a b s t r a c t

Objective: The study sought to evaluate changes in mortality and resource utilization in patients with low level Troponin elevations following a reduction in the cutoff for normal troponin I from 0.5 ng/mL to the 99th per- centile (0.06 ng/mL).

Methods: This was an interrupted time series comparing emergency department (ED) patients with possible acute coronary syndrome (ACS) and TnI values 0.06-0.5 ng/mL before and after an institutional decrease in the TnI cutoff. The primary outcome was overall mortality at 90 days. Secondary outcomes included rates of re- hospitalization, subsequent ACS, and coronary intervention within 90 days, as well as rates of anticoagulation, cardiology consultation, cardiac testing, and coronary intervention during the index visit. Outcomes for the pre-cutoff change group (control) and post-cutoff change group (post) were compared using tests of proportions and odds ratios.

Results: The study included a total of 1058 subjects with 529 in each cohort. No significant differences in 90 day outcomes were observed between groups, including mortality (13.2% post vs 14.1% control, OR 0.93 [95% CI: 0.65-1.34], p = 0.705). During the index visit, the post-group demonstrated higher rates of cardiology consulta- tion (55.4% vs 41.2%, OR 1.77 [1.39-2.26], p b 0.0001) and cardiac Stress testing (16.4% vs 10.6%, OR 1.66 [1.16- 2.38], p = 0.006), but no significant differences in coronary intervention or short-term mortality were observed. Conclusion: A reduction in the TnI cutoff to the 99th percentile did not change mortality or rates of coronary in- tervention in ED patients with low level Troponin elevations, but significantly increased the use of cardiology resources.

(C) 2018

Introduction

Background

Acute coronary syndrome (ACS) is a critical entity in emergency and acute care that carries significant morbidity and mortality. ACS consti- tutes N780,000 cases annually in the US, almost 75% of which are further classified as non-ST elevation myocardial infarctions (NSTEMI). Elevat- ed cardiac troponin levels define the diagnosis of MI, but there is consid- erable heterogeneity among the various troponin assays commercially available [1,2]. As early as 2000, consensus statements and clinical prac- tice guidelines began recommending a standardized Upper reference limit for normal cardiac troponin defined as the 99th percentile of a normal reference population for a given assay [2,3].

? Conflicts of Interest/Funding/Disclosures: None.Prior Presentations:

SAEM Mid-Atlantic Regional Meeting; Philadelphia, PA (February 2014).

SAEM Annual Meeting; Dallas, TX (April 2014).

* Corresponding author.

E-mail address: [email protected] (B.A. Becker).

Adoption of the recommended 99th percentile URL creates a category of low level or “gray zone” troponin elevations with values falling above the 99th percentile URL, but below the previously used cutoff value. Com- pared to patients with definitive troponin elevations above the predefined cutoff URL, ACS patients with low level troponin elevations re- ceive fewer therapeutic interventions, experience more recurrent MIs, and exhibit greater mortality [4,5]. Evidence suggests a lower diagnostic threshold for troponin improves outcomes in these patients. Mills et al. demonstrated a decrease in troponin cutoff from 0.2 ng/mL to 0.05 ng/ mL significantly reduced the risk of death and recurrent MI in patients with suspected ACS, particularly in those with a peak troponin level in the study’s “gray zone” range (0.05-0.19 ng/mL) [6].

In accordance with Guideline recommendations the diagnostic threshold for the troponin I assay in use at our institution was lowered from 0.5 ng/mL to 0.05 ng/mL, generating a “gray zone” TnI level of 0.06-0.5 ng/mL. Based on previously published research, we hy- pothesized that patients with TnI levels 0.06-0.5 ng/mL would receive increased cardiovascular testing/intervention and demonstrate de- creased mortality compared to patients with similar troponin levels prior to the reduction in TnI cutoff.

https://doi.org/10.1016/j.ajem.2018.02.001

0735-6757/(C) 2018

Goals of this investigation

The purpose of this study was to assess 90 day outcomes and in-hos- pital rates of cardiology consultation, diagnostic testing, and cardiovas- cular intervention in emergency department (ED) patients presenting with symptoms concerning for possible acute coronary syndrome (ACS) and “gray zone” TnI values (0.06-0.5 ng/mL) following an institu- tional decrease in the diagnostic threshold for a normal TnI.

Materials and methods

Study design and setting

The study was a retrospective interrupted time series comparing two cohorts of consecutive adult patients presenting to the ED with po- tential ACS and a TnI value in the “gray zone” range (0.06-0.5 ng/mL) before and after a change in the diagnostic threshold for TnI. On Novem- ber 11, 2011, the URL for the ADVIA Centaur CP TnI-Ultra assay (Sie- mens Medical Solutions USA, Inc., Malvern, PA) was decreased from

0.5 ng/mL to the prior 99th percentile (0.06 ng/mL) across the institu- tion [7]. Patients were identified from 3 discrete time intervals: (1) The 6-month period before the change in normal TnI threshold (May 2011 to October 2011), (2) the 3-month period following the change in TnI threshold (November 2011 to February 2012) and (3) a subse- quent 3-month period starting one year after the change in TnI thresh- old (November 2012 to January 2013). The control cohort comprised patients presenting prior to November 11, 2011 (Interval 1) and the post-cohort consisted of patients presenting on or after November 11, 2011 (Intervals 2 and 3). We chose an interrupted time series design to assess both the immediate and sustained effects of the TnI cutoff change on resource utilization and outcomes.

