Acute cerebrovascular accident (CVA) after percutaneous coronary intervention (PCI) for acute coronary syndrome and coronary artery disease is associated with high rates of morbidity and mortality. Nationwide Inpatient Sample from 1998 to 2008 was used to identify 1,552,602 PCIs performed for acute coronary syndrome and coronary artery disease. We assessed temporal trends in the incidence, predictors, and prognostic impact of CVA in a broad range of patients undergoing PCI. The overall incidence of CVA was 0.56% (95% confidence interval [CI] 0.55 to 0.57). The incidence of CVA remained unchanged over the study period (adjusted p for trend = 0.2271). The overall mortality rate in the CVA group was 10.76% (95% CI 10.1 to 11.4). The adjusted odds ratio (OR) of CVA for in-hospital mortality was 7.74 (95% CI 7.00 to 8.57, p <0.0001); this remained high but decreased over the study period (adjusted p for trend <0.0001). Independent predictors of CVA included older age (OR 1.03, 95% CI 1.02 to 1.03, p <0.0001), disorder of lipid metabolism (OR 1.31, 95% CI 1.24 to 1.38, p <0.001), history of tobacco use (OR 1.21, 95% CI 1.10 to 1.34, p = 0.0002), coronary atherosclerosis (OR 1.56, 95% CI 1.43 to 1.71, p <0.0001), and intra-aortic balloon pump use (OR 1.39, 95% CI 1.09 to 1.77, p = 0.0073). A nomogram for predicting the probability of CVA achieved a concordance index of 0.73 and was well calibrated. In conclusion, the incidence of CVA associated with PCI has remained unchanged from 1998 to 2008 in face of improved equipment, techniques, and adjunctive pharmacology. The risk of CVA-associated in-hospital mortality is high; however, this risk has decreased over the study period.
The field of interventional cardiology has advanced tremendously since the advent of coronary angioplasty in 1977. Despite improvements in techniques, equipment, and adjunctive pharmacology, acute cerebrovascular accidents (CVA) after percutaneous coronary intervention (PCI) remain one of the most devastating adverse complications with high rates of mortality and morbidity. The incidence of CVA in patients undergoing PCI ranges from 0.07% to 1.4%. Patients with CVA complications after PCI have a markedly high in-hospital mortality rate in the range of 10% to 37%. The Healthcare Cost and Utilization Project database encompasses an extensive collection of longitudinal hospital care data in the United States; this enables research on a broad range of health-care policy issues. The purpose of this study was to assess the temporal trends in the incidence, predictors, and prognostic impact of CVA in a broad range of patients with acute coronary syndrome and coronary artery disease undergoing PCI from 1998 to 2008 and to build a nomogram for predicting the likelihood of CVA with available patient profile information. The data in this study are a representation of real-world clinical practice in the United States in the last decade.
Methods
The Nationwide Inpatient Sample (NIS) is the largest all-payer United States inpatient care database that contains over a hundred clinical and nonclinical data elements from approximately 8 million hospital stays each year. From 1998 to 2008, a total of 1,552,602 PCI procedures performed in patients for symptomatic coronary artery disease and acute myocardial infarction (AMI) diagnoses, which encompass ST elevation myocardial infarction and non–ST elevation myocardial infarction, were identified. The Clinical Classifications Software developed by Healthcare Cost and Utilization Project was used in analyzing our dependent and independent variables. CVA was defined as any new focal neurologic deficit lasting ≥24 hours and transient cerebral ischemia as any new deficits lasting <24 hours. The specific single-level Clinical Classifications Software diagnosis categories used to define CVA in this study were “109: acute cerebrovascular disease and 112: transient cerebral ischemia.” The subtypes of CVA were defined as follows: “hemorrhagic CVA: International Classification of Diseases, Ninth Revision codes 430, 431, and 432; ischemic CVA: International Classification of Diseases, Ninth Revision codes 433 and 434; and transient cerebral ischemia: Clinical Classifications Software code 112.” CVA was the dependent variable. Additional data on covariates were collected and are listed in Table 1 .
