Comparison of new implantation of cardiac implantable electronic device between tertiary and non-tertiary hospitals: a Korean nationwide study

This study compared the characteristics and mortality of new implantation of cardiac implantable electronic device (CIED) between tertiary and non-tertiary hospitals. From national health insurance claims data in Korea, 17,655 patients, who underwent first and new implantation of CIED between 2013 and 2017, were enrolled. Patients were categorized into the tertiary hospital group (n = 11,560) and non-tertiary hospital group (n = 6095). Clinical outcomes including in-hospital death and all-cause death were compared between the two groups using propensity-score matched analysis. Patients in non-tertiary hospitals were older and had more comorbidities than those in tertiary hospitals. The study population had a mean follow-up of 2.1 ± 1.2 years. In the propensity-score matched permanent pacemaker group (n = 5076 pairs), the incidence of in-hospital death (odds ratio [OR]: 0.76, 95% confidence interval [CI]: 0.43–1.32, p = 0.33) and all-cause death (hazard ratio [HR]: 0.92, 95% CI 0.81–1.05, p = 0.24) were not significantly different between tertiary and non-tertiary hospitals. These findings were consistently observed in the propensity-score matched implantable cardioverter-defibrillator group (n = 992 pairs, OR for in-hospital death: 1.76, 95% CI 0.51–6.02, p = 0.37; HR for all-cause death: 0.95, 95% CI 0.72–1.24, p = 0.70). In patients undergoing first and new implantation of CIED in Korea, mortality was not different between tertiary and non-tertiary hospitals.

Statistical analysis. All baseline patient characteristics and comorbid conditions were summarized as mean ± standard deviation or frequency (percentage) for continuous or categorical variables, respectively. We evaluated whether there are differences for in-hospital mortality and all-cause death rates between tertiary and non-tertiary hospital group. We conducted separate analyses of the PPM and ICD groups. We also conducted subgroup analyses for ICD group according to subtypes of ICD (single/dual chamber ICD and CRT-D). Baseline patient characteristics were compared between tertiary and non-tertiary hospital group using the two sample T-test or the Fisher's exact test. For the in-hospital mortality, we used the logistic regression model, while we used the Cox proportional hazards regression model for the all-cause mortality rate analysis. In all regression analyses, the reference was the non-tertiary hospital group. We employed the propensity-score matching analysis to reduce the impact of potential confounding effects on the mortality risk comparison. The propensity-scores were derived nonparametrically using the variables of age, gender, hypertension, diabetes, diabetes with chronic complications, dyslipidemia, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, moderate to severe liver disease, renal disease, cancer, rheumatic disease, atrial fibrillation, ventricular tachyarrhythmia, aborted cardiac arrest, Charlson comorbidity index, type of CIED, and indication of ICD. We used the nearest neighbor matching approach with a caliper size of 0.2 and evaluated the matching quality by measuring the covariate balancing in the matched set. We computed standardized differences in means between the two groups and considered the covariate balance achieved as long as the absolute standardized difference is less than 0.2. All standardized differences in the covariates were less than 0.05. To account for the matched pairs, we used the generalized estimating equations for the in-hospital mortality, as well as the Cox

