Introduction

As the pressure on the U.S. health care system to reduce costs increases (1,2), hospitals are increasingly challenged to deliver higher quality care at lower costs (1–9). The Patient Protection and Affordable Care Act via the Bundled Payments for Care Improvement (BPCI) initiative applies direct pressure to hospitals to have both financial and performance accountability for care (3,4,10–15). These and future health care delivery models will provide strong incentives for hospitals to improve the efficiency of the care they provide. Percutaneous coronary intervention (PCI) is a prototypical example of shifting payment strategies. PCI was once primarily conducted, and paid for, as an inpatient procedure, until the Recovery Audit Contractor program began scrutinizing the clinical justification for paying for PCI as an inpatient, as opposed to an outpatient, procedure. Recently the Centers for Medicare and Medicaid Services (CMS) created and implemented the “2 midnight rule” to pay an inpatient fee only for PCI patients who spend at least 2 nights in the hospital, creating a strong incentive for hospitals to consider same-day discharge (SDD) to decrease their costs (13–19). In the face of decreasing reimbursements (1,2), existing care pathways need to be critically examined to define the safest and most efficient strategies.

PCI procedures are an important contributor to hospital costs in the United States. PCI procedures are performed in 600,000 patients annually (20) and have the highest aggregate costs of all cardiovascular procedures (including cardiac surgery) and the third highest aggregate cost of any surgical procedure, estimated at approximately $10 billion annually (10–14,21–24). Recently both transradial intervention (TRI) and SDD after PCI have been introduced as approaches that can substantially reduce hospitals’ cost to perform PCI. However, no study to date has examined the independent impact of various care pathways, including those involving TRI and SDD after PCI, on hospital costs. To address this gap in knowledge, we used data from Medicare claims files linked to the American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR) CathPCI Registry to examine the costs and outcomes associated with TRI and SDD from a hospital’s perspective.

Methods

Population

The study population consisted of Medicare fee-for-service beneficiaries undergoing PCI between July 1, 2009, and December 31, 2012. Both inpatient and outpatient PCI procedures were included in this analysis. The inpatient PCI population was identified using the CMS Inpatient Research Identifiable Files (RIF) (International Classification of Diseases-Ninth Revision codes 00.66, 36.06, and 36.07). Outpatient procedures were identified using the CMS Outpatient RIF and the corresponding Current Procedural Terminology codes 92982, 92980, G0290, 92981, 92984, G0290, and G0291. A waiver of the requirement to obtain informed consent and Institutional Review Board approval was obtained from the Saint Luke’s Hospital Institutional Review Board.

Linkage of medicare claims data to ACC NCDR CathPCI registry

The time period for this analysis was the third quarter of 2009 through the third quarter of 2012. To create a national dataset with detailed clinical and claims information, we linked the Medicare claims data to the ACC NCDR CathPCI Registry (25). Unique patient identifiers, such as Social Security number, were not available for use in the ACC NCDR CathPCI Registry to perform exact matching. We therefore performed a probabilistic match using the following identifiers: hospital National Provider Identifier number, date of birth, sex, and admission date (25). We required exact matches for hospital National Provider Identifier number, date of birth, and sex. The date of admission could vary by ±1 day. There were 831,357 CMS claims available for matching. Using a deterministic matching strategy, 749,366 of the 831,357 possible CMS claims (93%) were matched to the CathPCI Registry procedures. Of these 749,366 PCI procedures, the following procedures were excluded: procedures clearly indicated for inpatient stay (ST-segment elevation myocardial infarction [n = 90,050, 12.02%], non–ST-segment elevation myocardial infarction [n = 380,051, 49.16%], chronic total occlusion [n = 7,565, 0.98%], intra-aortic balloon pump [n = 1,254, 0.16%], cardiogenic shock [n = 1,002, 0.08%], and cardiac arrest within 24 h [n = 590, 0.08%]), salvage PCI (n = 49 [0.01%]), patients who died in-hospital within 24 h of PCI (n = 53 [0.01%]), beneficiaries with missing variables for whom propensity scores could not be calculated (n = 10,380 [1.34%]), and procedures for which cost variables were missing (n = 2,071 [0.27%]) (Figure 1). The final cohort available for analysis comprised 279,987 procedures.

