Trends and Outcomes After Same-Day Discharge After Percutaneous Coronary Interventions
Background—Despite the demonstrated safety of the same-day discharge (SDD) after percutaneous coronary intervention (PCI), uptake of this program has been relatively poor in the United States. We evaluated the temporal trends and variations in the utilization of SDD after PCI during the contemporary era. In addition, we evaluated the predictors of SDD (compared with next-day discharge) and the causes of readmission in these 2 patient cohorts.
Methods and Results—Data were extracted from State Ambulatory Surgical Database and State Inpatient Database from Florida and New York ranging from 2009 to 2013. All adults undergoing PCI in an outpatient setting were included. Data were merged with the directory available from the American Hospital Association to obtain detailed information on hospital-related characteristics. Unplanned readmissions within 7 and 30 days constituted the coprimary outcomes. There was modest increase in the proportion of SDD after PCI from 2.5% in 2009 to 7.4% in 2013 (P-trend <0.001). SDD was more frequently used among male and younger patients with fewer comorbidities. There were considerable differences in the discharge practices among the different hospital types. Larger hospitals, teaching hospitals, and high PCI volume hospitals had higher utilization of SDD compared with their respective counterparts. SDD and next-day discharge cohorts had similar rates of unplanned readmissions, in-hospital mortality, and acute myocardial infarction during follow-up. Furthermore, uninsured patients had significantly lower odds of SDD along with higher incidence of unplanned readmission within 30 days after PCI compared with insured patients.
Conclusions—During 2009 to 2013, there has been a modest increase in SDD after PCI. Several demographic and clinical characteristics play critical role in determination of SDD after PCI. There were significant disparities in discharge practices between different sex, racial, and insurance-based strata.
WHAT IS KNOWN
Observational studies and randomized control trials have clearly demonstrated the safety of same-day discharge (SDD) compared with next-day discharge after percutaneous coronary intervention (PCI); benefits of SDD after PCI include improved patient satisfaction, decreased length of stay, and cost savings.
Despite the demonstrated safety of the SDD after PCI, uptake of SDD has been relatively poor in the United States.
WHAT THE STUDY ADDS
There was modest increase in the proportion of SDD after PCI from 2.5% in 2009 to 7.4% in 2013.
There were considerable differences in the discharge practices among the different hospital types: larger hospitals, teaching hospitals, and high PCI volume hospitals had higher utilization of SDD compared with their respective counterparts.
In the real-world setting, SDD and next-day discharge cohorts had similar rates of unplanned readmissions, in-hospital mortality, and acute myocardial infarction during follow-up.
Overnight observation has been the standard of care after percutaneous coronary intervention (PCI) for coronary artery disease in the United States. The basis of this practice stems from the earlier days of balloon angioplasty when acute ischemic events from abrupt vessel closure or vascular access complications were common. Observational studies and randomized control trials have clearly demonstrated the safety of same-day discharge (SDD) compared with next-day discharge (NDD) after PCI.1 Benefits of SDD after PCI include improved patient satisfaction, decreased length of stay (LOS), and cost savings.2–4
Despite the demonstrated safety of the SDD after PCI, uptake of SDD has been relatively poor in the United States. According to an analysis of data from the CathPCI Registry, the prevalence of SDD after PCI during 2004 to 2008 was only 1.25%.5 This has been attributed to physician inertia, concerns of patient safety, lack of training in newer PCI techniques including transradial catheterization, or use of vascular closure devices and unclear payer expectations for hospital reimbursement.6,7 Since the publication of the CathPCI report in 2011, there have been no newer reports of trends in the utilization of SDD after PCI.5 The past decade has witnessed significant improvements in several aspects of PCI including increasing utilization of transradial catheterization, increasing use of femoral closure devices, use of second-generation drug-eluting stents, and reliable provision of dual antiplatelet therapy to our patients. The impact of these measures on uptake of SDD after PCI is currently unknown.
