Mortality in sepsis: Comparison of outcomes between patients with demand ischemia, acute myocardial infarction, and neither demand ischemia nor acute myocardial infarction

Abstract

Introduction

Elevation in cardiac troponins is common with sepsis despite unclear impact.

Hypothesis

We investigated whether demand ischemia(DI) resulted in variable outcomes compared to acute myocardial infarction(AMI) and those with neither DI nor AMI in sepsis.

Methods

We analyzed data from the 2011‐2014 National Inpatient Sample among patients admitted for sepsis. We compared outcomes among patients with DI i) versus AMI and ii) versus neither DI nor AMI, respectively using propensity matching. Primary study end‐point was in‐hospital mortality.

Results

We studied 666,154 patients, with mean age 63.7 years and 50.8% female participants. Overall, 94.7% of the included patients had neither DI nor AMI, 4.4% had AMI and 0.83% had DI. Between 2011 and 2014, we observed an increasing trend for DI but decreasing trend for AMI in sepsis. Patients with DI experienced higher rates of atrial and ventricular arrhythmias, had longer length of stay and higher cost of stay compared to patients with neither demand ischemia nor AMI. Despite higher hospital mortality at baseline with DI, post‐propensity matching revealed no difference in hospital mortality between patients with DI and those with neither (26.9% vs. 27.0%, adjusted odds ratio 0.99, 95% confidence intervals 0.92‐1.07;p=0.87). Patients with DI experienced lower hospital mortality compared to those with AMI pre (28.5% vs. 48.3%;p<0.001) and post‐propensity matching (41.1% vs. 29.1%, aOR 0.58, 95% CI 0.54‐0.63;p<0.001).

Conclusion

Among patients with sepsis, those with DI had similar adjusted in‐hospital mortality compared to those with neither DI nor AMI. Patients with AMI had the highest in‐hospital mortality among all groups.

1. INTRODUCTION

Sepsis remains a global health challenge and poses a significant economic burden on healthcare, accounting for $20 billion in annual hospitalization‐related costs within the United States.1 Elevation in cardiac troponins (Tn) has been observed in a wide range (31%–80%) of patients with systemic inflammatory response syndrome, sepsis, or septic shock.2, 3 These elevations can occur among those with significant coronary artery disease (CAD) without an acute myocardial infarction (AMI) but is noted even among patients without CAD or coronary thrombosis.4 Several explanations have been postulated for the rise of Tns accompanying sepsis, such as coronary artery spasm, coronary embolism, anemia, arrhythmias, trauma, heart failure, pulmonary embolism, hypertension (HTN), and hypotension, among others.2, 3, 4, 5 Evidence from several studies conducted in both healthy and high‐risk populations suggests a predilection toward higher prevalence of baseline cardiovascular (CV) comorbidities and greater incidence of future CV events among patients with high‐normal and above‐normal Tn measurements.5, 6

The term demand ischemia (DI) represents a state of mismatch between the oxygen demand and supply to the myocardium, resulting in rising Tn levels, in the absence of acute coronary syndrome (ACS).7 Research is limited on how DI affects short‐ or long‐term outcomes among patients with sepsis and management strategies that would benefit this patient subset. In our study, we compare the short‐term outcomes in 3 subgroups of patients admitted with a primary diagnosis of sepsis: those with DI, those with an AMI, and those with neither DI nor AMI.

2. METHODS

2.1. Data source and study population

We conducted our analysis on unweighted hospital discharge data from the Healthcare Cost and Utilization Project—National (Nationwide) Inpatient Sample (HCUP‐NIS) from 2011 to 2014. Annually, the NIS is composed of discharge‐level data from roughly 8 million hospitalizations and approximates a stratified sample of 20% of community hospitals in the United States. Each hospitalization within the database contains clinical and resource‐use information. Core hospital stay files contain details on patient demographics (eg, age, sex, race), International Classification of Diseases, 9th Revision, Clinical Modification (ICD‐9‐CM) diagnosis, Elixhauser comorbidities, length of hospital stay, discharge status, in‐hospital mortality, and total charges, among other variables. Patients with a primary diagnosis of sepsis (ICD‐9‐CM codes 995.91, 995.92, 996.64, 790.7, 785.52, 038.x, 038.xx) comprised the study population (see Supporting Information, Table 1, in the online version of this article). Patients age < 16 years and those with missing data or transferred out of the hospital were excluded from study. Presence of DI among the sepsis patients was identified using the ICD‐9‐CM code 411.89 (represents DI/coronary insufficiency and myocardial dysfunction) as a secondary diagnostic code. Similarly, AMI was identified using ICD‐9‐CM codes 410.xx and 411.1 among the secondary diagnostic codes. Patients with diagnostic codes for both DI and AMI (<0.2% of total study population) were excluded from analysis. Death was defined in the NIS as in‐hospital mortality. Different comorbidities were identified by using ICD‐9‐CM diagnoses and diagnosis‐related groups (see Supporting Information, Table 2, in the online version of this article).