Ascertainment of key data elements

Determination of arterial access, transfemoral or transradial, was derived from the ACC NCDR Cath PCI Registry. When more than 1 access site was used, the primary access site for the PCI was classified as the access site used to perform the majority of the PCI. Determination of SDD was made from the ACC NCDR Cath PCI Registry when discharge occurred on the date of the PCI procedure.

Estimation of cost

The primary outcome was hospital costs associated with PCI. When possible, we used hospital-level, cost center–specific cost-to-charge ratios to estimate the costs associated with PCI episodes of care. Cost was estimated using data from RIF using standard methodology described in detail previously (26). Briefly, using inpatient and outpatient RIF, beneficiary claims were collapsed into single PCI episodes of care. Charges were derived from hospital revenue codes as reported to Medicare. Cost-to-charge ratios for each revenue code are not publicly available for each hospital. However, cost-to-charge ratios are publically available by cost center using forms 2552-96 and 2552-10, worksheet C. “Cost centers” are groupings of similar resources used for services during an episode of care ( https://www.cms.gov/Research-Statistics-Data-and-Systems/Downloadable-Public-Use-Files/Cost-Reports/Hospital-2010-form.html). When hospital revenue codes were mapped to specific cost centers, those overall cost-to-charge ratios were used to estimate costs. A crosswalk of revenue codes to cost centers was obtained from the 2009–2012 documents ( http://www.dartmouthatlas.org/tools/downloads.aspx?tab=39). When revenue codes could not be mapped to cost centers, the overall hospital cost-to-charge ratio was used to estimate costs. Once the appropriate cost-to-charge ratio was determined, the charges for an episode of care for each cost center were multiplied by the cost-to-charge ratio. These hospital costs were summed over the admission and normalized across hospitals to the national average by applying the Dartmouth Atlas of Health Care 2012 Hospital Referral Region level price, age, sex, and race adjustment ( http://www.dartmouthatlas.org/tools/downloads.aspx). Costs were inflated to 2014 dollars by using the medical consumer price index. To remove extreme outliers, individual hospital costs were Winsorized by trimming to the 99th percentile (27,28).

Clinical outcomes

Secondary outcomes of interest included post-PCI bleeding, vascular complications, blood transfusions, and length of stay (LOS). Post-PCI bleeding was defined according to the CathPCI Registry bleeding definition and consisted of: 1) suspected bleeding with transfusion; 2) a decrease in hemoglobin of >3.0 g/dl; or 3) a procedural intervention to correct the bleeding event. Vascular complications collected by the CathPCI Registry included access-site occlusion, peripheral embolization, dissection, pseudoaneurysm formation, arteriovenous fistula, or other vascular access-site complications requiring intervention.

