|Home | About | Journals | Submit | Contact Us | Français|
Substantial heterogeneity in hospital length of stay (LOS) exists among patients admitted with non-ST-segment elevation myocardial infarction (NSTEMI). Furthermore, little is known about the factors that impact LOS and our ability to modify them.
We examined 39,107 NSTEMI patients admitted to 351 ACTION Registry®-GWTG™ hospitals from 1/1/07–3/31/09 who underwent cardiac catheterization and survived to discharge. Length of stay was categorized into four groups (≤2, 3–4, 5–7, and ≥8 days), where prolonged LOS was defined as >4 days.
The overall median (25th, 75th) LOS was 3 (2, 5) days. Patients with a LOS of >2 days were older with more comorbidities, but were less likely to receive evidence-based therapies or percutaneous coronary intervention (PCI). Among the factors associated with prolonged LOS >4 days were delay to cardiac catheterization >48 hours, heart failure or shock on admission, female gender, insurance type, and admission to the hospital on a Friday afternoon or evening. Hospital characteristics such as number of beds, academic versus non-academic, or urban versus rural setting, were not associated with prolonged LOS.
Patients with longer LOS have more comorbidities and in-hospital complications, yet paradoxically, are less often treated with evidence-based medications and are less likely to receive PCI. Hospital admission on a Friday afternoon or evening and delays to catheterization appear to significantly impact LOS. A better understanding of factors associated with LOS in patients with NSTEMI is needed to promote safe and early discharge in an era of increasingly restrictive healthcare resources.
More than a half million patients are hospitalized annually for non-ST-segment elevation myocardial infarction (NSTEMI) in the United States (U.S.).1 Among these patients, there is considerable variability in hospital length of stay (LOS) for reasons that are not well defined. Previous studies have demonstrated that patients with uncomplicated ST-segment elevation myocardial infarction (STEMI) treated with fibrinolysis could be safely discharged as early as three days after presentation.2 Studying the heterogeneity of LOS in STEMI patients has provided important insights into opportunities for safe earlier hospital discharge.3–7 However, to date, no studies have investigated LOS for patients with NSTEMI.
Increasingly, hospitals and healthcare providers are pressured to reduce medical resource expenditures and shorten hospital LOS. The importance of containing healthcare costs underscores the need to better understand the factors associated with longer hospital LOS and to explore opportunities for safe early discharge in patient with NSTEMI. Current treatment guidelines recommend a risk-tailored early invasive approach for the treatment of NSTEMI and advocate for early discharge—especially in those patients who are considered low-risk.8 The effect of patient characteristics and treatment strategies on LOS for NSTEMI patients has not been well studied.
This analysis utilized data from the National Cardiovascular Data Registry® (NCDR) Acute Coronary Treatment Intervention Outcomes Network Registry®-Get With The Guidelines™ (ACTION Registry-GWTG) to explore and identify patient characteristics and clinical factors associated with hospital LOS in patients admitted with NSTEMI who underwent cardiac catheterization. The variables analyzed in this study include patient demographics, treatment strategies, hospital characteristics, medication use, and clinical outcomes as they relate to hospital LOS.
The ACTION Registry-GWTG is a result of the merger between the American College of Cardiology Foundation’s (ACCF’s) NCDR ACTION Registry and the American Heart Association’s (AHA) GWTG-Coronary Artery Disease (CAD) Registry. The ACTION Registry-GWTG is a voluntary national quality improvement registry that began collecting data on January 1, 2007 on hospitalized STEMI and NSTEMI patients. This registry is now the largest national quality improvement initiative focusing on patients with myocardial infarction (MI).9
The full details of the registry operations, quality assurance, and data collected have been previously reported.10 Briefly, data were entered at each site by a trained data collector. The NCDR employs a standard data set with uniform data entry and data quality checks.9 The variables recorded included pre-hospital data, acute (within 24 hours) and discharge therapies, medical comorbidities, treatments administered, in-hospital procedures and outcomes, admission time and date, discharge time and date, and contraindications to evidence-based therapy. Institutional review board approval was obtained for data collection according to individual site policies and regulations as related to quality improvement data collection processes.
Between January 1, 2007 and December 31, 2009, 98,545 NSTEMI patients at 384 hospitals were enrolled into the ACTION Registry-GWTG. To focus on a more homogeneous population of patients relative to treatment strategy, patients were excluded if: (1) they did not undergo cardiac catheterization or had missing catheterization status data (n=24,899); (2) underwent coronary artery bypass grafting (CABG) or had missing CABG status data (n=10,581); (3) were transferred out (n=2,336) or transferred in to the reporting hospital (n=20,815); (4) had a LOS equal to zero or missing LOS data (n=59); or (5) died during their hospitalization (n=748). After exclusions, our final study population included 39,107 NSTEMI patients from 351 hospitals.
