, no previous analysis has evaluated the full range of possible in-hospital adverse outcomes or the importance of the type of ICD implanted on the volume–outcome relationship. We therefore analyzed the experience of the ICD Registry to examine the relationship between hospital annual ICD implantation volume and in-hospital procedural complications in a large contemporary cohort population.
The ICD Registry was initiated in 2005. Hospitals are required to submit data on ICD implantations for primary prevention among Medicare patients, and 80% of hospitals submit data on all ICD implantations performed, irrespective of the payer or clinical indication. As described in detail elsewhere (22–24), clinical, demographic, and procedural data are recorded in a standardized format based on established data definitions, and the quality of submitted data is monitored. Data on any in-hospital adverse events related to the procedure are recorded in the ICD registry; long-term follow-up data are not available.
For the purposes of this study, we examined all patients who had an ICD implanted between January 2006 and December 2008 at a participating hospital whose data met the data quality standards established by the NCDR (National Cardiovascular Data Registry) (24). Hospitals must achieve >95% completeness of specific data elements identified as “core fields” to be included in the registry's data warehouse for analysis. In addition, there is a site auditing program that consists of annual on-site chart review and data abstraction for at least 5% of participating sites. For a given reporting period (quarter), we included ICD implantations that were performed at a hospital that reported all ICD implantations for that reporting period, irrespective of payer or indication. We excluded patients who had a previous ICD or who required epicardial lead placement. The primary outcome for this study was any adverse event that occurred during the implantation or before hospital discharge. Major adverse events included cardiac arrest, cardiac perforation, valve injury, coronary venous dissection, hemothorax, pneumothorax, deep vein thrombosis, transient ischemic attack, stroke, myocardial infarction, pericardial tamponade, and arteriovenous fistula. Any adverse event included these major adverse events as well as drug reactions, conduction block, hematoma, lead dislodgement, peripheral embolus, superficial phlebitis, peripheral nerve injury, and device-related infection.
We determined hospital annual volume by dividing the total number of ICD implantations by the number of quarters the hospital contributed data and multiplying by 4. We ranked hospitals by annual ICD implantation volume and divided them into quartiles of increasing procedure volume for descriptive analysis.
We evaluated the baseline patient clinical characteristics, implanting physician certification, and hospital characteristics among hospitals in different quartiles of annual procedure volume with analysis of variance for continuous variables and the chi-square test for categorical variables. In the primary analysis, we used hierarchical logistic regression to test for a declining rate of any adverse event according to quartiles of increasing hospital annual ICD volume, unadjusted for other characteristics, and the Cochran-Armitage trend test was used to generate the p-trend. We repeated the analysis after sequentially adjusting for patient clinical characteristics, then physician certification (electrophysiology [EP] board certified, EP fellowship, surgery boards, pediatric cardiology boards, Heart Rhythm Society guidelines, none) in addition to patient characteristics, and finally, hospital characteristics (number of beds, teaching status, urban location) in addition to patient and physician characteristics. In a secondary analysis, we tested for a declining rate of adverse events associated with implantation of each of the ICD subtypes (single-chamber, dual-chamber, and biventricular) according to increasing hospital annual ICD volume. We conducted this secondary analysis first with overall hospital annual ICD volume and second with hospital annual volume for each subtype of ICD.
There were 333,993 patients who underwent an ICD implantation between January 2006 and December 2008 at 1 of 1,356 hospitals participating in the ICD Registry. After exclusion of 31,527 patients treated at 1 of the 147 hospitals that did not report all ICD implantations for a given reporting period, 5,310 patients who had epicardial leads implanted, and 72,923 patients who had previously undergone ICD implantation, the study population consisted of 224,233 patients from 1,201 hospitals. The median length of stay was 1 day (interquartile range 1 to 6 days).
