Sources of Hospital Variation in Short-Term Readmission Rates After Percutaneous Coronary Intervention
Background—Risk-standardized all-cause 30-day readmission rates (RSRRs) after percutaneous coronary intervention (PCI) have been endorsed as a national measure of hospital quality. Little is known about variation in the performance of hospitals on this measure, and whether high hospital rates of readmission after PCI are due to modifiable deficiencies in quality of care has not been assessed.
Methods and Results—We estimated 30-day, all-cause RSRRs for all nonfederal PCI-performing hospitals in Massachusetts, adjusted for clinical and angiographic variables, between 2005 and 2008. We assessed if differences in race, insurance type, and PCI and post-PCI characteristics, including procedural complications and discharge characteristics, could explain variation between hospitals using nested hierarchical logistic regression models. Of 36 060 patients undergoing PCI at 24 hospitals and surviving to discharge, 4469 (12.4%) were readmitted within 30 days of discharge. Hospital RSRRs ranged from 9.5% to 17.9%, with 8 of 24 hospitals being identified as outliers (4 lower than expected and 4 higher than expected). Differences in race, insurance, PCI, and post-PCI factors accounted for 10.4% of the between-hospital variance in RSRRs.
Conclusions—We observed wide variation in hospital 30-day all-cause RSRRs after PCI, most of which could not be explained by identifiable differences in procedural and postprocedural factors. A better understanding of etiologies of hospital variation is necessary to determine whether this measure is an actionable assessment of hospital quality, and, if so, how hospitals might improve their performance.
Hospital readmissions have been identified as an important contributor to both patient morbidity and rising healthcare expenditures across a number of disease processes.1,2 Because readmissions may be preventable,3,4 risk-adjusted rates of readmission within 30 days of discharge are currently used as publicly reported measures of hospital quality for 3 disease processes: pneumonia, heart failure, and myocardial infarction (MI).5 In 2007, the Medicare Payment Advisory Commission (MedPAC) recommended the use of these measures for financially penalizing those hospitals with elevated readmission rates, as a means of incentivizing improvements in quality of care while reducing costs.6 Under the Affordable Care Act, these financial penalties would go into effect in fiscal year 2012 for the current publicly reported diagnoses, with potential expansion of the policy to additional conditions.7
Percutaneous coronary interventions (PCI) are associated with the highest rates of subsequent short-term readmission, accounting for $350 million in healthcare expenditures per year.6 Having been commissioned by the Department of Health and Human Services to evaluate new hospital performance measures, the National Quality Forum endorsed using risk-standardized, all-cause, 30-day readmission rates (RSRR) after PCI, as a new measure of hospital quality.8 Although there is evidence of significant variation in unadjusted hospital 30-day readmission rates after PCI, it is unknown whether this variation persists after accounting for differences in hospital case mix and whether the factors responsible for this variation are modifiable.9
Understanding the reasons that patients are readmitted after PCI, identifying characteristics of patients at high risk, and determining the reasons for variability among hospital readmission rates are necessary first steps in predicting the impact of such a measure and developing implementable strategies to reduce readmissions. We therefore estimated a series of regression models to assess hospital RSRRs after PCI. We sought to characterize variation in hospital performance on this measure and to determine whether differences in risk-standardized rates among hospitals were explained by measurable factors including traditional measures of PCI quality, such as adherence to guideline-recommended discharge medications, and periprocedural complications.
Readmission within 30 days of discharge after percutaneous coronary intervention (PCI) is common and accounts for a large share of potentially preventable healthcare expenditures.
The National Quality Forum has endorsed hospital risk-standardized, 30-day readmission rates after PCI as a publicly reported quality measure, which may be used to determine hospital reimbursement.
This study is the first to examine variation in risk-standardized, 30-day readmission rates after PCI among hospitals.
Wide variation in these rates was observed among Massachusetts PCI-performing hospitals, very little of which was explained by differences in assessed measures of hospital quality including procedural complications and discharge medications.
Modifiable and unmodifiable factors responsible for the wide variation in short-term readmission rates after PCI between hospitals are largely unknown.
