Author + information
- Received January 13, 2015
- Accepted January 27, 2015
- Published online April 14, 2015.
- Mohammed Salim Al-Damluji, MD, MPH∗,
- Kumar Dharmarajan, MD, MBA†,‡,
- Weiwei Zhang, MPH†,
- Lori L. Geary, MPH†,
- Erik Stilp, MD∗,§,
- Alan Dardik, MD, PhD‖,¶,#,
- Carlos Mena-Hurtado, MD∗,§ and
- Jeptha P. Curtis, MD∗,†,§∗ ()
- ∗Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut
- †Center for Outcome Research and Evaluation, Yale University, New Haven, Connecticut
- ‡Division of Cardiology, Columbia University Medical Center, New York, New York
- §Cardiovascular Division, Department of Medicine, Yale University School of Medicine, New Haven, Connecticut
- ‖Yale University Vascular Biology and Therapeutics Program, New Haven, Connecticut
- ¶Department of Surgery, Yale University School of Medicine, New Haven, Connecticut
- #Veterans Affairs Connecticut Healthcare System, West Haven, Connecticut
- ↵∗Reprint requests and correspondence:
Dr. Jeptha P. Curtis, Cardiovascular Division, Department of Medicine, Yale University School of Medicine, P.O. Box 208056, 333 Cedar Street, New Haven, Connecticut 06520-8056.
Background In appropriately selected patients with severe carotid stenosis, carotid revascularization reduces ischemic stroke. Prior clinical research has focused on the efficacy and safety of carotid revascularization, but few investigators have considered readmission as a clinically important outcome.
Objectives The aims of this study were to examine frequency, timing, and diagnoses of 30-day readmission following carotid revascularization; to assess differences in 30-day readmission between patients undergoing carotid endarterectomy (CEA) and carotid artery stenting (CAS); to describe hospital variation in risk-standardized readmission rates (RSRR); and to examine whether hospital variation in the choice of procedure (CEA vs. CAS) is associated with differences in RSRRs.
Methods We used Medicare fee-for-service administrative claims data to identify acute care hospitalizations for CEA and CAS from 2009 to 2011. We calculated crude 30-day all-cause hospital readmissions following carotid revascularization. To assess differences in readmission after CAS compared with CEA, we used Kaplan-Meier survival curves and fitted mixed-effects logistic regression. We estimated hospital RSRRs using hierarchical generalized logistic regression. We stratified hospitals into 5 groups by their proportional CAS use and compared hospital group median RSRRs.
Results Of 180,059 revascularizations from 2,287 hospitals, CEA and CAS were performed in 81.5% and 18.5% of cases, respectively. The unadjusted 30-day readmission rate following carotid revascularization was 9.6%. Readmission risk after CAS was greater than that after CEA. There was modest hospital-level variation in 30-day RSRRs (median: 9.5%; range 7.5% to 12.5%). Variation in proportional use of CAS was not associated with differences in hospital RSRR (range of median RSRR across hospital groups 9.49% to 9.55%; p = 0.771).
Conclusions Almost 10% of Medicare patients undergoing carotid revascularization were readmitted within 30 days of discharge. Compared with CEA, CAS was associated with a greater readmission risk. However, hospitals' RSRR did not differ by their proportional CAS use.
Carotid revascularization is a commonly performed class of vascular interventions that in appropriately selected patients reduces the risk for ipsilateral stroke (1). Prior research has focused on the safety and efficacy of carotid revascularization, but few investigators have considered short-term readmission as a clinically important outcome (2). However, the potential importance of readmissions following carotid revascularization has increased dramatically with the development of a measure of hospitals' 30-day risk-standardized readmission rates (RSRRs) following vascular procedures (3), with implications for potential financial penalties for hospitals with excess readmissions.