Hospital annual ICD procedure volume varied widely, with a median of 57 patients/year (interquartile range 24 to 109 patients/year). The 125 hospitals in the lowest decile of hospital annual ICD volume performed a median of 8 implants/year, whereas the 120 hospitals in the highest decile of hospital annual ICD volume performed a median of 248 implants/year. Clinical characteristics of patients varied somewhat by hospital annual ICD procedure volume (Table 1).Most differences were small in absolute terms (<2%), yet were statistically significant because of the large sample size in the study. More of the ICD implantations at the higher-volume hospitals were performed by board certified cardiac EP physicians or by fellows training in cardiac EP. In addition, higher-volume hospitals were more likely to be teaching hospitals and academic centers and were more likely to be located in urban areas.
Overall, 7,151 patients (3.2%) had an adverse event after ICD implantation, 2,701 (1.2%) of whom experienced a major adverse event (Table 2).The rate of adverse events was lower among patients who received a single-chamber ICD (2.0%) than among patients who received a dual-chamber ICD (3.0%) or a biventricular ICD (4.3%). The most common adverse events were lead dislodgement (1.0%), hematoma (0.9%), pneumothorax (0.5%), and cardiac arrest (0.3%); this pattern was consistent across quartiles of hospital annual ICD volume and types of ICD placed.
The rate of any adverse event declined significantly (p trend < 0.0001) with increasing hospital annual procedure volume (Table 2, Fig. 1).The inverse relationship between hospital ICD volume and clinical outcomes remained significant when the analysis was restricted to major adverse events (Table 2). The inverse relationship between hospital annual procedure volume and in-hospital death was not significant among all ICD types, but was significant among single-chamber ICDs (p trend < 0.03) (Table 2). The overall inverse relationship between hospital annual ICD volume and any adverse event remained statistically significant after adjustment for patient clinical characteristics, operator characteristics, and hospital characteristics (Table 3).
The adjusted odds of any adverse event after ICD implantation were significantly higher in the lowest-volume quartile compared with the highest-volume quartile (odds ratio [OR]: 1.26, 95% confidence interval [CI]: 1.05 to 1.52) (Table 3). This relationship was statistically strongest for single-chamber ICDs (OR: 1.69, 95% CI: 1.22 to 2.35) and dual-chamber ICDs (OR: 1.32, 95% CI: 1.03 to 1.68) (Table 4).For biventricular ICDs the adjusted OR for the lowest quartile compared with the highest quartile was not statistically significant (OR: 0.99, 95% CI: 0.73 to 1.37), but the inverse trend relationship between hospital annual ICD volume and outcome was significant when tested over all 4 quartiles of volume (p trend = 0.02) (Table 4).
To further investigate whether the volume–outcome relationship for biventricular ICDs meaningfully differed from the other ICD subtypes, we evaluated the correlation between the subgroups and the overall group. We demonstrated strong correlation between hospital annual volume of all ICDs and hospital annual volume of biventricular ICDs, with a correlation coefficient of 0.93. In a fully adjusted regression analysis, the interaction between ICD subtype and hospital annual ICD volume was not statistically significant (p = 0.39), excluding a strong effect modification of the volume–outcome relationship by ICD subtype.
This study demonstrates an inverse relationship between the rate of procedure-related adverse events and the annual volume of ICDs implanted at a hospital. This inverse volume–outcome relationship was statistically significant for all ICDs and for each of the ICD subtypes. The lowest annual volume quartile of hospitals in this study had 26% higher odds of any adverse event than the highest annual volume quartile of hospitals. This effect of procedure volume on outcomes confirms the findings of Al-Khatib et al. (21) in their study of mechanical complications and infections after ICD implantation in Medicare beneficiaries. Our study extends these findings by including patients of all ages, including a broader array of adverse events, and examining the type of ICD implanted.
A strength of our analysis is the availability of detailed baseline data on patient characteristics, operator characteristics, and hospital characteristics. The inverse volume–outcome relationship was unaffected by systematic adjustment for these baseline characteristics. The patients enrolled at higher annual volume centers had more adverse clinical characteristics, but there was little evidence that referral bias led to the volume–outcome relationship we observed. There were very significant differences in implanting operator characteristics; at the highest annual volume hospitals the implanters were markedly more likely to be board certified or board eligible in EP. However, the inverse relationship between hospital annual volume and outcome remained significant after adjusting for physician characteristics. Finally, the highest annual volume centers were significantly more likely to be teaching hospitals, academic centers, and geographically urban, but again adjustment for these factors did not affect the magnitude of the volume–outcome relationship. Thus, the most likely explanation of our findings is that high-volume hospitals are more likely to have better outcomes as a result of their greater experience.