To that end, we evaluated temporal trends and variations in the utilization of SDD after PCI during the contemporary era across various types of healthcare facilities in Florida and New York. In addition, we evaluated factors associated with SDD after PCI compared with NDD after PCI. Furthermore, we quantified the incidence and the reasons for unplanned readmission within 7 days and within 30 days among SDD patients in comparison to NDD patients.
Data for this study were obtained from the State Ambulatory Surgical Database (SASD) to include all elective procedures performed on an outpatient basis. In addition, we used the State Inpatient Database (SID), comprising all in-hospital admissions in a specific state. We used data from the states of Florida and New York spanning the years of 2009 to 2013, because these states provided comprehensive data on repeat admissions. The SID and SASD are sponsored by the Agency for Healthcare Research and Quality as a part of Healthcare Cost and Utilization Project. The SID for the states of Florida and New York includes a visit linkage variable (VisitLink) that can be used in tandem with the timing variable (DaysToEvent) to study multiple hospital visits and outpatient procedural encounters for the same patient across hospitals and time while adhering to strict privacy regulations.
The SASD and SID provide the list of diagnoses and procedures for each hospitalization record. All diagnoses and procedures have been coded using the standard International Classification of Diseases, NinthRevision, Clinical Modification codes in the SID. Although the diagnoses have been coded using similar International Classification of Diseases, Ninth Revision, Clinical Modification codes in the SASD, the procedures have been coded using Current Procedural Terminology codes. All adult hospitalizations or procedural encounters (>18 years of age) with a procedure code corresponding to PCI were included in our study. The list of procedural codes used to identify patients with PCI is shown in Table I in the Data Supplement. The first diagnosis in the database is referred to as the principal diagnosis and is considered the primary reason for admission to the hospital. We used the Charlson comorbidity index to quantify the burden of comorbidities in the patient population based on 17 categories of diagnoses.8 In addition, the SID and SASD provide 29 Elixhauser comorbidities on each hospital admission, based on standard International Classification of Diseases, Ninth Revision, codes.9 These were used to derive the prevalence of hypertension, diabetes mellitus, obesity, and chronic kidney disease in our population.
From the entire cohort of PCI encounters, we considered the outpatient elective PCI (from the SASD data source) for inclusion. All outpatient elective PCIs that had a LOS≤1 day were included. Patients with LOS=0 days were referred to as the SDD cohort, and patients with LOS=1 day were referred to as the NDD cohort. Outpatient elective PCI that resulted in an admission to the hospital for >1 day or those who died as a result of the procedure were not included in the analysis for study outcomes.
Our study aimed to evaluate the temporal trend in the utilization of SDD after outpatient elective PCI during 2009 to 2013. This was expressed as a proportion of all PCIs performed during a specific year. Besides this, we quantified the incidence and the reasons for unplanned readmissions within 7 and 30 days among the SDD patients in comparison to the NDD patients. All readmissions that were tagged as emergent, urgent, or unscheduled were considered to be unplanned readmissions. The secondary outcome of the study included a composite of in-hospital mortality or acute myocardial infarction (MI) within 30 days after PCI. Furthermore, we evaluated the independent predictors for SDD among patients who were discharged from the hospital in ≤1 day after elective outpatient PCI. Outpatient PCIs that resulted in an admission to the hospital for >1 day were not included in our analysis. All available demographic, clinical, and hospital-related characteristics were considered. To evaluate the impact of hospital type on study outcomes, we procured the hospital-specific data from the American Hospital Association for the corresponding years.10 Four specific variables were studied location (rural versus urban), teaching affiliation (nonteaching, minor, and major), hospital size (small, medium, and large), and annual PCI volume quartile.