At present, many aspects of readmissions following carotid revascularization are poorly understood. Specifically, the timing and reasons for readmission, as well as the extent of variation in hospitals' 30-day readmission rates, have not been described. Furthermore, like many vascular procedures, carotid revascularization can be accomplished through an open or endovascular approach (carotid endarterectomy [CEA] and carotid artery stenting [CAS], respectively). Although prior research has shown that patients undergoing CAS are at increased risk for readmission compared with patients undergoing CEA (2), it is not known whether differences in hospitals' proportional use of CAS among all carotid revascularization procedures are associated with differences in hospital 30-day readmission rates.
To address these gaps in knowledge, we used administrative claims data to identify readmissions occurring within 30 days of the end of hospitalization during which a carotid revascularization procedure was performed. We examined the timing and diagnoses associated with readmission and characterized variation in RSRRs across hospitals. Furthermore, we assessed whether readmission rates varied by the choice of the revascularization strategy used (CAS vs. CEA) and whether hospitals' 30-day RSRRs differed by the proportional use of CAS among all carotid revascularization procedures.
We used Medicare fee-for-service administrative claims data to identify hospitalizations with an associated carotid revascularization procedure performed between January 1, 2009 and December 31, 2011. We identified patients using International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) codes or Current Procedural Terminology codes, previously reported by the Centers for Medicare & Medicaid Services (CMS) vascular procedure readmission measure (3). As with this measure, our study sample consisted of patients who underwent either CEA (ICD-9-CM codes 00.61, 00.63, 00.64, 39.72, 39.74; Current Procedural Terminology codes 37215, 37216, 0075T) or CAS (ICD-9-CM codes 38.02, 38.12, 38.32, 38.42 or Current Procedural Terminology codes 35201, 35005, 35231, 35301, 35701, 34001). Whether patients were symptomatic (i.e., had symptoms attributable to their carotid disease) was determined using the following ICD-9-CM codes previously described in published reports (2,4,5): 362.30, 362.31, 362.32, 362.33, 362.34, 362.35, 362.36, 362.37, 362.84, 433.11, 433.31, 434.01, 434.91, 435.0, 435.1, 435.2, 435.3, 435.8, 435.9, and 781.4. The primary outcome of interest was hospital readmission within 30 days of discharge from a hospitalization during which a carotid revascularization procedure had been performed. Among patients who had been readmitted, we examined the timing of the readmission by day after discharge (0 to 30). We categorized readmission diagnoses using the CMS condition category (CC) codes to group patients' principal discharge diagnoses. Each of the 189 CMS CCs describes a disease entity or medical condition. However, because more than 90% of these CCs contributed <1% of all readmissions, we further consolidated the 189 CMS CCs into 30 modified CCs. The categorization of CCs is described in Online Tables 1 and 2.
We used descriptive statistics to illustrate baseline characteristics. We compared differences in baseline characteristics between CEA and CAS patients using the Mantel-Haenszel chi-square test and the Student t test, as appropriate. We calculated crude readmission rates following carotid revascularization procedures by dividing the number of patients readmitted within the 30-day period following carotid revascularization by the total number of patients undergoing carotid revascularization. To examine the timing of readmission following carotid revascularization, we calculated the proportion of readmissions during each day over the 30-day period. We also identified the 10 most common diagnoses associated with readmission by modified CCs and presented their distribution over the following consecutive time periods after hospital discharge: days 0 to 7, 8 to 15, and 16 to 30, reflecting the time periods in which follow-up visits to ambulatory care providers frequently occur.
Procedure-specific readmission rates: CAS versus CEA
To assess the difference in readmission rates between patients undergoing CEA versus CAS, we plotted Kaplan-Meier readmission-free survival curves and fitted a generalized estimating equation logistic regression model. These models adjusted for patient and hospital characteristics. In these models, we considered all variables included in the models as fixed effects except the hospital variable, which was considered as a random effect to account for within-hospital correlation.
