Key Points
Question
What is the contemporary US practice of same-day discharge (SDD) after elective percutaneous coronary intervention (PCI) with respect to the incidence, variation, trends, costs, and safety outcomes?
Findings
In this cohort study that included 672 470 elective PCIs across 493 US hospitals from 2006 to 2015, SDD occurred infrequently (3.5%) with a substantial (382%) hospital variation. However, SDD was safe in the short and long term and was associated with large savings (>$5000); additionally, cost savings were attributed to reduced supply and room and board costs.
Meaning
Greater and more consistent use of SDD could increase the overall value of PCI care and save US hospitals approximately $577 million in costs if adopted in the United States in the bundled payment era.
Abstract
Importance
Same-day discharge (SDD) after elective percutaneous coronary intervention (PCI) is associated with lower costs and preferred by patients. However, to our knowledge, contemporary patterns of SDD after elective PCI with respect to the incidence, hospital variation, trends, costs, and safety outcomes in the United States are unknown.
Objective
To examine (1) the incidence and trends in SDD; (2) hospital variation in SDD; (3) the association between SDD and readmissions for bleeding, acute kidney injury (AKI), acute myocardial infarction (AMI), or mortality at 30, 90, and 365 days after PCI; and (4) hospital costs of SDD and its drivers.
Design, Setting, and Participants
This observational cross-sectional cohort study included 672 470 patients enrolled in the nationally representative Premier Healthcare Database who underwent elective PCI from 493 hospitals between January 2006 and December 2015 with 1-year follow-up.
Exposures
Same-day discharge, defined by identical dates of admission, PCI procedure, and discharge.
Main Outcomes and Measures
Death, bleeding requiring a blood transfusion, AKI and AMI at 30, 90, or 365 days after PCI, and costs from hospitals’ perspective, inflated to 2016.
Results
Among 672 470 elective PCIs, 221 997 patients (33.0%) were women, 30 711 (4.6%) were Hispanic, 51 961 (7.7%) were African American, and 491 823 (73.1%) were white. The adjusted rate of SDD was 3.5% (95% CI, 3.0%-4.0%), which increased from 0.4% in 2006 to 6.3% in 2015. We observed substantial hospital variation for SDD from 0% to 83% (median incidence rate ratio, 3.82; 95% CI, 3.48-4.23), implying an average (median) 382% likelihood of SDD at one vs another hospital. Among SDD (vs non-SDD) patients, there was no higher risk of death, bleeding, AKI, or AMI at 30, 90, or 365 days. Same-day discharge was associated with a large cost savings of $5128 per procedure (95% CI, $5006-$5248), driven by reduced supply and room and boarding costs. A shift from existing SDD practices to match top-decile SDD hospitals could annually save $129 million in this sample and $577 million if adopted throughout the United States. However, residual confounding may be present, limiting the precision of the cost estimates.
Conclusions and Relevance
Over 2006 to 2015, SDD after elective PCI was infrequent, with substantial hospital variation. Given the safety and large savings of more than $5000 per PCI associated with SDD, greater and more consistent use of SDD could markedly increase the overall value of PCI care.
This cohort study examines the incidence, trends, and hospital variation in same-day discharge and its association with readmissions for bleeding, acute kidney injury, acute myocardial infarction, or mortality in the United States.
Introduction
Elective percutaneous coronary intervention (PCI) is common in the United States, performed in approximately half of the 600 000 PCI procedures that are conducted annually.1 With the increasing pressure on hospitals to improve the quality and value of their services, reducing the costs of elective PCI is an important opportunity to explore. In fact, alternative payment models, such as the Centers for Medicare and Medicaid Services episode payment models, commonly known as “bundled payments,” are influencing hospitals to prepare for the shift in reimbursements from “payment for volume” to “payment for value.”2
Same-day discharge (SDD) after elective PCI is a potential strategy for improving the value of PCI as it is associated with greater patient satisfaction while simultaneously reducing costs.3,4,5,6,7 Despite observational and randomized data that demonstrate the safety of SDD, prior studies from 2004 to 2008 and 2009 to 2013 suggest a relatively modest uptake of this practice in the United States.8,9 These results are not surprising, as, to our knowledge, there have been few systematic efforts made toward implementing SDD after elective PCI, although emerging payments models may create an urgency to adopt this practice if it is safe and financially beneficial to hospitals. While our prior work has shown that the cost savings from SDD can be substantial, a contemporary analysis of the incidence, hospital variation, trends, and costs, the source of the cost savings and safety outcomes that are associated with SDD is needed to define the potential missed opportunity of adopting SDD and for improving the value of PCI. Therefore, we designed this large, nationally representative study with the following objectives: (1) to identify the contemporary incidence and temporal trends in SDD after elective PCI, (2) to identify the hospital variation in the practice of SDD after accounting for hospital case mix, (3) the hospital costs associated with SDD, and the sources of cost savings attributable to SDD, and (4) to compare the rate of readmissions for bleeding, acute kidney injury (AKI), acute myocardial infarction (AMI), and mortality at 30, 90, and 365 days after index PCI among SDD and non-SDD (NSDD) patients.
