Cost and Resource Utilization Associated with Use of Computed Tomography to Evaluate Chest Pain in the Emergency Department: The ROMICAT Study

Abstract

Background

Coronary computed tomography angiography (cCTA) allows for rapid non-invasive exclusion of obstructive coronary artery disease (CAD). However, concern exists whether implementation of cCTA in the assessment of patients presenting to the emergency room with acute chest pain will lead to increased downstream testing and costs compared to alternative strategies. Our aim was to compare observed actual costs of usual care (UC) with projected costs of a strategy including early cCTA in the evaluation of patients with acute chest pain in the Rule Out Myocardial Infarction Using Computed Tomography (ROMICAT I) study.

Methods and Results

We compared cost and hospital length of stay of UC observed among 368 patients enrolled in the ROMICAT I trial with projected costs of management based on cCTA. Costs of UC were determined by an electronic cost accounting system. Notably, UC was not influenced by cCTA results, as patients and caregivers were blinded to the cCTA results. Costs after early implementation cCTA were estimated assuming changes in management based on cCTA findings of presence and severity of CAD. Sensitivity analysis was used to test influence of key variables on both outcomes and costs.

We determined that in comparison to UC, cCTA-guided triage whereby patients with no CAD are discharged, could reduce total hospital costs by 23%, p < 0.001. However, when the prevalence of obstructive CAD increases, index hospitalization cost increases such that when the prevalence of ≥50% stenosis is greater than 28–33%, the use of cCTA becomes more costly than UC.

Conclusion

cCTA may be a cost saving tool in acute chest pain populations that have a prevalence of potentially obstructive CAD lower than 30%. However, increased cost would be anticipated in populations with higher prevalence of disease.

Background

Coronary artery disease is the leading cause of death in the United States, and emergency evaluation of acute coronary syndromes (ACS) alone accounts for 10 billion dollars annually in the United States.¹ Contributing to this cost is the fact that a misdiagnosis involving acute coronary syndrome is the number one cause of litigation for Emergency Room practitioners.²

Current clinical practice or usual care (UC) for evaluating patients with suspected ACS typically involves an approximate 12 hour stay on a hospital ward or chest pain unit to allow for serial evaluation of cardiac biomarkers and electrocardiograms (ECG), often followed by an array of diagnostic tests. Coronary CT angiography (cCTA) is an accurate non-invasive diagnostic test for assessment of symptoms of possible angina^{3, 4}, which has proven prognostic value in thousands of patients.^{5, 6} Recently several studies, including one single center⁷ and three multicenter trials ^8–10, have suggested that use of cCTA in the ER represents a rapid, and efficient method to evaluate patients with acute chest pain.^{8, 10–15}

When examining the cost of care, the CT-STAT (Coronary Computed Tomographic Angiography for Systematic Triage of Acute Chest Pain Patients to Treatment) trial⁸ demonstrated that cCTA reduced time to diagnosis and was associated with a 38% lower ED (not total) cost. When total hospital cost was examined in a subset of patients from the ROMICAT II study¹⁰, the mean cost of care between the cCTA group and SOC care were similar, although there was a trend toward more downstream testing among patients randomized to cCTA. Thus, while cCTA evaluation of acute chest pain may reduce ER cost and hospital stay, concern exists that cCTA may increase post-test resource utilization and downstream cost.^{8–10, 16–18}

The Rule out Myocardial Infarction by Computer Assisted Tomography (ROMICAT I) study utilized a unique design by which all patients and caregivers were blinded to cCTA results, thus actual observed patient care and costs were based upon UC and were not influenced by cCTA results. This design – which has not been used in any other cCTA studies – provides actual UC costs, yet contains real cCTA data for each patient which can be used to simulate the impact of various cCTA guided algorithms on costs, length of stay and resource utilization. Therefore, our aim was to compare actual costs of usual care (if cCTA was not available) with projected costs of a strategy of early cCTA in the evaluation of patients with acute chest pain.