To account for baseline differences between CEA and CAS, we used propensity score matching to create a new matched dataset (N = 64,238) on the basis of age, sex, race, symptomatic status, and other comorbidities identified in the multivariable model. After matching, the standardized difference in any variable between the 2 groups did not exceed 10% (Online Table 3), indicating successful matching with respect to the chosen matching variables. We then fit a generalized estimating equation model using this propensity-matched cohort.
Hospital-specific risk-standardized readmission rates
We estimated hospital-specific 30-day RSRRs using a hierarchical mixed-effects logistic regression model adjusted for demographics and comorbidities. To account for within-hospital correlation, we estimated a random intercept for each hospital by including the hospital variable as a random effect variable. We estimated the hospital-specific RSRR by calculating the ratio of the predicted 30-day readmission rate within each hospital to the expected 30-day readmission rate within each hospital. We used this ratio to standardize an unadjusted national (U.S.) average of readmission rates following carotid revascularization procedures. We estimated the predicted readmission rate of each hospital using the random intercept specific to that hospital and the population mix within that hospital. We calculated the hospital expected readmission rate from a mean hospital intercept, calculated from our sample of hospitals and the population mix of that specific hospital.
We also examined differences in hospital-specific RSRRs by the utilization of a carotid revascularization strategy. In order to obtain a stable estimate of hospitals' proportional use of CAS, we limited this analysis to data from hospitals that performed at least 25 carotid revascularization procedures during the observation period. To examine whether hospital RSRR varied by the proportion of revascularization procedures performed using CAS, we stratified hospitals into 5 groups by their proportional use of CAS among all carotid revascularization procedures (0%, >0% to 10%, >10% to 20%, >20% to 30%, and >30%). We utilized descriptive statistics to summarize hospital-specific RSRRs within each hospital group. We then compared the medians of these 5 hospital groups using a Kruskal-Wallis test. The study protocol was approved by the Human Investigation Committee of Yale University School of Medicine. We used SAS version 9.3 (SAS Institute, Cary, North Carolina) for statistical analyses and defined statistical significance as 2-tailed p < 0.05.
We identified 168,323 patients who underwent a total of 180,059 carotid revascularization procedures (93.2% patients underwent 1 carotid intervention; 6.8% underwent >1 carotid intervention) at 2,287 hospitals between January 1, 2009 and December 31, 2011. Of these 180,059 procedures, 81.5% (146,831) were CEA and 18.5% (33,228) were CAS. The mean age of the cohort was 76.3 years, and 93.2% of included patients were Caucasian (Table 1). Symptomatic carotid stenosis, diabetes, hypertension, and congestive heart failure were present in 9.7%, 43.3%, 89.3%, and 19.7% of the study population, respectively. A total of 43.1% of patients had other peripheral vascular diseases, and 17.6% had renal failure.
The unadjusted 30-day readmission rate following carotid revascularization was 9.6% (17,202 procedures). One-fourth of all readmissions occurred within 3 days of hospital discharge, 44.3% occurred within the first 7 days of discharge, and 67.8% occurred within the first 14 days of discharge (Figure 1). The unadjusted 30-day mortality rate following carotid revascularization was 1.2% (2,247 procedures). The reasons for readmission following carotid revascularization procedures were broad (Figure 2), with the 10 most common readmission diagnoses contributing only 50.0% of all 30-day readmissions. The most common reasons for readmission were complications of care (8.6%), heart failure (6.6%), and pneumonia (including aspiration pneumonitis) (5.2%). The most common complications of care included hematoma formation (2.3%), iatrogenic stroke/intracranial hemorrhage (0.8%), and non–central nervous system bleeding (0.5%). Additionally, cerebral complications including ischemic stroke, transient ischemic attack, and cerebral hemorrhage were collectively associated with 10.7% of readmissions. Some readmission diagnoses varied over time (Figure 3). For instance, complications of care, heart failure, and acute stroke were more often encountered during the first week following discharge. In contrast, other peripheral vascular diseases, chronic angina, and coronary artery disease were more common in the third and fourth weeks following discharge.