Statistical analysis

Demographic data are described as mean ± SD for continuous variables and as numbers for categorical variables. To obtain independent costs for each arterial access site and SDD, and to mitigate selection bias and confounding, we developed 3 separate propensity score models as follows: 1) a logistic regression model to predict use of TRI (vs. transfemoral intervention [TFI]); 2) a logistic regression model to predict SDD (vs. non-same-day discharge [NSDD]); and 3) a multinomial logistic regression model to obtain predicted probabilities of TRI SDD, TFI SDD, and TRI NSDD, with TFI NSDD as the comparator for all (29–34). All 3 propensity score models included the following patient-level variables: age, sex, race, smoking status, diabetes, hypertension, dyslipidemia, a family history of coronary artery disease, prior myocardial infarction, prior congestive heart failure, estimated glomerular filtration rate, prior valve surgery, prior PCI, prior coronary artery bypass graft surgery, current dialysis, prior cerebrovascular disease, prior peripheral vascular disease, chronic lung disease, PCI indication (elective or unstable angina), procedure status (urgent or elective), severity of angina (angina class), current heart failure (heart failure symptoms in the prior 2 weeks), left ventricular dysfunction, perioperative evaluation, prior cardiac arrest, number of diseased vessels, probability of bleeding according to the NCDR bleeding risk model (a composite of several variables), probability of mortality according to the NCDR mortality risk model (which includes several additional variables), probability of restenosis according to the NCDR target vessel revascularization risk prediction model (which also includes several additional variables), hospital type (teaching hospital, public hospital, urban vs. suburban vs. rural hospital, government hospital, private vs. community hospital, university hospital), and regional census divisions. Variables such as use of anticoagulant agents and LOS were not included in the propensity model, because they are directly related to access site and in the causal pathway of costs. To estimate the independent costs of access site (TRI vs. TFI), a generalized linear cost model was developed using inverse probability weighting (IPW) (13), with total hospital cost as the dependent variable and TRI (vs. TFI) as the independent variable. To estimate the independent costs of SDD versus NSDD, a second, separate generalized linear cost model was developed using IPW (13), with total hospital cost as the dependent variable and SDD (vs. NSDD) as the independent variable. These 2 models yielded the adjusted costs of TRI versus TFI and SDD versus NSDD. To estimate the independent costs of the various combinations of PCI care pathways, we developed a third generalized linear cost model, with total hospital cost as the dependent variable and the various combination of PCI care pathways as the independent variables, with predicted probabilities of these care pathway variables (obtained from the multinomial propensity model) included as adjustment covariates. This modeling approach allowed us to obtain independent costs associated with the various care pathways while adjusting for factors that influence access-site selection and SDD status selection.

In the IPW models, individual weights were stabilized to account for the effect of extreme weights in the model (13). To ascertain the adequacy of the model with IPW, standardized differences in covariate imbalances with and without the IPW were calculated for the first 2 models (Figures 2 and 3, Table 1) (14–16). The following variables remained unbalanced after IPW adjustment, as defined by a maximal standard difference of >10%: PCI indication, census division, and mortality risk; these were included as additional covariates in the final cost and outcome models. Costs were trimmed back to the 99th percentile to remove extreme outliers (27,28). Then bootstrapping using 1,000 replicates with replacement was applied to the models to account for skewness in the data (35) as follows: we sampled from 1 group, calculated that group’s mean and its 95% confidence interval [CI], sampled from a comparator group, calculated that group’s mean and its 95% CI, then calculated the difference between those means. This was repeated 1,000 times. The differences obtained themselves had a distribution and a variance, from which we estimated the 95% CI. We report the mean and 95% CI for costs of various care pathways and the cost differences between the means of various care pathways.

Secondary outcomes

Other outcomes of interest included post-PCI bleeding, vascular complications, blood transfusion, and LOS. Adjusted differences were estimated using an IPW generalized linear regression model for LOS and an IPW logistic regression model for bleeding.

Budget impact analysis

We used the mean costs of the various care pathways to estimate the impact of converting from TFI NSDD (the current most common care pathway) to TRI SDD, TRI NSDD, and TFI SDD pathways for a hypothetical hospital performing 1,000 elective PCI procedures annually and for the country, performing 600,000 PCI procedures annually (36).

Statistical significance for all analyses was defined as p < 0.05. All analyses were performed at Saint Luke’s Mid America Heart Institute using SAS version 9.3 (SAS Institute, Cary, North Carolina).

Results

Demographic and descriptive statistics

Of the 279,987 PCI procedures, TRI was performed in 25,301 (9.0%), and SDD occurred in 14,812 (5.3%). TRI and SDD occurred in 3,424 (1.2%), TFI and SDD occurred in 11,388 (4.1%), TRI and NSDD occurred in 21,877 (7.8%), and TFI and NSDD was the most common post-PCI care pathway, occurring in 243,298 (86.9%). The study population is depicted in Figure 1. Clinical demographics and hospital and procedural characteristics of the cohort are described in Table 1. In general, TRI and SDD patients were younger, were more often white men, and had a lower prevalence of cardiovascular risk factors and comorbidities. Unfractionated heparin was more frequently used in TRI and SDD patients, and bivalirudin was used less often. Procedural and fluoroscopy times were slightly longer with TRI, while the contrast volume used was lower with TRI.

Propensity score model and IPW adjustment

Figures 2 and 3 depict the standardized differences in all variables included in the propensity score model after IPW adjustment for both TRI and SDD.

