Author + information
- Received December 20, 2010
- Revision received March 4, 2011
- Accepted April 18, 2011
- Published online August 1, 2011.
- Chadwick D. Miller, MD, MS⁎,⁎ (, )
- Wenke Hwang, PhD†,
- Doug Case, PhD‡,
- James W. Hoekstra, MD⁎,
- Cedric Lefebvre, MD⁎,
- Howard Blumstein, MD⁎,
- Craig A. Hamilton, PhD§,
- Erin N. Harper, BS⁎ and
- W. Gregory Hundley, MD‖,¶
- ↵⁎Reprint request and correspondence:
Dr. Chadwick D. Miller, Department of Emergency Medicine, Wake Forest School of Medicine, Medical Center Boulevard, Winston-Salem, North Carolina 27357
An abstract of this work was presented at ACC.11/i2 Summit in New Orleans in April 2011.
Objectives This study sought to compare the direct cost of medical care and clinical events during the first year after patients with intermediate risk acute chest pain were randomized to stress cardiac magnetic resonance (CMR) observation unit (OU) testing versus inpatient care.
Background In a recent study, randomization to OU-CMR reduced median index hospitalization cost compared with the cost of inpatient care in patients presenting to the emergency department with intermediate risk acute chest pain.
Methods Emergency department patients with intermediate risk chest pain were randomized to OU-CMR (OU care, cardiac markers, stress CMR) or inpatient care (admission, care per admitting provider). This analysis reports the direct cost of cardiac-related care and clinical outcomes (myocardial infarction, revascularization, cardiovascular death) during the first year of follow-up subsequent to discharge. Consistent with health economics literature, provider cost was calculated from work-related relative value units using the Medicare conversion factor; facility charges were converted to cost using departmental-specific cost-to-charge ratios. Linear models were used to compare cost accumulation among study groups.
Results We included 109 randomized subjects in this analysis (52 OU-CMR, 57 inpatient care). The median age was 56 years; baseline characteristics were similar in both groups. At 1 year, 6% of OU-CMR and 9% of inpatient care participants experienced a major cardiac event (p = 0.72) with 1 patient in each group experiencing a cardiac event after discharge. First-year cardiac-related costs were significantly lower for participants randomized to OU-CMR than for participants receiving inpatient care (geometric mean = $3,101 vs. $4,742 including the index visit [p = 0.004] and $29 vs. $152 following discharge [p = 0.012]). During the year following randomization, 6% of OU-CMR and 9% of inpatient care participants experienced a major cardiac event (p = 0.72).
Conclusions An OU-CMR strategy reduces cardiac-related costs of medical care during the index visit and over the first year subsequent to discharge, without an observed increase in major cardiac events. (Cost Comparison of Cardiac Magnetic Resonance Imaging [MRI] Use in Emergency Department [ED] Patients With Chest Pain; NCT00678639)
Chest pain observation units reduce costs of the index hospital visits for patients with chest pain (1,2) and are recommended for use in the American College of Cardiology/American Heart Association (ACC/AHA) guidelines (3). However, patients with intermediate risk chest pain and no definite evidence for acute coronary syndrome (ACS) are commonly admitted for inpatient care due to the difficulty of managing patients with prior coronary heart events, and the higher acuity of their presenting illness.
A recent randomized clinical trial in emergency department (ED) patients with intermediate risk chest pain demonstrated that observation unit (OU) care coupled with stress cardiac magnetic resonance testing (CMR) was associated with a median reduction in cost of $588 during the index hospitalization when compared the costs of inpatient hospital care (4). However, the cost of an index hospitalization may not adequately reflect resource consumption for a particular illness; also, the care delivered during the index hospitalization may influence downstream care patterns.
The objective of this study was to compare the cost of medical care, measures of resource utilization, and report the clinical outcomes in ED patients with intermediate risk acute chest pain in the year following randomization to either inpatient care or OU-CMR. We anticipated that the reduction in index hospital visit cost seen with OU-CMR would be followed by similar utilization patterns among groups after hospital discharge. We suspected that reductions seen with OU-CMR would occur without an increase in the rates of major cardiac events.