Length of stay was defined as the time interval between date of admission and date of discharge. Exact patient discharge times were not available, so all discharge times were recorded as 12:00pm. For patients with partial days in their LOS, we rounded to the nearest whole number. Patients were categorized into four groups based on hospital LOS: (1) ≤2 days, (2) 3–4 days, (3) 5–7 days, or (4) ≥8 days. Patients were also grouped into the following six categories, based on the day of the week they were admitted: (1) Monday–Wednesday; (2) Thursday; (3) Friday morning (12:00am–<12:00pm); (4) Friday afternoon or evening (12:00pm–<12:00am); (5) Saturday; or (6) Sunday. Major bleeding was defined as an absolute hemoglobin (Hgb) drop of ≥4 g/dL, intracranial hemorrhage (ICH), witnessed retroperitoneal bleeding, red blood cell (RBC) transfusion in a patient with a baseline Hgb ≥9 g/dL, or an RBC transfusion with a witnessed bleeding event in a patient with a baseline Hgb of <9g/dL. In-hospital complications were defined as post-admission reinfarction, cardiogenic shock, heart failure, stroke, major bleeding, or blood transfusion. Early cardiac catheterization and percutaneous coronary intervention (PCI) were defined as occurring within 48 hours. All other variables are defined on the NCDR web site (www.ncdr.com/WebNCDR/ACTION/Elements.aspx).
Demographic information, co-existing medical comorbities on admission, presenting signs and symptoms, electrocardiogram (ECG) findings, acute (≤24 hour of hospital arrival) and discharge medications received, time from hospital presentation to ECG, time from hospital presentation to cardiac catheterization and PCI, in-hospital procedures, in-hospital clinical events, and hospital characteristics, were stratified by four LOS categories. Treatments such as acute and discharge medications and procedures were only considered among those without a documented contraindication. Continuous variables were reported as median values (with 25th, 75th percentiles) and categorical values were reported as percentages. Chi-square tests were used to compare categorical variables and the Kruskal-Wallis test was used for continuous variables. Distribution of LOS was displayed in a histogram (Figure 1).
To explore factors associated with prolonged LOS (defined as >4 days vs. ≤4 days), a logistic generalized estimating equations (GEE) method was used with exchangeable working correlation matrix to account for within-hospital clustering. This method produced estimates similar to those from logistic regression, but variances were adjusted for the correlation of outcomes within a hospital.11 Variables in the model were selected based on clinical input.
Statistical significance was defined as p<0.05. All analyses were conducting using SAS software version 9.2 (SAS Institute, Cary, NC).
For the 39,107 patients with NSTEMI in our analysis cohort, the median age overall was 64 (54, 75) years. Women made up 37.0% of the population and 82.1% of patients were Caucasian. The median LOS (25th, 75th) for the overall cohort in our study population was 3 (2, 5) days and the distribution of LOS is displayed in Figure 1. There were 13,047 (33.4%) patients with LOS ≤2 days; 15,870 (40.6%) with LOS 3–4 days; 6,325 (16.1%) with LOS 5–7 days; and 3,865 (9.9%) with LOS ≥8 days. Twenty-six percent of patients had a LOS >4 days.
The patient baseline and hospital characteristics of our population categorized by LOS are displayed in Table 1. As LOS increased, patients were older, more often female, and less frequently Caucasian. In addition, as LOS increased, patients more frequently presented with heart failure symptoms and ST-segment depression on the initial ECG. They also had an increased number of co-morbidities, including a history of prior MI, heart failure, hypertension, atrial fibrillation, CABG, stroke, or peripheral arterial disease. Patients with longer hospital LOS presented with higher serum creatinine levels, had higher peak serum creatinine levels, and were more likely to be on hemodialysis. These patients were also less likely to be current smokers.
Patients with higher LOS were more likely to be treated at an academic medical center, and at centers with larger bed volumes. Urban versus rural setting of the hospital did not seem to be associated with LOS (Table 1).
In the absence of documented contraindications, the administration of medications to treat acute coronary syndromes within 24 hours is shown in Table 2 for each category of LOS, as well as the use of anticoagulants at any time during the hospitalization. Table 3 shows discharge medications by LOS category. In general, patients with longer LOS were less likely to be treated with evidence-based medications during the hospitalization and at the time of discharge.