Continuous variables have been presented as mean±SD, and categorical variables have been presented as proportions. The Student t test was used for comparing means of continuous variables between the 2 categories. The χ2 test was used for comparison of categorical variables. The trend in utilization of SDD after PCI was expressed as a proportion of all PCIs performed during a specific year. The nonparametric test for trend across ordered groups by Cuzick11 was used to determine the significance of differences in SDD utilization across the study period. The incidence of unplanned readmission was calculated at 7 and 30 days for all patients in the SDD and NDD cohorts. Multivariable hierarchical logistic regression analysis with exchangeable matrix (clustered by unique patients) was used to determine independent predictors of SDD among patients undergoing elective PCI with LOS≤1 day (SDD or NDD). Covariates included age, sex, race, insurance status, Charlson comorbidity index, and clinical comorbidities along with hospital characteristics (location, teaching affiliation, size, and annual PCI volume). Furthermore, we performed a comparison of unplanned readmission rates, 30-day in-hospital mortality, and 30-day acute MI among propensity-matched subpopulations of SDD and NDD patients. Propensity matching was performed using a 1:1 greedy matching algorithm with no replacement feature and a caliper width of 0.00001. All available clinical, demographic, and hospital-based characteristics were included in the propensity score calculations. To examine the impact of hidden bias between the matched groups, we calculated the Rosenbaum bounds for average treatment effects using the statistical package mhbounds.12 The procedure then calculates Mantel–Haenszel test statistics that give bound estimates of significance levels at given levels of hidden bias under the assumption of either systematic over- or underestimation of treatment effects. Comparison of outcomes between the matched pairs was performed using McNemar test for paired nominal data.13
All statistical tests were 2-tailed; a P value <0.05 was considered significant. All data available from the Healthcare Cost and Utilization Project have been deidentified, and hence, the analysis is exempt from the federal regulations for the protection of human research participants. The data set was obtained from the Agency for Healthcare Research and Quality after completing the data use agreement with Healthcare Cost and Utilization Project. Therefore, institutional review board approval was not necessary for the conduct of this study.
Trend of SDD After PCI
A total of 467 253 PCI encounters were available for analysis during 2009 to 2013 from the states of FL and NY. Of these, 55229 were outpatient elective PCI encounters. There was a progressive increase in the proportion of outpatient elective PCI during 2009 to 2013 (Figure 1). Of all elective PCI encounters, 17 089 were SDDs and 34 295 were NDDs. Figure 2 demonstrates the trend of SDD PCI expressed as a proportion of all PCI. As evident in this figure, there has been a modest increase in the proportion of SDD PCI from 2.5% in 2009 to 7.4% in 2013 (P-trend <0.001). The increase in the proportion of SDD was significantly higher in New York (2.9% in 2009 to 10.3% in 2013) compared with Florida (2.1% in 2009 to 4.8% in 2013).
Table 1 demonstrates the baseline characteristics of SDD compared with NDD PCI encounters. There was a significant difference in the demographic profile and clinical characteristics between the 2 groups. SDD patients tended to be younger, were more likely to be men, and had fewer comorbidities compared with the patients discharged the day after PCI. Significant differences were found in the distribution of these cohorts among various hospital strata. Figure 3 demonstrates the proportion of SDD after PCI among elective PCI encounters stratified according to hospital type. Compared with small hospitals, medium- and large-sized hospitals had significantly higher SDD rates (P<0.001 for both comparisons). Stratified by teaching affiliation, hospitals with minor and major teaching affiliation had higher proportion of SDDs compared with nonteaching hospitals (P<0.001 for both comparisons). With respect to annual PCI volumes, increase in the annual PCI volume was associated with increasing utilization of SDD across the hospitals.
Predictors of SDD
Table 2 demonstrates the predictors of SDD among patients undergoing elective PCI, obtained using multivariable hierarchical logistic regression analysis. Increasing age and female sex were associated with lower odds of SDD, compared with their respective counterparts. In addition, black or Hispanic race was associated with significantly higher odds of SDD compared with white race. Uninsured patients had lower odds of SDD compared with insured patients. On further analysis, there were significant differences in the hospitals where uninsured and insured patients were treated. A significantly higher proportion of uninsured patients were treated at teaching hospitals compared with insured patients (70.6% versus 59.2%; P=0.001). The presence of comorbidities was associated with decreased odds of SDD. With respect to hospital types, there were significant differences in the odds of SDD between different strata. These differences on adjusted analysis were similar to the unadjusted differences detailed above.