Hospital costs

The total unadjusted PCI costs associated with TRI were $14,316 ± 9,089 versus $15,866 ± 10,140 for TFI, which is a difference of $1,550 favoring TRI (p < 0.001). After risk adjustment, the difference was $916 (95% CI: $778 to $1,035; p < 0.001) favoring TRI.

The total unadjusted PCI costs for SDD were $12,449 ± 8,028 versus $15,909 ± 10,129 for NSDD, a difference of $3,460 favoring SDD (p < 0.001). After risk adjustment, the difference was $3,502 (95% CI: $3,347 to $3,648; p < 0.001) favoring SDD.

Adjusted costs of PCI by various care pathway groups are presented in Table 2. The differences in the adjusted PCI costs across these care pathways are presented in Table 3. A TRI SDD pathway was consistently associated with cost savings when compared with any other pathway (Table 3). The largest difference in adjusted costs was observed between a TRI SDD pathway, which was the least frequently used pathway, and a TFI NSDD pathway (the current most common pathway), which resulted in $3,689 in cost savings to hospitals favoring TRI SDD (Table 3).

Bleeding and LOS outcomes

In-hospital bleeding occurred overall in 2.8% of patients, 1.4% in TRI and 3.0% in TFI procedures (p < 0.001). Similarly, the rates of transfusion and other vascular complications were also lower in the TRI versus TFI groups. After propensity adjustment, these differences remained statistically significant (Table 4).

LOS was 2.2 ± 4.4 days in the TRI group compared with 2.4 ± 3.5 days in the TFI group, a difference of 0.2 days favoring TRI (p < 0.001), and persisted after risk adjustment.

Budget impact analysis

For a hypothetical hospital performing 1,000 elective PCI procedures annually, the impact of converting from TFI NSDD (the current most common pathway of care) to TRI SDD, TRI NSDD, and TFI SDD pathways is shown in Figure 4. A relatively small 30% conversion from the current pathway of TFI NSDD to TRI SDD could potentially save $1 million annually. Hospitals shifting practice by even 100 cases every year may save up to $350,000 (Figure 4). Approximately 600,000 PCI procedures are performed in the United States annually (20,36), one-half of which are elective PCI procedures, eligible for SDD. A 30% conversion from the current pathway of TFI NSDD to TRI SDD could potentially save U.S. hospitals approximately $332 million annually.

Discussion

Summary of findings

In this linked analysis from the CMS and ACC NCDR CathPCI Registry, we have identified the hospital costs of PCI associated with alternative clinical care pathways in a nationally representative population. We found that: 1) TRI costs were $916 lower compared with TFI; 2) SDD costs were $3,500 lower than NSDD costs; 3) a strategy of TRI access coupled with SDD cost nearly $3,700 less than the current, most frequently used pathway of TFI NSDD; 4) TRI access was associated with less bleeding, fewer vascular complications, and less transfusions than TFI; and 5) conversion from the existing most common care pathway of TFI NSDD to TRI SDD could potentially yield large cost savings to U.S. hospitals.

Clinical implications

Our results have important cost implications for hospitals and the U.S. health system. Hospitalization costs for Medicare beneficiaries are on the rise. The hospitalization cost for Medicare beneficiaries is higher ($12,200) than that of privately insured patients ($9,700) (23), and the number of hospitalizations billed to Medicare has been increasing at a greater rate than that of privately insured patients (22). For PCI hospitalization specifically, increasing intensity of services (resources and cost per day) and increasing LOS were major drivers for growth in the aggregate cost (21). Recent data from 2012 demonstrate that Medicare paid $5.9 billion for short inpatient stays, an average of $5,142 per stay, while it paid only $2.6 billion for observation stays, an average of $1,741 per stay (13). A PCI procedure was ranked as the second most common reason contributing to the high costs of short inpatient stays, the first being chest pain (13). The average payment difference between a short inpatient and an observation stay for PCI was $2,267, which was among the highest differences in cost for any diagnosis between a short inpatient stay and an observation stay (13). CMS has created time-based rules for determining inpatient status, and the final rule was published in August 2013 (15,19). Under this “2-midnight rule,” inpatient admission is appropriate only if a physician expects a patient’s stay to span at least 2 midnights (13–15,19). With the implementation of the 2-midnight rule (13–15,19), more PCI patients will be reclassified as outpatients (13) and billed under outpatient observation rather than inpatient status (13,15–19). The current practice of classifying a PCI patient as an outpatient but not discharging the patient on the same hospital day is anticipated to result in higher cost in the range of $2,000 to $4,000 per PCI to hospitals (13,15–19). By providing a detailed analyses of hospital costs across various care pathways for PCI, we have identified a mechanism to reduce hospital losses with the implementation of new rules. Our findings show that TRI combined with SDD may offer a tremendous potential for hospitals across the country to reduce PCI costs.