Since 2002, the Massachusetts Department of Public Health has systematically collected data on all PCI admissions performed in adults aged ≥18 years at all acute care nonfederal Massachusetts hospitals. The data are collected by trained hospital personnel using the National Cardiovascular Data Registry (NCDR) Cath-PCI data collection instrument and are submitted electronically to the Massachusetts Data Analysis Center at Harvard Medical School. Selected covariates and outcomes are audited, adjudicated, and verified.10 For this study, we initially included all patients undergoing PCI between October 1, 2005, and September 30, 2008. To obtain data on patients subsequent to discharge, including information on readmissions, we then linked data from the Massachusetts Data Analysis Center to hospital-discharge billing data collected by the Massachusetts Division of Health Care Finance and Policy.11 We excluded patients who could not be matched between the 2 datasets (n=2726), were not Massachusetts residents (n=2998), or had incomplete data (n=57). Patients not surviving to discharge were excluded from the analysis (n=589).
The remaining 36 060 adults (≥18 years of age) undergoing PCI were included in the analysis. Patients were assigned to the hospital performing the PCI. Readmissions at separate facilities occurring within 1 calendar day of discharge were considered to be transfers and were linked to form single episodes of care. These episodes are referred to as the index hospitalization.
We identified readmissions for any cause within 30 days after discharge from the index hospitalization, counting only 1 readmission for each discharge. We tabulated the primary discharge diagnosis and primary procedure associated with each readmission, based on billing codes, to identify the most common reasons for readmission, as categorized by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. Readmissions associated with any repeat revascularization, as well as revascularization of the target vessel from the prior PCI, were also identified.11
Definitions for the data elements for version 3.04 of the Cath-PCI registry are available at http://www.ncdr.com/WebNCDR/Elements.aspx. We identified a list of variables to be included in the regression model to standardize 30-day readmission rate, based on clinical relevance. Variables were categorized in the following manner: Sociodemographic information included age, sex, admission status (eg, outpatient referral, admission from emergency room, transfer), and smoking status. Medical history included body mass index, hypertension, diabetes with and without insulin use, glomerular filtration rate (GFR), and chronic lung disease. Cardiovascular history included prior MI, cerebrovascular disease, peripheral vascular disease, prior PCI, prior coronary artery bypass graft, prior congestive heart failure, and Canadian Cardiovascular Society/New York Heart Association Classification. Clinical status variables at admission included disease presentation (ST-segment–elevation MI, non–ST-segment–elevation MI, unstable angina, stable angina or atypical chest pain, no chest pain) and PCI status (elective, urgent, emergent/salvage). Estimated GFR was calculated using the abbreviated Modification of Diet in Renal Disease formula, based on creatinine assessed at the time of admission.12 We also identified certain relevant angiographic variables, including number of disease vessels, saphenous vein graft lesion, high risk lesion, and bifurcation lesion.
Post-PCI variables, as well as patient race/ethnicity and insurance status, would typically be excluded in the assessment of hospital quality to avoid giving “credit” to hospitals for inferior quality of care delivered to at-risk subgroups.13 However, these factors may be highly predictive of readmission and explain differences in hospital readmission rates. Consequently, we identified a set of “explanatory variables” as follows: Insurance status was categorized as Medicare including Medicare Advantage, state insurance including Medicaid, health maintenance organization, private insurance, and unknown. Race/ethnicity included black, Hispanic, white, and other. Device selection included stent type (drug-eluting stent versus other) and attempted use of a vascular closure device. Although stent type is unlikely to be causally related to 30-day readmission risk, because selection of stent type is a marker of unmeasured patient comorbidity, we included it as a “proxy” variable.14,15 Direct thrombin inhibitor use was defined as any use of bivalirudin, argatroban, or lepirudin before or during the PCI. Complete revascularization was defined as occurring when PCI was performed in all vessels with significant disease (>50% left main stenosis or >70% stenosis in the left anterior descending, left circumflex artery, or right coronary artery). Complications included access site and non–access site bleeding and vascular complications, periprocedural MI, lesion dissection, perforation or abrupt vessel closure, postprocedure cardiogenic shock, tamponade or heart failure, cerebrovascular accident, renal failure, and requirement for unanticipated post-PCI coronary bypass surgery. Discharge variables included use of guideline-recommended medications, such as aspirin, statins, β-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers and thienopyridines, discharge disposition (eg, home, nursing home, extended care, etc), and length of hospital stay.
Characteristics of readmitted and nonreadmitted patients were compared, using the χ2 test or t test for dichotomous and continuous variables, respectively. We first generated a nonparsimonious hierarchical logistic regression model to identify predictors of 30-day readmission and estimated individual patient risk for readmission, adjusted for demographic, clinical, and angiographic variables but excluding “explanatory” variables (hospital quality model). Model discrimination was assessed using the c-statistic and graphical displays. To account for within-hospital correlation, we estimated random-effects models that included hospital-specific intercepts. A fully Bayesian approach was adopted to account for uncertainty in all parameter estimates.