Methods
Study Participants
We used the Premier Healthcare Database (https://www.premierinc.com/), which is an administrative claims database representing approximately 20% of all acute care hospitalizations in the United States for more than 15 years and contains sociodemographics, comorbidities, interventional procedures, medications, costs, and outcomes that are based on International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) codes for diagnoses and procedures as reported by the contributing hospitals. Institutional review board approval was granted by Washington University, and informed consent was waived. To assess time trends, we included PCI procedures that were performed during a 10-year period starting on January 1, 2006, and ending December 31, 2015. During this period, 1 443 297 PCIs were available, from which we defined elective PCIs using a Consolidated Standards of Reporting Trials diagram (Figure 1). To ensure an “all-comer” elective PCI population, we included patients with a discharge status of “outpatient,” patients with a discharge status of “inpatient” but were admitted as “elective,” or patients who were admitted directly from home, clinic, or primary care or referred by a health maintenance organization without an admission diagnosis of an acute coronary syndrome, but who were admitted as “nonelective” and discharged as an “inpatient” (Figure 1). Lastly, we also recognized that a few patients with chest pain or unstable angina are occasionally directly referred from clinics or an emergency department (ED) visit for an elective cardiac catheterization and ad hoc PCI. Therefore, we also included patients marked as “elective” status on admission and referred with an admission diagnosis of unstable angina from 1 of the following sources: home, clinic, primary care, or referred by a health maintenance organization or ED and “inpatients” at discharge. We believe that these inclusions allow the capture of the full spectrum of “real-world, all-comer” elective patients who have undergone PCI in the United States.
Figure 1. Consolidated Standards of Reporting Trials Flowchart to Identify Elective Percutaneous Coronary Intervention (PCI) Population in Premier.
ACS indicates acute coronary syndrome; ED, emergency department; HMO, health maintenance organization; UA, unstable angina.
SDD
Same-day discharge was identified when the date of admission, date of PCI procedure, and date of discharge were identical. Based on this, patients were categorized into 2 groups: those who underwent SDD and those who did not (NSDD).
Study Outcomes, Comorbid Conditions, and Confounders
Information on death, bleeding requiring a blood transfusion, AKI, and AMI following discharge after the index PCI was available at 3 points: 30, 90, and 365 days from the date of PCI. The follow-up information (within 30, 90, and 365 days) was limited to survivors from the index hospitalization and therefore excluded deaths during the index hospitalization. Moreover, we included information on the following potential site-level and patient-level confounders: the number of beds in the hospital, hospital teaching status, hospital location, the primary payer, sociodemographics, procedural characteristics, and a history of 24 comorbidities (Table 1).
Table 1. Characteristics of the 672 470 Patients Included in the Study.
| Characteristic | No. (%) | SDD Rate (%) | |
|---|---|---|---|
| SDD (n = 60 920) |
NSDD (n = 611 550) |
||
| Hospital characteristics | |||
| Total number of beds at hospital | |||
| 0-99 | 1615 (2.65) | 6161 (1.01) | 20.77 |
| 100-199 | 2228 (3.66) | 45 740 (7.48) | 4.64 |
| 200-299 | 7924 (13.01) | 79 188 (12.95) | 9.10 |
| 300-399 | 12 374 (20.31) | 129 119 (21.11) | 8.75 |
| 400-499 | 8015 (13.16) | 110 995 (18.15) | 6.73 |
| ≥500 | 28 764 (47.22) | 240 347 (39.30) | 10.69 |
| Hospital-teaching | |||
| No | 29 943 (49.15) | 309 533 (50.61) | 8.82 |
| Yes | 30 977 (50.85) | 302 017 (49.39) | 9.30 |
| Hospital-urban/rural | |||
| Rural | 3608 (5.92) | 49 482 (8.09) | 6.80 |
| Urban | 57 312 (94.08) | 562 068 (91.91) | 9.25 |
| Patient and hospitalization characteristics | |||
| Agea | 65.30 (10.80) | 65.55 (11.42) | NA |
| Female | 17 775 (29.18) | 204 222 (33.39) | 8.01 |
| Marital status “married” | 34 492 (56.62) | 341 992 (55.92) | 9.16 |
| Hispanic ethnicity | 4070 (6.68) | 26 641 (4.36) | 13.25 |
| Race | |||
| African American | 3790 (6.22) | 48 171 (7.88) | 7.29 |
| Other | 11 074 (18.18) | 103 544 (16.93) | 9.66 |
| Unknown | 96 (0.16) | 579 (0.09) | 14.22 |
| White | 44 457 (72.98) | 447 366 (73.15) | 9.04 |
| Insurance payer | |||
| Medicare, traditional | 24 627 (40.43) | 265 433 (43.40) | 8.49 |
| Managed care | 17 040 (27.97) | 151 239 (24.73) | 10.13 |
| Medicare, managed care | 7995 (13.12) | 68 835 (11.26) | 10.41 |
| Commercial, indemnity | 3757 (6.17) | 38 607 (6.31) | 8.87 |
| Medicaid, traditional | 1547 (2.54) | 19 214 (3.14) | 7.45 |
| Self-pay | 890 (1.46) | 16 335 (2.67) | 5.17 |
| Prior history | |||
| Diabetes | 24 289 (39.87) | 256 479 (41.94) | 8.65 |
| Dyslipidemia | 49 102 (8.60) | 504 439 (82.49) | 8.87 |
| Hypertension | 50 628 (83.11) | 522 231 (85.39) | 8.84 |
| Smoking | 27 536 (45.20) | 268 956 (43.98) | 9.29 |
| Congestive heart failure | 8915 (14.63) | 118 726 (19.41) | 6.98 |
| CABG | 2475 (4.06) | 20 294 (3.32) | 10.87 |
| AMI | 7944 (13.04) | 84 123 (13.76) | 8.63 |
| TIA | 1175 (1.93) | 12 227 (2.00) | 8.77 |
| Hemorrhagic stroke | 4496 (7.38) | 45 049 (7.37) | 9.07 |
| Ischemic stroke | 1115 (1.83) | 12 483 (2.04) | 8.20 |