Unadjusted 30-day all-cause readmission rates after SDD and NDD were 7.2% and 7.4%, respectively (P=0.38; Table 3). All-cause readmission rates within 7 days after SDD and NDD were 2.1% and 2.0%, respectively (P=0.56). The rates of unplanned readmissions within 7 or 30 days were also similar in the SDD and NDD cohorts (Table 3). Figure 4 demonstrates the causes for unplanned readmission within 7 and 30 days in the SDD and NDD cohorts. The most common cause for 30-day unplanned readmission were coronary artery disease including repeat PCI accounting for 41.9% and 41.4% of all unplanned readmissions in the SDD and NDD cohorts, respectively. Similarly, coronary artery disease also constituted the most common cause for unplanned readmission within 7 days after PCI accounting for 33.7% and 33.1% of all unplanned readmissions in the SDD and NDD cohorts, respectively. The second most common cause for unplanned readmission within 7 days was procedural complications accounting for 13.3% and 10.7% of all unplanned readmissions in the SDD and NDD cohorts, respectively.
The independent predictors of unplanned 30-day readmission in our study cohort are shown in Table 4. Adjusted odds of readmission were similar in the SDD and NDD cohorts (odds ratio, 0.94; 95% confidence interval, 0.87–1.03). Uninsured individuals had higher odds of unplanned readmission within 30 days compared with insured patients (odds ratio, 1.49; 95% confidence interval, 1.18–1.89). Increasing comorbidity burden was associated with increasing odds of readmission. In addition, major teaching affiliation hospitals had significantly higher odds for unplanned readmission compared with nonteaching hospitals (odds ratio, 1.15; 95% confidence interval, 1.02–1.29). On the contrary, hospital location, hospital size, and annual PCI volume did not impact the odds of readmission in the study population.
In-Hospital Mortality/Acute MI
The incidence of 30-day in-hospital mortality or acute MI was 0.46% in the SDD cohort, which was similar to 0.50% in the NDD cohort (P=0.46; Table 3). On adjusted analysis, there was no significant difference in the incidence of composite primary outcome between the 2 groups (odds ratio, 0.96; 95% confidence interval, 0.71–1.28; Table II in the Data Supplement).
We performed a comparison of the incidence of primary and secondary outcomes among propensity-matched subpopulations of SDD and NDD patients. All available clinical, demographic, and hospital-based characteristics were included in the propensity score calculations. Propensity matching yielded 10 957 matched pairs of SDD and NDD patients. Figure I in the Data Supplement demonstrates the extent of covariate imbalance in terms of standardized mean difference or standardized percentage differences using dot charts. As seen in the figure, there was a good covariate balance achieved in the matched populations. In addition, the matching was found to be robust without evidence of significant unobserved heterogeneity or hidden bias. In the propensity-matched population, the incidence of 30-day in-hospital mortality or acute MI was similar between the 2 groups (0.3% in SDD cohort; 0.3% in the NDD cohort; P=0.81). Similarly, the incidence of 7-day unplanned readmission was 1.6% and 1.7% in the SDD and NDD cohorts, respectively (P=0.71). Furthermore, the incidence of 30-day unplanned readmission was 5.2% and 5.6% in the SDD and NDD cohorts, respectively (P=0.16).
In the current study, we have evaluated the trends of SDD after PCI prevalent in the United States during the contemporary era using a combination of large real-world administrative databases. We have several important findings. First, there has been a modest increase in SDD after PCI during 2009 to 2013 in the states of Florida and New York. Second, in a real-world setting, SDD has been typically used in a younger population with fewer comorbidities. Third, compared with NDD, SDD was not associated with a higher risk of unplanned readmission within 7 or 30 days after PCI. Fourth, the incidence of in-hospital mortality and acute MI within 30 days after PCI was similar between the SDD and NDD cohorts. Fifth, there were considerable differences in the discharge practices among the different hospital types. Finally, uninsured patients had a significantly lower rate of SDD along with higher incidence of unplanned readmission within 30 days after PCI compared with insured patients.