Using this model, we then generated RSRRs, based on previously defined methods.16 These rates represent a specific hospital's performance on the measure relative to an average hospital with similar case mix. We examined the variability of the point estimates of hospital risk-standardized readmission rates and the number of outlier hospitals. Outliers were defined as hospitals with 95% posterior intervals that excluded the state overall readmission rate, consistent with the definition used for other Massachusetts hospital quality measures, as well as those used elsewhere.17,18 We subsequently introduced explanatory variables into the model (explanatory model) and then reestimated hospital-adjusted readmission rates, the variance in the revised hospital-adjusted readmission rate point estimates, and the number of outlying hospitals. We calculated the relative change in variance in hospital-adjusted readmission rates between the hospital quality and the explanatory models by subtracting the variance of the estimates derived from the explanatory model from the variance of the estimate derived from the hospital quality model and then dividing the difference by the variance of the estimates derived from the hospital quality model. This value can be interpreted as the percentage of between-hospital variation in risk-standardized rates attributable to factors introduced in the explanatory model. Likewise, a change from outlier to nonoutlier status would imply that outlier status in the hospital quality model was attributable to differences in explanatory variables.
Because some readmission after PCI may have been scheduled readmissions for intervention on additional diseased coronary vessels—“staged PCIs”—we conducted a sensitivity analysis in which we attempted to exclude such readmissions. We identified staged PCI admissions as those hospitalizations during which PCI was performed and the principal discharge diagnosis did not include an acute cardiovascular diagnosis as defined in the proposed Centers for Medicare and Medicaid Services performance measure.19 We considered the following diagnoses as acute cardiovascular conditions: congestive heart failure (402.01, 402.11, 402.91, 404.01, ICD-9-CM 404.03, 404.11, 404.13, 404.91, 404.93, 428.xx), acute MI (410.xx except 410.x2), unstable angina (411.xx), arrhythmia (427.xx, except 427.5), or cardiac arrest (427.5). We then recalculated hospital risk-standardized readmission rates and estimated between-hospital variance in rates after excluding PCIs identified as being staged by this definition. We elected to consider this analysis secondary, rather than primary, due to the inability to validate the staged PCI definition within our dataset.
All statistical analyses were performed by the Massachusetts Data Analysis Center at Harvard Medical School, using the SAS Version 9.2 and WinBUGS 1.4.3 for the Bayesian models. Model convergence was confirmed for the hierarchical models by running 3 chains and computing the ratio of the variance of the estimates between chains to the variance of the estimates within a chain, with a value of 1 being indicative of convergence. Parameter estimates are based on 5000 draws, after an initial burn-in of 5000 draws.
A total of 36 060 patients undergoing PCI between September 1, 2005, and August 31, 2008, treated at 24 hospitals who survived to discharge, were included in the study. Of these, 4469 (12.4%) patients were readmitted within 30 days of discharge, at a median of 11 days. The most common principal discharge diagnoses associated with readmission were ischemic heart disease (24.5%), chest and respiratory symptoms (12.3%), heart failure (8.5%), acute MI (4.8%), procedural complications (4.3%) and cardiac dysrhythmias (3.8%), with the 10 most common diagnoses comprising 65% of all readmissions (Figure 1). Among readmitted patients, 796 (17.8%) underwent PCI and 83 (1.9%) underwent coronary artery bypass graft during the readmission hospitalization. Revascularization of the target vessel of the original PCI occurred in 8.1% of readmitted patients.
Readmitted patients differed significantly from nonreadmitted patients, including being older, more likely to be female, nonwhite, and Medicare-insured and having a higher prevalence of most cardiovascular and noncardiovascular comorbidities (Table 1). Readmitted patients also had significantly more procedural complications after the index PCI, including periprocedural MI and combined bleeding and vascular complications. They were more often transferred to other hospitals after PCI or discharged to extended-care facilities or nursing homes and were less often discharged on antiplatelet medications, β-blockers, or statins (Table 2).
Hospital Quality Model
Multivariable-adjusted predictors of readmission included GFR <30 mL/min per 1.73 m2 (odds ratio [OR], 1.88; 95% confidence interval [CI], 1.59–2.19), admission through the emergency department (OR, 1.54; 95% CI, 1.35–1.75), emergent or salvage PCI (OR, 1.48; 95% CI, 1.22–1.77), 3-vessel coronary disease (OR, 1.47; 95% CI, 1.32–1.62), and chronic lung disease (1.39; 95% CI, 1.27–1.51) (Table 3). The hospital quality risk adjustment model had moderate discrimination (c=0.666; 95% CI, 0.657–0.674).
