Outcomes and Temporal Trends of Inpatient Percutaneous Coronary Intervention at Centers With and Without On-site Cardiac Surgery in the United States

Key Points

Question  Is percutaneous coronary intervention safe at centers without on-site cardiac surgery?

Findings  In this cross-sectional study of 6.9 million percutaneous coronary interventions performed in the United States from 2003 to 2012, there was no significant difference in the adjusted in-hospital mortality between centers with and without on-site cardiac surgery.

Meaning  Percutaneous coronary interventions at centers without on-site cardiac surgery may be safe for all indications.

Importance  There are concerns whether percutaneous coronary intervention (PCI) at centers without on-site cardiac surgery is safe outside of a tightly regulated research environment.

Objective  To analyze the outcomes and temporal trends of inpatient PCI at centers without on-site cardiac surgery in an unselected and nationally representative population of the United States.

Design, Setting, and Participants  A national inpatient sample (N = 6 912 232) was used to identify patients who underwent inpatient PCI in the United States from January 1, 2003, to December 31, 2012. Hospitals that performed 1 or more coronary artery bypass graft surgeries in a given calendar year were classified as centers with on-site cardiac surgery, and weighted sampling of all inpatient hospitalizations was performed. Data analysis was performed from February to May 2016.

Exposures  Inpatient PCI.

Main Outcomes and Measures  In-hospital mortality.

Results  Of the 6 912 232 inpatient PCIs performed, 2 336 334 patients (33.8%) were women and 4 575 898 (66.2%) were men; their mean (SD) age was 64.5 (12.3) years. Of these PCIs, 396 741 (5.7%) were conducted at centers without on-site cardiac surgery. The rate of in-hospital mortality was significantly lower at centers with on-site cardiac surgery compared with centers without on-site cardiac surgery (1.4% vs 1.9%; unadjusted odds ratio [OR], 0.74; 95% CI, 0.72-0.75). After adjustment, there was no significant difference in in-hospital mortality between centers with and without on-site cardiac surgery (OR, 1.01; 95% CI, 0.98-1.03; P = .62) for acute coronary syndromes and elective procedures requiring inpatient hospitalization. In addition, there were no significant differences in the risk-adjusted, in-hospital mortality between the 2 groups in prespecified subgroups after adjusting for multiple comparisons, including ST-elevation myocardial infarction (OR, 0.99; 95% CI, 0.96-1.03; P = .65), non–ST-elevation acute coronary syndrome (OR, 0.99; 95% CI, 0.93-1.05; P = .66), and elective PCI (OR, 0.93; 95% CI, 0.84-1.03; P = .17). There was a significant increase in the proportion of PCIs at centers without on-site cardiac surgery within the study period (from 1.8% to 12.7%; P < .001 for trend by Cochrane-Armitage test) reflected across all the indications.

Conclusions and Relevance  There was a 7-fold increase in the proportion of PCIs at centers without on-site cardiac surgery from 2003 to 2012 in the United States, with the adjusted in-hospital mortality after inpatient PCI being similar at centers with and without on-site cardiac surgery. These data provide evidence that PCI at centers without on-site cardiac surgery may be safe in the modern era.

Percutaneous coronary intervention (PCI) centers without on-site cardiac surgery (CS) backup were established to reduce the door-to-balloon time and increase the availability of primary PCI. Small single-center studies,¹ registry-based studies,^2-4 and meta-analyses^5,6 have reported no significant differences in the rate of in-hospital or long-term mortality after primary PCI at centers with and without on-site CS. Randomized clinical trials concluded that nonprimary PCI at centers without on-site CS was noninferior to PCI at centers with on-site CS.^7,8 However, these trials represent highly selected patient populations and excluded a large number of high-risk patients.

A previous study reported an increase in the number of PCIs performed at centers without on-site CS from 2001 to 2004 in the United States.⁹ More recently, data from the United Kingdom showed an increase in the number of PCIs at centers without on-site CS from 2006 to 2012.³ A similar increase would be expected in the United States as well^10,11; however, to our knowledge, no study has evaluated these trends in the contemporary era.

The objective of this study was to analyze the safety and dissemination of PCI at centers without on-site CS outside of a tightly regulated research environment in the United States. We utilized the National Inpatient Sample (NIS)¹² database to assess the outcomes and temporal trends of PCI at centers with and without on-site CS in the United States from January 1, 2003, to December 31, 2012.