2.2. Statistical analysis

Demographics and baseline characteristics were summarized using descriptive statistics. Continuous variables was expressed as mean ± standard deviation (SD) and analyzed using the Student t test or analysis of variance. The Fisher exact test or Pearson χ² test was used for analysis of categorical variables. Trend analyses were performed using the Mantel–Haenszel test of trend. Propensity‐score matching was used to identify cohorts of patients with similar baseline characteristics in 2 separate populations. The first population comprised patients with DI and those with neither DI nor AMI. Propensity‐score matching was performed using 1:2 (1 case with DI and 2 cases with neither AMI nor DI) matching protocol without replacement and caliper width 10⁻³ of SD of the logit of propensity score using DI as treatment variable. Matching variables included age, sex, race, weekend admission, admission year, HTN with and without complications, diabetes mellitus, dyslipidemia, current or past smoking, obesity, renal failure, liver disorders, prior MI, previous stroke or transient ischemic attack, presence of shock, vasopressor use, and ventilator use. The second population included patients with sepsis and DI compared to those with AMI and comparisons were made using similar 1:2 propensity matching. Our primary outcome for comparison was in‐hospital mortality. Secondary study endpoints included rate of atrial arrhythmias, ventricular arrhythmias, cost of hospitalization, and length of hospitalization. Standard statistical analyses were implemented as performed in previously published data using the NIS.8, 9, 10, 11 Results were considered statistically significant for P values <0.05. SPSS statistics version 23.0 (IBM Corp., Armonk, NY) was used to perform data analysis.

3. RESULTS

A total of 666  154 patients with sepsis satisfied the inclusion criteria, out of which 631  328 (94.7%) had neither DI nor AMI, 5537 (0.83%) had DI, and 29  289 had AMI (4.4%).

3.1. Trends in incidence and in‐hospital mortality

From 2011 to 2014, we observed that the proportion of patients with a diagnosis of DI increased from 0.3% to 1.3% (Figure; P _trend < 0.01). During the same study period, the rate of AMI diagnosis decreased from 4.9% in 2011 to 4.0% by 2014 (P _trend < 0.01). In‐hospital mortality for the entire study group decreased from 22.1% to 16.8% between 2011 and 2014 (P _trend < 0.01). This pattern of a reduction in‐hospital mortality was seen within the subgroups of patients with AMI (50.9% to 46.1%; P _trend < 0.001) and those with neither AMI nor DI (20.6% to 15.4%; P _trend < 0.001). Patients with DI saw no significant reduction in hospital mortality (29.0% to 28.1%, P _trend = 0.4; see Supporting Information, Figure 1, in the online version of this article).

Figure 1.

Trends in frequency of AMI and DI diagnoses among patients with sepsis from 2011 to 2014 using the National Inpatient Sample. Abbreviations: AMI, acute myocardial infarction; DI, demand ischemia

3.2. Baseline characteristics for study populations

The mean age of patients included in the study was 63.7 ±18.0 years (median, 66 years; interquartile range, 52–78 years). The study was composed of 50.8% females, and a majority of the patients within the study were Caucasian (66.0%), followed by African American (12.3%). Average length of stay (LOS) was 6.4 days (median, 5 days; interquartile range, 3–8 days), and the hospital cost of stay was US$57  765. More than one‐quarter of all admissions occurred during the weekend, and a majority of the hospital admissions occurred at urban institutions.

Baseline differences between the 3 study groups can be seen in Table 1. Patients with sepsis and AMI were more likely to be male, were older, and had a greater prevalence of underlying comorbidities such as diabetes mellitus, prior stroke, and renal failure. On the other hand, patients with sepsis and DI had the highest prevalence of HTN, dyslipidemia, prior MI, peripheral vascular disease, valvular heart disease, atrial arrhythmias, and obstructive sleep apnea. Patients with neither AMI nor DI had a higher rate of smoking and underlying liver disease. Patients with AMI had higher disease‐severity scores and the highest rate of overall incidence of shock (including cardiogenic), acute cardiorespiratory failure, ventricular arrhythmias, cardiac arrest, ventilator use, and in‐hospital mortality compared to the other study groups. Mean cost of stay and LOS were also significantly higher among patients with AMI.