A number of explanatory factors were associated with subsequent readmission at 30 days (Table 3). Among these, discharge to nursing home (OR, 2.09; 95% CI, 1.51–2.79) and lack of insurance/unknown insurance (OR, 2.22; 95% CI, 1.63–2.93) had strong associations with readmission risk. Other predictors of readmission included Medicare (OR, 1.51; 95% CI, 1.36–1.66) or state insurance including Medicaid (OR, 1.42; 95% CI, 1.24–1.63) and periprocedural complications, such as periprocedural MI (OR, 1.29; 95% CI, 1.05–1.56) and bleeding or vascular access complications (OR, 1.26; 95% CI, 1.08–1.46). Among discharge medications, prescription of statins (OR, 0.83; 95% CI, 0.75–0.93) and β-blockers (OR, 0.88; 95% CI, 0.77–0.99) were associated with decreased likelihood of 30-day readmission. Longer length of stay was associated with a higher readmission risk (OR, 1.02 per additional day; 95% CI, 1.02–1.03). Complete revascularization during index PCI was associated with a lower readmission risk (OR, 0.72; 95% CI, 0.64–0.81). Model discrimination after introduction of explanatory variables improved modestly (c=0.685; 95% CI, 0.678–0.694).
Hospital Readmission Rates
Unadjusted 30-day readmission rates varied among hospitals, ranging from 0% to 20.7% (median, 12.4%; interquartile range, 10.7–13.4%). After risk standardization based on the hospital quality model, variation in readmission rates ranged from 9.5% to 17.9%. There were 8 (33.3%) “outlier” hospitals (Figure 2). Of these, 4 (16.7%) hospitals had risk-standardized readmission rates that were significantly higher than expected, and 4 (16.7%) had risk-standardized rates that were significantly lower than expected.
Differences in race, insurance status, procedural complications, discharge medications, and discharge disposition explained the outlier status of 2 of the 8 outlying hospitals, 1 with higher than expected and 1 with lower than expected readmission rates. Only 10.4% of the observed variation in hospital risk-standardized readmission rates was attributable to differences in these factors.
We identified 709 readmissions associated with PCI that were not associated with an acute cardiovascular diagnosis and thus may have been readmissions for staged procedures. When such readmissions were excluded, there continued to be substantial variation in risk-standardized readmission rates, varying from 7.1% to 15.2%, with 6 outlying hospitals. Differences in race, insurance status, procedural complications, and discharge medications explained the outlier status of only 1 of the 6 outlying hospitals and accounted for only 14.8% of observed variation attributable to differences in explanatory variables. Predictors of readmission were similar, apart from complete revascularization, which was no longer associated with readmission risk after exclusion of staged PCIs (OR, 0.94; 95% CI, 0.82–1.09).
In a large observational study of patients undergoing PCI, inclusive of a broad sociodemographic spectrum, we have shown substantial variation in risk-standardized rates of readmission within 30 days of discharge among hospitals. Although a number of post-PCI factors, including procedural complications and adherence to guideline-recommended medications at discharge, were found to be associated with short-term readmission, these variables explained between 10% and 15% of the variation in risk-standardized readmission rates seen among hospitals and did not markedly improve model discrimination.
Hospital readmissions are common after discharge for a number of disease processes and are a substantial contributor to the high costs associated with hospital care.1,2,5,6,20 In a 2007 report to Congress, MedPAC found that PCI was associated with among the highest 30-day readmission rates among all hospitalizations within the Medicare population, accounting for $359 million in potentially preventable medical expenditures.6 Curtis et al9 reported that the overall readmission rate within 30 days of PCI was 14.6% within the Medicare fee-for-service population in 2005 and that there was substantial variation in crude hospital readmission rates.9 Two recent analyses, the first among residents of New York State and the second among patients at a single center in Minnesota, also described high rates of 30-day readmission after PCI and identified a number of important predictors of readmission.21,22 These included female sex, lung disease, renal failure, and long length of stay, among other factors, similar to what was found in multivariable analysis in our population. However, whether variation in hospital crude short-term readmission rates after PCI was due to differences in patient case mix or hospital quality has not been previously examined. Our results suggest that in Massachusetts, whereas some hospital variation in readmission rates is associated with differences in patient case mix, the marked variation in 30-day RSRRs adjusted for patient characteristics remains.