The study was conducted at a university-affiliated, community teach- ing hospital with chest pain center accreditation and an annual ED census of approximately 81,000 patients. Patients with potential ACS were hos- pitalized under the internal medicine, Family Medicine or medical intensivist services. The cardiology consultation service was comprised exclusively of board-certified attending cardiologists. The decision to ob- tain cardiology consultation or cardiovascular testing was left to the dis- cretion of the admitting physician. Options for non-invasive provocative cardiac testing for hospitalized patients at the institution include exercise and chemical stress tests with echocardiography and nuclear imaging, as indicated. Stress testing was typically ordered by the admitting service with or without consultant input, while invasive cardiac testing and in- terventions were initiated exclusively by a cardiologist. Coronary com- puted tomography angiography (CCTA) was not performed.

The study team consisted of two attending EM physicians, five EM residents, one clinical research manager and six undergraduate research assistants. Study design and execution were performed in accordance with previously published guidelines regarding retrospective and ob- servational studies [8-10]. Institutional review board approval was ob- tained with waiver of consent.

Troponin I assay

Blood TnI concentrations were measured using the ADVIA Centaur CP TnI-Ultra assay (Siemens Medical Solutions USA, Inc., Malvern, PA) with analytic range 0.006-50 ng/mL and a 10% coefficient of variation of 0.03 ng/mL. Prior to November 11, 2011 the clinical decision point was 0.5 ng/mL. Starting November 11, 2011, the clinical decision point was decreased to 0.06 ng/mL based on a 99th percentile cutoff [7].

Selection of participants

An initial query of electronic laboratory records generated a master list of consecutive ED patients with a TnI of 0.06-0.5 ng/mL between May 2011 and February 2012 and between November 2012 and

February 2013. Patients were eligible for inclusion if they were at least 18 years old and demonstrated a TnI value 0.06-0.5 ng/mL. Patients were excluded for active DNR, DNI or “comfort care only” status at the time of ED assessment, initial presentation consistent with ST-elevation myocardial infarction , lack of an initial EKG or a chief complaint other than chest pain, dyspnea, generalized weakness, syncope/near syncope, or upper abdominal pain. These specific complaints were cho- sen to include atypical ACS presentations while excluding patients with conditions for which TnI is often obtained, but ACS is not a primary clin- ical concern, such as sepsis and altered mental status.

Measurements

Data were abstracted from the EMR utilizing a standardized closed- response data collection form. We conducted a comprehensive chart re- view of eligible patients, including a review of physician and nursing notes, laboratory values, diagnostic testing reports, operative notes, discharge summaries and subsequent follow up documentation. Basic patient demographics were collected, including age, gender, race, baseline creatinine (Cr), and TnI value. Relevant comorbidities were re- corded, including chronic kidney disease (CKD), coronary artery disease (CAD), diabetes mellitus , hyperlipidemia, hypertension (HTN), and history of current or prior smoking. Data for continuous variables (age, Cr, TnI) were collected via manual input and data for categorical variables (all others) were collected using a multiple choice, closed re- sponse format.

To assess outcomes at 90 days, charts were reviewed for subsequent visits to outpatient facilities, EDs, and hospitals within our institutional health system. As determination of outcomes at 90 days relied on docu- mentation in the EMR, we anticipated that in some cases this informa- tion would be missing or otherwise unobtainable. In such instances, the corresponding 90 day outcomes were deemed “unknown”. Missing mortality data triggered a manual search for the corresponding patient in the Social Security Death Index (SSDI). Mortality was ultimately deemed unknown if death was not reported in the SSDI within one year of the index visit.

All members of the research team participated in chart review and data collection after one-on-one instruction by the principal investiga- tor (PI). Following abstraction of 10-20 charts, each abstractor met with the PI and reviewed the collected data to ensure accuracy, consis- tency and proper adherence to definitions. A random selection of charts underwent redundant abstraction by a second different study team member to assess for Interobserver agreement. In cases of disagreement between abstractors, final data were determined via adjudication by the primary author (BB) and clinical research manager (BS).

Outcomes

The primary outcome of the study was mortality at 90 days. Second- ary outcomes consisted of additional outcomes at 90 days, including re- hospitalization following discharge from the index visit, as well as a di- agnosis of MI/ACS, percutaneous intervention (PCI) or coronary bypass grafting (CABG) during subsequent Hospital visits. MI/ACS was defined as a discharge diagnosis of ACS, acute MI, or unstable angina, as docu- mented by the treating physician in the discharge summary.

