Trends in Characteristics and Outcomes of Hospital Inpatients Undergoing Coronary Revascularization in the United States, 2003-2016

1Coronary artery revascularization has affected millions of patients with coronary artery disease (CAD) worldwide. Both surgical and percutaneous revascularization strategies have evolved from experimental stages to routine procedures that can safely tackle complex coronary anatomic features and high-risk patients.,, Several studies,,, have documented a significant decrease in coronary artery bypass grafting (CABG) operations after the emergence of percutaneous coronary interventions (PCI) in the 1990s. However, the annual volumes of both PCI and CABG decreased significantly in more recent years possibly because of advances in medical therapy, the emergence of data questioning the benefit of PCI in stable CAD, and the increasing implementation of appropriate use criteria.,,, Whether these temporal changes in procedural volume were associated with changes in the risk profiles of patients referred for percutaneous or surgical coronary revascularization and the outcomes of these procedures remain unknown. This study used a nationwide, representative sample hospital inpatients in the United States to assess the temporal changes in baseline characteristics of patients undergoing PCI or CABG and crude and risk-adjusted in-hospital mortality after PCI or CABG stratified by clinical indication.

2We conducted a retrospective cohort study using the Nationwide Inpatient Sample (NIS) database to derive patient-relevant information from January 1, 2003, to December 31, 2016. Data analysis was performed from July 15 to October 4, 2019. The West Virginia University Institutional Review Board exempted the study from board approval and waived the requirement for informed consent because the NIS is a publicly available deidentified database. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

3The NIS is part of the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality, and is the largest publicly available all-payer claims-based database in the United States. The database contains hospital inpatient stays derived from billing data submitted by hospitals to statewide data organizations across the United States. These data include clinical and resource use information typically available from discharge abstracts. Researchers and policy makers use the NIS to make national estimates of health care utilization, access, charges, quality, and outcomes. The NIS sampling frame includes data from 47 statewide data organizations, covering more than 97% of the US population. The annual sample encompasses approximately 8 million discharges, which represent 20% of inpatient hospitalizations across different hospital types and geographic regions. The national estimates of the entire US hospitalized population are calculated using a standardized sampling and weighting method provided by the HCUP. The NIS has been used extensively to assess national trends in the utilization, disparities, and outcomes of coronary artery interventions.,,,,,,

4Patients aged 18 years or older who underwent PCI or CABG between 2003 and 2016 were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes (eTable 1 in the Supplement). We further classified PCIs into those performed for ST-segment elevation myocardial infarction (STEMI), non–ST-segment elevation myocardial infarction (NSTEMI), or unstable angina or stable ischemic heart disease (UA-SIHD). Given the rarity of CABG performed in the context of STEMI, we classified CABG operations into 2 groups: CABG in the context of acute myocardial infarction (AMI) and CABG performed for UA-SIHD.

5Our study investigated trends in clinical risk profile among hospital inpatients undergoing PCI and CABG divided into 3 eras (2003-2007, 2008-2012, and 2013-2016). These eras were selected to provide relatively equal periods and to illustrate the global and not the year-to-year change in baseline characteristics among patients undergoing coronary revascularization. The study also investigated trends in the crude and adjusted in-hospital mortality associated with PCI and CABG, stratified by clinical indication.

6Weighted data were used for all statistical analyses. Descriptive statistics are presented as numbers with percentages for categorical variables. Means (SDs) are used to report continuous measures. To evaluate changes in baseline characteristics by calendar year, we used the Mantel-Haenszel test of trend for categorical variables and linear regression for continuous variables. To assess whether in-hospital mortality improved over time, multivariable logistic regression models were constructed to estimate the odds ratios and 95% CIs. To directly estimate rate ratios, a modified Poisson regression approach was used that included a robust variance estimate in the models. Calendar year was included as a categorical variable, with 2003 as the reference year. All the multivariable regression models used in risk-adjusted estimates were fitted with generalized estimating equations to account for clustering of outcomes within hospitals. Adjusted risk ratios and P values for trend were determined with a model evaluating calendar year as a continuous variable. Variables included in the regression models included demographic characteristics (age, sex, and race/ethnicity), socioeconomic factors (primary expected payer and median household income), Elixhauser comorbidity index score, and clinically relevant comorbidities (eTable 2 in the Supplement). The trend weight files were merged onto the original NIS files by year and hospital identification number. For years before 2012, the trend weight was used to create national estimates for trend analysis. For 2012 and after, no trend weight was needed, and the regular discharge weight was used, consistent with the redesigned NIS trend analysis.