The NIS is the largest publicly available all-payer database of hospitalized patients in the United States and is sponsored by the Agency for Healthcare Research and Quality as a part of the Healthcare Cost and Utilization Project.¹² The NIS includes deidentified data on primary and secondary discharge diagnoses and procedures from more than 7 million hospitalizations annually. The NIS database covers all nonfederal US hospitals (excluding rehabilitation and long-term acute care hospitals) in the participating states (n = 44) and represents more than 95% of the US general population. The NIS does not include robust data on interhospital transfers; thus, the number of patients who were transferred from centers without on-site CS could not be determined. Data quality assessments are performed annually to maintain internal validity of the database. Furthermore, estimates from the NIS database have been externally validated against data from the American Hospital Association Annual Survey Database, the National Hospital Discharge Survey, and the Medicare Provider Analysis and Review inpatient data from the Centers for Medicare & Medicaid Services.¹³ Further sampling details are presented in the eMethods in the Supplement. This study was deemed exempt by the institutional review board guidelines of Mayo Clinic and New York Medical College because the NIS is a publicly available database that contains deidentified patient information.

From 2003 to 2012, a total of 79 185 729 records were included in the NIS, corresponding to a national estimate of 388 442 328 hospital discharges in the United States. We used the International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) procedure codes 00.66, 36.01, 36.02, 36.05, 36.06, 36.07, and 17.55 to identify all patients 18 years or older who underwent PCI (n = 6 914 256). The use of the NIS database for PCI numbers in the United States has been validated by Epstein et al,¹⁴ who reported a mean difference of only 0.2% in quarterly PCI counts between Medicare claims and the NIS. Patients with missing data on in-hospital death, length of stay, or both (n = 2024) were excluded, giving us our final study cohort of 6 912 232 (eFigure 1 in the Supplement). Hospitals that performed 1 or more coronary artery bypass graft (CABG) surgeries (ICD-9-CM code, 36.1 × for CABG) in a given calendar year were classified as centers with on-site CS. All other hospitals that did not perform any CABG in a calendar year were classified as centers without on-site CS.

The following patient characteristics were extracted from the NIS database: demographics (age, sex, race, primary expected payer, admission day, and median annual household income), clinical risk factors associated with increased mortality in patients undergoing PCI (smoking, dyslipidemia, family history of coronary artery disease, prior myocardial infarction [MI], prior PCI, prior CABG, atrial fibrillation, and carotid artery disease),^15,16 and 29 Elixhauser comorbidities. The Elixhauser comorbidity index is a validated measure of comorbidities in large administrative databases as defined by the Agency for Healthcare Research and Quality.^17,18 Patients undergoing PCI for acute coronary syndromes (ACSs) were identified using respective ICD-9-CM codes for ST-elevation myocardial infarction (STEMI) and non–ST-elevation ACS (NSTE-ACS). Those without diagnosis codes for ACS were presumed to have undergone elective PCI for stable ischemic heart disease. The ICD-9-CM codes and Clinical Classification Software codes used to identify patient and procedural characteristics are provided in eTable 1 in the Supplement.

The primary outcome was all-cause, in-hospital mortality defined as died in the NIS database. Secondary outcomes were hemorrhage requiring blood transfusion, transient ischemic attack or stroke, vascular complications, and mean length of stay. The ICD-9-CM codes used to identify these conditions are provided in eTable 1 in the Supplement.

Weighted estimates were obtained by applying trend weights to the unweighted discharge data and were used for all statistical analyses. Between-group differences were analyzed using Pearson χ² test for categorical variables and t test for continuous variables. To examine differences in in-hospital mortality and in-hospital complications between patients undergoing PCI at centers with and without on-site CS, multivariate logistic regression models were constructed with the use of generalized estimating equations to account for clustering of outcomes within hospitals. Variables included in the regression models were demographics, hospital characteristics, 29 Elixhauser comorbidities, clinical risk factors (smoking, dyslipidemia, family history of coronary artery disease, prior MI, prior PCI, prior CABG, atrial fibrillation, and carotid artery disease), and indication for PCI. Given the positively skewed distribution of length of stay, we used log transformation of length of stay as the dependent variable. All patients were included in primary regression models except those with missing data on primary expected payer (0.1%), socioeconomic status (2.3%), hospital characteristics (0.4%), and Elixhauser comorbidities (0.7%). A sensitivity analysis including the entire study population was conducted for both the primary and secondary outcomes by replacing the missing values for these variables with the dominant category. Race was also included in the sensitivity analysis, and the missing values for race were treated as a separate category in the regression models.^19,20 A priori subgroup analysis was performed for in-hospital mortality by age, sex, indication of PCI, clinical risk factors, and procedural and hospital characteristics. In the generalized estimating equations models for in-hospital mortality in STEMI and NSTE-ACS subgroups, we additionally adjusted for cardiogenic shock.