| Acute renal failure | 3128 (5.13) | 48 100 (7.87) | 6.11 |
| Chronic renal disease | 6086 (9.99) | 86 072 (14.07) | 6.60 |
| Atrial fibrillation | 6502 (10.67) | 80 425 (13.15) | 7.48 |
| COPD | 9118 (14.97) | 109 778 (17.95) | 7.67 |
| Alcohol abuse | 475 (0.78) | 6818 (1.11) | 6.51 |
| Drug abuse | 269 (0.44) | 3950 (0.65) | 6.38 |
| Any type of cancer | 6965 (11.43) | 74 639 (12.20) | 8.54 |
| Heart transplant | 1 (0) | 25 (0) | 3.85 |
| Medications during index PCI | |||
| LMWH given on day of PCI | 1037 (1.70) | 86 985 (14.22) | 1.18 |
| Any GP2B3A given on day of PCI | 6339 (10.41) | 123 589 (20.21) | 4.88 |
| PCI characteristics | |||
| Drug-eluting stent used | 44 961 (73.80) | 476 717 (77.95) | 8.62 |
| Bare-metal stents used | 8116 (13.32) | 102 891 (16.82) | 7.31 |
| Radial access | 5424 (8.90) | 20 864 (3.41) | 20.63 |
| Bifurcation during PCI | 1035 (1.70) | 15 177 (2.48) | 6.38 |
| FFR during PCI | 3477 (5.71) | 17 116 (2.80) | 16.88 |
| IVUS used | 7012 (11.51) | 59 405 (9.71) | 10.56 |
| Rotational atherectomy | 57 (0.09) | 435 (0.07) | 11.59 |
| LASER atherectomy | 972 (1.60) | 6714 (1.10) | 12.65 |
Abbreviations: AMI, acute myocardial infarction; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; FFR, fractional flow reserve; GP2B3 A, glycoprotein IIb/IIIa; IVUS, intravascular ultrasonography; LASER, light amplification by stimulated emission of radiation; LMWH, low-molecular-weight heparin; NA, not applicable; NSDD, non–same-day discharge; PCI, percutaneous coronary intervention; SDD, same-day discharge; TIA, transient ischemic attack.
Columns show mean and standard deviation and not No. and percentage.
Costs
Premier uses a microcosting approach to report department-wise and total costs associated with PCI and hospitalization. Costs were reported as total-fixed, total-variable, and total costs. We adjusted the costs for inflation using the medical consumer price index10 inflation rates at the end of 2016.
Statistical Analyses
Descriptive statistics included means (SD) or medians for continuous variables and frequencies (percentages) for categorical variables as appropriate. In all multivariable analyses, we used hierarchical, mixed-effects regression models with hospital as the random effect. This strategy not only allowed a more robust estimation of the standard errors but also permitted an assessment of the across-hospital variation. The incidence was estimated using mixed-effects Poisson regression models. Time trends were assessed using regression models with calendar year as a covariate. The association of SDD with the study outcomes was determined using mixed-effects logistic regression models, and cost differences associated with SDD were determined using a mixed-effects linear regression. Interhospital variation was quantified as follows: from linear regression models we estimated the intraclass correlation coefficient as the contribution of the hospital-level variance to the overall variance,11 and from Poisson regression models we estimated the median incidence rate ratio (MIRR) using the methods of Larsen and Merlo12 and Rabe-Hesketh and Skrondal.13 The MIRR quantifies the average (median) likelihood that a statistically identical patient presenting at one hospital vs another would undergo SDD. If the MIRR was equal to 1, there would be no differences between hospitals in the likelihood of undergoing SDD. Confidence intervals around the MIRR were generated to quantitatively define the significance of the variation in SDD across hospitals.12
To ensure that the association of SDD with outcomes and costs was robust, we conducted propensity score–matching analyses. A multivariable propensity score was generated using a single nearest-neighbor matching method. This propensity score model predicting SDD, adjusted for the confounders of age, female sex, Medicare/Medicaid, the number of hospital beds, teaching hospital, urban hospital, history of diabetes, hypertension, chronic obstructive pulmonary disease, peripheral arterial disease, cerebrovascular disease, acute myocardial infarction, prior coronary artery bypass grafting, prior PCI, current heart failure, shock, cardiac arrest, multivessel disease, intra-aortic balloon pump (IABP) use, bare metal stent use, atherectomy performed, bifurcation lesion PCI, and chronic total occlusion PCI. The variable-level balance before and after matching was examined using the standardized difference of means, in which a difference of less than 10% was considered good balance, while model-level balance was examined using the Rubin B and R statistics14 (eFigure 1 in Supplement 1). All association analyses of the association of SDD with outcomes used the propensity score as a covariate in hierarchical models. Finally, to ensure that our observations and inferences were not influenced by likely confounders, we conducted 3 additional sets of sensitivity analyses. We repeated all analyses by excluding (1) low PCI–volume hospitals (<50 PCIs/y), (2) transradial PCI, and (3) “high-cost” patients who either decompensated during their PCI and required hemodynamic support with Impella (Abiomed) or IABP, mechanical ventilation, or rotational, orbital, or light amplification by stimulated emission of radiation atherectomy. All statistical analyses were conducted using Stata, version 12.0 (StataCorp). We used the user-defined programs xtmrho15 to quantify interhospital variation and psmatch216 for the propensity score analyses. Significance was tested at a 2-sided type 1 error rate of .05.