Similarly, the BPCI initiative, begun in January 2013 by CMS through its Innovation Center authority, which was created by the Patient Protection and Affordable Care Act (37–40), seeks to improve health care delivery and ultimately to reduce costs by allowing providers to enter into pre-negotiated payment arrangements that include financial and performance accountability for a clinical episode in which a risk-and-reward calculus must be determined. As participation of hospitals and practices under these alternative payment models increases, providers will need to embrace these changes and identify areas of opportunity to maintain a competitive advantage. This study of index PCI hospitalization cost across various PCI care pathways should guide hospitals in reducing costs of index PCI care and conserve resources that could be reallocated more efficiently for downstream costs of care in the era of bundled payments. For U.S. hospitals, our budget impact analysis showed that even a relatively small 30% conversion from the current pathway of TFI NSDD to TRI SDD could result in large savings. These potential savings are meaningful and offer insight to hospitals to help them modify staffing patterns or initiate SDD protocols.

The association of TRI with fewer complications is consistent with previously published studies. Our study notes fewer bleeding complications associated with TRI compared with TFI. Both the direction and magnitude of reduced bleeding in this study are also consistent with prior randomized and observational studies comparing TRI with TFI (41–44). Particularly noteworthy, the observed bleeding reduction in the TRI group was achieved despite frequent use of other bleeding avoidance strategies (e.g., 51% bivalirudin use, 58% closure devices) in the TFI group. Although the feasibility of TRI in older patients has been challenged, several studies have established the success of TRI in older patients and those with complex coronary anatomy, including bifurcations, bypass grafts, left main disease, long lesions, and calcified vessels. Our findings confirm lower complication rates in a large national population and establish that better outcomes may be achieved at a lower cost.

Study limitations

First, the observational nature of these data may be subject to selection bias and unmeasured confounding, despite rigorous exclusions and propensity score methods with IPW to adjust for confounding associated with TRI use and SDD. Propensity score methods do not account for unmeasured confounding or selection bias (45–47).

Second, the costs demonstrated here are from the hospital perspective. A broader societal-level perspective may be preferable, as downstream costs or savings associated with TRI are possible.

Third, costs were obtained from claims data by applying hospital and cost center–specific cost-to-charge ratios. No detailed costing of actual resources, direct and indirect costs of care have been obtained.

Fourth, CMS claims were linked to the CathPCI Registry via matching techniques, which was not a perfect match. However, we were able to match 95% of the observations. It should be noted that the CathPCI Registry only captures data on the percutaneous entry site used for the majority of the procedure. Therefore, some variation may exist in reported rates of TRI and TFI when an operator switched entry sites during a given procedure, although we believe this occurred infrequently.

Conclusions

In this nationally representative dataset of Medicare beneficiaries, both TRI and SDD were associated with significantly lower costs compared with TFI and NSDD when examined from a hospital perspective. The magnitude of cost savings, exceeding $900 for TRI PCI and $3,500 for TRI PCI with SDD, may be appealing to hospitals that consider adopting TRI and SDD programs for PCI.

• 1. Banham R. Relieving the pressure. Healthcare Finance. October 2, 2014. Available at: http://www.healthcarefinancenews.com/news/relieving-pressure?page=0. Accessed December 19, 2016.