As previously described (4), we conducted a single-center randomized clinical trial. All participants provided written informed consent. The study was approved by the Institutional Review Board of Wake Forest University School of Medicine and complied with the Health Insurance Portability and Accountability Act.
From January 2008 through April 2009, participants were recruited from the ED of a tertiary care referral center that has an annual volume of 96,000 visits per year and is a level 1 trauma center. The ED is staffed by board-prepared/board-certified emergency physicians supervising care provided by residents in training. The observation unit is under the direction of the ED and is staffed by nurse practitioners and physician assistants with care provided jointly with emergency physicians. Cardiovascular medicine consultation is available at the discretion of the emergency physicians.
We enrolled individuals presenting to the ED with intermediate risk acute chest pain. Intermediate risk was defined as at least an intermediate probability that the patient's symptoms were related to ACS or a TIMI (Thrombolysis In Myocardial Infarction) risk score ≥2 calculated in the ED. Care providers were encouraged to use the ACC/AHA framework to formulate their clinical impression. Additional inclusion criteria were age ≥18 years, care provider impression that an inpatient assessment was indicated, and care provider assessment that the patient could be discharged if cardiac disease was excluded. Exclusion criteria were an initial troponin I level above the diagnostic threshold for myocardial infarction (MI), ST-segment elevation (≥1 mV) or depression (≥2 mV), contraindications to CMR, systolic blood pressure <90 mm Hg, inability to lie flat, refusal of follow-up procedures, <3 months life expectancy, pregnancy, renal insufficiency (such that the estimated glomerular filtration rate was <45 ml/min), chronic liver disease, or solid organ transplant.
To find eligible participants, the study staff screened ED patients with chest pain or related symptoms from 8 am to 11 pm Monday through Thursday and 8 am to 11 am on Friday. Consented participants were then randomized to 1 of 2 study arms: OU-CMR or inpatient care. OU-CMR participants were placed into the OU, received serial cardiac markers at 4 and 8 h from the initial marker, and had orders placed for a stress CMR study. One OU-CMR participant left against medical advice prior to OU placement. Inpatient care participants were consulted for admission following standard admission practices at the study institution. Study patients were admitted to the services of cardiologists (n = 29), internists (n = 17), family medicine physicians (n = 8), with 3 leaving against medical advice prior to admission.
OU-CMR imaging protocol
Imaging in OU-CMR participants has been previously described (4) and was similar to imaging used for clinical care at the study institution. In general, imaging was conducted with a 1.5-T Siemens Magnetom Avanto system (Siemens Medical Solutions, Munich, Germany). Initial orders were placed for an adenosine CMR, which included the imaging sequences noted in Table 1. Dobutamine stress was available as an alternative in the event adenosine was contraindicated. Images were interpreted by board-certified cardiology or radiology faculty with at least level II training in CMR (5).
Outcomes and definitions
The primary outcome of this analysis is the direct cost of cardiac-related health care through 1 year after randomization. Utilization measures used to calculate cost were cardiac-related hospital admissions, cardiac procedures, and cardiac-related office visits, further defined in Table 2. If a hospitalization or office visit met the definition of cardiac-related, the entire cost of the visit was considered cardiac-related. The secondary outcome is major cardiac events, composed of MI, revascularization, and cardiovascular death.
Myocardial infarction occurring at the study institution was defined as a troponin I >1.0 ng/ml in the presence of ischemic symptoms. Troponin I was measured in the central lab using either the TnI-Ultra assay using the ADVIA Centaur platform (Siemens) or the Access AccuTnI Troponin I Assay using the dxi800s platform (Beckman Coulter, Brea, California). Patients reporting major cardiac events outside the study institution were considered to have had the event.
Data collection and outcome measurement
After randomization, patients were followed during their index hospitalization by record review. Study staff then conducted a structured record review and telephone follow-up at approximately 30 days, 3 months, 6 months, and 1 year after randomization. Participants reporting hospitalizations occurring at another facility that were possibly cardiac-related had those records requested from the facility and reviewed.