Results
Study Participants
From 1 443 297 PCIs, we included 672 470 patients (46.6%) who underwent elective PCI from 493 US hospitals (Figure 1). A total of 62 920 patients (9.1%) underwent SDD. Among those undergoing radial access, the rate of SDD was also low; approximately 1 in 5 patients who underwent elective radial PCI underwent SDD. The patient characteristics by SDD are detailed in Table 1. Briefly, the mean (SD) age was 65.5 (11.36) years, 450 473 (67%) were men, and 491 823 (73.1%) were white. Most hospitals (269 111 [40.0%]) had 500 or more beds; 619 380 (92.1%) were urban and 332 994 (49.5%) were teaching hospitals. Medicare/Medicaid accounted for 402 864 admissions (59.9%). Bare-metal stents were used in 111 007 cases (16.5%), while 134 305 cases (20%) were for multiple vessels. A few patients who underwent elective PCI decompensated during the procedure and required hemodynamic support with Impella or IABP, mechanical ventilation, or rotational, orbital, or light amplification by stimulated emission of radiation atherectomy (all <1%). Hospitals with fewer than 100 beds and use of transradial access were associated with a crude SDD rate that exceeded 20% while using low-molecular-weight heparin and glycoprotein IIb/IIIa, and hospitals with 100 to 199 beds were associated with crude SDD rates that were less than 5% (Table 1).
Incidence, Trends, and Variation in SDD Across Hospitals
Figure 2A shows the adjusted annual rate of SDD in elective PCIs estimated using a mixed-effects Poisson regression model with hospital as a random effect. The unadjusted, overall SDD rate (9.1%) was corrected to 3.5% (95% CI, 3.0%-4.0%) after accounting for the significant interhospital variation, suggesting that the higher unadjusted rate is attributable to a few larger centers performing a larger number of SDD procedures, while most smaller centers had lower rates of SDD. The MIRR was 3.82 (95% CI, 3.48-4.23) implying that on average (median) a patient with an identical clinical profile was 382% more likely to undergo SDD at one hospital vs another hospital in the sample. We observed that the adjusted incidence steadily increased from 0.4% in 2006 to 6.3% in 2015, corresponding to a 19% annual increase over time, which was significant (P for trend <.001) (Figure 2B). Also, transradial access was significantly associated with a higher likelihood of SDD (incidence rate ratio, 1.45; 95% CI, 1.40-1.50; P < .001). We observed marked variation in the SDD rate that ranged from 0% to 83% (Figure 2C). Over time, the MIRR declined from 6.66 in 2006 to 3.57 in 2015 (Figure 2D). Despite reductions over time, the variability across hospitals remained large in 2015 (MIRR, 3.57; 95% CI, 3.18-4.04).
Figure 2. Temporal Trends and Hospital Variation in the Practice of Same-Day Discharge (SDD) After Elective Percutaneous Coronary Intervention (PCI) in the United States.
A, Temporal trend in adjusted SDD annual rates. B, A magnified, scaled graph of the temporal trend for SDD, with a regression coefficient of 1.19, implying an increase of 19% annually over the baseline rate in 2006. C, Graph of the rate of SDD by hospitals performing more than 50 PCIs annually. D, Temporal trend in the median incidence rate ratio for SDD for hospitals across the study years, implying a substantial but decreasing variation in SDD practices across hospitals.
Association of SDD With Outcomes
In a series of mixed-effects, hierarchical, logistic regression models, we examined the association of SDD with each study outcome, first without propensity adjustment and then with adjustment for the propensity score (Table 2). From these results, we observed that SDD was not associated with a higher rate of rehospitalization for bleeding, AKI, AMI, or mortality after discharge.