Results
Study population and baseline characteristics. Between July 2013 and June 2017, a total of 17,655 patients undergoing first and new implantation of CIED were identified from the claims database of HIRA ( Fig. 1). The mean age of study participants was 68.5 ± 13.6 years and 8596 (48.7%) were male. Hypertension, diabetes, dyslipidemia, congestive heart failure, and atrial fibrillation were observed in 13,311 (75.4%), 6310 (35.7%), 10,268 (58.2%), 5232 (29.6%), and 4147 (23.5%) patients, respectively. The annual number of CIED procedures was had significantly increased during the study period (Table 1). According to the hospital categories, 11,560 (65.5%) patients were in the tertiary hospital group and 6095 (34.5%) patients were in the nontertiary hospital group. The number of hospitals providing CIED (including PPM, CRT-P, ICD, or CRT-D) in the tertiary hospital group and non-tertiary hospital group were 47 and 127, respectively. Tertiary hospital group performed more CIED procedures (61.9 ± 53.4 per year) than non-tertiary hospital group (11.9 ± 19.0 per year).
Tertiary hospital versus non-tertiary hospital in PPM group. According to the CIED procedures, the study participants who underwent PPM or CRT-P implantation were classified as PPM group (n = 13,891). Among them, patients were categorized into the tertiary hospital group (n = 8792) and non-tertiary hospital group (n = 5099). Patients in non-tertiary hospital group were older and had more comorbidities than those in tertiary hospital group (Table 1). Tertiary hospital group performed more implantation of dual chamber and epicardial system pacemaker compare to non-tertiary hospital group.
In-hospital mortality of tertiary and non-tertiary hospital group was 0.4% and 0.6%, respectively. There was no significant difference in the in-hospital mortality between tertiary and non-tertiary hospital group (unadjusted odds ratio [OR] of tertiary hospital: 0.68, 95% confidence interval [CI]: 0.41-1.12, p = 0.13). During a mean follow-up of 2.2 ± 1.2 years, non-tertiary hospital group was more likely to have all-cause death than tertiary hospital group (unadjusted hazard ratio [HR] of tertiary hospital: 0.85, 95% CI 0.76-0.96, p = 0.008). Figure 2a shows the unadjusted cumulative incidence rates for all-cause deaths of the two groups.
After propensity-score matching, there were 5076 matched pairs. In the matched cohort, there were no other significant differences between tertiary and non-tertiary hospital group for any of the covariates ( Table 2). There was no significant difference in terms of the incidence of in-hospital mortality between tertiary and non-tertiary  Tertiary hospital versus non-tertiary hospital in ICD group. We also analyzed the patients with ICD group including single/dual chamber ICD or CRT-D (n = 3764). Among them, patients were categorized into the tertiary hospital group (n = 2768) and non-tertiary hospital group (n = 996). Patients in non-tertiary hospital group were older and had more comorbidities than those in tertiary hospital group (Table 1). Tertiary hospital group performed more implantation of dual chamber ICD compare to non-tertiary hospital group. In addition, tertiary hospital group performed more implantation of ICD for secondary prevention compare to non-tertiary hospital group.
In-hospital mortality of tertiary and non-tertiary hospital group was 0.6% and 0.4%, respectively. There was no significant difference in the in-hospital mortality between tertiary and non-tertiary hospital group (unadjusted OR of tertiary hospital: 1.44, 95% CI 0.48-4.32, p = 0.51). During a mean follow-up of 2.1 ± 1.2 years, there was no significant difference in all-cause death between the two groups (unadjusted HR of tertiary hospital: 0.85, 95% CI 0.67-1.06, p = 0.15). Figure 2b shows the unadjusted cumulative incidence rates for all-cause deaths of the two groups.
After propensity-score matching, there were 992 matched pairs. In the matched cohort, no significant differences were observed in terms of covariates between the two groups ( Table 2). There was no significant difference in terms of the incidence of in-hospital mortality between the two groups (OR of tertiary hospital: 1.76, 95% CI 0.51-6.02, p = 0.37). In addition, the incidence of all-cause death (HR of tertiary hospital: 0.95, 95% CI 0.72-1.24, p = 0.70) was not different between tertiary and non-tertiary hospital group during the follow-up period (mean, 2.0 ± 1.2 years) ( Table 3).
We also conducted subgroup analyses for ICD group according to subtypes of ICD (single/dual chamber ICD and CRT-D). Baseline characteristics of both ICD subtypes were shown in Supplementary Table S1 online. After propensity-score matching, there were no significant differences between tertiary and non-tertiary hospital group for any of the covariates in both ICD subtypes (Supplementary Table S2 online). In both subtypes of ICD, there were no significant differences in the incidence of in-hospital mortality and all-cause death between tertiary and non-tertiary hospital group (Supplementary Table S3 online).