| Variable | CVA | p-Value | |
|---|---|---|---|
| No (n = 1,543,858) | Yes (n = 8,744) | ||
| Age, mean ± SD (years) | 64 ± 12 | 70 ± 12 | <0.0001 |
| Women | 34.1% | 47.6% | <0.0001 |
| White | 81.9% | 79.0% | |
| Black | 6.7% | 9.2% | |
| Hispanic | 6.1% | 6.9% | <0.0001 |
| Asian | 1.7% | 1.7% | |
| Native American | 0.4% | 0.3% | |
| Other | 3.2% | 2.9% | |
| Died during hospitalization, total | 11,798 (0.8%) | 939 (10.8%) | <0.0001 |
| Length of stay, mean ± SD (days) | 2.8 ± 3.1 | 7.9 ± 8.0 | <0.0001 |
| Total charge, mean ± SD ($) | 39,952 ± 29,948 | 68,516 ± 63,381 | <0.0001 |
| Hypertension | 975,346 (63.2%) | 5,623 (64.3%) | 0.0288 |
| Dyslipidemia | 839,967 (54.4%) | 3,434 (39.3%) | <0.0001 |
| Diabetes mellitus | 445,889 (28.9%) | 2,683 (30.7%) | 0.0002 |
| Chronic renal failure | 44,201 (2.9%) | 448 (5.1%) | <0.0001 |
| Chronic obstructive pulmonary disease and bronchiectasis | 150,545 (9.8%) | 1,254 (14.3%) | <0.0001 |
| Pulmonary heart disease | 21,603 (1.4%) | 200 (2.3%) | <0.0001 |
| Tobacco use disorder | 259,563 (16.8%) | 974 (11.1%) | <0.0001 |
| History of tobacco use | 148,677 (9.6%) | 543 (6.2%) | <0.0001 |
| Peripheral and visceral atherosclerosis | 104,620 (6.8%) | 942 (10.8%) | <0.0001 |
| Occlusion or stenosis of precerebral arteries | 22,879 (1.5%) | 637 (7.3%) | <0.0001 |
| Other and ill-defined cerebrovascular disease | 3,152 (0.2%) | 137 (1.6%) | <0.0001 |
| Congestive heart failure | 171,936 (11.1%) | 2,131 (24.4%) | <0.0001 |
| Heart valve disorders | 109,748 (7.1%) | 1,050 (12.0%) | <0.0001 |
| Aortic valve disorder | 7,695 (0.5%) | 145 (1.7%) | <0.0001 |
| Acute myocardial infarct | 531,130 (34.4%) | 4,471 (51.1%) | <0.0001 |
| Coronary atherosclerosis and other heart disease | 1,484,506 (96.2%) | 7,644 (87.4%) | <0.0001 |
| Atrial Fibrillation/Flutter | 121,235 (7.9%) | 1,609 (18.4%) | <0.0001 |
| Peri- endo- and myocarditis, cardiomyopathy | 45,167 (2.9%) | 432 (4.9%) | <0.0001 |
| Aortic/peripheral/visceral artery aneurysm, embolism or thrombosis | 23,296 (1.5%) | 233 (2.7%) | <0.0001 |
| Cardiac and circulatory congenital anomalies | 4,264 (0.3%) | 64 (0.7%) | <0.0001 |
| Shock | 13,574 (0.9%) | 361 (4.1%) | <0.0001 |
| Malignancy, any | 95,696 (6.2%) | 598 (6.8%) | 0.0133 |
| Headache, including migraine | 8,620 (0.6%) | 115 (1.3%) | <0.0001 |
| Systemic lupus erythematosus/connective tissue disease | 46,831 (3.0%) | 575 (6.6%) | <0.0001 |
| Coagulation and hemorrhagic disorders | 21,142 (1.4%) | 292 (3.3%) | <0.0001 |
| Anemia | 85,394 (5.5%) | 892 (10.2%) | <0.0001 |
| Intra-aortic balloon pump | 8,143 (0.5%) | 147 (1.7%) | <0.0001 |
Our primary analysis was to assess the incidence of CVA and its trend over the study period. Our secondary analysis looked at independent predictors of CVA, mortality rate in patients with CVA, and temporal trend in mortality rate in patients with CVA over the study years. In addition, we built a nomogram that predicts the likelihood of developing CVA for patients undergoing PCI with readily available demographic and clinical characteristic variables.
The study population was separated into 2 groups—those with CVA and without CVA. The summary statistics with baseline characteristics were generated for the entire population separated into the 2 groups and for the subpopulations stratified by the year.
All tests were 2-tailed, and a p value of <0.05 was considered significant for all tests. Univariate analysis was initially conducted to summarize the data. The Pearson chi-square tests were used to test for categorical variables and are presented as percentages. The nonparametric Wilcoxon rank sum test was used to test for all continuous variables and is presented as mean ± SD.
Logistic regressions were fit to the data to evaluate the trend for incidence of CVA over the years 1998 to 2008. The Wald test was used to test the null hypothesis of no trend. The logistic regression model was then used to assess predictors of CVA after adjusting for the observed baseline demographic and clinical characteristics. The logistic regression model was also used to investigate the trends for incidence of in-hospital mortality with and without CVA and to assess the trends for the adjusted and unadjusted odds ratio (OR) for the association between death and CVA over the study years. We used the propensity score method to evaluate the association between CVA and the mortality rate. Propensity scores were estimated using a logistic regression model with CVA as the outcome and all the observed baseline demographic and clinical characteristic variables. We used the method of regression adjustment by the estimated propensity scores to estimate the association between CVA and the mortality rate, taking into account all the other observed baseline demographic and clinical characteristic variables. Advantage of this 2-step propensity score procedure is that this allows us to fit a complicated propensity score model with interactions and higher order terms for more accurate estimation of CVA probability.