Several approaches to reducing readmission have been previously suggested, including reducing adverse events such as procedural complications, nosocomial infections, and bleeding; providing appropriate prescriptions to patients the time of discharge; and improving communication with patients and outpatient providers before and after discharge.6,23 Adverse safety events after surgical procedures have been demonstrated to be associated with increased rates of short-term readmission.23 Consistent with this, in our study, procedural complications—including vascular complications, bleeding, and periprocedural MI—were associated with increased rates of 30-day readmission after PCI. Various strategies known to reduce such complications exist and can be used by hospitals to reduce 30-day readmission rates. Likewise, the association between the receipt of β-blockers and statins on discharge suggests that the improved coordination of discharge medications may be an effective means to improve performance on the measure.
Although addressing procedural complications and discharge medications might reduce a given hospital's readmission rate, our analysis suggests that these factors explained less than 15% of existing differences in hospital 30-day readmission rates. For example, among the 4 hospitals identified as having worse than expected readmission rates, only 1 of those hospitals would have been reclassified as a nonoutlier if their post-PCI complication rate and discharge medication prescribing had improved to the average level of other hospitals. Defining those factors that explain between-hospital differences thus becomes a critical step to ultimately improving the value of risk-standardized readmission rates as a useful and actionable quality measure. Possible explanations for the continued between-hospital differences in RSRRs include differences in other unmeasured markers of hospital quality of care (postprocedure care, discharge coordination) as well as issues that may be beyond the control of hospitals (availability of outpatient follow-up within an appropriate time frame, differing thresholds for seeking emergency room treatment, patient willful noncompliance). Implementation of this quality measure, backed by strong financial incentives, likely would lead hospitals to devote resources to understanding reasons for readmission after PCI, develop strategies to prevent them, and ultimately improve quality of care. However, at present, without a clear understanding of the spectrum of causes for readmission or the etiology of variation in readmission rates, such quality measures may also result in inappropriate penalties for hospitals and lead to unintended and adverse consequences for patients at high risk for readmission.24 These unintended consequences could include unnecessary prolongation of initial hospital stay to minimize the chance—however small—of readmission, or a rise in the threshold for readmission, even when medically indicated, because of associated penalties. Accordingly, the impact of 30-day, all-cause RSRR as a quality metric will require further analyses as it is implemented. Such analyses should include, among others, understanding specific hospital factors that are associated with readmission risk after PCI (eg, availability of on-site cardiac surgery, operator and hospital PCI volume, percentage of PCI population that is minority or underserved) and differences in preventive and ambulatory care resources in the community, in addition to the specific processes leading up to and beyond discharge. Further efforts to understand cause-specific readmission rates rather than all-cause readmission may also lend greater insight into explaining hospital variation in readmission rates.
Our study has a number of important limitations. In these analyses, we defined outlier hospitals as those for whom the 95% posterior interval for the estimated 30-day RSRR did not include the state-wide average. Although this definition is common to other performance measures,17,18 we recognize this to be an inherently arbitrary threshold, and it is unclear what the ideal readmission rate should be. Identification of the reasons for readmission relied on administrative claims data, thus limiting our ability to clearly elucidate the circumstances leading to readmission. Next, the overall model discrimination for both the hospital quality and the explanatory model were modest, and unobserved factors may be associated with 30-day readmission and with hospital choice. We included all readmissions within the outcome measure, and while we performed a sensitivity analysis eliminating potential “staged” readmissions for repeat procedures, we were not able to validate whether our identification of staged procedures was accurate. However, it should be noted that wide variation in hospital RSRRs was noted in analyses whether or not these readmissions were included and are unlikely to be explained by misclassification of staged procedures. In addition, our finding that complete revascularization at index was no longer associated with 30-day readmission after exclusion of staged PCIs may suggest validity to the definition. Among modifiable targets for performance improvement, we were only able to include periprocedural complications and discharge medications and did not have access to other important data on transitions of care and outpatient care, including cardiac rehabilitation, that are likely to influence readmission risk. Finally, because PCI outcomes are a mandatory publicly reported measure in Massachusetts, our population is inclusive of the complete age, race, and insurance spectrum. However, the findings observed may not be generalizable to other settings, and we were not able to include readmissions that may have occurred outside of Massachusetts.
In summary, we have shown marked variation in 30-day, hospital risk-standardized readmission rates after PCI. Although we identified a number of modifiable factors associated with patient readmission including periprocedural complications and prescription of recommended medications at discharge, we were unable to account for the majority of observed variation in hospital readmission rates. Further studies to determine the factors that explain differences in hospital readmission rates and to identify which factors are modifiable are necessary to evaluate the validity of short-term readmission after PCI as a useful quality measure.