Table 2. Short-term and Long-term Outcomes After Same-Day Discharge.
| Outcome | Incidence (95% CI)a | Strength of Association (95% CI)b | ||||
|---|---|---|---|---|---|---|
| SDD Incidence | NSDD Incidence | Unadjusted | P Value | Propensity Adjusted | P Value | |
| 30 d | ||||||
| Death | 0.29 (0.14-0.63) | 1.82 (1.68-1.98) | 0.26 (0.18-0.37) | <.001 | 0.33 (0.23-0.47) | <.001 |
| Transfusion for bleeding | 4.23 (3.51-5.11) | 6.90 (6.25-7.61) | 0.46 (0.40-0.52) | <.001 | 0.53 (0.46-0.60) | <.001 |
| Acute kidney injury | 5.14 (4.39-6.02) | 9.94 (9.40-10.52) | 0.44 (0.39-0.50) | <.001 | 0.53 (0.47-0.59) | <.001 |
| Acute myocardial infarction | 4.74 (4.01-5.61) | 7.59 (7.13-8.08) | 0.56 (0.49-0.63) | <.001 | 0.62 (0.54-0.70) | <.001 |
| 90 d | ||||||
| Death | 1.60 (1.20-2.12) | 3.99 (3.74-4.26) | 0.39 (0.32-0.48) | <.001 | 0.48 (0.39-0.59) | <.001 |
| Transfusion for bleeding | 8.91 (7.58-10.48) | 14.02 (12.71-15.47) | 0.48 (0.44-0.53) | <.001 | 0.56 (0.51-0.61) | <.001 |
| Acute kidney injury | 11.20 (9.87-12.72) | 20.21 (19.21-21.27) | 0.51 (0.47-0.55) | <.001 | 0.60 (0.55-0.65) | <.001 |
| Acute myocardial infarction | 9.31 (8.18-10.59) | 14.49 (13.76-15.27) | 0.58 (0.53-0.64) | <.001 | 0.65 (0.59-0.71) | <.001 |
| 1 y | ||||||
| Death | 5.39 (4.63-6.28) | 10.74 (10.17-11.33) | 0.45 (0.40-0.51) | <.001 | 0.54 (0.48-0.61) | <.001 |
| Transfusion for bleeding | 21.10 (18.57-23.98) | 30.66 (27.88-33.71) | 0.55 (0.52-0.58) | <.001 | 0.63 (0.59-0.66) | <.001 |
| Acute kidney injury | 30.61 (27.96-33.50) | 49.35 (47.03-51.79) | 0.57 (0.54-0.60) | <.001 | 0.66 (0.63-0.70) | <.001 |
| Acute myocardial infarction | 23.17 (21.17-25.35) | 33.31 (31.86-34.82) | 0.64 (0.60-0.68) | <.001 | 0.70 (0.66-0.74) | <.001 |
Abbreviations: NSDD, not same-day discharge; OR, odds ratio; PCI, percutaneous coronary intervention; SDD, same-day discharge.
Incidence rates are shown per 1000 PCIs and are estimated using hierarchical, mixed-effects Poisson regression model that used hospitals as the random effects.
All results are from hierarchical logistic regression models that used hospital site as the random effect.
Association of SDD With Hospital Costs
Next, we determined the association of SDD with hospital costs and their components. Figure 3A shows that SDD was associated significantly with reduced fixed, variable, and total costs. The total hospital costs were $5128 (95% CI, $5006-$5248) less in SDD patients compared with NSDD patients even after accounting for the interhospital variation in case mix and the propensity score (Figure 3A). We next divided our cohort into 2 groups of top-decile SDD hospitals (median SDD rate, 44.5%; interquartile range, 35.2%-55.6%; n = 75 694) and non–top-decile SDD hospitals (median SDD rate, 2.2%; interquartile range, 0.5%-5.5%; n = 596 776). If the non–top-decile hospitals increased their SDD rate from a median of 2.2% to match the top-decile SDD hospitals' SDD rate of 44.5%, we estimated that the annual savings would be $129 million across Premier hospitals and $433 828 annually for an average hospital that performs 200 elective PCIs annually. With approximately 300 000 elective PCIs being performed in the United States annually, and assuming a shift in practice from 2.2% SDD to 44.5% SDD among the non–top decile hospitals (where 88.74% PCIs were performed), the projected cost savings would be approximately $577 million annually. Assuming a more conservative shift from 22.3% SDD (National Cardiovascular Data Registry CathPCI institutional report, quarter 4, 2017, obtained from Barnes Jewish Hospital) to 44.5% SDD, the projected cost savings would still be substantial at $341 million annually. Interestingly, the rates of adverse outcomes after SDD in the top–SDD decile hospitals compared with the remaining hospitals were not significantly different (eFigure 2 in Supplement 1), supporting the conjecture that the previously mentioned shift in SDD practices may be achieved without an additional burden of adverse outcomes. Finally, when we investigated the department-wise components of costs, we found that the major drivers of the reduced costs were central supply and room and board costs (Figure 3B).
Figure 3. Cost Savings Associated With Same-Day Discharge (SDD) and Drivers of Cost Savings Attributable to SDD.