Methods

Study Population

The study design and results of the ROMICAT I study have been described in detail previously.¹⁸ The study was approved by Partners Healthcare Institutional Review Board and all subjects enrolled in the ROMICAT study provided informed consent. Briefly, this single center, double-blinded observational cohort study enrolled 368 consecutive patients with acute chest pain, unremarkable initial troponin and ECG, and no known coronary artery disease who were awaiting hospital admission between May 2005 and May 2007. All patients underwent 64-slice contrast-enhanced coronary cCTA prior to hospital admission with caregivers and patients blinded to the results of the examination. Immediate (within the index hospitalization) and long term (2 year) outcomes included: additional noninvasive imaging tests during the index hospitalization, invasive coronary angiography, percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), ST-elevation myocardial infarction (STEMI), non-STEMI, and cardiovascular death. ACS was defined as either an acute myocardial infarction (i.e. patients developed a positive troponin during serial testing (6 hour or 9 hours after ED presentation) or unstable angina pectoris (UAP) according to the AHA/ACC/ESC guidelines.^19–21 UAP was defined as clinical symptoms suggestive of ACS (unstable pattern of chest pain - at rest, new onset, or crescendo angina), optimally with objective evidence of myocardial ischemia such as a positive stress test. Clinical events were ascertained using a standardized phone interview at 6 months and 2 years. Additionally, records were reviewed. A combined endpoint of MACE (cardiac death, MI, or coronary revascularization) was adjudicated by two physicians blinded to the cCTA data. Disagreements were resolved by consensus with a third experienced clinician.^{19, 22}

Cost Data

To determine all hospital related costs for each patient in our study we utilized the Eclipsys Sunrise Decision Support Manager (SDSM) system. SDSM is a cost accounting that tracks of all costs associated with each patient’s hospital admission at a line item level. From this system we obtained detailed costs for all services provided to the patient except for the actual physician fees. Compared to an approach solely based on Medicare reimbursement values, this approach is more accurate in describing actual hospital costs of all direct and indirect services provided and thus allows for a valid comparison between different treatment strategies.

Cost and Resource Utilization Under Usual Care Strategy

For each of the 368 patients included in the ROMICAT I study, cost under UC was derived by adding all line item costs for the entire hospital admission. Costs included costs in all units (ED, observation unit, and inpatient wards). For each patient, total cost was calculated by adding all direct and indirect costs incurred by each department. While direct costs represent all expenses that can be directly traced back to direct patient care (i.e. procedure, medications, nursing care) indirect costs represent overhead costs that are associated with each charge (e.g. equipment). Such costs are important to consider as they are used during the treatment of patients although may not necessarily relate to any one procedure. Cost of all cardiac imaging tests for each admission was calculated by adding the cost of each test that the patient underwent. All costs were converted to 2007 US$ using the consumer price index.

Length of stay (LOS) was calculated as the time difference from presentation to the ED until discharge and thus included both ED LOS, observation unit LOS (when applicable), and ward LOS.

Estimated Cost and Resource Utilization Using Coronary CT Strategy

Based on results of cCTA, patients were divided into the following groups:

• Negative Coronary CT -- no plaque and no stenosis (n=183)

• Plaque with no significant stenosis (n=117)

• Indeterminate stenosis (n=34) – stenosis could not be excluded

• Significant stenosis (n=34)

The presence of any plaque and the presence of significant coronary stenosis, which was defined as ≥50% luminal narrowing, was determined in a consensus reading by two experienced cardiac imaging investigators. All patients with presence of stenosis were further assessed using semi-automated software and subdivided in those with < and ≥70% luminal narrowing.²³

Proposed management and subsequent projected costs were determined for each subgroup as detailed below in Figure 1. Key underlying assumptions with supporting literature are summarized in Table 1.

Figure 1.

Alternate Scenarios. CAD = coronary artery disease; cCTA = coronary computed tomography angiography; D/C = discharge; ICA = invasive coronary angiography; MPI = myocardial perfusion imaging; Neg = negative; Trp = troponin.

Table 1.

Evidence from Prior cCTA Studies That Provides The Basis for Modeled Scenarios and Assumptions.