Procedure-specific readmission rate: CAS versus CEA
The characteristics of patients undergoing CEA and CAS procedures varied modestly (Table 1). The mean ages of CAS and CEA patients were 76.1 and 76.3 years, respectively. Male patients and Caucasian patients were more likely to have undergone CEA than CAS. Greater proportions of patients undergoing CEA had diabetes, hypertension, and coronary atherosclerosis (all, p < 0.001). In contrast, greater proportions of CAS patients had symptomatic carotid stenosis, congestive heart failure, coronary syndrome, other peripheral vascular diseases, and chronic renal failure (all, p < 0.001). Crude 30-day readmission rates following CEA and CAS were 9.0% (13,222 of 146,831) and 12.0% (3,980 of 33,228), respectively. On multivariate analysis, the adjusted risk for readmission was significantly greater among patients undergoing CAS compared with that in patients undergoing CEA (adjusted odds ratio [AOR]: 1.13; 95% confidence interval [CI]: 1.08 to 1.18; p < 0.001) (Central Illustration, upper panel). Within the propensity-matched subgroup, CAS patients also had a significantly greater rate of readmission compared with that in patients undergoing CEA (AOR: 1.18; 95% CI: 1.07 to 1.23; p < 0.001).
Symptomatic patients were significantly more likely to have been readmitted compared with asymptomatic patients (AOR: 1.18; 95% CI: 1.12 to 1.24; p < 0.001) (Figure 4). Other notable predictors of readmissions were age >65 years (AOR: 1.01; 95% CI: 1.01 to 1.02; p < 0.001), female sex (AOR: 1.12; 95% CI: 1.08 to 1.15; p < 0.001), and non-Caucasian race (AOR: 1.12; 95% CI: 1.06 to 1.19; p < 0.001).
In stratified analyses, compared with CEA, CAS was associated with greater rates of readmission in both patients with symptomatic (AOR: 1.24; 95% CI: 1.12 to 1.37; p < 0.001) and asymptomatic (AOR: 1.10; 95% CI: 1.05 to 1.14; p < 0.001) carotid stenosis. An interaction term (symptomatic status × procedural approach) was not significant. Similarly, compared with CEA, CAS was associated with greater rates of readmission among patients age >80 years (AOR: 1.12; 95% CI: 1.04 to 1.20; p < 0.001), patients age <80 years (AOR: 1.12; 95% CI: 1.07 to 1.18; p < 0.001), male patients (AOR: 1.13; 95% CI: 1.07 to 1.19; p < 0.001), female patients (AOR: 1.13; 95% CI: 1.07 to 1.20; p < 0.001), Caucasian race (AOR: 1.11; 95% CI: 1.06 to 1.16; p < 0.001) and non-Caucasian race (AOR: 1.25; 95% CI: 1.10 to 1.42; p < 0.001).
Hospital-specific risk-standardized readmission rates
There was modest hospital-level variation in 30-day RSRRs (Figure 5). The median hospital 30-day RSRR was 9.5%, with a range of 7.5% to 12.5% and an interquartile range of 9.2% to 10.0%. The use of CAS varied significantly across hospitals (Figure 6). Of the 2,287 hospitals, 1,500 (65.6%) performed more than 25 carotid revascularization procedures and were included in our analysis of hospitals' proportional use of CAS. Variation in the proportional use of CAS was not associated with significant differences in hospital RSRR (p = 0.771) (Central Illustration, lower panel).
Among Medicare patients undergoing carotid revascularization, 9.6% were readmitted within 30 days of discharge (2,6–9). A high proportion of readmissions occurred within the first 7 days of hospital discharge, and readmissions were associated with a wide range of principal diagnoses. Furthermore, we observed that patients undergoing CAS were at greater risk for readmission compared with patients undergoing CEA. Yet hospitals performing a greater proportion of revascularization via CAS did not have greater hospital 30-day RSRRs.