Further secondary outcomes assessED resource utilization during the index visit, including full anticoagulation with cardiac-dose heparin, cardiology consultation, stress testing, cardiac catheterization, PCI, and CABG. Stress tests were considered positive in the presence of inducible ischemia, as documented by the interpreting cardiologist.

Sample size

Based on prior research, the reduced TnI threshold was expected to decrease mortality at 90 days by 70% in patients with a “gray zone” TnI value [6]. Pilot data at our institution suggested a baseline mortality of

4% in this population. Using these assumptions, it was calculated a priori that 522 patients would be required per study cohort to detect the an- ticipated difference in mortality at 90 days with a type I error (?) of

0.05 and power (1-?) of 80%. This sample size also permits 80% power

to assess for the anticipated difference in subsequent diagnosis of MI or ACS at 90 days [6].

Data analysis

Analysis of baseline characteristics was descriptive with continuous and categorical variables reported as means with standard deviations (SD) and percentages, respectively. All outcomes were dichotomous, categorical variables and proportions between the study cohorts were compared using the ?² test. Odds ratios with associated 95% confidence intervals were calculated with the control cohort serving as the refer- ence. Unknown or missing 90 day outcome data that could not be ob- tained from the SSDI were excluded from statistical analysis. All tests were two-tailed and p-values b 0.05 were considered statistically signif- icant. Data compilation was performed using Remark Office OMR 7 (Gravic, Malvern, PA) and Microsoft Excel 2010 (Microsoft, Redmond, WA). Data analysis was performed using IBM SPSS Statistics for Win- dows version 22.0 (IBM Corp, Armonk, NY) and MedCalc Statistical Soft- ware version 12.5.0 (MedCalc Software bvba, Ostend, Belgium; https:// www.medcalc.org; 2017).

Results

Characteristics of study subjects

We reviewed a total of 1905 patient charts and excluded 847. Of the 1058 included patients, each cohort contained 529. The control cohort consisted of 529 patients from interval 1 and the post-cohort consisted of 336 patients from interval 2 and 193 patients from interval 3 (Fig. 1). Baseline characteristics across each cohort and all three intervals were similar; although patients in the control cohort were more likely to be female. Rates of hospitalization between cohorts were equivalent. Preexisting CKD and a final diagnosis of ACS/MI were slightly more common in the post-cohort (Table 1).

Outcomes for index visit/hospitalization (Table 2)

Overall, the post-cohort demonstrated significantly higher rates of anticoagulation (14.2% vs 8.3%, OR 1.82 [1.23-2.70], p = 0.003) and car- diology consultation (55.4% vs 41.2%, OR 1.77 [1.39-2.26], p b 0.001). These differences were largely maintained at one year. The use of stress testing was significantly higher in the post-cohort (16.4% vs 10.6%, OR 1.66 [1.16-2.38], p = 0.006), while the percentage of positive tests was lower (28.7% vs 35.7%). Despite increased stress testing and lower

Total Charts Reviewed: 1905 patients

Exclusions: 847 patients

DNR/DNI: 325

No EKG: 28

Non-cardiac complaint: 397

STEMI: 56

Multiple: 41

Post Cohort: 529 patients

193 patients

(Interval 3)

336 patients

(Interval 2)

Control Cohort: 529 patients

(Interval 1)

Fig. 1. Study flowchart.

diagnostic yield in the post-cohort, there were similar overall rates of positive stress tests between groups (4.7% vs 3.8%, OR 1.26 [0.66- 2.30], p = 0.447). Mortality during the index visit was similar between cohorts (4.0% vs 4.2%, OR 0.95 [0.52-1.75], p = 0.876) and there was no significant difference between cohorts with regards to cardiac catheter- ization, PCI or CABG.

Outcomes at 90 days (Table 3)

No significant difference in the primary outcome of mortality at 90 days was seen between the post and control cohorts (13.2% vs 14.1%, OR 0.93 [0.65-1.34], p = 0.705). Fig. 2 demonstrates the survival plots for each cohort. Secondary outcomes at 90 days were also similar be- tween groups with no observed difference in rates of rehospitalization, recurrent MI/ACS or subsequent PCI/CABG.

There were 108 patients (10.2%) with at least one unknown out- come at 90 days. The control and post-cohorts contained 57 (10.8%) and 51 (9.6%) such patients, respectively. Table 4 quantifies the number of patients with missing data by outcome. Mortality data was initially unknown for 58 (11.0%) and 51 (9.6%) patients in the control and post-cohorts, respectively, which decreased to 24 (4.5%) and 23 (4.4%) after review of the SSDI.