7Statistical analysis was performed accounting for data changes in trend analysis and avoiding use of nonspecific secondary diagnosis codes to infer in-hospital events. Methodologic standards in research using the NIS were met as recommended. A 2-sided P <.05 was considered to be statistically significant. All statistical analyses were performed with SPSS software, version 24 (IBM Corp).

8A total of 12 062 081 revascularization hospitalizations were identified: 8 687 338 PCIs (72.0%; mean [SD] patient age, 66.0 [10.8] years; 66.2% male) and 3 374 743 CABGs (28.0%; mean [SD] patient age, 64.5 [12.4] years; 72.1% male). The annual PCI volume decreased from 777 780 in 2003 to 440 505 in 2016 (eTable 3 in the Supplement). This volume corresponded to a decrease in the PCI rate from 366 to 180 per 100 000 US adults between 2003 and 2016 (Figure 1). Similarly, the annual CABG volume decreased from 337 444 in 2003 to 201 840 in 2016, corresponding to a decrease in the CABG rate from 159 to 82 per 100 000 US adults between 2003 and 2016 (Figure 1 and eTable 3 in the Supplement). Significant temporal changes occurred in the demographic characteristics, socioeconomic status, prevalence of risk factors, and clinical presentations of patients undergoing PCI and CABG and in the characteristics of the procedures.

9In the PCI group, a temporal increase occurred in the proportions of older and male patients, nonwhite patients, and patients with lower socioeconomic status. There was also a significant increase in the prevalence of atherosclerotic and nonatherosclerotic risk factors (Table 1). The proportion of patients with an Elixhauser comorbidity index score of 3 or greater increased from 24.7% in 2003 to 2007 to 52.3% in 2012 to 2016. The proportion of women among all patients undergoing PCI decreased from 34.0% in 2003 to 2006 to 32.8% in 2012 to 2016 (P <.001). The proportion of women among all patients undergoing CABG decreased from 29.0% in 2003 to 2006 to 26.0% in 2012 to 2016 (P <.001). The percentage of PCI for AMI among all PCIs increased from 22.8% in 2003 to 2007 to 53.1% in 2012 to 2016. The characteristics of PCIs changed as well. Patients who underwent PCI between 2012 and 2016 (vs those who underwent PCI between 2003 and 2007) had fewer multivessel PCIs (16.2% vs 17.9%) and used bare metal stents less (15.5% vs 27.4%) but had more PCIs for chronic total occlusion (3.4% vs 0.1%) and cardiogenic shock (5.0% vs 1.8%) and had greater use of intravascular ultrasonography and/or fractional flow reserve (9.2% vs 2.5%) and circulatory support devices (4.6% vs 2.5%) (P <.001 for all).

10The CABG group also had a temporal increase in the proportion of male, elderly, and nonwhite patients and patients with lower socioeconomic status. Similar to what was observed in the PCI cohort, the prevalence of clinical risk factors increased significantly over time (Table 2). The proportion of patients with an Elixhauser comorbidity index score of 3 or greater increased from 29.8% in 2003 to 2007 to 52.2% in 2012 to 2016. The indications for CABG and surgical techniques also evolved over time. Compared with the 2003 to 2006 era, in 2012 to 2016, CABG was performed in a greater proportion of patients with AMI (28.2% vs 19.5%) and cardiogenic shock (6.1% vs 2.8%); however, these CABGs were more likely to be limited to 1 to 2 vessels (65.3% vs 55.6%), use arterial conduits (87% vs 82.2%), use double mammary conduits (3.7% vs 3.0%), or be isolated (86.9% vs 84.5%) but were less likely to use off-pump techniques (24.0% vs 19.3%) (P <.001 for all). Perioperative intra-aortic balloon pump use decreased from 9.7% to 8.9% (P <.001).

11In-hospital mortality after PCI increased between 2003 and 2016 (eTable 4 in the Supplement). However, after risk adjustment for patient- and hospital-level characteristics, in-hospital mortality only modestly increased after PCI for STEMI (4.9% to 5.3%; P <.001 for trend) or UA-SIHD (0.8% to 1.0%; P <.001) but remained stable after PCI for NSTEMI (1.6% to 1.6%; P =.18) (Figure 2A). In contrast, in-hospital mortality after isolated or combined CABG decreased significantly between 2003 and 2016 (eTable 5 in the Supplement). This temporal improvement in CABG mortality persisted after risk adjustment in both patients undergoing CABG in the context of AMI (5.6% to 3.4%; P <.001 for trend) or for UA-SIHD (2.8% to 1.7%; P <.001 for trend) (Figure 2B). Similar trends were observed when the analysis was limited to patients who underwent isolated CABG (Figure 2C and eTable 6 in the Supplement) or when we excluded patients who underwent both PCI and CABG during the same admission (eTable 7 in the Supplement). Length of stay after revascularization decreased across all groups, revascularization methods, and indications except among patients who underwent PCI for UA-SIHD (eTable 8 in the Supplement).