CMS has indicated its intent to publicly report hospitals' 30-day readmission rates in patients undergoing vascular procedures, which includes carotid revascularization procedures (3). A measure of vascular readmissions has been submitted to the National Quality Forum for evaluation, and in the future, hospitals with greater-than-expected readmission rates may be subject to payment penalties (3,10). In this context, our analyses of hospital RSRRs have important implications. We have characterized both current national performance and the extent of hospital variation in readmissions following carotid revascularization. This information provides a benchmark to which hospitals can evaluate efforts to reduce readmission rates in this population.
We found that although patients undergoing carotid revascularization were at a disproportionately higher risk for readmission in the first week after discharge, they remained at substantial risk for readmission throughout the 30-day post-discharge period. Furthermore, patients were readmitted with a wide range of diagnoses, including cardiovascular, pulmonary, neurological, and renal disorders, as well as complications of care, with no single diagnosis contributing to more than 10% of readmissions. Indeed, the 10 most common readmission diagnoses were collectively involved in only one-half of readmissions. Yet, almost one-third of readmission diagnoses were potentially due to procedural complications, including cerebral events (10.7%), complications of care (8.6%), acute coronary syndrome (5.0%), and arrhythmias (4.0%). These complications may represent high-yield areas for targeted efforts to reduce readmission. However, the broad range of diagnoses is consistent with the theory that patients experience a period of increased risk for a variety of illnesses following hospital discharge (11,12). As such, interventions targeting a specific readmission diagnosis or time period may be relatively ineffective in reducing readmission rates. Instead, more general strategies leading to improvement in discharge planning, medication reconciliation, and early follow-up after discharge may better address underlying vulnerabilities in the transitional care process (13).
In our sample of Medicare patients, the rates of readmission attributed to stroke and myocardial infarction were 0.5% (883 of 180,059) and 0.5% (861 of 180,059), respectively. These rates are less than the rates of similar events reported by comparable randomized controlled trials (stroke: CREST [Carotid Revascularization Endarterectomy versus Stenting Trial] (1) 3.2%, SAPPHIRE [Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy] (14) 3.3%; myocardial infarction: CREST 1.7%, SAPPHIRE 4.2%). However, direct comparisons are challenging given that the present study included only events that occurred after hospital discharge, whereas the numbers reported in the clinical trials included events that occurred during both hospitalization and follow-up. Our results suggest that the majority of these complications occur during hospitalization and are relatively rare following discharge.
Consistent with prior investigators, we found that CAS patients were more likely to have been readmitted than were CEA patients (2). The difference in the risk for readmission is likely driven in part by patient selection. Currently, CEA is the mainstay treatment of symptomatic and asymptomatic carotid artery stenosis, and CAS is typically reserved for patients with comorbidities or anatomy that put them at increased risk for adverse outcomes following CEA (1,15,16). Nonetheless, the risk for readmission with CAS was greater than that with CEA even after observed differences in patient characteristics were accounted for. This finding may reflect the limitations of administrative data, but additional research examining whether the excess risk could be mitigated is warranted. CAS is still a relatively new procedure, and the excess risk for readmission may reflect the learning curve associated with incorporating CAS into routine clinical practice. Within a hospital, CAS may be performed by physicians in different specialties with a range of experience and expertise, and researchers have shown that most physicians who perform CAS are low-volume operators with variable use of recommended strategies, such as the use of embolic protection (17,18). Therefore, there may be an opportunity for improving care by consolidating experience within a smaller team. Such an approach could have ancillary benefits with regard to reducing readmissions by facilitating a smooth transition of care from the inpatient to the outpatient setting.
In our study, we identified a wide range of demographic characteristics (e.g., advancing age, female sex, and non-Caucasian race) and baseline comorbidities (e.g., symptomatic carotid stenosis, diabetes, congestive heart failure, acute coronary syndrome, and ischemic stroke) associated with an increased risk for readmission. These factors may serve as important markers for identifying patients at risk for readmission.