During each telephone follow-up, a modified version of a previously described script was used to identify potential cardiac events (6). Participants were then asked to identify all healthcare encounters during the follow-up interval, which were then categorized as an office visit, an ED visit, a hospitalization, or a procedure. These patient reports were cross-referenced to those found in the medical record at the time of the phone contact. After completion of follow-up, itemized physician and facility billing records were obtained for each participant for the year following enrollment. Billing information was then manually paired with the encounters previously identified by record review and telephone follow-up. Encounters discovered through any of the 3 methods of identifying utilization (phone, medical record review, and billing record review) and classified as cardiac-related (Table 2) were included in the primary analysis. When available, exact dates were used for each component of resource consumption. If the date was not available, the date between the last contact and event reporting was used.
Follow-up was conducted with medical record and billing record review on all participants through 1 year. In addition, telephone attempts continued through the patient's enrollment in the study regardless of the success of the prior attempt. Follow-up time was included through 1 year for those patients with more than 1 year of follow-up. For those patients who could not be contacted at 1 year, we used data through the longer of either the last telephone contact or when telephone and medical records (including billing records) no longer demonstrated accumulation of cardiac-related services.
Cost comprised 2 elements: provider and facility cost. Provider cost was calculated by obtaining work-related relative value units associated with each charge and converting to dollars using the Medicare conversion factor. Facility cost was determined by applying to each charge the departmental-specific cost to charge ratios used to file cost reports with Centers for Medicare and Medicaid Services annually. Costs for events occurring at other institutions (8 of 246 events) were calculated using the mean cost for similar services consumed among other study participants at the study institution.
The sample size for this study was determined to provide adequate power for assessing the effect of OU-CMR on the cost of the index hospital visit, which was the primary objective of this trial (4). A cost model was constructed to assist in estimating the possible treatment effect and the variance of the index cost was estimated from pilot data. Based on this information, it was calculated that 110 participants would provide 90% power to detect a $2,000 difference in the index hospital visit cost between study groups, allowing for an 8% attrition rate.
Baseline demographics and clinical outcomes were compared using Fisher exact tests for proportions; medians of continuous variables were compared using Wilcoxon rank sum tests. The primary analysis compared the first-year cardiac-related costs for patients randomized to OU-CMR with those costs for patients randomized to usual inpatient care. A year was defined as the first 360 days following discharge, allowing us to look at costs accrued over 12 equal “months” of 30 days. Linear regression was used to assess group differences in total costs and costs following discharge, unadjusted and adjusted for patient characteristics. Covariates included the 2 stratification variables (known coronary disease [Yes/No] and time of day of presentation [6 am to 3 pm or 3 pm to 6 am]), age at enrollment, sex, confirmed prior heart failure, confirmed prior MI, confirmed prior coronary revascularization, and chest pain at ED arrival. The cost data were highly skewed so log transformations were used to reduce the skewness and to stabilize the variances between groups. A small constant of 1 was added to follow-up costs so that log transformations could be used when the costs were 0. The p values for group differences in cost are based on analyses of the log-transformed data. We then divided costs up into 12 monthly periods of 30 days and repeated measures models were used to assess changes over time and to determine if the changes over time varied by group. An unstructured covariance matrix was used to account for the within-patient correlation in costs over time.
Study population, study intervention, and follow-up
Enrollment was conducted over 15 months during which 180 patients appearing to meet inclusion and exclusion criteria were approached to consent for participation in the study. Twenty-five exhibited unrecognized exclusion criteria, and 46 refused to participate in the randomization process. The remaining 109 formed the study population accounting for 110 participant encounters. One participant was enrolled twice and the second encounter was excluded from this analysis, although the costs of the second encounter are included along with the costs associated with the initial counter for that participant. This analysis included 57 participants randomized to inpatient care and 52 randomized to OU-CMR. The baseline characteristics of study participants are shown in Table 3 and are not statistically different between groups.
Of the 52 OU-CMR participants, 51 were managed in the OU and 1 left against medical advice. Stress CMR testing was conducted in 48 of 52 participants over a median scan time of 53 min. Hospital admission was avoided in 41 OU-CMR participants. In the inpatient care group, 3 of 57 left against medical advice prior to admission and the remaining 54 were admitted. Once admitted, 9 inpatient participants received stress CMR as part of their clinical evaluation, 31 received stress echocardiography, 9 received cardiac catheterization, and 10 received no stress testing or catheterization. Several patients in both groups received multiple procedures.