Table 5 reports baseline characteristics for the subset of patients with missing data in two different ways: The first two columns compare the baseline characteristics of patients with unknown 90 day outcomes to those of patients with complete 90 day outcome data. The second two columns compare the subset of patients in the control cohort with unknown 90 day outcomes to the subset of patients in the post-co- hort with unknown 90 day outcomes. Baseline characteristics of pa- tients with missing data were generally similar to those patients with complete data, although less likely to be diagnosed with ACS/MI. Pa- tients with at least one unknown 90 day outcome underwent signifi- cantly fewer Cardiology consultations, stress tests and cardiac catheterizations during the index visit compared to those patients with complete 90 day outcome data. Patients with unknown 90 day outcomes in the control group were very similar to patients with un- known 90 day outcomes in the post-group and there were no signifi- cant differences with regards to any index visit outcome.

Interrater reliability

Sixty (5.6%) charts were selected at random and independently ab- stracted by a different member of the study team. Interobserver agree- ment for categorical variables yielded calculated Cohen’s ? values ranging from 0.594 to 0.769.

Discussion

In ED patients with a TnI in the ‘gray zone’ range (0.06-0.5 ng/mL) and possible ACS we found no difference in mortality during the index visit or at 90 days following a decrease in the URL of TnI from 0.5 ng/ mL to the 99th percentile (0.06 ng/mL). No differences were seen in other 90 day outcomes, including rates of rehospitalization, subsequent ACS/MI diagnosis or revascularization. During the index visit, resource utilization was significantly increased with regards to anticoagulation, cardiology consultation and non-invasive cardiac stress testing; howev- er, there was no change in overall rates of inducible ischemia, cardiac catheterization, or revascularization. A discharge diagnosis of ACS or MI was more common following the decrease in TnI URL. These differ- ences were observed immediately following the change in TnI URL and persisted at one year.

The change in TnI URL at our institution yielded additional healthcare utilization for patients with suspected ACS without concurrent improve- ment in outcomes. This contrasts with prior research by Mills et al. which demonstrated a significant decrease in mortality and subsequent diagnosis of MI following a similar reduction in TnI URL [6]. This may be

Table 1

Patient demographics and clinical characteristics.

Characteristics	Control cohort		Post-cohort
	Interval 1		Total	p	Interval 2	Interval 3
Patients, no. (%)	529 (50.0)		529 (50.0)	-	336 (31.8)	193 (18.2)
Age, u years (SD)	71.0 (14)		70.3 (14)	0.420	70.4 (15)	70.1 (13)
Female, no. (%)	246 (46.5)		214 (40.5)	0.047	134 (39.9)	80 (41.5)
Race, no. (%) Black	42 (7.5)		46 (8.7)	0.656	27 (8.0)	19 (9.8)
Hispanic	4 (0.8)		7 (1.3)	0.363	2 (0.6)	5 (2.6)
White	460 (87.0)		455 (86.0)	0.653	296 (88.1)	159 (82.4)
Other/unknown	23 (4.3)		21 (4.0)	0.758	11 (3.3)	10 (5.2)
Cr, u mg/dL (SD)	1.86 (2.0)		1.86 (1.9)	0.988	1.88 (1.8)	1.82 (2.1)
TnI, u ng/mL (SD) Comorbidity, no. (%) CKD	0.14 (0.1) 166 (31.4)		0.15 (0.1) 202 (38.2)	0.189 0.020	0.15 (0.1) 129 (38.4)	0.15 (0.1) 73 (37.8)
CAD	260 (49.1)		254 (48.0)	0.712	159 (47.3)	95 (49.2)
DM	204 (38.6)		208 (39.3)	0.801	138 (41.1)	70 (36.3)
Hyperlipidemia	293 (55.4)		319 (60.3)	0.106	223 (66.4)	96 (49.7)
HTN	386 (73.0)		377 (71.3)	0.537	229 (68.2)	148 (76.7)
smoking history	218 (41.2)		235 (44.4)	0.291	136 (40.5)	99 (51.3)
Hospitalized, no. (%)	493 (93.2)		496 (93.8)	0.709	314 (93.5)	182 (94.3)
Discharge Dx, no. (%) ACS/MI	68 (12.9)	94 (17.8)		0.027	55 (16.4)	39 (20.2)
CP, UE	25 (4.7)	32 (6.0)		0.341	19 (5.7)	13 (6.7)
CP, MSK	3 (0.6)	2 (0.4)		0.654	1 (0.3)	1 (0.5)
AFib	35 (6.6)	40 (7.6)		0.549	29 (8.6)	11 (5.7)
CHF	107 (20.2)	126 (23.8)		0.159	78 (23.2)	48 (24.9)
COPD	20 (3.8)	13 (2.5)		0.216	7 (2.1)	6 (3.1)
GERD	2 (0.4)	5 (0.9)		0.256	2 (0.6)	3 (1.6)
Pneumonia	37 (7.0)	38 (7.2)		0.905	24 (7.1)	14 (7.3)
PE	17 (3.2)	8 (1.5)		0.069	4 (1.2)	4 (2.1)
Other	215 (40.6)	171 (32.3)		0.005	117 (34.8)	54 (28.0)