	Assumptions	Rationale/References	Limitations/ Counterpoint
1	Patients with no plaque or stenosis following cCTA can be safely discharged home and do not require any further testing.	Low event rate among normal cCTA, see Table 2.	N/A
2	Patients with non- obstructive (<50%) plaque can be safely discharged home if negative cardiac biomarkers.	- Hollander et al24: Among 481 patients with non- obstructive CAD (i.e. stenosis<50%) who were discharged home from ED, no patients had any cardiac death, MI, or coronary revascularization at 1year follow-up. - Goldstein et al8: CT STAT trial (multicenter randomized clinical trial comparing cCTA and MPI) utilized a study design that encouraged patients to be discharged home if stenosis < 25%. - Litt et al9: In the ACRIN-PA trial, 0 of 640 patients with cCTA 0 to < 50% stenosis had a cardiac event during 30 day follow-up.	A small proportion of patients with non- obstructive plaque may have ACS. Notably, among patients with negative serial cardiac enzymes in ROMICAT I, all such events are due to unstable angina.
3	Patients with moderate stenosis (50–70%) will undergo noninvasive myocardial perfusion imaging to determine if flow limiting disease; patients with ischemia will then be referred for invasive angiography.	- Hollander; Blankstein et al8, 25: Moderate stenosis on cCTA may be associated with ischemia and/or the presence of an acute coronary syndrome. MPI would be useful for further risk stratification. -	N/A
4	Among patients with moderate stenosis referred for MPI, 30% will have ischemia and will require further invasive testing; of those referred for invasive angiography 70% will undergo coronary revascularization.	- Goldstein et al 8: In CT-STAT, 4 out of 37 (11%) patients with intermediate / non-diagnostic cCTA were abnormal ; 3 of them had invasive angiography. - In current study 18 out of the 33 referred for invasive angiography underwent coronary revascularization; however when excluding the 6 patients that had no disease on cCTA, 18 out 27 (67%) patients referred for ICA required coronary revascularization.	Blankstein et al25: The number of patients with moderate stenosis that will have ischemia is variable in trials. Hachamovitch; Tonino 26, 27: Not all patients with ischemia require further invasive assessment as individuals with only mild ischemia may benefit from optimization of medical therapies.
5	All patients with severe stenosis (>70%) will be referred for invasive angiography	- Most patients with stenosis >70% will benefit from further testing. - Goldstein et al8: In CT-STAT it was recommended that patients with severe stenosis undergo invasive angiography (although only 9 out of 13 patients with stenosis > 70% were actually referred.	Referring all patients with >70% stenosis may result in over utilization of invasive angiography, and possibly PCI.7; some of these patients may benefit from initial MPI testing followed by selective use of invasive angiography.
6	50% of patients with stenosis on cCTA>70% who are referred for invasive angiography will undergo coronary revascularization	- In the current study, 8 out of the 9 patients with stenosis on cCTA > 70% referred for invasive angiography underwent coronary revascularization.
7	Patients with un- interpretable cCTA will be treated as >50% stenosis	- Consistent with analysis of ROMICAT I19 - Patients with lesions uninterpretable for stenosis will require at least some additional testing and are likely to undergo ICA or MPI	Will be compared to usual care

No Plaque, no stenosis on Coronary CT Angiography

In the ROMICAT I population^{19, 22}, the absence of plaque and stenosis by cCTA had a 100% negative predictive value for ACS and was associated with no MACE after two years. Based on these results as well as the findings from others ^{2, 28–30} one may confidently conclude that this sub-group of patients can be safely discharged home and do not require any further testing.

For each patient who had a negative cCTA, total cost (direct and indirect) was estimated based on SDSM data and using the following adjustments: (a) shorter length of stay as discussed below and (b) elimination of any costs related to cardiac imaging tests (exercise EKG stress tests, nuclear perfusion imaging studies, echocardiography, cardiac catheterizations). Therefore, costs for patients with normal cCTA results were calculated by using the real world costs from SDSM and subtracting any additional imaging or catheterization cost.