A indicates adjusted; EKG, electrocardiogram; ICC, intraclass correlation coefficient; ICU, intensive care unit; U, unadjusted.
Sensitivity Analyses
To ensure that the results were not swayed by confounders, we conducted 3 additional sensitivity analyses. First, low–PCI volume hospitals (<50 PCIs per year) could affect the variation in SDD rates across hospitals. After excluding low–PCI volume hospitals, the adjusted SDD rate remained unchanged at 3.50% (95% CI, 2.97-4.12), with a highly significant and unchanged interhospital variation in SDD rate (MIRR, 3.84; 95% CI, 3.44-4.33) (eTable 1 in Supplement 1).
Second, because transradial PCI is associated with reduced costs and better outcomes, and patients with transfemoral access were less likely to undergo SDD (55 496 [8.6%]) vs transradial PCI (5424 [20.6%]), we excluded transradial PCI and examined if costs and outcomes associated with SDD among the subset of transfemoral PCI were influenced by this exclusion. In patients undergoing transfemoral PCIs (646 182 [96.1%]), the associations between SDD and the study outcomes were unchanged from the overall analyses (eTable 2 in Supplement 1) but had slightly lower (but statistically nonsignificant) adjusted cost savings of $5095 (95% CI, $4966-$5224) in patients who underwent transfemoral PCI than the overall cost savings of $5128 (95% CI, $5006-$5248).
Third, the association of SDD with costs could have been skewed by high-cost patients who decompensated during PCI and required hemodynamic support, mechanical ventilation, or atherectomy. Excluding these patients (7909 [1.2%]) did not significantly influence the cost savings (eTable 3 in Supplement 1). After excluding these patients, the total cost savings associated with SDD were reduced to $4813 (95% CI, $4714-$4912) in all hospitals and to $4790 (95% CI, $4690-$4891) in high–PCI volume (≥50) hospitals. Together, our sensitivity analyses demonstrate that the study findings are unlikely to have been confounded by hospital PCI volume, transradial access, and patients who decompensated and required hemodynamic support, mechanical ventilation, or atherectomy.
Discussion
As hospitals face increasing pressure to provide safe and effective health care at a lower cost, SDD has been described as one strategy to improve the value of PCI.17 To our knowledge this is the first and only study of contemporary SDD practice in the United States that builds on prior studies of SDD with 3 novel observations. First, not only was the rate of SDD low with a weakly increasing trend, but there was also substantial variation in the practice of SDD across US hospitals, indicating that SDD practices in the United States are essentially random and likely driven by the local culture rather than evidence-based practices. Second, in this era of bundled payments, our study highlights the economic opportunity of SDD and the source of the cost savings. The costs savings that are attributable to SDD were substantially more than $5000 per case and were driven by reducing central supply and room and board costs. Third, SDD was safe after discharge. Not only were the 30-, 90-, and 365-day adverse outcomes similar for SDD vs NSDD patients, but these outcomes were also similar among patients undergoing SDD at top-decile hospitals vs other hospitals, indicating the sustained safety of SDD across time and hospitals and supporting the conjecture that a shift in practice may be achieved without the additional burden of adverse outcomes.
Our study and prior studies8 indicate that while SDD is increasing perhaps because of a greater adoption of radial access, SDD is still performed only in few patients undergoing elective PCI, and the magnitude of the increase has been modest and the room for improvement is substantial. While a radial approach facilitates SDD, there are cases in which the femoral access remains the procedural of choice. In a recent study from Barnes Jewish Hospital (St Louis, Missouri) in which we observed a cost savings of approximately $7000 per case of SDD, more than half of the SDD patients actually underwent femoral access, using 85% vascular closure devices.17 In this study, SDD after femoral access resulted in slightly lower, but still substantial cost savings of $5095 per case.
A unique aspect of Premier is that the costs reflect actual resource-use costs that were obtained directly from each hospital’s financial department. The cost savings associated with SDD were large, exceeding $5000 per case, because of supply and room and board costs that were averted. Increasing SDD from the existing low rates to even modestly higher rates could result in a large savings for hospitals, and adopting SDD could be an important strategy for hospitals participating in the Centers for Medicare and Medicaid Service’s Bundled Payments for Care Improvement Advanced.2
It is unclear if outdated hospital policies, physician inertia, or concerns regarding patient safety limit the uptake of SDD after elective PCI. While complications are generally rare after elective PCI, when they do occur, they usually do so in the first few hours after PCI, facilitating the identification of patients who are unsafe for SDD.3,4,8,18,19,20 The practice of overnight observation for all patients after elective PCI for the concern for patient safety is not evidence based. Several randomized clinical trials have confirmed the safety of SDD vs NSDD.5,6,21 A meta-analysis of 30 observational studies and 7 randomized clinical trials validated the comparable safety of SDD and NSDD.6 Our study also did not find any excess risk of short-term or long-term outcomes, such as bleeding, AKI, AMI, or death among patients who were undergoing SDD vs NSDD groups. Even more powerful is the signal of sustained safety in the 44.5% of patients undergoing SDD at top-decile hospitals who did not have greater 30-, 90-, and 365-day adverse outcomes than the 2.2% of SDD patients at non–top decile SDD hospitals.