The missing data were omitted as follows: in the no CVA group (n = 1,543,858), age (n = 36, 0.002%), length of stay (n = 22, 0.001%), mean total charge (n = 20,721, 1.3%), female gender (n = 137, 0.009%), race (n = 419,744, 27.2%), and death (n = 315, 0.02%) and in the CVA group (n = 8,744), mean total charge (n = 141, 1.6%), female gender (n = 1, 0.01%), race (n = 2,283, 26.1%), and death (n = 20, 0.23%).
A multivariate logistic regression model was built to link the demographic and clinical characteristic variables with CVA, which served as the basis of the nomogram for predicting the probability of developing CVA. To relax the common modeling assumption that the association between risk factors and outcome is linear, we applied restricted cubic splines to continuous variables. The nomogram was internally validated with 1,000 bootstrap resamples to objectively evaluate the predictive performance after correcting overfit bias. First, model discrimination ability was quantified using the C-index, which is equivalent to the area under the receiver operating characteristic curve and, ranges from 0.5 to 1, with 0.5 indicating no difference from chance and 1 for perfect prediction. In addition, the agreement between the observed and the predicted outcomes was visually checked with a calibration plot.
All analyses were performed using SAS statistical software, version 9.2 (SAS Institute Inc., Cary, North Carolina), or the open source statistical package R-2.15.2 (R Core Team, 2012, Vienna, Austria).
All investigators have read and agree to the manuscript as written. The investigators are solely responsible for the design and conduct of this study, all study analyses, and the drafting and editing of the report and its final contents. This study has been approved by the University of Illinois at Chicago Institutional Review Board.
Results
From 1998 to 2008, there were 1,552,602 PCI procedures performed for AMI and coronary artery disease diagnoses. Patients’ baseline characteristics and clinical presentation are listed in Table 1 . CVA during PCI hospitalization occurred in 8,744 patients (0.56%, 95% confidence interval [CI] 0.55 to 0.57). The overall incidence of CVA associated with PCI remained unchanged during the study period from 0.60% in 1998 to 0.66% in 2008 (p for trend = 0.2308 univariate analysis, p = 0.2271 multivariate analysis; Figure 1 ). Incidences of various types of CVA are listed in Table 2 .
| Acute Cerebrovascular Accidents (n = 8,744) | |
|---|---|
| Hemorrhagic CVA | 715 (8.2%) |
| Ischemic CVA | 4669 (53.4%) |
| Transient cerebral ischemia | 2889 (33.0%) |
The predictors of CVA after adjusting for covariates were identified ( Figure 2 ). Important predictors in this multivariate analysis included older age (adjusted OR 1.03, 95% CI 1.02 to 1.03, p <0.0001), disorder of lipid metabolism (OR 1.31, 95% CI 1.24 to 1.38, p <0.001), history of tobacco use (OR 1.21, 95% CI 1.10 to 1.34, p = 0.0002), coronary atherosclerosis or other heart disease (OR 1.56, 95% CI 1.43 to 1.71, p <0.0001), and intra-aortic balloon pump use (IABP; OR 1.39, 95% CI 1.09 to 1.77, p = 0.0073). Female gender, diabetes mellitus, hypertension, chronic renal failure, AMI, congestive heart failure, aortic valve disorder, atrial fibrillation or flutter, peripheral and/or visceral atherosclerosis, and shock were not significant predictors for CVA in multivariate analysis. The length of stay (p <0.0001) and mean total charge (p = 0.0463) were significantly higher and associated with CVA after adjusting for covariates.
The overall mortality rate in the CVA group was 10.8% (95% CI 10.1 to 11.4). Patients who died were 15.66 times (95% CI 14.60 to 16.80, p <0.0001) more likely to have CVA than the patients who were alive; this OR decreased to 7.74 (95% CI 7.00 to 8.57, p <0.0001) after adjusting for covariates. The in-hospital mortality in patients with CVA (adjusted p for trend <0.0001) and without CVA (adjusted p for trend <0.0001) decreased from 1998 to 2008 ( Figure 3 A and B, respectively). Both the adjusted and unadjusted OR for in-hospital mortality associated with CVA were high for each individual year compared with in-hospital mortality without CVA (p <0.0001 for all years). The temporal trend of the unadjusted and adjusted OR for in-hospital mortality associated with acute CVA (p for trend = 0.0026 and <0.0001, respectively) decreased over the study period ( Table 3 ).