Estimated LOS was determined based on the following constraints and assumptions: (1) cCTA service was available at the time of ED triage, (2) Patients modeled as being safe for discharge following cCTA would have a 54% reduction in LOS, similar to the time reduction seen in cCTA trials.⁸

Presence of plaque, but no stenosis on Coronary CT Angiography

Given results of ROMICAT I, patients who have plaque on their cCTA – even if this does not result in significant stenosis –cannot completely be ruled out for ACS (e.g., spasm, , lysis of thrombus spontaneously or due to medical therapy). For the purposes of estimating resource utilization and cost, we modeled this subgroup by assuming they would undergo UC. Subsequent decisions on testing and patient management would then be based on actual care that took place without knowledge of the cCTA results.

Presence of stenosis on Coronary CT Angiography

Among patients with stenosis and acute chest pain, clinical reasoning would suggest that if positive enzymes are identified or if there are any other high-risk features (e.g. persistent symptoms, hemodynamic instability) then the patient should be referred for invasive coronary angiogram with possible PCI (Fig 1). If, on the other hand, cardiac markers are negative, further imaging testing should be obtained. An underlying assumption, which is supported by results from all studies to date, is that among patients with significant stenosis, some at lower risk may be evaluated with non-invasive testing while those at higher risk would proceed to invasive angiography. One potential strategy is to offer invasive angiography to all patients who are found to have a stenosis based on CT. The limitation of this approach is that patients with stenosis that is not hemodynamically significant would be referred for angiography / interventions. Such a strategy is unlikely to be economically attractive as it may result in increased use of coronary angiography when compared to UC.⁷ Therefore, the current study assumes that there will be a mix of both invasive angiography and non-invasive functional testing, as described by the scenarios described below. (Figure 1):

Baseline Scenario

• All patients undergo cCTA and those with normal scans (i.e., no plaque and no stenosis) are discharged home without additional testing, and accordingly a decreased LOS.

• All patients with plaque but no stenosis remain in the ER/observation unit for a total of 3 sets of cardiac biomarkers (12 additional hours in hospital). If biomarkers are normal, patients are discharged home with cost determined by usual care (any real world non-cCTA imaging costs are applied). If positive (n=3), admit for ICA with additional costs determined by usual care.

• All patients with ≥50% but <70% stenosis undergo MPI. If MPI is abnormal, ICA is performed and cost is determined by usual care.

• All those with ≥70% stenosis undergo selective ICA at a rate of what was performed in the CT-STAT⁸ trial (9 of 13 patients with ≥70% stenosis were referred to ICA; 4 patients were medically managed per physician’s discretion) [this data has not been published for the ACRIN-PA or ROMICAT II studies]

Alternate Scenario A

• All patients undergo cCTA and those with normal or <50% stenosis are discharged home without additional testing, and accordingly a decreased LOS.

• Management of ≥50% and ≥70% stenosis follow the baseline scenario

Alternate Scenario B

• All patients undergo cCTA and those with normal scans (i.e. no plaque and no stenosis) are discharged home without additional testing, and accordingly a decreased LOS.

• All patients with <50% stenosis remain in the ER/observation unit for a total of 3 sets of cardiac biomarkers (12 additional hours in hospital). If biomarkers are normal, patients are discharged home with cost determined by usual care (any real world non-cCTA imaging costs are applied). If positive (n=3), admit for ICA with additional costs determined by usual care.

• All patients with ≥50% stenosis remain in the ER/observation unit for a total of 3 sets of cardiac biomarkers (12 additional hours in hospital). If biomarkers are normal, patients undergo MPI. If MPI is normal, patients are discharged home. If MPI or biomarkers are abnormal, ICA is performed and cost is determined by usual care.

Alternate Scenario C

• All patients undergo cCTA and those with normal or <50% stenosis are managed according to the baseline scenario. Those with No CAD are discharged home without additional testing, and accordingly a decreased LOS.

• All indeterminate or >50% stenosis undergo MPI. Cost is determined by subtracting non-cCTA imaging cost from usual care except for cost of MPI.