We found marked variation, with more than 300% variation in the likelihood of SDD across hospitals. This degree of variation suggests that hospitals’ practices for SDD are essentially random and not explained by patient characteristics nor case mix, and it implies that (1) some hospitals are more comfortable than others in performing SDD and (2) the evidence base for SDD is not strong, hence SDD practices across hospitals are culturally derived rather than evidence-based. In a recent study from Barnes Jewish Hospital, we found that developing a “patient-centered” protocol for SDD based on patients’ predicted risks of complications, such as bleeding and AKI, led to a rapid adoption of SDD in more than 70% of patients undergoing elective PCI and was associated with a $1.8 million cost savings annually in hospitalization costs.17
Limitations
Our study should be interpreted in the context of several limitations. First, our data until 2015 are behind the current practice by 3 years. More contemporary National Cardiovascular Data Registry CathPCI registry institutional reports show a substantially higher rate of SDD (22.3% in the last quarter of 2017). Notwithstanding this, increasing the SDD rate from 22.3% to the top-decile rate of 44.5% would still represent substantial cost savings (an estimated $341 million). However, it should be noted that the 22.3% unadjusted SDD rate in the CathPCI registry (or 16.96% in 2015 in Premier) does not account for the substantial interhospital variation, and the resulting cost savings would be underestimated assuming a similar pattern of interhospital variation in CathPCI. Second, angiographic details and procedural complexity are not captured in our data and there is potential for unmeasured confounding. Third, outcomes such as bleeding, AKI, and mortality have been ascertained via ICD-9 codes, which could result in a misclassification of outcomes. Fourth, the cost savings associated with SDD in the study are resource use costs from a hospital’s perspective. They do not capture the opportunity costs and underestimate the true cost savings. Fifth, our elective population included a few patients with unstable angina and those who decompensated during PCI and required hemodynamic support, atherectomy, or mechanical ventilation. Their inclusion does not imply they were eligible for SDD; rather, their inclusion is important to capture the entire spectrum of the “real-world, all-comer” elective PCI population in the United States. Sixth, the association of SDD with 30-,90-, and 365-day outcomes may have a strong likelihood of confounding by indication. Nonetheless, the rates of events are still instructive, and because they are low, it appears that SDD in the patients who were selected does not compromise safety. Seven, as exact time stamps of PCI and discharge were unavailable, we were unable to identify the patients who were treated late in the day who otherwise would have been eligible for SDD but were kept overnight because of the late hour. Eight, this study is unable to identify the specific criteria different hospitals chose for SDD nor their angiographic nor PCI characteristics. Lastly, based on the association between transradial access and SDD, it is noteworthy that the rising trend in SDD may, in part, be associated with an increasing trend in the practice of transradial access.
Conclusions
In this large, contemporary, and nationally representative study of SDD practices in the United States, we found that from January 2006 to December 2015, despite reduced costs and sustained safety, SDD was used for few patients, and variation in the practice of SDD among hospitals was marked. Given the safety and large savings that exceed $5000 per case, greater and more consistent use of SDD could increase the value of PCI and save US hospitals approximately $577 million. Taken together, our findings underscore a potentially large missed opportunity of SDD in the United States.
eTable 1. Estimated incidence of same-day discharge without any adjustment and adjusted for inter-hospital variation. All hierarchical models included hospitals as random effects.
eTable 2. Association of SDD with study outcomes in patients treated with transfemoral intervention (Nhospitals = 493, Npatients = 646,182). All models used hospitals as random effects.
eTable 3. Estimated per-patient cost savings associated with SDD after excluding patients who required hemodynamic support (HDS), atherectomy or mechanical ventilation (MV). All models used hospitals as random effects.
eFigure 1. Test balance of the variables before and after propensity score matching.
eFigure 2. Comparison of study outcomes across the top SDD decile hospitals and other hospitals.
Data sharing statement.
References
- 1.Mozaffarian D, Benjamin EJ, Go AS, et al. ; Writing Group Members; American Heart Association Statistics Committee; Stroke Statistics Subcommittee . Heart disease and stroke statistics-2016 update: a report from the American Heart Association. Circulation. 2016;133(4):e38-e360. doi: 10.1161/CIR.0000000000000350 [DOI] [PubMed] [Google Scholar]
- 2.The Centers for Medicare & Medicaid Services (CMS) BPCI advanced. https://innovation.cms.gov/initiatives/bpci-advanced. Accessed February 15, 2018.