Clopidogrel use in the first 24 hours after presentation was 77.1% in the ≤2 days group, 66.6% in the 3–4 days group, 56.7% in the 5–7 days group, and 51.8% in the ≥8 days group (p<0.0001 for all comparisons). This pattern was also generally true for aspirin, beta-blockers, angiotensin converting enzyme inhibitors (ACE-I) or angiotensin receptor blockers (ARBs), and statins. The use of anticoagulants such as unfractionated heparin, low molecular weight heparin, bivalirudin, or fondaparinux differed to a lesser degree (Table 2).
Patients who had longer LOS were also less likely to be discharged home on evidence-based medications. Those with the longest LOS (≥8 days) were the least likely to be discharged home on aspirin (95.8%), clopidogrel (76.0%), or a statin (83.0%); p<0.0001 for all comparisons among LOS groups for each medication (Table 3).
Longer LOS was associated with higher rates of clinical complications, less use of PCI, and longer delays to cardiac catheterization and PCI when they were performed (Table 4). Patients with LOS ≥8 days had the highest rates of MI (2.5%), shock (7.6%), heart failure (18.3%), stroke (2.1%), or major bleeding (38.8%) compared with all other groups; p<0.0001. Nevertheless, patients with LOS ≥8 days were the least likely to receive early cardiac catheterization or PCI (53.7%), and were most likely to receive a noninvasive stress test; p<0.0001 for all comparisons. Furthermore, if they did receive cardiac catheterization or PCI, they had the longest delays in time from presentation to cardiac catheterization (49.3 hours, interquartile range [IQR] 19.2, 110.0) or PCI (55.0 hours, IQR 18.4, 131.6) (Table 4).
Delays to cardiac catheterization and the presence of complications were both associated with longer LOS (Figure 2). Almost 85% of patients (n=33,010) had no complications. Less than 2% had three or more complications. Moreover, approximately 90% of patients (n=35,037) went to the catheterization lab within three days. The shortest LOS was among those patients with no complications who went to the catheterization lab within three days. The longest LOS was in those patients with at least one complication who went to the catheterization lab after three days. Irrespective of complications, LOS was longer in the cohort who got catheterized after three days, as compared to those who got catheterized within three days (6.66 vs. 2.89, respectively).
The risk of mortality calculated by the ACTION Registry-GWTG mortality risk score was also positively associated with hospital LOS.12 Figure 3 demonstrates that as the risk increases, there is a consistent increase in LOS.
Depicted in Table 5 is the time to cardiac catheterization and hospital LOS based on the day of the week admitted to the hospital. Admission to the hospital on a Friday afternoon or evening was studied for its association with prolonged LOS as a way to describe possible delays in care from the upcoming weekend. Those admitted on a Friday afternoon or evening had the longest delays to catheterization (58.5 [13.8–67.1] hours) and longest LOS (3.8 [2.8–4.9] days). While those admitted on Friday morning had a very short time to catheterization (10.6 [4.5–72.0] hours), but this did not translate to earlier discharge (3.5 [2.2–5.1] days).
Baseline demographic, clinical factors, and hospital features associated with prolonged LOS (>4 days vs. ≤4 days) are displayed in Table 6. The factors most strongly associated with LOS >4 days in the adjusted model were undergoing cardiac catheterization >48 hours after presentation (odds ratio [OR] 4.87; 95% confidence interval [CI]: 4.54–5.22), shock at presentation (OR 7.82; 95% CI: 6.01–10.17), or heart failure without shock (OR 2.09; 95% CI: 1.93–2.26). Older age was also strongly associated with LOS >4 days with an OR of 1.14 (95% CI: 1.13–1.16) per every 5-year increase in age.
As compared to admission to the hospital on a Monday–Wednesday, admission on a Friday afternoon or evening (OR 1.13; 95% CI: 1.02–1.25) or on a Thursday (OR 1.50; 95% CI: 1.40–1.61) was associated with LOS >4 days. Those patients admitted to the hospital on a Saturday were less likely to have a prolonged LOS (OR 0.88; 95% CI: 0.81–0.96) when compared to those admitted to the hospital Monday–Wednesday.
Other comorbidities such as anemia, prior MI, diabetes, heart failure, stroke, peripheral arterial disease, and hypertension all were linked to LOS >4 days. Patients with a history of prior PCI were less likely to have prolonged LOS (OR 0.82; 95% CI: 0.77–0.88). Additionally, patients with private/HMO insurance were also less likely to have a prolonged LOS as compared to patients with all other types of insurance (Medicare, Medicaid), or no insurance at all (none/self-pay). Treatment at an academic medical center versus a non-academic center, a rural hospital versus an urban setting hospital, race of the patient, smoking status, and a history of CABG were variables tested and not associated with a prolonged LOS >4 days (Table 6).