ACS/MI, acute coronary syndrome or myocardial infarction; AFib, atrial fibrillation/flutter; CAD, coronary artery disease; CKD, chronic kidney disease; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; CP, chest pain; Cr, creatinine; Dx, diagnosis; DM, diabetes mellitus; GERD, gastroesophageal reflux disease; HTN, hypertension; MSK, musculoskel- etal; PE, pulmonary embolism; SD, standard deviation; TnI, troponin I; UE, unknown etiology.

partly due to the lower prevalence of MI/ACS in our Control population compared to that of Mills et al. (13% vs 20%) and an increased risk of false positive TnI values.

Despite this disparity regarding mortality and recurrent MI, there were similar findings with regards to resource utilization: Rates of car- diac catheterization increased without any significant change in

Table 2

Outcomes during index visit.

Outcomes	Control		Post
	Interval 1 (n = 529)		Interval 2 + 3 (n = 529)			Interval 2 (n = 336)			Interval 3 (n = 193)
	n		n	OR		n	OR		n	OR
	(%)		(%)	[95% CI]		(%)	[95% CI]		(%)	[95% CI]
Anticoagulantion	44		75	1.82		50	1.93		25	1.64
Cardiology consult	(8.3) 218		(14.2) 293	[1.23-2.70] p = 0.003 1.77		(14.9) 190	[1.25-2.96] p = 0.003 1.86		(13.0) 103	[0.97-2.76] p = 0.063 1.63
	(41.2)		(55.4)	[1.39-2.26] p b 0.001		(56.5)	[1.41-2.45] p b 0.001		(53.4)	[1.17-2.27] p = 0.004
Stress test	56		87	1.66		50	1.48		37	2.00
Positive stress test	(10.6) 20		(16.4) 25	[1.16-2.38] p = 0.006 1.26		(14.9) 16	[0.98-2.22] p = 0.062 1.27		(19.2) 9	[1.27-3.15] p = 0.003 1.24
Cardiac catheterization	(3.8) 91		(4.7) 106	[0.69-2.30] p = 0.447 1.20		(4.8) 65	[0.65-2.49] p = 0.482 1.15		(4.7) 41	[0.56-2.78] p = 0.594 1.30
PCI	(17.2) 40		(20.0) 41	[0.88-1.65] p = 0.237 1.03		(19.3) 23	[0.81-1.64] p = 0.425 0.90		(21.2) 18	[0.86-1.96] p = 0.215 1.26
CABG	(7.6) 6		(7.8) 10	[0.65-1.62] p = 0.908 1.68		(6.8) 7	[0.53-1.53] p = 0.693 1.85		(9.3) 3	[0.70-2.25] p = 0.441 1.38
Mortality	(1.1) 22		(1.9) 21	[0.60-4.65] p = 0.319 0.95		(2.1) 13	[0.62-5.57] p = 0.271 0.93		(1.6) 8	[0.34-5.56] p = 0.654 1.00
	(4.2)		(4.0)	[0.52-1.75] p = 0.876		(3.9)	[0.46-1.87] p = 0.833		(4.1)	[0.44-2.28] p = 0.994

CABG, coronary artery bypass graft; OR, odds ratio; PCI, percutaneous coronary intervention.

Table 3

Outcomes at 90 days.

Outcomes	Control		Post
	Interval 1		Interval 2 + 3			Interval 2			Interval 3
	n		n	OR		n	OR		n	OR
	(%)		(%)	[95% CI]		(%)	[95% CI]		(%)	[95% CI]
Mortality	71/505a		67/506	0.93		41/318	0.90		26/188	0.98
	(14.1)		(13.2)	[0.65-1.34]		(12.9)	[0.60-1.37]		(13.8)	[0.60-1.59]
Readmissionb	208/469		206/467	p = 0.705 0.99		133/295	p = 0.635 1.03		73/172	p = 0.938 0.92
	(44.3)		(44.1)	[0.77-1.28]		(45.0)	[0.77-1.38]		(42.4)	[0.65-1.32]
				p = 0.942			p = 0.842			p = 0.666
ACS/MI	86/475		108/481	1.31		66/301	1.27		42/180	1.38
	(18.1)		(22.5)	[0.95-1.80]		(21.9)	[0.89-1.82]		(23.3)	[0.91-2.09]
				p = 0.095			p = 0.192			p = 0.133
PCI/CABG	53/475		63/483 (13.0)	1.19		39/303	1.18		24/180	1.22
	(11.2)			[0.81-1.76]		(12.9)	[0.76-1.83]		(13.3)	[0.73-2.05]
				p = 0.372			p = 0.471			p = 0.441

ACS, acute coronary syndrome; CABG, coronary artery bypass graft; MI, myocardial infarction; OR, odds ratios; PCI, percutaneous coronary intervention.