  Google Scholar

• 2. Brino A. 2015 Medicare rates increase financial pressure on hospitals. Healthcare Finance. May 1, 2014. Available at: http://www.healthcarefinancenews.com/news/2015-medicare-rates-add-more-pressure. Accessed December 19, 2016.

  Google Scholar

• 3. Centers for Medicare and Medicaid Services. Hospital Value-Based Purchasing Program. March 1, 2013. Available at: http://www.cms.gov/Outreach-and-Education/Medicare-Learning-Network-MLN/MLNProducts/downloads/Hospital_VBPurchasing_Fact_Sheet_ICN907664.pdf. Accessed December 19, 2016.

  Google Scholar

• 4. Centers for Medicare and Medicaid Services. Bundled Payments for Care Improvement initiative fact sheet. July 31, 2014. Available at: http://www.cms.gov/Newsroom/MediaReleaseDatabase/Fact-sheets/2014-Fact-sheets-items/2014-07-31.html. Accessed December 19, 2016.

  Google Scholar

• 5. Committee on the Learning Health Care System in America, Institute of Medicine : Best care at lower cost: the path to continuously learning health care in America . Washington, District of Columbia: National Academies Press2014.

  Google Scholar

• 6. The Commonwealth Fund. Better care at lower cost: is it possible? October 1, 2014. Available at: http://www.healthcarefinancenews.com/news/relieving-pressure?page=0. Accessed December 19, 2016.

  Google Scholar

• 7. Reinke T. : "Medicare gets serious about payment cuts". Manag Care 2010; 19: 40.

  MedlineGoogle Scholar

• 8. Steele J.R. and Reilly J.D. : "Bundled payments: bundled risk or bundled reward?". J Am Coll Radiol 2010; 7: 43.

  CrossrefMedlineGoogle Scholar

• 9. "CMS continues to shift emphasis to quality of care". Hosp Case Manag 2012; 20: 150.

  MedlineGoogle Scholar

• 10. "Medicare program; hospital inpatient prospective payment systems for acute care hospitals and the long-term care hospital prospective payment system and fiscal year 2013 rates; hospitals’ resident caps for graduate medical education payment purposes; quality reporting requirements for specific providers and for ambulatory surgical centers. final rule". Fed Regist 2012; 77: 53257.

  MedlineGoogle Scholar

• 11. "Medicare program; hospital inpatient prospective payment systems for acute care hospitals and the long-term care hospital prospective payment system and fiscal year 2014 rates; quality reporting requirements for specific providers; hospital conditions of participation; payment policies related to patient status. Final rules". Fed Regist 2013; 78: 50495.

  MedlineGoogle Scholar

• 12. "Medicare program; hospital inpatient prospective payment systems for acute care hospitals and the long-term care hospital prospective payment system and fiscal year 2015 rates; quality reporting requirements for specific providers; reasonable compensation equivalents for physician services in excluded hospitals and certain teaching hospitals; provider administrative appeals and judicial review; enforcement provisions for organ transplant centers; and electronic health record (EHR) incentive program. Final rule". Fed Regist 2014; 79: 49853.

  MedlineGoogle Scholar

• 13. U.S. Department of Health and Human Services, Office of Inspector General. Hospitals’ use of observation stays and short inpatient stays for Medicare beneficiaries. OEI-02-12-00040. July 29, 2013. Available at: http://oig.hhs.gov/oei/reports/oei-02-12-00040.asp. Accessed December 19, 2016.

  Google Scholar

• 14. Medicare Payment Advisory Commission. Hospital inpatient and outpatient services. In: MedPAC report to the Congress: Medicare payment policy. March 2014. Available at: http://www.medpac.gov/docs/default-source/reports/mar14_ch03.pdf?sfvrsn=0. Accessed December 19, 2016.

  Google Scholar

• 15. Centers for Medicare and Medicaid Services inpatient prospective payment system 1599-F. Fiscal year 2014 final rule. August 19, 2013. Available at: http://www.gpo.gov/fdsys/pkg/FR-2013-08-19/pdf/2013-18956.pdf 2014. Accessed December 19, 2016.

  Google Scholar

• 16. Sheehy A.M. : "Dedicated observation unit for patients with “observation status”—reply". JAMA Intern Med 2014; 174: 301.