Temporal trend analyses were performed using the Cochrane-Armitage test. Statistical analysis was conducted using SPSS Statistics, version 21.0 (IBM Corp). All P values are 2-sided with a significance threshold <.05. For subgroup analyses, the P value threshold for significance was <.002 after Bonferroni adjustment for multiple subgroup comparisons. Data analysis was performed from February to May 2016.

A total of 6 912 232 weighted records of inpatient PCIs were detailed in the NIS database from 2003 to 2012. Of these, 396 741 procedures (5.7%) were performed at hospitals without on-site CS, and 6 515 491 procedures (94.3%) were performed at hospitals with on-site CS. Table 1 and eTable 2 in the Supplement summarize the differences in demographics, risk factors, comorbidities, and procedural and hospital characteristics between centers with and without on-site CS. A total of 2 336 334 patients were women (33.8%) and 4 575 898 were men (66.2%); their mean (SD) age was 64.5 (12.3) years.

Centers without on-site CS performed a significantly higher proportion of PCIs for STEMI (34.4% vs 20.0%) and a significantly lower proportion of elective PCIs (24.0% vs 37.7%) compared with centers with on-site CS (both P < .001). Use of drug-eluting stents, percutaneous ventricular assist devices, and multivessel PCI was significantly higher at centers with on-site CS centers compared with centers without on-site CS (Table 1). A total of 80.7% of the PCIs at on-site CS centers were performed at hospitals doing 400 or more annual PCIs compared with only 9.0% of the PCIs at centers without on-site CS (P < .001) (Table 1).

In the overall study cohort, unadjusted in-hospital mortality was lower at PCI centers with on-site CS (1.4% vs 1.9%; odds ratio [OR], 0.74; 95% CI, 0.72-0.75). Figure 1B-D presents the unadjusted in-hospital mortality at centers with and without on-site CS by indication of inpatient PCI. However, after multivariate adjustment, there was no significant difference in the rate of in-hospital mortality between centers with and without on-site CS (OR, 1.01; 95% CI, 0.98-1.03) for ACSs and elective PCIs requiring inpatient hospitalization. The incidence of in-hospital transient ischemic attack or stroke was the same in both centers (0.8%; adjusted OR, 1.03; 95% CI, 0.99-1.07). Although the incidence of vascular injury was higher at centers with on-site CS (1.1% vs 0.9%; adjusted OR, 1.31; 95% CI, 1.26-1.35), there was no difference in the incidence of hemorrhage requiring blood transfusion (0.7% vs 0.8%; adjusted OR, 1.02; 95% CI, 0.98-1.06) (Table 2). Sensitivity analysis was performed accounting for missing values and showed largely similar results (eTable 3 in the Supplement). More patients were discharged home at centers with on-site CS (91.2% vs 87.1%), whereas a significantly higher proportion of patients were discharged to either a short-term hospital (0.5% vs 3.6%) or a skilled care facility at centers without on-site CS (3.6% vs 4.4%), respectively (P < .001) (eFigure 2 in the Supplement).

There was an inverse association between procedural volume and in-hospital mortality both at centers with and those without on-site CS (eTable 4 in the Supplement). Among the centers without on-site CS, patients undergoing PCIs at hospitals with annual volumes of 201 to 400 PCIs annually (adjusted OR, 0.87; 95% CI, 0.82-0.92) or more than 400 PCIs annually (adjusted OR, 0.84; 95% CI, 0.75-0.93) had a significantly lower in-hospital mortality rate compared with hospitals performing 200 or fewer PCIs annually. A similar association between annual PCI volume and in-hospital mortality results was noted at centers with on-site CS (eTable 4 in the Supplement).

There was no significant difference in adjusted in-hospital mortality between centers with and without on-site CS for different indications of PCI, including STEMI, NSTE-ACS, or stable ischemic heart disease. After adjusting for multiple comparisons (P < .002 threshold), in-hospital mortality was similar at centers with and without on-site CS in various subgroups detailed in Table 3.

Overall, the total number of PCIs being performed annually declined from 776 388 in 2003 to 535 890 in 2012 (P < .001 for trend). Of the total number of PCIs being performed every year, the proportion of PCIs at centers without on-site CS increased significantly from 1.8% in 2003 to 12.7% in 2012 (P < .001 for trend). This increase in the proportion of PCIs at centers without on-site CS was reflected in all of the 3 clinical subgroups (P < .001 for trend for all) (Figure 2).