^a Denominators represent total number of patients with known data for the given outcome.

^b Readmission data does not include subjects who died during the index visit.

revascularization procedures performed [6]. Our results also parallel more recent findings by Bjurman et al., in which a decrease in the TnI URL to the 99th percentile yielded a significant increase in rates of cor- onary angiography (2.8 vs 3.3%), but no significant change in overall mortality or rates of coronary revascularization [11].

The static rates of PCI and CABG seen in these studies suggest that a lower TnI URL fails to identify any additional ACS patients necessitating intervention. As these studies also demonstrate higher rates of MI diag- nosis, it is logical to presume that patients with low-level TnI elevations are likely manifesting demand-related myocardial ischemia (type 2 MI), as opposed to a primary coronary artery process (type 1 MI) [1]. This is supported by further analysis of the original Mills’ study cohort by Shah et al., which revealed the reduced TnI cutoff resulted in a disproportion- ate increase in the diagnosis of type 2 MI. Furthermore, patients with “gray zone” troponin values and a type 2 MI diagnosis demonstrated a higher rate of resource utilization without any reduction in recurrent MI or death [12]. Other research has demonstrated that only about 25% of ED patients with elevated troponins are actually suffering from a type 1 MI, yet reiterate mortality rates between patients with type 1 MI and type 2 MI are equivalent [13].

More sensitive troponin assays and lower Diagnostic thresholds have been shown capable of identifying a population with overall

Fig. 2. Kaplan-Meier curve for overall survival.

greater morbidity and mortality, speeding the MI “rule out” process in ED patients, and even reducing Hospital admission rates [4,11,14,15]. Despite promising research, uncertainty persists regarding the optimal role of increasingly sensitive troponin assays in accurately identifying patients with ACS [2]. Cost analysis research suggests that the improved outcomes associated with more sensitive troponin assays may come at a disproportionately higher overall rate of healthcare expenditures [16]. Our findings further emphasize the difficulty associated with the Clinical interpretation of low-level troponin elevations and the resulting ten- dency towards heightened use of Healthcare resources. Further research is needed to optimize the clinical use of highly sensitive troponin assays, clarify the interpretation of lower troponin URL and balance potential increased Healthcare costs associated with reduced adverse outcomes.

Limitations

There are the standard limitations inherent to the retrospective study design and execution at a single center. Only patients with low level TnI elevations (0.06-0.5 ng/mL) were included, so conclusions cannot be drawn regarding patients with TnI elevations of lesser or greater magnitude; however, the classification of these TnI values as normal or abnormal was not affected by the change in TnI cutoff. Thus, management of these populations should have remained con- stant. As the study focused on ED patients, only TnI values obtained in the ED were used to identify patients for inclusion. Peak TnI and subse- quent trends in TnI values following hospitalization were not assessed. The intrinsic relationship between abnormal TnI and the diagnosis of MI likely affected the relative rates of ACS/MI between cohorts.

Table 4

Unknown outcomes at 90 days.

Outcomes	Control (n = 529)		Post (n = 529)
	n		n
	(%)		(%)
Mortality	24		23
	(4.5)		(4.4)
Readmission	38		41
	(7.2)		(7.8)
ACS/MI	54		48
	(10.2)		(9.1)
PCI/CABG	54		46
	(10.2)		(8.7)

ACS, acute coronary syndrome; CABG, coronary artery bypass graft; MI, myocardial infarc- tion; OR, odds ratios; PCI, percutaneous coronary intervention.

Table 5

Patient demographics, clinical characteristics and index visit outcomes of patients with unknown outcomes at 90 days.

Characteristics All patients (n = 1058)

Patients with >=1 unknown outcome (n = 108)