In the current era, SDD after PCI may be considered among patients with excellent procedural outcomes, rapid and reliable stabilization of the access site permitting safe ambulation, reliable provision of dual antiplatelet therapy along with availability of social support, and a routine early follow-up.14 For successful implementation of this practice, it is imperative to accurately assess patients for suitability for SDD after successful PCI. It is important to identify high-risk features that may be patient and procedure related, which should caution physicians against SDD. It is also important to identify factors like frailty and health literacy along with the presence of social support, which may be important for successful SDD.
Our study demonstrated a progressive increase in the number and the proportion of outpatient elective PCI during the study period. This reflects the ongoing change in practice of PCI, wherein greater numbers of patients are starting to be treated on an outpatient basis rather than on inpatient basis. This might also be secondary to healthcare facilities being more stringent about what qualifies as inpatient admission. Our study clearly demonstrated modest increase in utilization of SDD after PCI during 2009 to 2013. The utilization of SDD was relatively higher in New York compared with Florida—the definitive reason for this regional difference is unclear from our study. The utilization of SDD after PCI in the current age is considerably higher than what has been reported in older studies.5 During 2004 to 2008, use of SDD after PCI was only 1.25% as reported from the CathPCI registry.5 The rise of SDD after PCI could be attributed to greater adoption of radial access, refinement in femoral access, and use of vascular closure devices and reduction in the size of arterial access.
Traditionally, physician inertia and concerns about patient safety have been the most important barriers in implementation of SDD after PCI. However, several studies including randomized control trials have now demonstrated the safety of SDD in comparison to NDD after PCI.1 A large meta-analysis of 30 observational studies and 7 randomized control trials demonstrated comparable safety of SDD and NDD after PCI.1 In a similar systematic review and meta-analysis, it was pointed out that the clinical risk profile of patients in the SDD cohort was similar to patients who were hospitalized overnight.15 This was contrary to our findings, where we found that the patients in the NDD cohort were significantly older and had greater burden of clinical comorbidities compared with the SDD cohort. With an increasing emphasis on growth of SDD programs in various hospitals, it is likely that physicians are more inclined to discharge lower risk patients the same day after PCI and observe the older, higher risk patients in the hospital overnight. Our study provides some unique perspectives on SDD in the current era. We did not find a significant difference in the 30-day in-hospital mortality or 30-day acute MI incidence between the SDD and NDD cohorts. These findings are similar to earlier studies.1,14,16 Because of use of administrative data sets in our study, it was not possible to discern the rates of target lesion revascularization or target vessel revascularization. However, we feel that major complications after PCI often necessitate longer hospital stays than just 1 day. Hence, we could speculate that the success of PCI and the immediate postprocedural complications were likely similar between the SDD and NDD cohorts. It was reassuring to note that both 7- and 30-day incidence of unplanned readmissions was similar in the SDD and NDD cohorts.
The other unique aspect of our study was characterization of hospitals with respect to SDD. Large hospitals tended to have a higher proportion of patients discharged same day after PCI compared with smaller hospitals. Similarly, higher volume PCI centers had the higher utilization of SDD compared with lower quartiles. Similarly, hospitals with minor or major teaching affiliation had the higher SDD practice compared with the nonteaching hospitals. Whether this difference is attributable to the risk profile of the patients treated at these hospitals or this is secondary to actual differences in discharge practices at these facilities is not completely clear in our study but is likely because of a combination of these 2 factors.