We demonstrated a consistent volume–outcome relationship for all of the ICD subtypes (interaction p value = 0.39), but the relationship was not as strong for biventricular ICDs. The confidence limits for the OR for any adverse event comparing the lowest-volume quartile with the highest-volume quartile for biventricular devices included 1.0, even though the overall trend between volume and outcome was significant (p trend = 0.02). There are multiple possible mechanisms that could explain a weaker volume–outcome relationship for biventricular ICDs. The patients undergoing biventricular ICD implantation were markedly older (mean age 70 years vs. 63 years for single-chamber) and more ill (86% with New York Heart Association functional class III/IV vs. 32% for single-chamber), and the increased risk of adverse events due to poorer overall health status might have overwhelmed any advantage conferred by higher implanting volume. In addition, it is possible that only more-skilled operators attempted the more-complicated biventricular ICD implantations. This physician self-selection would minimize the variation in outcomes associated with the less-complicated device implantations. Future investigation is needed to further evaluate these hypotheses.
First, we only included procedures performed between 2006 and 2008. Consequently, we were unable to assess the overall long-term experience of centers with ICD implantation. Second, the hospital annual volume of ICD implantations might not have been uniform over the 3 years of the study, and our analysis did not account for this possible variation. Third, we only had data on in-hospital adverse events and were unable to assess the impact of hospital procedure volume on long-term adverse events. Fourth, we only included hospitals whose data met the data quality standards of the NCDR, which might have biased our results. Finally, we did not evaluate the effect of individual physician implantation volume on the rate of adverse events. Hospital volume and physician volume might have independent effects on adverse event rates as well as effect interaction, and these complicated relationships will need to be evaluated in future studies.
Our findings demonstrate that patients who have an ICD implanted at a high volume hospital are less likely to have an adverse event associated with the procedure than a patient who has an ICD implanted at a low-volume center. This volume–outcome relationship was statistically significant, clinically important, and consistent across all subtypes of ICDs, suggesting that implantation of these devices might be preferentially performed at high-volume centers.
Dr. Freeman is supported in part by the American Heart Association Pharmaceutical Roundtable/Spina Cardiovascular Outcomes Research Centers Program. Dr. Heidenreich is a paid member of the steering committee for the Altitude Registry of Boston Scientific and has also received research funding from Medtronic. Dr. Curtis receives significant salary support from the National Cardiovascular Data Registryand has stock holding in Medtronic. All other authors report that they have no relationships to disclose. Bruce D. Lindsay, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- confidence interval
- implantable cardioverter-defibrillator
- odds ratio
- percutaneous intervention
- Received April 10, 2010.
- Revision received June 11, 2010.
- Accepted July 6, 2010.
- American College of Cardiology Foundation
- Hannan E.L.,
- Wu C.,
- Ryan T.J.,
- et al.
- Vakili B.A.,
- Kaplan R.,
- Brown D.L.
- Nallamothu B.K.,
- Saint S.,
- Ramsey S.D.,
- Hofer T.P.,
- Vijan S.,
- Eagle K.A.
- Malenka D.J.,
- McGrath P.D.,
- Wennberg D.E.,
- et al.,
- Northern New England Cardiovascular Disease Study Group
- Parsonnet V.,
- Bernstein A.D.,
- Lindsay B.
- Parsonnet V.,
- Crawford C.C.,
- Bernstein A.D.
- Lee D.S.,
- Krahn A.D.,
- Healey J.S.,
- et al.
- Al-Khatib S.M.,
- Lucas F.L.,
- Jollis J.G.,
- Malenka D.J.,
- Wennberg D.E.
- ↵National Cardiovascular Data Registry: American College of Cardiology Foundation, 2010. http://www.ncdr.com. Accessed June 1, 2010.