There are 4 main findings of this study. First, there was a 7-fold increase in the proportion of PCIs being performed at centers without on-site CS from 2003 to 2012 in the United States. Second, there was no significant difference in the rate of adjusted in-hospital mortality after inpatient PCI between centers with and without on-site CS in the United States. Third, this finding was consistent among patients with ACS, those undergoing elective inpatient PCI, and other prespecified subgroups. Fourth, procedural volume was inversely associated with in-hospital mortality, both at centers with and those without on-site CS.

Our study has several strengths. To our knowledge, this is the largest study to date evaluating the in-hospital outcomes of PCI at centers with and without on-site CS including the analysis of 6 912 232 PCIs. Previous evaluations have been limited by small sample sizes,¹ registry-based studies in countries other than the United States,³ or US-based registry studies that evaluated outcomes before 2006 that may not be applicable to the contemporary era.^4,21 Two randomized trials have been performed in nonprimary PCI patient populations; however, these were highly selected patients and hospitals.^7,8 The present study provides data from a nationally representative population of patients undergoing PCI in the United States.

The rate of in-hospital mortality in the present study was similar to that of other studies in patients undergoing PCI, thus validating the overall results.^22,23 Unadjusted in-hospital mortality was significantly lower at centers with on-site CS compared with centers without on-site CS, with an absolute risk reduction of 0.5%. After adjusting for patient-level, procedural, and hospital characteristics, there was no significant difference in the rate of in-hospital mortality between centers with and those without on-site CS after PCI for ACS or elective PCI requiring inpatient hospitalization. Similar in-hospital mortality after risk adjustment was most likely due to a small absolute risk difference in the unadjusted analysis and a significantly higher percentage of patients with STEMI at centers without on-site CS who had much higher in-hospital mortality rates (4.6%) compared with those undergoing PCI for NSTE-ACS (0.9%) or stable ischemic heart disease (0.4%). Our findings are similar to those of previous studies,^1-8,21 including a recent meta-analysis of 23 studies reporting no difference in the rate of in-hospital mortality for primary and nonprimary PCI between centers with and without on-site CS.⁵ Two randomized trials have evaluated the safety of nonprimary PCI at centers without on-site CS.^7,8 Approximately one-third of the patients in these trials presented without ACS; however, no subgroup analysis was provided. These trials had stringent criteria for inclusion of centers without on-site CS (eg, hospitals should have been capable of performing >200 PCIs per year and perform >300 diagnostic cardiac catheterizations per year).⁸ In our study, 60% of PCIs at centers without on-site CS were performed at hospitals with an annual PCI volume of less than 200 and would have been excluded in these trials. The trials also excluded high-risk patients, such as those with ejection fractions of less than 20%, high-risk coronary anatomy, left main disease, or saphenous vein graft interventions. Of the patients who initially provided consent in the Cardiovascular Patient Outcomes Research Team trial, 75% were excluded and the final study population was a low-risk cohort.⁸ Although the NIS does not have data pertaining to hemodynamics or coronary angiography, it is an unselected population.

The incidence of vascular complications was higher at centers with on-site CS compared with centers without on-site CS, as was noted in a previous study using the National Cardiovascular Database Registry database.⁴ Patients undergoing PCI at centers with CS were older, had more vascular risk factors, and had a higher prevalence of peripheral vascular disease, which could potentially explain this finding. However, the rate of hemorrhage requiring transfusion was similar at both centers.

A previous study from the US-based National Cardiovascular Database Registry showed that there was a significant increase in the number of elective and nonelective PCIs at centers without on-site CS from 2001 to 2004.⁹ Coronary revascularization has changed significantly in recent years with a decline in the total number of PCIs,²⁴ availability of second-generation drug-eluting stents, and worsening risk profile of patients.^25,26 However, to our knowledge, no contemporary data are available regarding the performance of PCI at centers without on-site CS in the United States. We found that there was a significant increase in the proportion of PCIs being performed at centers without on-site CS: from 1.8% in 2003 to 12.7% in 2012. Centers performing PCI without on-site CS were initially developed for the STEMI population to reduce the door-to-balloon time. With experience, the procedural indications have been expanded to include NSTE-ACS and elective PCI. Our data show that the highest surge in the proportion of PCIs at centers without on-site CS over the 10-year period was for elective PCIs followed by NSTE-ACS and STEMI. Although the proportion of PCI for STEMI was significantly higher than centers with on-site CS, NSTE-ACS still constituted the highest proportion of all PCIs performed at centers without on-site CS.