Shown in Figure 4 are the differences in hospital LOS among four regions in the U.S. The Northeast appears to have the highest percentage of patients with an LOS of at least five days. On the other hand, the West appears to have the lowest percentage of those with LOS ≥5 days and the highest percentage of those with the shortest LOS ≤2 days. The median (25th, 75th) LOS were significantly different among the regions, although only to a mild degree. The shortest LOS was in the West region at 2.77 days (1.98–4.14), while the longest LOS was in the Northeast at 3.34 (2.52–5.14). The South had a median LOS of 3.10 days (2.22–4.63) and the Midwest had a median LOS of 2.97 (2.08–4.56); p<0.0001 for all comparisons.
This study examined national patterns of hospital LOS for over 39,000 NSTEMI patients in the ACTION Registry-GWTG. In our analysis, we found that, as expected, patients with longer LOS were older, more likely to have comorbidities and in-hospital complications, and were at a higher calculated baseline risk of death. However, these patients were less often treated with evidence-based medications despite a lack of documented contraindications and were less likely to receive cardiac catheterization or PCI within 48 hours, or PCI in general. We also found that hospital admission on a Friday afternoon or evening and delays to cardiac catheterization were among the factors most highly associated with prolonged LOS. Interestingly, hospital characteristics such as an academic medical center versus non-academic center, and a rural versus urban setting had no association with LOS after adjusting for other variables in the model. However, there does appear to be a geographic variation in LOS for reasons that are not clearly defined.
Despite higher incidences of heart failure, cardiogenic shock, and in-hospital MI, those with the longest LOS were less likely to receive cardiac catheterization or PCI within the first 48 hours of hospital admission and were more likely to undergo noninvasive stress testing. Paradoxically, the Timing of Intervention in Patients with Acute Coronary Syndromes (TIMACS) study, which evaluated an early intervention therapy versus a delayed intervention strategy, demonstrated that those patients with the highest mortality risk derive the most benefit from an early interventional strategy.13 In this study, the one-third of patients with the highest risk (Global Registry of Acute Coronary Events [GRACE] score >140) comprised the only subgroup that showed a significant reduction in the primary composite endpoint of death, MI, or stroke at six months when early intervention was implemented. These studies highlight the risk-treatment mismatch that is common in medicine.
This analysis suggests that those patients admitted on a Friday after 12:00pm have the longest delays to cardiac catheterization and longest LOS. This fact is presumably due to the upcoming weekend and reduced services available in most U.S. hospitals on Saturday and Sunday. Interestingly, those admitted on Friday morning had the shortest delays to cardiac catheterization, but also had the second longest LOS, behind only the Friday afternoon or evening group. This may reflect that there is a push to perform the cardiac catheterization before the upcoming weekend when a patient is admitted on a Friday morning, but that discharges are still delayed because of the weekend. A better understanding of the impact of diminished weekend services on discharge patterns could provide important insights into areas for improved healthcare resource utilization. For example, would routinely performing cardiac catheterizations during the weekend reduce hospital LOS? Would this be cost-effective? Although our study raises this possibility, this important question remains to be answered.
There is substantial heterogeneity in the patient population admitted for NSTEMI, which was related to wide variations in hospital LOS. Little is known about the factors that influence hospital LOS in NSTEMI. In an era of increasing pressure to save costs and focus on efficient use of resources, there is an important need to better understand factors that contribute to prolonged hospital stays and to appropriately limit their influence. Identifying the characteristics that predict LOS may provide insight into methods to reduce them in a safe manner. Older age was an important predictor of longer LOS in our population, and may in part be explained by the delays to catheterization and less use of PCI. Other studies have demonstrated similar trends with the use of cardiac catheterization among older patients. In an Italian population, the BLITZ-2 study surveyed the epidemiology and management strategies of non-ST-segment elevation acute coronary syndrome (NSTE ACS) in the Italian cardiological network and demonstrated that older patients were significantly less likely to undergo cardiac catheterization.14 Analysis of data from the international trials Sibrafiban versus Aspirin to Yield Maximum Protection from Ischemic Heart Events Post-acute Coronary Syndromes (SYMPHONY) and 2nd SYMPHONY found similar results. Both trials demonstrated that patients over 75 years old were significantly less likely to undergo cardiac catheterization, when compared to younger patients.15
These findings raise the question as to whether a strategy of earlier cardiac catheterization could result in a reduction in LOS for patients admitted with NSTEMI. There are many reasons why older patients with prolonged LOS and increases in in-hospital complications may not receive cardiac catheterization at all. Some of these reasons may include patient or family preference, increased risk of the procedure, or perceived lack of benefit. It is unclear whether or not a more aggressive strategy to perform cardiac catheterization on a higher portion of these patients would translate into a decreased LOS. However, for those in whom cardiac catheterization is felt to be appropriate, decreasing the time from admission to cardiac catheterization may result in shorter hospital stays.