	No unknown outcomes at 90 d	>=1 unknown outcome at 90 d	p		Control cohort	Post-cohort	p
Patients, no. (%)	950 (89.8)	108 (10.2)	-		57 (52.8)	51 (47.2)	-
Age, u years (SD)	70.2 (14.1)	74.5 (15.0)	0.003		75.2 (15.2)	73.8 (14.6)	0.608
Female, no. (%)	413 (43.5)	47(43.1)	0.993		29 (50.8)	18 (35.3)	0.105
Race, no. (%)
Black	81 (8.5)	7 (6.5)	0.466		5 (8.8)	2 (3.9)	0.309
Hispanic	10 (1.1)	1 (0.1)	0.902		0 (0.0)	1 (2.0)	0.290
White	820 (86.3)	95 (88.0)	0.635		48 (84.2)	47 (92.2)	0.207
Other/unknown	39 (4.1)	5 (4.6)	0.796		4 (7.0)	1 (2.0)	0.214
Cr, u mg/dL (SD)	1.88 (1.99)	1.68 (1.78)	0.327		1.50 (1.00)	1.89 (2.34)	0.261
TnI, u ng/mL (SD)	0.15 (0.10)	0.13 (0.09)	0.074		0.12 (0.08)	0.14 (0.09)	0.336
Comorbidity, no. (%)
CKD	336 (35.4)	31 (28.7)	0.168		15 (26.3)	16 (31.4)	0.564
CAD	467 (49.2)	47 (43.5)	0.267		25 (43.9)	22 (43.1)	0.940
DM	376 (39.6)	36 (33.3)	0.207		17 (33.3)	19 (33.3)	0.416
Hyperlipidemia	567 (59.7)	45 (41.7)	b0.001		20 (35.1)	25 (49.0)	0.145
HTN	696 (73.3)	67 (62.0)	0.014		42 (73.7)	25 (49.0)	0.009
Smoking hx	412 (43.4)	41 (38.0)	0.282		21 (36.8)	20 (39.2)	0.801
Hospitalized, no. (%)	891 (93.8)	99 (91.7)	0.394		52 (91.2)	47 (92.2)	0.800
Discharge Dx, no. (%)
ACS/MI	159 (16.7)	3 (2.8)	b0.001		1 (1.8)	2 (3.9)	0.496
CP, UE	48 (5.1)	9 (8.3)	0.153		4 (7.0)	5 (9.8)	0.603
CP, MSK	5 (0.5)	0 (0.0)	0.450		0 (0.0)	0 (0.0)	-
AFib	69 (7.3)	4 (3.7)	0.167		1 (1.8)	3 (5.9)	0.255
CHF	210 (22.1)	21 (19.4)	0.526		8 (14.0)	13 (25.5)	0.135
COPD	27 (2.8)	6 (5.6)	0.124		6 (10.5)	0 (0.0)	0.018
GERD	4 (0.4)	1 (0.9)	0.469		0 (0.0)	1 (1.9)	0.290
Pneumonia	65 (6.8)	9 (8.3)	0.565		3 (5.3)	6 (11.8)	0.225
PE	23 (2.4)	2 (1.9)	0.712		2 (3.5)	0 (0.0)	0.179
Other	334 (35.2)	53 (49.1)	0.005		32 (56.1)	21 (41.2)	0.122
Index visit outcomes, no. (%)
Anticoagulation	111 (11.7)	8 (7.4)	0.183	5 (8.8)		3 (5.9)	0.569
Cardiology consult	474 (49.9)	37 (34.3)	0.002	16 (28.1)		21 (41.2)	0.154
Stress test	139 (14.6)	4 (3.7)	0.002	1 (1.8)		3 (5.9)	0.259
Positive stress test	43 (4.5)	2 (1.9)	0.192	1 (1.8)		1 (2.0)	0.937
Cardiac catheterization	188 (19.8)	9 (8.3)	0.004	3 (5.3)		6 (11.8)	0.225
PCI	77 (8.1)	4 (3.7)	0.103	1 (1.8)		3 (5.9)	0.259

ACS/MI, acute coronary syndrome or myocardial infarction; AFib, atrial fibrillation/flutter; CAD, coronary artery disease; CI, confidence intervals; CKD, chronic kidney disease; CHF, con- gestive heart failure; COPD, chronic obstructive pulmonary disease; Cr, creatinine; Dx, diagnosis; DM, diabetes mellitus; GERD, gastroesophageal reflux disease; HTN, hypertension; MSK, musculoskeletal; PE, pulmonary embolism; SD, standard deviation; TnI, troponin I; UE, unknown etiology.

Although rates of ACS/MI were not significantly different at 90 days, there is the possibility of ascertainment bias.

While the included baseline characteristics of each cohort were sim- ilar, other potentially relevant data were not collected, including Thrombolysis in Myocardial Infarction score, other risk stratifica- tion scores, medication use, and the type of MI.

Follow up was limited to 90 days and our results cannot be general- ized to outcomes outside this interval. Increased cardiology consultation following the decrease in TnI cutoff may have resulted in more intensive outpatient follow up and related interventions, such as greater risk fac- tor modification; however, this study was not designed or powered to detect the long-term impact of these factors.

Outcomes at 90 days relied on documentation in the EMR and at least one data element was unknown in approximately 10% of the study population; however, mortality data was missing in b5%. Baseline characteristics and index visit outcomes in patients with missing data were similar between cohorts. After excluding patients with unknown data the total patients included in the analysis fell slightly below the cal- culated sample size needed for assessment of the primary outcome of mortality at 90 days; however, this was based on an underestimate of expected 90 day mortality (4%). A more accurate assumption closer to the overall observed 90 day mortality (13%) would have yielded a sig- nificantly smaller sample size requirement of 167 patients per cohort.

CCTA did not play a significant role in assessing patients with poten-

tial ACS and this may limit generalizability of our results. Lastly, our study involves data from 2011 to 2012 following the change in TnI cut- off to an older 99th percentile. The cut point was subsequently reduced

to the current 99th percentile of 0.04 ng/mL. While this study was not designed to directly assess the effects of this additional change, we feel our results meaningfully reflect on the general clinical effects of lowering the cutoff value for cardiac troponins.