All patients were followed through 360 days using medical record review and billing record analysis. Follow-up information after discharge was obtained in 98% of participants through review of medical and billing records, and phone interviews designed to identify or confirm the absence of events. Phone contact at 1 year or beyond was established in 81%, each of whom contributed 360 days of information. Remaining participants contributed follow-up data up to the last date of known consumption of cardiac-related services or the last date of phone follow-up. The mean follow-up time was 320.5 days for the inpatient care participants and 309.0 for the OU-CMR participants.
Cardiac-related healthcare utilization
After discharge from the index hospital visit, 15% of OU-CMR participants experienced at least 1 cardiac-related ED visit during the subsequent year compared with 37% in the inpatient care group (p = 0.02) (Table 4). Similarly, cardiac-related hospital admissions were significantly lower in the OU-CMR group (12%) compared with 35% in the inpatient care group (p = 0.01). No differences were seen in the number of cardiac-related outpatient visits per group (p = 0.30) with 83% of participants having between 0 and 2 visits over the year.
Cardiac testing procedures by study group are displayed in Table 4. During the index visit, cardiac catheterizations were performed in 13% of the OU-CMR group compared with 16% of the inpatient care group (p = 0.79). After discharge, no difference in the rates of stress testing were detected (OU-CMR 10% vs. inpatient care 16%, p = 0.40), but the inpatient care group had a higher rate of cardiac catheterization (19% vs. 2%) in the OU-CMR group (p = 0.01). During the incident visit and the subsequent 1-year follow-up, more OU-CMR participants received stress testing (96% vs. 79%, p = 0.01), whereas fewer received cardiac catheterization (15% vs. 33%, p = 0.04).
From randomization to completion of follow-up, 8 participants experienced 9 major cardiac events. In the OU-CMR group, 2 participants had major cardiac events during the index visit: 1 experienced an MI before CMR testing and had a percutaneous coronary intervention (PCI) and 1 had inducible ischemia on CMR testing and received coronary artery bypass graft surgery. At follow-up, 1 additional patient had MI without revascularization. In the inpatient group, 5 participants had major cardiac events during the index hospitalization; all had PCI, 1 had MI. At follow-up, 1 of these patients had an additional PCI. There were no statistical differences among groups in the occurrence of major cardiac events.
Cardiac-related healthcare costs
The distributions of cost by study group from enrollment through 1 year are shown in Figure 1, the cumulative accrual of cost after discharge is shown in Figure 2, and the data are summarized in Table 5. Costs were remarkably variable, with overall costs ranging from $616 to $34,084 and follow-up costs ranging from $0 to $17,698. Considering all cardiac-related costs from enrollment through 360 days, the geometric mean (95% confidence interval) cost was $3,101 ($2,519 to $3,817) among OU-CMR participants and $4,742 ($3,888 to $5,783) among inpatient care participants (p = 0.004). The difference remains significant once adjustment is made for covariates (p = 0.018). In the adjusted model, prior coronary artery disease (p = 0.047), prior heart failure (p = 0.049), no prior revascularization (p = 0.029), and pain at ED arrival (p = 0.046) were associated with higher costs. When only considering the cost of care after discharge from the index hospital visit, the geometric mean cost was $29 ($11 to $74) for OU-CMR participants and $152 ($63 to $366) among inpatient care participants (p = 0.012). This difference remains significant after adjustment is made for covariates (p = 0.049). In the adjusted model, only pain at ED arrival (p = 0.024) was associated with higher costs.
We then looked at the cardiac-related costs over time in both groups. The repeated measures analysis of covariance models showed a significant time (p < 0.001) and group (p = 0.012) effect with the costs for first month being significantly higher than for the other months and the average costs for the inpatient group being higher than for the OU-CMR group. Although it appears that the biggest difference between the 2 groups occurs in the first few months following the discharge, the group × time interaction was nonsignificant (p = 0.388). The only covariate that was significantly associated with higher cost in this model was having prior heart failure (p = 0.001).