Cardiogenic shock and heart failure on presentation were the factors most strongly associated with prolonged LOS, and keeping this high-risk population in the hospital for longer may be entirely appropriate. In part, prolonged LOS for this subgroup of patients is due to the heterogeneity of heart failure and shock in NSTEMI, which is rarely due to focal coronary lesions and is confounded by age and end organ dysfunction.
A better understanding is needed of our observation that there is a decreased use of evidence-based medications in patients with longer LOS who had more comorbidities and in-hospital complications. This finding is not inconsistent with previously published data. Wang et al. demonstrated that patients who experience bleeding complications after PCI are less likely to be discharged home on appropriated antiplatelet medications, and that this discrepancy continues well after the bleeding event for at least six months.16 Cohen et al. evaluated the racial and ethnic differences in the treatment of MI in 443 hospitals participating in the GWTG-CAD program and found that not only was the use of evidence-based medicines high in this group, but it continued to improve over the 5-year duration of the study.17 Importantly, studies have demonstrated that participation in systematic processes (such as the ACTION Registry-GWTG from which this study is conducted) is associated with improved adherence to evidence-based medicines. Nevertheless, efforts to better characterize medication use in this particularly vulnerable cohort, as well as increase the use of evidence-based medications where appropriate, are needed.
The ACTION Registry-GWTG provides retrospective observational data and, therefore, has a number of limitations. First, data are collected during acute hospitalizations only and do not provide longitudinal follow-up. Thus, we cannot examine the association of LOS with post-hospital outcomes. Second, exact discharge times were not available. Since most patients are discharged during daytime business hours, we assumed a discharge time of 12:00pm for all patients. As opposed to using strictly the dates of admission and discharge, it was felt that this would minimize the amount of error in calculating LOS for those admitted late in the day. Third, as in any registry, there is an inherent selection bias. Data comes from a heterogeneous mixture of hospitals that vary in the services and types of procedures they can provide. In an attempt to overcome this limitation, we restricted our analysis to only patients who received cardiac catheterization at the facility to which they initially presented, with the understanding that this would exclude all patients medically managed for NSTEMI and those transferred for cardiac catheterization. We used statistical methods to account for hospital characteristics and within hospital correlation. Restricting our population to patients undergoing cardiac catheterization also served to focus on patients for whom the care team was more convinced that the patient had true acute coronary syndrome, but with enough functionality to justify an invasive approach. Finally, the data collected in this registry captures contraindications to treatment and allows for exclusion of those patients in these analyses; however, relative or undocumented contraindications are not captured and cannot be accounted for.
This is one of the largest studies ever performed to examine the characteristics of patients admitted with NSTEMI and their association with hospital LOS. Patients with longer LOS were older and had more comorbidities and in-hospital complications, but were less likely to receive evidence-based medications and had delays in cardiac catheterization. Additionally, hospital admission on a Friday afternoon or evening and delays to cardiac catheterization were associated with prolonged LOS. A considerable amount of the heterogeneity was not accounted for by these variables, supporting the concept that there may be the opportunity to reduce LOS by simply applying consistent practice. Applying the results of this study to drive the development of processes to reduce heterogeneity in hospital LOS and associated treatment disparities in NSTEMI patients, as well as to promote safe and early discharge in an era of increasingly restrictive healthcare resources, should lead to more appropriate and efficient care.
Funding Sources: The ACTION Registry®-GWTG™ is an initiative of the American College of Cardiology Foundation and the American Heart Association, with partnering support from Society of Chest Pain Centers, The Society of Hospital Medicine, and The American College of Emergency Physicians. The registry is sponsored by Bristol-Myers Squibb/Sanofi Pharmaceuticals.
This project received infrastructure support from grant number U18HS010548 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
The authors would like to thank Erin LoFrese for her editorial contributions to this manuscript. Ms. LoFrese did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted.
Author Verification Statement: All authors had access to the data and a role in writing the manuscript.