Conclusion

In summary, a reduction in the troponin I cutoff to the 99th percen- tile did not change mortality or rates of coronary intervention in ED pa- tients presenting with suspected ACS and low level troponin elevations. There was a significant increase in the use of healthcare resources, in- cluding cardiology consultation and cardiac stress testing, which persisted one year following the change in troponin cutoff value.

Acknowledgements

We would like to thank Theodore Bell MS, Nicole Battaglioli MD, Leah Deter, Daniel Goodberry MD, Rodney Grim PhD, Thuyvi Luong MD, Steven Manzella PhD, April Miller DO, Crystal Miller DO, Megan Newnam, Durand Richards MD, Micah Richardson, Chance Rohrbaugh and Emily Siegel for their assistance with the study.

References

1. Amsterdam EA, Wenger NK, Brindis RG, Casey Jr DE, Ganiats TG, Holmes Jr DR, et al. 2014 AHA/ACC guideline for the management of patients with non-ST-elevation acute coronary syndromes: a report of the American College of Cardiology/American

   Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2014;64: e139-28.

   Newby LK, Jesse RL, Babb JD, Christenson RH, De Fer TM, Diamond GA, et al. ACCF 2012 expert consensus document on practical clinical considerations in the interpre- tation of troponin elevations: a report of the American College of Cardiology Foun- dation task force on Clinical Expert Consensus Documents. J Am Coll Cardiol 2012; 60:2427-63.

2. Thygesen K, Alpert JS, White HD, Joint ESCAAHAWHFTFftRoMI, Jaffe AS, Apple FS, et al. Universal Definition of Myocardial Infarction. Circulation 2007;116:2634-53.
3. Bonaca M, Scirica B, Sabatine M, Dalby A, Spinar J, Murphy SA, et al. Prospective eval-

   uation of the prognostic implications of improved assay performance with a sensi- tive assay for cardiac troponin I. J Am Coll Cardiol 2010;55:2118-24.

   Kim J, Gaddam S, Wu WC, Behera V, Sharma S, Choudhary G. Stratified reporting of high sensitivity troponin I assay is associated with suboptimal management of pa- tients with acute coronary syndrome and intermediate troponin elevation. J Clin Lab Anal 2013;27:402-6.

4. Mills NLC, D. AM, Lee KK, Anand A, Gamble D, Shah ASV, et al. Implementation of a sensitive troponin I assay and risk of recurrent myocardial infarction and death in patients with Suspected acute coronary syndrome. JAMA 2011;305:1210-6.
5. van de Kerkhof D, Peters B, Scharnhorst V. Performance of the Advia Centaur sec- ond-generation troponin assay TnI-ultra compared with the first-generation cTnI assay. Ann Clin Biochem 2008;45:316-7.
6. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008; 61:344-9.
7. Kaji AH, Schriger D, Green S. Looking through the retrospectoscope: reducing bias in emergency medicine chart review studies. Ann Emerg Med 2014;64:292-8.
8. Gilbert EH, Lowenstein SR, Koziol-McLain J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1995; 27:305-8.
9. Bjurman C, Zywczyk M, Lindahl B, Carlsson T, Johanson P, Petzold M, et al. Decreased admissions and hospital costs with a neutral effect on mortality following lowering of the troponin T cutoff point to the 99th percentile. Cardiol J 2017;24(6):612-22.
10. Shah AS, McAllister DA, Mills R, Lee KK, Churchhouse AM, Fleming KM, et al. Sensi- tive troponin assay and the classification of myocardial infarction. Am J Med 2015; 128:493-501 [e3].
11. Meigher S, Thode HC, Peacock WF, Bock JL, Gruberg L, Singer AJ. Causes of elevated cardiac troponins in the emergency department and their associated mortality. Acad Emerg Med 2016;23:1267-73.
12. Neumann JT, Sorensen NA, Schwemer T, Ojeda F, Bourry R, Sciacca V, et al. Diagnosis of myocardial infarction using a High-sensitivity troponin I 1-hour algorithm. JAMA Cardiol 2016;1:397-404.
13. Mueller M, Celik S, Biener M, Vafaie M, Schwoebel K, Wollert KC, et al. Diagnostic and prognostic performance of a novel high-sensitivity cardiac troponin T assay compared to a contemporary sensitive cardiac troponin I assay in patients with acute coronary syndrome. Clin Res Cardiol 2012;101:837-45.
14. Kaambwa B, Ratcliffe J, Horsfall M, Astley C, Karnon J, Coates P, et al. Cost effective- ness of high-sensitivity troponin compared to conventional troponin among pa- tients presenting with Undifferentiated chest pain: a trial based analysis. Int J Cardiol 2017;1:144-50.