Discussion
This study aimed to compare the characteristics and mortality of first and new implantation of CIED between tertiary and non-tertiary hospitals. The major findings of the present study using NHI claims data in South Korea are as follows: (1) patients in non-tertiary hospital group were older and had more comorbidities than those in tertiary hospital group; (2) after propensity-score matching, the incidences of in-hospital death and all-cause death were not significantly different between tertiary and non-tertiary hospitals in PPM and ICD group.
In this study, patients in non-tertiary hospitals were older and had more comorbidities compare to those in tertiary hospitals. This finding could be explained by the location of the tertiary (mostly urban region) and non-tertiary (mostly suburban and rural region) hospitals. In this study, 70.2% of tertiary hospitals were located in urban region. In Korea, the mean age of the population in suburban and rural region was higher than that in urban region and comorbidities (hypertension, diabetes, and congestive heart failure) were more prevalent in suburban and rural regions 15 . This difference between urban and rural region were also reported in previous studies [16][17][18] .   Figure 2. Unadjusted cumulative incidence rates for all-cause deaths in the study population. Cumulative incidence curves are shown for all-cause death in patients with PPM group (a) and in those with ICD group (b). The numbers in each figure represent the cumulative incidence rates at each time point. All p values were calculated with the use of the log-rank test. ICD, implantable cardioverter-defibrillator; PPM, permanent pacemaker.  www.nature.com/scientificreports/ Tertiary hospital is a medical facility which provides a high degree of subspecialty expertise for patients 19 and is generally larger and provide more procedure compare to non-tertiary hospital. Consequently, tertiary hospital performed more CIED procedure than non-tertiary hospital (62 versus 12 per year) in this study. Several previous studies demonstrated that high procedure volume hospital is less likely to have an adverse outcome after CIED implantation. Nowak et al. reported that hospital annual PPM volume was inversely related to surgical complications and atrial or ventricular dislocation 20 . Similar trends were also observed in ICD 21 . Unlike previous studies, however, we did not observe a better prognosis among patients in the tertiary hospital, high procedure volume hospital, compare to non-tertiary hospital. This might be explained by the difference in the definition of the study outcomes compared to previous studies. Most previous studies defined the study outcomes to include not only death but also traumatic complications (pneumothorax, hemothorax, pericardial effusion, and pericardial tamponade), lead-related complications, and device-related infection 20,21 . When the definition of study outcome was confined to mortality, however, procedure volume did not have a predictive value. Freeman et al. reported that adverse events (including cardiac arrest, cardiac perforation, pneumothorax, hemothorax, lead dislodgement, and device-related infection) were significantly higher in the lowest-procedure volume quartile compared to the highest-procedure volume quartile (OR 1.26, 95% CI 1.05-1.52, p < 0.0001) 21 . However, in-hospital death was not different according to procedure volume. Another study also did not show a consistent procedure volumemortality relationship 22 . In addition to procedure volume, tertiary hospitals are teaching hospital and have more medical personnel, facility, and equipment compared to non-tertiary hospitals because designation criteria for tertiary hospitals are consist of these hospital characteristics 11 . However, we did not observe a better prognosis among patients in the tertiary hospital compare to non-tertiary hospital. This is consistent with previous study 22 .
Several previous studies have shown that operator characteristics are associated with clinical outcomes. Al-Khatib et al. showed that although there was no correlation between operator volume of ICD implantation and 90-day mortality, 90-day rates of mechanical complication (OR 1.47, 95% CI 1.09-1.99) and ICD infection (OR 2.47, 95% CI 1.18-5.17) were significantly higher among operator in the lowest volume quartile compared with those in the highest volume quartile 23 . Curtis et al. reported that ICD implantations by a non-electrophysiologist were associated with a higher risk of in-hospital procedural complications compared to ICD implantation by an electrophysiologist 24 . Our study did not capture the risk associated with the operator characteristics. Therefore, further clinical studies are needed to evaluate it.
Concentration of patients in tertiary hospitals causes not only the economic problem in terms of waste of health care resource and the loss of opportunity cost for other hospitals but also the delay of timely treatment. Previous studies showed that patients tend to make choices based on hospital size, facility, modernity, and professional credibility 7,8 . The phenomenon of the concentration in tertiary hospitals might derive due to the lack of information (such as clinical outcomes). The results of the present study could be helpful for patients' decision to choose the hospitals and might alleviate the concentration of patients to tertiary hospitals in CIED procedures.
This study had several limitations. First, this study was a retrospective, observational study. Although we rigorously adjusted for baseline covariates using propensity-score matching, there are inherent limitations of a non-randomized study. Second, our study was based on administrative data from the HIRA in South Korea. In patients with CIED, procedural and device-related complications, such as traumatic-, lead-related complications, lead dislodgement, and device-related infection, are important clinical outcomes in addition to mortality. Similar to previous study using administrative databases 22 , this study could not identify these specific types of complications. In addition, we did not specify the cause of death. However, mortality is the most powerful hard endpoint and unbiased endpoint. Furthermore, to the best of our knowledge, this study is the largest study for mortality in Asian patients undergoing first and new implantation of CIED. Finally, this study only included the Korean population, and this may limit the applicability of our findings to other countries. However, considering the paucity of data concerning Asian populations, we believe that this study may have clinical implications.
In conclusion, this nationwide study suggested that mortality was not different between tertiary and non-tertiary hospitals in Korean patients undergoing first and new implantation of CIED. Further prospective, national cohort studies are needed to confirm these findings.

Data availability
The present study analyzed the NHI claims data in South Korea. Data of the NHI claims are accessible to researchers after permission of the HIRA in South Korea. Qualified, interested researchers may request access to these data from the HIRA (http://opend ata.hira.or.kr/home.do). The authors do not have any special access privileges to these data.