• Those with >70% stenosis undergo ICA and cost are determined by adding catheterization cost to usual care, if not performed.

Sensitivity analyses

In ROMICAT I study, the prevalence of indeterminate or ≥50% stenosis was 68/368 (18.5%). A sensitivity analysis was performed to assess the impact of cost upon each scenario after varying the prevalence of ≥50% stenosis from 0 to 100%, in 1% increments.

Because some patients had a prolonged hospital stay for reasons not related to their chest pain or heart disease, a separate sensitivity analysis was performed to exclude patients with lengths of hospital stay greater than the 95% percentile for the cohort (5.083 days). Among these 18 patients, there were 5 patients with no CAD, 5 patients with <50% stenosis, and 8 patients with indeterminate or ≥50% stenosis.

Statistical analysis

Continuous variables were compared with ANOVA. Categorical variables were compared with chi-squared. P-values are two-sided with an alpha of 0.05. All statistics were performed with Stata 12.0 (College Station, Texas).

Costs were indexed as a percent of total cost observed under UC (real-world) as determined from the line item hospital costs discussed above. The purpose of indexing costs to UC was to evaluate how differences in the management strategies compare to UC. Actual cost data was not reported as per institutional policies. Costs were compared by medians (Kruskal-Wallis test).

Results

cCTA Exam Results

cCTA findings in the ROMICAT I study included 183 (50%) possible ACS patients with normal cCTA, 117 (32%) with <50% stenosis, 34 (9%) indeterminate, and 34 (9%) with ≥50% stenosis. Among the 34 patients with ≥50% stenosis, 18 had ≥70% stenosis.¹⁹ Patients without significant CAD were younger and had fewer cardiac risk factors (Table 2).

Table 2.

Demographics.

	No CAD	<50% CAD	Indeterminate or ≥50% CAD	All Patients	p-value*
n	183	117	68	368
Age	47(9)	56(10)	62(12)	53(12)	< 0.001
% Male	55%	66%	71%	226(61)	0.006
Hypertension	26%	43%	71%	145(39)	< 0.001
Hyperlipidemia	22%	49%	54%	135(37)	< 0.001
Diabetes	7.1%	7.7%	27%	40(11)	<0.001
Family History	21%	22%	35%	89(24)	0.04
Current Smoker	25%	26%	25%	93(25)	0.87
Aspirin Therapy	27%	32%	46%	117(32)	0.02
Statin Therapy	17%	34%	47%	103(28)	< 0.001
Anti-hypertensive	25%	43%	71%	143(39)	< 0.001
ACS
STEMI	0	0	1	1	0.04
NSTEMI	0	3	4	7	0.001
UA	0	4	19	23	<0.001

^*p-value for trend across all groups.

Observed Resource Utilization under Usual Care

Under usual care in the ROMICAT I study, providers (who were blinded to cCTA findings) performed 68 SPECT, 72 ETT, 20 transthoracic echo, and 6 ICA on those (n=183 patients) with no CAD by cCTA; 50 SPECT, 34 ETT, 9 transthoracic echo, and 10 ICA on those (n=117) with <50% stenosis by cCTA; 35 SPECT, 11 ETT, 15 transthoracic echo, 1 transesophageal echo, 1 exercise stress echo (ESE), and 17 ICA on those with ≥50% stenosis or indeterminate scans. Table 3 summarizes the actual use of diagnostic testing according to cCTA finding versus the modeled use (only baseline scenario shown).

Table 3.

Diagnostic Test Utilization in Usual Care and Baseline Scenario.