12This study documents 3 major findings. First, a decrease in the number of percutaneous and surgical coronary revascularization procedures conducted among hospital inpatients in the United States was found between 2003 and 2016. Second, significant changes were found in the demographic characteristics, socioeconomic status, risk profile, and clinical presentation of hospital inpatients undergoing coronary revascularization over time, as well as a significant change in the characteristics of the revascularization procedures. Third, a temporal decrease was found in in-hospital mortality after CABG but not after PCI across various indications.

13Several studies,,,,,, have found a decrease in coronary revascularization procedures in the United States in the past 2 decades. However, most of these studies,,,,,, were not contemporaneous, included only certain subsets of patients (eg, patients with Medicare insurance or AMI), or examined trends of surgical or percutaneous revascularization procedures. Although our primary objective was to assess the temporal changes among hospital inpatient risk profiles, procedural characteristics, and procedural mortality, the current study, to our knowledge, provides the most up-to-date nationwide analysis of the annual trends in coronary interventions conducted in hospitals. We documented a 40% decrease in CABG volume and a 43% decrease in PCI volume between 2003 and 2016. However, these downward trends appeared to stabilize at approximately 200 000 CABG procedures annually and 450 000 PCIs annually, with CABG volume reaching a steady level earlier than PCI volume (2010 vs 2014). Albeit speculative, reasons for these downward trends in the earlier years of our study may include the change in the management of stable CAD after the publication of landmark clinical trials reporting the effectiveness of medical management of stable CAD,,, the implementation of appropriate use criteria, and the improved efficacy of CAD preventive measures.,,, The increasing proportion of patients with AMI among all patients undergoing PCI (22.8% to 53.1%) and CABG (19.6% to 28.2%) over time further supports this notion.

14This study also found a temporal change in the demographic characteristics, socioeconomic status, and clinical risk profiles of hospital inpatients undergoing PCI or CABG and an evolution of the characteristics of these procedures. There was a modest increase in the number of elderly patients undergoing PCI or CABG but a more notable increase in the proportion of racial/ethnic minority patients and those with lower household income over time. Although this finding may reflect a change in the total population demographic characteristics and socioeconomic status during the same period, it may also partially reflect better penetration of coronary interventions to underserved populations. With regard to sex-related disparities in revascularization, not only did women remained underrepresented (approximately one-third overall) but also their proportion among all patients undergoing revascularization continued to decrease in both the CABG cohort (29.0% in 2003-2006 to 26.0% in 2012-2016, P <.001) and the PCI cohort (34.0% in 2003-2006 to 32.8% in 2012-2016, P <.001). Reasons for this disparity are likely multifaceted and deserve further investigations.

15There was a marked increase in the prevalence of clinical risk factors among hospital inpatients undergoing revascularization over time, which was reflected by the doubling of the proportion of patients with an Elixhauser comorbidity index score of 3 or greater between 2003 and 2016 in both the PCI and the CABG cohorts. This increase was global for atherosclerotic risk factors (eg, hypertension, hyperlipidemia, and diabetes), nonatherosclerotic risk factors (eg, lung, renal, and liver disease), and concomitant noncoronary atherosclerosis (eg, carotid stenosis and vascular disease). This finding may represent an increase in the prevalence of certain risk factors in the total US population,, the tendency to offer revascularization to sicker populations, the shift in risk resulting from performing fewer revascularization procedures in patients with stable CAD, or the database included in this study, which does not include PCIs conducted in outpatient settings, or a mixture of these factors. These findings have important implications for prerevascularization risk assessment and postrevascularization medical management. For example, the increasing prevalence of atrial fibrillation and anemia among hospital inpatients undergoing PCI may pose a challenge for a post-PCI antithrombotic and antiplatelet medical regimen.