The results of this study indicate that a CMR imaging strategy can be used in an OU to reduce the cost of care for patients typically managed in the inpatient setting. The reduction in costs seen with OU-CMR during the index visit was not associated with a “rebound” increase in consumption after discharge, as would be seen if testing or interventions were being deferred from the index visit. In contrast, the OU-CMR group continued to accumulate costs at a lower rate, due to fewer cardiac-related ED visits, hospitalizations, and catheterizations. In this study, 1-year costs are lower in those randomized to OU-CMR versus an inpatient admission at the time of their presentation to the ED.
Importance of considering care after hospital discharge
Within the 1-year period after the original ED visit, one can detect health-related expenditures that may be deferred from the index visit to the outpatient arena, account for changes in behavior resulting from the care pathway, and assess differences in clinical outcomes. This has particular importance in cardiac care as a previous study (7) has shown that PCIs are associated with an increase in subsequent care utilization. Our results suggest that OU-CMR reduces follow-up costs. The source of reduced cost relates to procedures and recidivism. Both groups had similar patterns of cardiac-related outpatient visits. However, the inpatient care group experienced higher rates of cardiac-related ED visits, rehospitalizations, and cardiac catheterizations. Nearly one-third of the inpatient care group received a cardiac catheterization by the trial's end, compared with only 15% of the OU-CMR group. These findings are consistent with those reported by Farkouh et al. (1) who also reported a reduction in cardiac-related utilization associated with OU care through 6 months of follow-up.
Observation unit care in patients with non–low risk chest pain
OU care is common for patients with chest pain, but is mostly used in patients with low pre-test probability for ACS. OU care is not commonly implemented in patients with non–low risk chest pain and the utility of OU care in this population is not well understood. Only 1 other trial (1) included patients with prior coronary events and randomized patients to OU care versus inpatient care. That trial by Farkouh et al. (1) allowed a variety of testing modalities in the OU group, including exercise electrocardiogram testing, stress echo, and stress nuclear imaging. They found that the cost savings associated with OU care at 6 months was dependent on using stress electrocardiogram testing without imaging; among the patients receiving stress imaging, there was no cost benefit to OU care (8). Our findings contradict their findings in that we demonstrated cost savings with OU-CMR care through the first year after presentation. This could mean that the cost reduction results from using stress CMR rather than OU care, or perhaps it is the combination of using CMR as an imaging modality within an OU that is lowering 1-year follow-up costs.
Despite a class I endorsement by the ACC/AHA guidelines (3), the safety of OU care is not well investigated in patients with non–low risk chest pain. In the OU arm of the trial by Farkouh et al. (1), 3 of 214 patients from the OU group died within 6 months of randomization, but 0 died in the inpatient group (p = NS). The OU group also experienced a total of 6 cardiac events between 30 days and 6 months from randomization compared with 1 in the inpatient group (p = NS). In our trial, 1 patient in each group had a major cardiac event, both between 6 months and 1 year following randomization. For OU-CMR participants, this is consistent with previous analyses demonstrating the excellent long-term prognosis after a negative stress CMR (9,10). The low event rates support our underlying assertion that an OU-CMR pathway should have similar safety to inpatient care because both pathways typically incorporate serial cardiac markers and objective cardiac testing. However, a larger trial should address the safety of OU care in this population.
Mechanisms for OU-CMR reducing cost
We propose several mechanisms that could, in part, contribute to the observed differences in costs between the treatment pathways. First, patients in the OU-CMR group could have been more stringently selected for revascularization during the index visit. As a result, the increase in post-revascularization cardiac-related care occurred less often. Second, patients in the OU-CMR group may have felt more reassured, reducing their likelihood of returning to the ED. Similarly, a recent CMR exam could have influenced physician behavior after discharge when deciding whether to pursue a patient's complaint of chest pain. This mechanism could also potentially account for the dramatic difference in catheterizations after discharge. Finally, it could be that inpatient care participants established relationships with subspecialists that in turn increased the likelihood that a subspecialist procedure would be performed.