	Usual Care n=368			Coronary CT n=368
cCTA Outcome	---	No CAD (N=183)		<50% CAD (N=117)		Indeterminate or ≥50% CAD (N=68)
Cardiac Testing	UC	UC	Baseline	UC	Baseline	UC	Baseline
Coronary CT	0	0	183	0	117	0	68
Invasive Angio	33	6	0	10	10	17	17
SPECT	151	68	0	50	50	35	35
Exercise Stress Test	117	72	0	34	34	11	11
Rest Echo	44	20	0	9	9	16	16
Stress Echo	1	0	0	0	0	0	0
No Testing	74	36	0	25	0	15	0
Hours Stay	26.2	24.4	11.2	25.9	25.9	34.8	34.8
Indexed Cost	100%	100%	82%	100%	82%	100%	114%

Projected Costs based on cCTA

Figure 2 compares modeled cost indexed as a percentage of actual spending based on different patterns of testing while holding LOS unchanged. Compared to real world costs, total costs for the baseline scenario were 92%, Alternate Scenario A 92%, Alternate Scenario B 91%, Alternate Scenario C 93% (all scenarios differed significantly from actual cost, p < 0.001; all scenarios differed from each other, p < 0.001).

Figure 2.

cCTA Cost (Projected) Indexed. Costs were indexed by dividing the real-world costs the projected costs from each scenario within each subgroup (i.e., no plaque/no stenosis, plaque without stenosis, plaque with indeterminate stenosis, or plaque with significant stenosis) and reporting the percent. CAD = coronary artery disease; <50% CAD = <50% worst coronary stenosis (non-obstructive CAD); ≥50% CAD = ≥50% worst coronary stenosis (obstructive CAD). All scenarios differed significantly from actual cost, p < 0.001; all scenarios differed from each other, p < 0.001.

When cost savings that are expected to result from decreased LOS are added to each model (Figure 3), more pronounced differences in cost are observed. Specifically, the cost of the baseline scenario was 85%, Alternate Scenario A 77%, Alternate Scenario B 84%, and Alternate Scenario C 85% (all scenarios differed significantly from actual cost, p < 0.001; all scenarios differed from each other, p < 0.001) (Figure 3).

Figure 3.

cCTA Cost (Projected) Discounted for Shorter ER Stay. A reduction in length of stay (LOS) has been modeled for patients with No CAD in all scenarios and for those with <50% CAD in Alternate Scenario A. CAD = coronary artery disease; <50% CAD = <50% worst coronary stenosis (non-obstructive CAD); ≥50% CAD = ≥50% worst coronary stenosis (obstructive CAD). All scenarios differed significantly from actual cost, p < 0.001; all scenarios differed from each other, p < 0.001.

For each scenario, significant cost savings occur in the patients with normal or <50% stenosis because of both a shorter LOS and the fact that under UC many of these patients with no or mild CAD underwent testing that could be averted under cCTA care. For example, under UC 68 patients with No CAD (n=183) by cCTA underwent SPECT and 6 underwent catheterization. On the other hand, cost for the patients with significant or indeterminate CAD was significantly higher in all scenarios since cCTA would increase test layering without reducing LOS.

After including modeled reductions in ED time (Figure 3), the relative cost for patients with no CAD was 63% of UC across all scenarios (p<0.05 versus UC) while for patients with non-obstructive CAD, the relative cost was reduced to 56% of UC (p<0.05) in Alternate Scenario A. However, among patients with obstructive CAD, cost was increased by 10–14% (p<0.05 versus UC), depending upon the modeled options for management.

Sensitivity Analysis

Figure 4 depicts the cost of each scenario indexed versus actual cost in the ROMICAT I study (100% value). With decreasing prevalence of obstructive CAD, cost savings occur. As prevalence of ≥50% stenosis increases beyond 28% for the baseline scenario, 33% for Alternate Scenario A, 30% for Alternate Scenario B, and 28% for Alternate Scenario C, the triage scenarios using cCTA for initial triage of possible ACS increase cost when compared to the observed UC costs from the ROMICAT I study. After excluding outliers for prolonged hospital stay (>5.083 days), there was no significant change in these findings (data not shown).

Figure 4.