Sources of Funding
Data analysis was supported by funding from the Massachusetts Department of Public Health. Dr Yeh was supported in part by a grant from the American Heart Association (12CRP9010016).
Dr Rosenfield reports the following disclosures: Research grant: Abbott Vascular, Bard Peripheral Vascular; Consulting/Advisory Board: Abbott Vascular, Angioguard (Cordis), Boston Scientific Corp, Complete Conference Manager, Harvard Clinical Research Institute; Equity: Lumen Biomedical, Medical Stimulation Corp, and VIVA Physicians Association. Dr Mauri receives institutional research support from Abbott, Boston Scientific, Cordis, Medtronic, Eli Lilly, Daiichi Sankyo, Bristol Myers Squibb, and Sanofi-Aventis and has consulted for Cordis and Medtronic. Dr Jacobs reports being the site principal investigator for a registry sponsored by Abbott.
Guest Editor for this article was Theodore A. Bass, MD.
- Received October 12, 2011.
- Accepted February 21, 2012.
- © 2012 American Heart Association, Inc.
- Krumholz HM,
- Merrill AR,
- Schone EM,
- Schreiner GC,
- Chen J,
- Bradley EH,
- Wang Y,
- Lin Z,
- Straube BM,
- Rapp MT,
- Normand SL,
- Drye EE
- Koehler BE,
- Richter KM,
- Youngblood L,
- Cohen BA,
- Prengler ID,
- Cheng D,
- Masica AL
- Ross JS,
- Chen J,
- Lin Z,
- Bueno H,
- Curtis JP,
- Keenan PS,
- Normand SL,
- Schreiner G,
- Spertus JA,
- Vidan MT,
- Wang Y,
- Krumholz HM
Medicare Payment Advisory Commission (US). Report to the Congress: promoting greater efficiency in Medicare. Washington, DC: MedPAC; 2007.
United States., United States. Congress. House. Office of the Legislative Counsel, United States. Congress. House. Committee on Ways and Means., United States. Congress. House. Committee on Energy and Commerce., United States. Congress. House. Committee on Education and Labor. Compilation of Patient Protection and Affordable Care Act: as amended through November 1, 2010 including Patient Protection and Affordable Care Act health-related portions of the Health Care and Education Reconciliation Act of 2010. Washington: US Government Printing Office; accessed October 1, 2011.
National Quality Forum Press Release. NQF endorses patient outcome measures for high-impact conditions. 2010. http://www.qualityforum.org/News_And_Resources/Press_Releases/2010/NQF_Endorses_Patient_Outcome_Measures_for_High-Impact_Conditions.aspx. Accessed October 5, 2011.
- Shahian DM,
- Silverstein T,
- Lovett AF,
- Wolf RE,
- Normand SL
- Krumholz HM,
- Brindis RG,
- Brush JE,
- Cohen DJ,
- Epstein AJ,
- Furie K,
- Howard G,
- Peterson ED,
- Rathore SS,
- Smith SC Jr.,
- Spertus JA,
- Wang Y,
- Normand SL
- Krumholz HM,
- Wang Y,
- Mattera JA,
- Han LF,
- Ingber MJ,
- Roman S,
- Normand SL
- Shahian DM,
- Wolf RE,
- Iezzoni LI,
- Kirle L,
- Normand SL
Massachusetts Data Analysis Committee. Adult Percutaneous Coronary Intervention in the Commonwealth of Massachusetts, Fiscal Year 2009, October 1, 2008 to September 30, 2009. http://www.massdac.org/reports/pci.html. Accessed October 6, 2011.
National Quality Forum. National Voluntary Consensus Standards for Patient Outcomes: Measure Summary, Measure OT1-008-09. http://www.qualityforum.org/Projects/n-r/Patient_Outcomes/OT1-008-09_Supplement,aspx. Accessed November 10, 2010.
- Lindenauer PK,
- Bernheim SM,
- Grady JN,
- Lin Z,
- Wang Y,
- Merrill AR,
- Han LF,
- Rapp MT,
- Drye EE,
- Normand SL,
- Krumholz HM
- Hannan EL,
- Zhong Y,
- Krumholz H,
- Walford G,
- Holmes DR Jr.,
- Stamato NJ,
- Jacobs AK,
- Venditti FJ,
- Sharma S,
- King SB III.