16The changes in hospital inpatient presentation and clinical risk profile were also accompanied by changes in coronary revascularization techniques over time. In the PCI cohort, there was an increasing uptake of intravascular ultrasonography and fractional flow reserve measurement and a downward trend in the use of bare metal stents. There was also an increasing representation of higher-risk patients (eg, cardiogenic shock and chronic total occlusion) but fewer multivessel PCIs. In the CABG cohort, there were fewer multivessel (>2) CABGs and off-pump CABGs but greater use of arterial conduits over time. These trends likely reflect the effect of the emerging data that suggest the incremental benefit of certain techniques or devices (eg, fractional flow reserve, drug-eluting stents, and arterial conduits) and the limited value of others (eg, off-pump bypass).,,,,,,,

17We hypothesized that the temporal decrease in PCI and CABG volume, as well as the accompanying changes in hospital inpatient risk profile and procedural characteristics, might have been associated with a change in procedural mortality over time. We thus evaluated crude and risk-adjusted rates of in-hospital mortality of both procedures stratified by indication. We found that crude and risk-adjusted CABG mortality decreased significantly over time despite the substantial decrease in annual volume and the increasing prevalence of key comorbidities. Reasons for this trend may include changes in surgical techniques, the wider adoption of quality improvement, the changes in case mix and patient selection in light of the advances in PCI techniques, and public reporting of surgical outcomes, as well as the limitations of the NIS database.,,

18Contemporary data on the trends in PCI mortality are limited to subanalysis of specific PCI indications or certain subgroups of hospital inpatients. In a large study from the National Cardiovascular Data Registry CathPCI Registry, risk-adjusted in-hospital mortality of primary PCI for STEMI increased from 4.7% in 2005 to 5.3% in 2011 (P =.06). In another analysis from the same CathPCI registry, in-hospital mortality after PCI for cardiogenic shock increased over time. Goel et al found that in-hospital mortality after PCI in nonagenarians remained stable between 2003 and 2014 in the context of STEMI or NSTEMI but increased in the context of UA-SIHD.

19To our knowledge, our study provides the largest contemporary analysis of the temporal trends of PCI mortality among hospital inpatients overall. In this analysis, we found that in-hospital mortality after PCI did not improve over time. These findings may seem counterintuitive because of the advances in PCI tools, technique, and quality (eg, drug-eluting stents, radial access, mechanical circulatory support, and door-to-balloon time); however, other factors could have offset the assumed mortality benefits of these tools and techniques (eg, decreased operator experience and inadequate adjustment for patient risk in our study’s database). These assumptions deserve further elaboration. The association between operator volume and outcomes after coronary revascularization has been both well-established historically and reconfirmed in contemporary analyses.,, Although the decrease in operator experience because of the decreasing volume of revascularization procedures applies to both CABG and PCI, we speculate that its association with outcomes might be greater with PCI because of the larger number of PCI operators. Similarly, the lack of granular anatomic, laboratory, and procedural data in the NIS may have influenced the robustness of our risk adjustment. Although this lack of data applies to both the PCI and CABG groups, it is possible that the addition of such data to the logistic regression models could have affected the PCI group more than the CABG group. For example, the complexity of coronary lesions (eg, bifurcation and calcification) may affect PCI outcomes more than CABG outcomes. Nonetheless, in light of these data, more studies are needed to identify effective strategies to further optimize PCI outcomes.

20This study has limitations. First, the NIS is an administrative database that collects data for billing purposes. Thus, it is subject to undercoding, overcoding, or erroneous coding. However, coding of major procedures is the main method of obtaining reimbursement, and thus systematic inaccuracy in coding for PCI and CABG is unlikely. In addition, the NIS database has been used extensively to examine PCI and CABG trends and outcomes. Epstein et al validated the accuracy of the national estimation of annual volume with NIS by reporting a mean difference of 0.2% in quarterly PCI counts between Medicare claims and the NIS. Second, angiographic findings, laboratory data, characteristics of the PCI or CABG culprit vessel(s), access site, and perioperative medications are not available in NIS. Thus, the association of these unmeasured confounders with postrevascularization outcomes cannot be assessed. Third, the NIS allows detailed assessment of in-hospital outcomes but does not track patients after discharge or account for procedures conducted among outpatients. Therefore, long-term outcomes after PCI or CABG could not be investigated with this database. Despite these limitations, the NIS affords the unique opportunity to comprehensively assess the national trends in the utilization and outcomes of both percutaneous and surgical revascularization procedures in the United States during a 14-year period.

21There were considerable changes in the demographic characteristics, risk profile, and clinical presentation of patients undergoing PCI and CABG in hospitals that accompanied the substantial decrease in the annual volume of both procedures in the United states between 2003 and 2016. Risk-adjusted in-hospital mortality decreased over time after CABG but not after PCI across various clinical indications.