Cost as a Function of Increasing Obstructive CAD Prevalence. With increasing prevalence of obstructive CAD (x-axis), the cost of care following imaging by cCTA increases due to increased diagnostic tests and medical therapies (y-axis). Depicted are the projected (modeled) costs for several large studies including the first randomized trial of cCTA for acute chest pain in the ED (Goldstein et al 2007⁷), ROMICAT I¹⁰, CT-STAT⁸, and the ACRIN-PA⁹ trial. cCTA findings of prevalence of obstructive CAD have not been reported yet for ROMICAT II. As prevalence of ≥50% increases beyond 28% for the baseline scenario, 33% Alternate Scenario A, 30% Alternate Scenario B, and 28% Alternate Scenario C, the triage scenarios using cCTA for initial triage of possible ACS increase cost when compared to the actual data under usual care from the ROMICAT I trial. CAD = coronary artery disease.

Discussion

In this cost analysis of the ROMICAT I study we found that despite use of more imaging tests, initial triage of possible ACS with cCTA would be expected to result in total hospital stay cost savings of approximately 15–23% (Figure 3). Given that 10 billion dollars¹ are spent each year in the U.S. on the evaluation of acute chest pain, on aggregate, these findings could translate into significant cost savings.

Few reports have compared the cost of a cCTA versus UC strategy in the ER for acute chest pain, and there are noteworthy limitations of these analyses. For instance, some of them only consider ER but not total hospital cost⁸ while others evaluate ER and total hospital cost.¹⁰ Similarly, simulation models of projected cost may be limited to ER or procedural costs and not inclusive of total hospital cost.³¹ The ROMICAT I study design is uniquely suited for observing actual cost of care since physicians were blinded to cCTA findings and subsequent patient management was based exclusively on the ER and subsequent care team discretion.

Because of the lack of trials comparing cost-effectiveness of various CAD diagnostic modalities, the United States National Institutes of Health (NIH) has invested millions of public health research dollars in large trials comparing outcomes of different testing options. For example, the PROspective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial³², which has nearly completed enrollment of 10,000 outpatients with stable symptoms, is randomizing patients to a strategy of anatomical testing with cCTA versus functional testing (including exercise ECG, stress echocardiography, and stress nuclear myocardial perfusion imaging). The hypothesis of this study is that information derived from cCTA will lead to a lower composite end-point of death, myocardial infarction, major complications from cardiovascular procedures, or unstable angina requiring hospitalizations. As a secondary end point, this study will also examine cost. However, the PROMISE trial is evaluating a population with stable angina, which is very different from the acute chest pain population that is the subject of our analysis. The enrollment of the PROMISE study is expected to complete in August 2014, although the results will not be reported until at least 2016.

The anticipated cost reductions of our study are based upon eliminating diagnostic tests that are unlikely to occur after cCTA (SPECT, ETT, invasive angiography) and expected reduction in LOS, particularly among patients with no or mild CAD. A fundamental assumption to our study, and to all studies conducted to date assessing the economic consequences of imaging in the emergency department, is that testing is only performed in patients for whom further imaging is deemed clinically necessary. Although it could be argued that some patients could be triaged with no testing^{33, 34}, given the malpractice risk of “missed MI,” avoiding testing across a large proportion of patients with acute chest pain may not be realistic in all practice environments.³⁵ When considering a strategy of no testing, one option might be to discharge all patients with three negative troponins. However, in the ROMICAT study, 123 (57%) of the patients with negative serial troponins had at least mild CAD and 53 (25%) had significant stenosis or indeterminate scans, suggesting that such a strategy would risk missing potentially significant CAD. Notably, 22 of these patients (18%) were diagnosed with unstable angina (blinded to cCTA). This strategy also would be burdened by the cost of a longer period of observation for complete assessment by serial troponins, unless highly sensitive troponins allow for earlier triage time.³⁶

Given that the strongest influence on cost of care is LOS, it is important to recognize that assumptions of any given scenario regarding LOS will have a significant influence on cost. Moreover, it is important to recognize that the true cost of LOS is highly variable among different hospital systems as well as within departments of a given hospital (ER versus intensive care unit, versus observation ward). Recognizing these facts, we aimed to use a very conservative method to estimate the projected LOS under cCTA care and only estimate a percent reduction based on the actual LOS. Another option could be to use a fixed LOS (based on prior published studies or data on time to diagnosis by cCTA that are available in ROMICAT I) for patients with no CAD on cCTA who would be discharged directly from the ER. However, this technique would likely bias the analysis significantly in favor of cCTA since some patients in ROMICAT I study had non-cardiac medical problems in addition to their cardiac evaluation. Thus, we chose to use a percent reduction in LOS in order to provide a more conservative estimate of the cost.