- 3.Amin AP, Patterson M, House JA, et al. Costs associated with access site and same-day discharge among Medicare beneficiaries undergoing percutaneous coronary intervention: an evaluation of the current percutaneous coronary intervention care pathways in the United States. JACC Cardiovasc Interv. 2017;10(4):342-351. doi: 10.1016/j.jcin.2016.11.049 [DOI] [PubMed] [Google Scholar]
- 4.Bertrand OF, Rodés-Cabau J, Larose E, et al. Early and late outcomes in patients excluded from same-day home discharge after transradial stenting and maximal antiplatelet therapy. Catheter Cardiovasc Interv. 2008;72(5):619-625. doi: 10.1002/ccd.21662 [DOI] [PubMed] [Google Scholar]
- 5.Abdelaal E, Rao SV, Gilchrist IC, et al. Same-day discharge compared with overnight hospitalization after uncomplicated percutaneous coronary intervention: a systematic review and meta-analysis. JACC Cardiovasc Interv. 2013;6(2):99-112. doi: 10.1016/j.jcin.2012.10.008 [DOI] [PubMed] [Google Scholar]
- 6.Brayton KM, Patel VG, Stave C, de Lemos JA, Kumbhani DJ. Same-day discharge after percutaneous coronary intervention: a meta-analysis. J Am Coll Cardiol. 2013;62(4):275-285. doi: 10.1016/j.jacc.2013.03.051 [DOI] [PubMed] [Google Scholar]
- 7.Shroff A, Kupfer J, Gilchrist IC, et al. Same-day discharge after percutaneous coronary intervention: current perspectives and strategies for implementation. JAMA Cardiol. 2016;1(2):216-223. doi: 10.1001/jamacardio.2016.0148 [DOI] [PubMed] [Google Scholar]
- 8.Agarwal S, Thakkar B, Skelding KA, Blankenship JC. Trends and outcomes after same-day discharge after percutaneous coronary interventions. Circ Cardiovasc Qual Outcomes. 2017;10(8):10. doi: 10.1161/CIRCOUTCOMES.117.003936 [DOI] [PubMed] [Google Scholar]
- 9.Rao SV, Kaltenbach LA, Weintraub WS, et al. Prevalence and outcomes of same-day discharge after elective percutaneous coronary intervention among older patients. JAMA. 2011;306(13):1461-1467. doi: 10.1001/jama.2011.1409 [DOI] [PubMed] [Google Scholar]
- 10.Agency for Healthcare Quality and Research Medical Expenditure Panel Survey Using appropriate price indices for analysis of health care expenditures or income, across multiple years. https://meps.ahrq.gov/about_meps/Price_Index.shtml. Accessed May 20, 2018.
- 11.Snijders TAB, Bosker RJ. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. Thousand Oaks, CA: Sage Publications; 1999. [Google Scholar]
- 12.Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol. 2005;161(1):81-88. doi: 10.1093/aje/kwi017 [DOI] [PubMed] [Google Scholar]
- 13.Rabe-Hesketh S, Skrondal A. Multilevel and Longitudinal Modeling Using Stata. 2nd ed College Station, TX: Stata Press Publication; 2008. [Google Scholar]
- 14.Rubin DB. Using propensity scores to help design observational studies. Health Services and Outcomes Res Methodology. 2001;2(3-4):169-188. doi: 10.1023/A:1020363010465 [DOI] [Google Scholar]
- 15.Hedburg EC. ICCVAR: Stata module to calculate intraclass correlations after xtmixed. https://ideas.repec.org/c/boc/bocode/s457468.html. Accessed May 20, 2018.
- 16.Leuven E, Sianesi B PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. https://ideas.repec.org/c/boc/bocode/s432001.html. Accessed May 20, 2018.
- 17.Amin AP, Crimmins-Reda P, Miller S, et al. Novel patient-centered approach to facilitate same-day discharge in patients undergoing elective percutaneous coronary intervention. J Am Heart Assoc. 2018;7(4):e005733. doi: 10.1161/JAHA.117.005733 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Gilchrist IC. If only the doctor will let me go home: same day discharge after PCI. Cardiovasc Revasc Med. 2017;18(4):231-232. doi: 10.1016/j.carrev.2017.05.017 [DOI] [PubMed] [Google Scholar]
- 19.Bundhun PK, Soogund MZ, Huang WQ. Same day discharge versus overnight stay in the hospital following percutaneous coronary intervention in patients with stable coronary artery disease: a systematic review and meta-analysis of randomized controlled trials. PLoS One. 2017;12(1):e0169807. doi: 10.1371/journal.pone.0169807 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Bertrand OF, Larose E, De Larochellière R, et al. Outpatient percutaneous coronary intervention: Ready for prime time? Can J Cardiol. 2007;23(suppl B):58B-66B. doi: 10.1016/S0828-282X(07)71012-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Abdelaal E, Bertrand OF. Transradial PCI and same day discharge: a winning team? Catheter Cardiovasc Interv. 2013;81(1):24-25. doi: 10.1002/ccd.24760 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. Estimated incidence of same-day discharge without any adjustment and adjusted for inter-hospital variation. All hierarchical models included hospitals as random effects.
eTable 2. Association of SDD with study outcomes in patients treated with transfemoral intervention (Nhospitals = 493, Npatients = 646,182). All models used hospitals as random effects.
eTable 3. Estimated per-patient cost savings associated with SDD after excluding patients who required hemodynamic support (HDS), atherectomy or mechanical ventilation (MV). All models used hospitals as random effects.
eFigure 1. Test balance of the variables before and after propensity score matching.
eFigure 2. Comparison of study outcomes across the top SDD decile hospitals and other hospitals.
Data sharing statement.