This large study shows that it is safe to perform PCI at hospitals without on-site CS for STEMI and NSTE-ACS. These results have important clinical and policy implications because they are applicable to the general US population requiring acute interventional care. Similar in-hospital mortality in patients with stable ischemic heart disease undergoing elective inpatient PCI should be interpreted with caution because the criteria for inpatient hospitalization at centers with and without on-site CS may be different and are biased by physician opinion, local hospital policies, and patient preferences. We also found that higher procedural volume was associated with lower in-hospital mortality at centers without on-site CS. Therefore, it is imperative to select the appropriate centers that can sustain high volumes for better outcomes. The present study also highlights the importance of using a large administrative database, such as the NIS, to answer relevant clinical questions across different practice sites and clinical presentations in a timely manner.

This study has important limitations that are related to the observational design and unavailability of detailed clinical and angiographic information. Data on PCIs that were performed on an outpatient basis were not available because the NIS records only inpatient hospitalizations. The Healthcare Cost and Utilization Project statistical briefs from 28 states (not all states collect these data) reported that the proportion of outpatient percutaneous transluminal coronary angioplasty procedures is low, increasing from 3.2% in 2003 to 11.9% in 2012.²⁷ There is a possibility that the in-hospital mortality rate at the centers without on-site CS might have been underestimated since sicker patients may have been transferred out and recorded as alive during that hospitalization in the NIS database. Based on prior studies, the number of transfers may be less than 0.1% and is unlikely to influence our findings.³ Another limitation of the present analysis is the lack of data on door-to-balloon time in patients with STEMI, procedural success, recurrent ischemia, target vessel revascularization, or long-term follow-up since the NIS includes only in-hospital outcomes.

In this nationally representative US population, there was no significant difference in the adjusted in-hospital mortality after PCI for ACSs and elective procedures requiring inpatient hospitalization between centers with and without on-site CS. Higher procedural volume was associated with lower in-hospital mortality at both types of centers. From 2003 to 2012, there was a 7-fold increase in the proportion of PCIs being performed at centers without on-site CS spread across all indications of PCI. These data provide support for further thoughtful and careful expansion of centers without on-site CS.

Corresponding Author: Charanjit S. Rihal, MD, Department of Cardiovascular Diseases, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (rihal@mayo.edu).

Accepted for Publication: September 14, 2016.

Published Online: November 23, 2016. doi:10.1001/jamacardio.2016.4188

Author Contributions: Drs Goel and Gupta contributed equally to this study, had full access to all the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Goel, Gupta, Fonarow, Rihal.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Goel, Gupta, Rihal.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Goel, Gupta, Kolte.

Administrative, technical, or material support: Goel, Gupta, Rihal.

Study supervision: Goel, Rihal.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Bhatt is a member of the advisory boards of Cardax, Elsevier Practice Update Cardiology, Medscape Cardiology, and Regado Biosciences; a member of the board of directors of Boston Veterans Affairs Research Institute, Society of Cardiovascular Patient Care; chair of the American Heart Association Quality Oversight Committee; a member of data monitoring committees for the Duke Clinical Research Institute, Harvard Clinical Research Institute, Mayo Clinic, and Population Health Research Institute; has received honoraria from the American College of Cardiology as senior associate editor of Clinical Trials and News, Belvoir Publications as editor in chief of Harvard Heart Letter, Duke Clinical Research Institute as a member of clinical trial steering committees, Harvard Clinical Research Institute as a member of the clinical trial steering committee, HMP Communications as editor in chief of Journal of Invasive Cardiology, Journal of the American College of Cardiology as a guest editor and associate editor, Population Health Research Institute as a member of the clinical trial steering committee, Slack Publications as chief medical editor of Cardiology Today’s Intervention, the Society of Cardiovascular Patient Care as secretary/treasurer, and WebMD as a member of the continuing medical education steering committees. He also reported being deputy editor of Clinical Cardiology and chair of the NCDR-ACTION (National Cardiovascular Database Registry–Acute Coronary Treatment and Intervention Outcomes Network) Registry Steering Committee and the VA CART (Veterans Affairs Cardiovascular Assessment, Reporting, and Tracking) Research and Publications Committee; receiving research funding from Amarin, Amgen, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Forest Laboratories, Ischemix, Medtronic, Pfizer, Roche, Sanofi, and The Medicines Company; receiving royalties from Elsevier; serving as a site co-investigator for Biotronik, Boston Scientific, and St Jude Medical; serving as a trustee of the American College of Cardiology; and performing unfunded research for FlowCo, PLx Pharma, and Takeda. No other disclosures were reported.

Disclaimer: Dr Fonarow is the associate editor for Health Care Quality and Guidelines, JAMA Cardiology but was not involved in the editorial review or the decision to accept the manuscript for publication.