Notwithstanding these findings, our study does have inherent limitations. First, since all patients received UC, we used projected rather than actual resource utilization following cCTA to inform our cost modeling. While our assumptions are based on real world data from the use of cCTA in this population as well as the results of other published trials, it is possible that even following a normal cCTA some clinicians may feel compelled to order additional testing^{7–10, 19}. Furthermore, the results of any scenario may differ from real world care. For instance, in the ROMICAT II trial, the potential in-hospital cost savings that would be predicted by our analysis were not realized, primarily due to a higher rate of ICA and revascularization in the cCTA arm. Second, the ROMICAT I study, like most studies of cCTA in the ED, enrolled hemodynamically stable patients with non-acute initial ECG and normal initial troponin. Importantly, these results should not be extrapolated to patients with hemodynamic instability or those patients presenting to the ED with acute chest pain who have not undergone initial ECG and cardiac biomarker laboratory testing. Third, we applied a model with 24 hour availability of cCTA and thus our findings are only applicable to patients who present to the hospital when cCTA is available. In systems where large numbers of patients present with acute chest pain after day time hours, the lack of available cCTA could negate some of the potential cost savings, as patients waiting to undergo cCTA would have an increased length of stay. Fourth, there was a 9% rate of indeterminate cCTA which is higher than that observed in other studies. We modeled these patients as having ≥50% disease (since many of these patients would be managed as if they have disease) although many may not have significant CAD. The net effect of this conservative assumption would be to increase cost of the cCTA strategies modeled versus UC, since the prevalence of indeterminate cases is generally lower in contemporary studies. Fifth, our cost analysis does not address radiation dose, and increased use of cCTA will increase radiation exposure to patients, the public health consequences of which are not known.¹⁰ Finally, we cannot account for the potential for cCTA to affect rates of downstream testing, particularly ICA^{37, 38}, since physicians were blinded to the cCTA results. Although study of a large Medicare database concluded that cCTA may increase downstream catheterization compared to SPECT or ESE¹⁷, this finding occurred in a population that has a mean age of 74, which is older than that published ED trials.³⁹ In such a population, the prevalence of CAD (and of non-evaluable scans) would be higher and, as our results demonstrate, is likely to be associated with increased cost.

Our analysis demonstrates potential for cost-saving particularly among the subgroups with no or mild CAD. However, it must be realized that there are many factors that drive hospital costs. While our analysis (which is based on the hospital’s point of view) focuses on minimizing unnecessary expenditures, due to potential perverse incentives of a fee-for-service system,⁴⁰ some payment structures may encourage more testing, therefore creating a potential barrier to cost-effective use of cCTA.

An important consideration remains that cCTA may increase cost when implemented for populations with high prevalence of ≥50% stenosis or among groups in which stenosis cannot be excluded. As shown on Figure 4, when the prevalence of ≥50% or indeterminate stenosis increases beyond 28–33% (depending on which scenario used) the cost of cCTA exceeds that of UC. Nevertheless, given that the current prevalence of obstructive CAD is low among most patients considered suitable for cCTA evaluation,⁴¹ our findings demonstrate that if used appropriately, in aggregate, cCTA is likely to be cost savings across most populations of acute chest pain patients in whom further testing is required and who are expected to have adequate image quality.

Abbreviations

ACS

acute coronary syndrome

CABG

coronary artery bypass graft surgery

CAD

coronary artery disease

cCTA

coronary computed tomography angiography

ECG

electrocardiogram

NSTEMI

non-ST elevation myocardial infarction

PCI

percutaneous coronary intervention

ROMICAT I

rule out myocardial infarction by computer assisted tomography I

STEMI

ST-elevation myocardial infarction

UC

usual care