Rationale for CMR use in observation unit care
CMR was felt by this study team to be the ideal imaging modality for OU implementation. The strength of CMR rests in its ability to provide a comprehensive evaluation without ionizing radiation. The comprehensive assessment in this protocol included a T2-weighted assessment for myocardial edema that often accompanies ACS, resting left ventricular wall motion, resting and stress perfusion to identify ischemia, and delayed enhancement to identify infarcts. Through this comprehensive assessment, clinicians are readily able to distinguish new infarcts or ischemia in the setting of pre-existing infarcts, left ventricular dilation, or hypertrophy. In addition, other causes of chest pain, such as myocarditis, or other processes that simulate infarcts (e.g., Takotsubo cardiomyopathy) can be readily identified.
Our study has some limitations. First, the data were collected from a single center and contained a modest number of participants. Future work will require a larger sample size from other sites to determine if these findings remain valid across healthcare systems and diverse patient populations. Second, the results were obtained in the tightly controlled setting of a clinical trial. The OU-CMR intervention may have a different effect when examined outside the clinical trial context. Third, the classification of healthcare encounters as cardiac-related or not cardiac-related was conducted without the immediate knowledge of subjects' study group assignments. However, the study group assignment could have been obtained, as this step required a record review. Objective definitions were used to standardize these assessments and minimize bias; however, this represents a potential threat to internal validity. Fourth, extensive medical and financial record reviews were performed to identify events after discharge, and further, most patients were contacted at 1 year to identify additional events. However, some participants did not contribute complete information. Follow-up did not differ by group and we feel it is unlikely that this introduced systematic error. Finally, there is the possibility that patient groups were unbalanced despite randomization. We stratified the randomization scheme based on prior coronary disease to prevent this from occurring, no statistical differences in baseline data were seen among study groups, and the findings remained significant after adjusting for covariates in our cost models. However, we cannot entirely exclude the possibility that differences in study groups contributed to our results.
Early findings demonstrate that an OU-CMR strategy is an efficient management strategy for ED patients with intermediate risk chest pain, but without definite ACS. In addition to reducing the index hospital visit cost, an OU-CMR strategy continues to reduce cost after hospital discharge leading to lower total cost at 1 year compared with inpatient care. The reduction in cost after discharge was the result of fewer cardiac catheterizations, cardiac-related ED visits, and cardiac-related hospitalizations.
Dr. Hwang is now affiliated with the Division of Health Services Research, Department of Public Health Sciences, Penn State Hershey College of Medicine, Hershey, Pennsylvania. This study was supported by grants from the Translational Science Institute of Wake Forest University School of Medicine; the National Institutes of Health (1 R21 HL097131-01A1 [to Dr. Miller] and 1 R01 HL076438 [to Dr. Hundley]); and the American Heart Association (0980008N [to Dr. Miller]). Dr. Miller has received research grants from EKR Therapeutics, Johnson & Johnson/Scios Inc., Chiron Corporation, and Novartis Pharmaceuticals; research support from Siemens; and has been an expert witness for the Law Offices of Wade E. Byrd, P.A., and Lewis and Oliver. Dr. Hoekstra has received consulting fees from Sanofi-Aventis and Verathon; and has served on the advisory boards of Merck & Co. Inc., Ortho-McNeil Pharmaceutical, Daiichi Sankyo, and AstraZeneca. Dr. Hamilton has stock in/ownership of Prova, Inc. Dr. Hundley has received research support from Bracco Diagnostics; and has stock in/ownership of Prova, Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Eike Nagel, MD, PhD, served as Guest Editor for this article.
- Abbreviations and Acronyms
- American College of Cardiology
- acute coronary syndrome
- American Heart Association
- cardiac magnetic resonance imaging
- emergency department
- myocardial infarction
- observation unit
- percutaneous coronary intervention
- Thrombolysis In Myocardial Infarction
- Received December 20, 2010.
- Revision received March 4, 2011.
- Accepted April 18, 2011.
- American College of Cardiology Foundation
- Anderson J.L.,
- Adams C.D.,
- Antman E.M.,
- et al.
- Budoff M.J.,
- Cohen M.C.,
- Garcia M.J.,
- et al.
- Curtis J.P.,
- Schreiner G.,
- Wang Y.,
- et al.
- Hundley W.G.,
- Morgan T.M.,
- Neagle C.M.,
- Hamilton C.A.,
- Rerkpattanapipat P.,
- Link K.M.
- Ingkanisorn W.P.,
- Kwong R.Y.,
- Bohme N.S.,
- et al.