In our sample of hospitals, a large proportion of the hospitals did not perform a CAS procedure, and among hospitals that performed both CEA and CAS, the proportional use of CAS varied widely. Of all hospitals that performed carotid revascularization procedures, only 135 (9%) performed more CAS than CEA. Importantly, although readmissions were more common after CAS than after CEA, the RSRRs of hospitals that used CAS more frequently were comparable to those of hospitals that performed CAS less frequently. This observation suggests that hospitals using CAS more frequently will not necessarily be disadvantaged if and when measures of vascular readmissions are publicly reported and included in payment programs.
First, we used administrative claims data to identify the sample population and associated comorbidities. Such databases are subject to coding errors, coding differences across providers and institutions, and changes in coding patterns over time (19,20). The accuracy of these coding systems in identifying carotid revascularization has not been reported; however, the specificity and positive predictive values in identifying comorbidities in cardiovascular patients are generally reasonable (21,22). Second, administrative claims data may not be suitable for identifying staged revascularization procedures, as some physicians may elect to perform the carotid repair over 2 or more consecutive sessions. These planned readmissions were included in our model despite being unrelated to the provided care quality and, consequently, may have inflated our readmission estimates. Third, administrative claims data do not provide detailed information about the procedure itself (11,23). Other sources of data such as clinical registries may provide insight into the association of intraprocedural practices and the risk for subsequent readmission. Finally, the patients in our cohort were not randomized to the carotid revascularization strategies, and as a result, residual confounds may be present in our observed association despite the use of propensity score matching.
One in 10 Medicare patients undergoing carotid revascularization is readmitted within 30 days. The present study is in large part foundational, raising awareness among clinicians, hospital administrators, and policymakers regarding this issue and illustrating the extent to which readmission rates differ by the type of revascularization procedure.
COMPETENCY IN MEDICAL KNOWLEDGE: Almost 10% of Medicare patients undergoing carotid revascularization are readmitted to the hospital within 30 days of discharge for a wide range of conditions, most of which occur within 7 days after discharge. Carotid artery stenting is associated with 30-day readmission rates greater than that with endarterectomy.
TRANSLATIONAL OUTLOOK: Further investigation is warranted for developing strategies that reduce the need for hospital readmission following carotid revascularization procedures.
This research was funded by the National Heart, Lung, and Blood Institute (U01 HL105270-03 [to the Center for Cardiovascular Outcomes Research at Yale University]). Dr. Curtis has received support from the Centers for Medicare & Medicaid Services to develop and maintain performance measures that are used for public reporting; and he holds stock options in Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- carotid artery stenting
- condition category
- carotid endarterectomy
- Centers for Medicare & Medicaid Services
- International Classification of Diseases-Ninth Revision-Clinical Modification
- risk-standardized readmission rate
- Received January 13, 2015.
- Accepted January 27, 2015.
- American College of Cardiology Foundation
- Galiñanes E.L.,
- Dombroviskiy V.Y.,
- Hupp C.S.,
- Kruse R.L.,
- Vogel T.R.
- ↵Bernheim SG, Li L, Li S-X, et al. Hospital 30-day all-cause risk-standardized readmission rate (RSRR) following vascular procedures—Measure Methodology report. 2011.
- Brinjikji W.,
- El-Sayed A.M.,
- Kallmes D.F.,
- Lanzino G.,
- Cloft H.J.
- Curran T.,
- Lo R.C.,
- Fokkema M.,
- et al.
- Medicare Payment Advisory Committee. Report to the Congress: Promoting Greater Efficiency in Medicare. 2007. Available at: http://www.medpac.gov/documents/reports/Jun07_EntireReport.pdf. Accessed March 2, 2015.
- Brott T.G.,
- Halperin J.L.,
- Abbara S.,
- et al.