  CrossrefMedlineGoogle Scholar

• 17. Sheehy A.M., Caponi B., Gangireddy S.et al. : "Observation and inpatient status: clinical impact of the 2-midnight rule". J Hosp Med 2014; 9: 203.

  CrossrefMedlineGoogle Scholar

• 18. Worth T. : "Two-midnight rule a double-edged sword". Healthc Fin News 2014; 8: 12.

  Google Scholar

• 19. 2 Midnight inpatient admission guidance & patient status reviews for admissions on or after October 1 2013. Available at: https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/Medical-Review/Downloads/QAsforWebsitePosting_110413-v2-CLEAN.pdf. Accessed June 1, 2016.

  Google Scholar

• 20. Go A.S., Mozaffarian D., Roger V.L.et al. : "Heart disease and stroke statistics—2014 update: a report from the American Heart Association". Circulation 2014; 129: e28.

  MedlineGoogle Scholar

• 21. Pfuntner A, Levit K, Elixhauser A. Components of cost increases for inpatient hospital procedures 1997–2009. Statistical Brief #133. Available at: https://www.hcup-us.ahrq.gov/reports/statbriefs/sb133.pdf. Accessed December 19, 2016.

  Google Scholar

• 22. Weiss AJ, Elixhauser A, Pfuntner A, Levit K, Elixhauser A. Overview of Hospital Stays in the United States 2012. Statistical Brief #180. Available at: http://www.hcup-us.ahrq.gov/reports/statbriefs/sb180-Hospitalizations-United-States-2012.pdf. Accessed December 19, 2016.

  Google Scholar

• 23. Moore B, Levit K, Elixhauser A, et al. Costs for hospital stays in the United States 2012. Statistical Brief #181. Available at: https://www.hcup-us.ahrq.gov/reports/statbriefs/sb181-Hospital-Costs-United-States-2012.jsp. Accessed December 19, 2016.

  Google Scholar

• 24. Weiss AJ, Elixhauser A, Andrews RM, et al. Characteristics of operating room procedures in U.S. hospitals 2011. Statistical Brief #170. Available at: https://www.hcup-us.ahrq.gov/reports/statbriefs/sb170-Operating-Room-Procedures-United-States-2011.jsp. Accessed December 19, 2016.

  Google Scholar

• 25. Brennan J.M., Peterson E.D., Messenger J.C.et al. : "Linking the National Cardiovascular Data Registry CathPCI Registry with Medicare claims data: validation of a longitudinal cohort of elderly patients undergoing cardiac catheterization". Circ Cardiovasc Qual Outcomes 2012; 5: 134.

  CrossrefMedlineGoogle Scholar

• 26. Dalton K, Freeman S, Bragg A. Refining cost to charge ratios for calculating APC and DRG relative payment weights. Interim report: prepared for the Centers for Medicare & Medicaid Services. RTI Project Number 0209853.008. Research Triangle Park, NC: RTI International; 2012.

  Google Scholar

• 27. Mihaylova B., Briggs A., O’Hagan A. and Thompson S.G. : "Review of statistical methods for analysing healthcare resources and costs". Health Econ 2011; 20: 897.

  CrossrefMedlineGoogle Scholar

• 28. Briggs A. and Gray A. : "The distribution of health care costs and their statistical analysis for economic evaluation". J Health Serv Res Policy 1998; 3: 233.

  CrossrefMedlineGoogle Scholar

• 29. Joffe M.M. and Rosenbaum P.R. : "Invited commentary: propensity scores". Am J Epidemiol 1999; 150: 327.

  CrossrefMedlineGoogle Scholar

• 30. Little R.J. and Rubin D.B. : "Causal effects in clinical and epidemiological studies via potential outcomes: concepts and analytical approaches". Annu Rev Public Health 2000; 21: 121.

  CrossrefMedlineGoogle Scholar

• 31. Rubin D.B. : "Estimating causal effects from large data sets using propensity scores". Ann Intern Med 1997; 127: 757.

  CrossrefMedlineGoogle Scholar

• 32. Imbens G.W. : "The role of the propensity score in estimating dose-response functions". Biometrika 2000; 87: 706.