In addition to detailed perspective on SDD after PCI, our study also provides some additional data that might form the basis of future studies. We observed a significantly higher proportion of SDD among men compared with women. Whether these sex-based outcome differences reflect true biological differences or are more indicative of different clinical profiles that render female sex a surrogate for poor outcome in the eyes of the treating physician is a matter of speculation. In addition, uninsured patients tended to have lower utilization of SDD compared with insured patients. Uninsured patients typically have more advanced disease, late presentation, risk factors, and health illiteracy. All of these factors make them poor candidates for SDD and may explain the higher incidence of 30-day unplanned readmission among uninsured individuals compared with insured individuals. In addition, there were differences in the hospitals where uninsured patients were treated compared with insured patients. The inherent differences in discharge practices between different hospitals might also account for differences in SDD among uninsured patients. Furthermore, there were significant race-based differences in utilization of SDD and outcomes after PCI. Hispanics and blacks had a significantly higher proportion of SDD compared with whites. The reasons for the differences are unclear; however, these may be secondary to different family structures and support systems among nonwhite races, which might alter the physician perception of safety for SDD among this population.
Our study has several limitations inherent in the administrative data sources. First, anatomic details and procedural characteristics were not available in this database, which are important for determination of suitability of SDD. In addition, data on access site (radial versus femoral), use of vascular closure devices, and choice of antithrombotics/glycoprotein inhibitors were not available. Second, mortality reported in the database only includes in-hospital mortality. Any mortality occurring outside of the hospital was not recorded and, hence, not available for inclusion in the analysis. Third, the SID and SASD both provide state-specific encounters. Any future encounters that did not occur in the respective state were not captured in the database. Fourth, our study included only outpatient elective PCI encounters in the 2 states, which might impact the generalizability of the results to other scenarios and regions. Despite these limitations, the administrative databases like SID and SASD are valuable in understanding temporal trends and variations in the utilization of hospital practices across the nation.
There has been a modest increase in SDD after PCI during 2009 to 2013 in the states of Florida and New York. In this real-world setting, SDD has been typically used in a younger population with significantly lower degree of comorbidities. SDD and NDD were associated with similar risk of unplanned readmissions within 7 or 30 days after PCI. In addition, the incidence of in-hospital mortality and acute MI within 30 days after PCI was similar between the SDD and NDD cohorts. We also noted considerable differences in the discharge practices among different hospital types. Furthermore, there were disparities in discharge practices and outcomes among different sex, racial, and insurance-based strata.
The Data Supplement is available at http://circoutcomes.ahajournals.org/lookup/suppl/doi:10.1161/CIRCOUTCOMES.117.003936/-/DC1.
- Received March 4, 2017.
- Accepted July 6, 2017.
- © 2017 American Heart Association, Inc.
- Brayton KM,
- Patel VG,
- Stave C,
- de Lemos JA,
- Kumbhani DJ
- Kim M,
- Muntner P,
- Sharma S,
- Choi JW,
- Stoler RC,
- Woodward M,
- Mann DM,
- Farkouh ME
- Heyde GS,
- Koch KT,
- de Winter RJ,
- Dijkgraaf MG,
- Klees MI,
- Dijksman LM,
- Piek JJ,
- Tijssen JG
- Rinfret S,
- Kennedy WA,
- Lachaine J,
- Lemay A,
- Rodés-Cabau J,
- Cohen DJ,
- Costerousse O,
- Bertrand OF
- Bottner RK,
- Blankenship JC,
- Klein LW
- Elixhauser A SC,
- Palmer L
- 10.↵American Hospital Association. DATAVIEWER. https://www.ahadataviewer.com/. Accessed August 4, 2017.
- Shroff A,
- Kupfer J,
- Gilchrist IC,
- Caputo R,
- Speiser B,
- Bertrand OF,
- Pancholy SB,
- Rao SV
- Abdelaal E,
- Rao SV,
- Gilchrist IC,
- Bernat I,
- Shroff A,
- Caputo R,
- Costerousse O,
- Pancholy SB,
- Bertrand OF
- Saad Y,
- Shugman IM,
- Kumar M,
- Pauk I,
- Mussap C,
- Hopkins AP,
- Rajaratnam R,
- Lo S,
- Juergens CP,
- French JK