  CrossrefGoogle Scholar

• 33. Rosenbaum P.R. and Rubin D.B. : "The central role of the propensity score in observational studies for causal effects". Biometrika 1983; 70: 41.

  CrossrefGoogle Scholar

• 34. Rosenbaum P.R. and Rubin D.B. : "Constructing a control group using multivariate matched sampling methods that incorporate the propensity score". Am Stat 1985; 39: 33.

  Google Scholar

• 35. Briggs A.H., Wonderling D.E. and Mooney C.Z. : "Pulling cost-effectiveness analysis up by its bootstraps: a non-parametric approach to confidence interval estimation". Health Econ 1997; 6: 327.

  CrossrefMedlineGoogle Scholar

• 36. Roger V.L., Go A.S., Lloyd-Jones D.M.et al. : "Heart disease and stroke statistics—2012 update: a report from the American Heart Association". Circulation 2012; 125: e2.

  MedlineGoogle Scholar

• 37. Greenwald A.S., Bassano A., Wiggins S. and Froimson M.I. : "Alternative reimbursement models: bundled payment and beyond: AOA critical issues". J Bone Joint Surg Am 2016; 98: e45.

  CrossrefMedlineGoogle Scholar

• 38. Centers for Medicare and Medicaid Services. Bundled Payments for Care Improvement (BPCI) initiative: general information. Available at: https://innovation.cms.gov/initiatives/bundled-payments/. Accessed December 19, 2016.

  Google Scholar

• 39. U.S. Department of Health and Human Services. Read the law: the Affordable Care Act, section by section. Title III, Part III, Sec. 3021. Establishment of Center for Medicare and Medicaid Innovation within CMS. Available at: https://www.hhs.gov/healthcare/about-the-law/read-the-law/. Accessed December 19, 2016.

  Google Scholar

• 40. U.S. Department of Health and Human Services. Better, smarter, healthier: in historic announcement, HHS sets clear goals and timeline for shifting Medicare reimbursements from volume to value. Available at: https://www.hhs.gov/about/news/2015/01/26/better-smarter-healthier-in-historic-announcement-hhs-sets-clear-goals-and-timeline-for-shifting-medicare-reimbursements-from-volume-to-value.html. Accessed December 19, 2016.

  Google Scholar

• 41. Amin A.P., House J.A., Safley D.M.et al. : "Costs of transradial percutaneous coronary intervention". J Am Coll Cardiol Intv 2013; 6: 827.

  View ArticleGoogle Scholar

• 42. Jolly S.S., Amlani S., Hamon M., Yusuf S. and Mehta S.R. : "Radial versus femoral access for coronary angiography or intervention and the impact on major bleeding and ischemic events: a systematic review and meta-analysis of randomized trials". Am Heart J 2009; 157: 132.

  CrossrefMedlineGoogle Scholar

• 43. Jolly S.S., Yusuf S., Cairns J.et al. : "Radial versus femoral access for coronary angiography and intervention in patients with acute coronary syndromes (RIVAL): a randomised, parallel group, multicentre trial". Lancet 2011; 377: 1409.

  CrossrefMedlineGoogle Scholar

• 44. Marso S.P., Amin A.P., House J.A.et al. : "Association between use of bleeding avoidance strategies and risk of periprocedural bleeding among patients undergoing percutaneous coronary intervention". JAMA 2010; 303: 2156.

  CrossrefMedlineGoogle Scholar

• 45. Stukel T.A., Fisher E.S., Wennberg D.E., Alter D.A., Gottlieb D.J. and Vermeulen M.J. : "Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods". JAMA 2007; 297: 278.

  CrossrefMedlineGoogle Scholar

• 46. Austin P.C. : "An introduction to propensity score methods for reducing the effects of confounding in observational studies". Multivariate Behav Res 2011; 46: 399.

  CrossrefMedlineGoogle Scholar

• 47. Austin P.C. : "The relative ability of different propensity score methods to balance measured covariates between treated and untreated subjects in observational studies". Med Decis Making 2009; 29: 661.

  CrossrefMedlineGoogle Scholar

Footnotes