Table 1.

Baseline characteristics within the study cohort

Characteristics	Sepsis With Neither DI Nor AMI, n = 631  328	Sepsis With DI, n = 5537	Sepsis With AMI, n = 29  289	P Value
Mean age, y	63.2 ±18.0	71.6 ±14.4	72.5 ±13.7	<0.001
Median age, y	65 (52–78)	74 (62–84)	74 (63–84)	NA
Female sex	51.0	48.9	48.8	<0.001
Race				<0.001
Caucasian	65.9	67.3	68.9
African American	12.5	10.0	10.4
Hispanic	10.0	7.6	8.4
Other or missing	11.7	15.2	12.3
Weekend admission	26.1	27.5	27.5	<0.001
Elective admission	4.3	1.6	3.9	<0.001
Hospital location: urban	88.7	97.6	90.3	<0.001
Hospital region				<0.001
Northwest	16.4	30.5	19.6
Midwest	20.7	18.6	19.4
South	37.7	16.4	37.8
West	25.2	34.4	23.2
HTN with and without complications	55.4	63.6	63.1	<0.001
DM with and without complications	33.0	34.8	38.6	<0.001
Dyslipidemia	28.5	35.8	33.6	<0.001
Chronic pulmonary disease	26.4	29.4	29.4	<0.001
Current or past smoker	27.3	24.9	21.9	<0.001
History of stroke or TIA	5.3	6.1	6.2	<0.001
History of MI	4.5	8.8	7.5	<0.001
Obesity	14.2	13.2	11.4	<0.001
Renal failure	21.1	30.0	32.9	<0.001
Vasopressor use	2.2	7.8	5.8	<0.001
Ventilator use	13.6	24.5	39.2	<0.001
All etiologies for shock	20.6	37.3	48.7	<0.001
Cardiogenic shock	0.7	2.8	8.5	<0.001
Peripheral vascular disorders	8.1	15.8	14.7	<0.001
Coagulopathy	13.1	19.7	19.7	<0.001
Depression	11.3	9.9	8.6	<0.001
Liver disease	6.5	5.6	5.1	<0.001
Valvular disease	6.0	14.1	13.9	<0.001
OSA	6	6.8	5	<0.001
NSTEMI	—	—	85.6	NA
STEMI	—	—	14.8	NA
CABG	<0.01	0.0	0.4	<0.001
PCI	0.03	0.12	3.77	<0.001
Blood product transfusion	12.8	15.5	20.7	<0.001
Acute cardiorespiratory failure	29.1	53.2	63.4	<0.001
Cardiac arrest	2.3	4.0	8.5	<0.001
APR‐DRG: risk of mortality subclass				<0.001
1	14.5	0.6	0.2
2	21.5	8.2	1.6
3	32.2	33.4	14.3
4	31.7	57.8	83.9
APR‐DRG: severity of illness subclass				<0.001
1	3.9	0.1	0.1
2	23.5	5.5	1.0
3	40.9	41.7	14.8
4	31.7	52.7	84.2
Study endpoints at baseline
Mean LOS, d	6.35	7.23	8.06	<0.001
Median LOS, d	4 (3–7)	5 (3–9)	6 (3–10)	NA
Mean cost of stay, USD	55  934	75  674	93  655	<0.001
Median cost of stay, USD	31  294 (17  085–60  761)	45  886 (25  699–86  260)	55  660 (28  694–109  859)	NA
AF/flutter	16.6	31.2	27.9	<0.001
VT or VF	1.7	3.9	7.2	<0.001
In‐hospital mortality	17.7	28.5	48.3	<0.001

Abbreviations: AF, atrial fibrillation; AMI, acute myocardial infarction; APR‐DRG, all patient refined–diagnosis‐related group; CABG, coronary artery bypass grafting; DI, demand ischemia; DM, diabetes mellitus; HTN, hypertension; IQR, interquartile range; LOS, length of stay; MI, myocardial infarction; NA, not applicable; NSTEMI, non–ST‐segment elevation myocardial infarction; OSA, obstructive sleep apnea; PCI, percutaneous coronary intervention; SD, standard deviation; STEMI, ST‐segment elevation myocardial infarction; TIA, transient ischemic attack; USD, United States dollars; VF, ventricular fibrillation; VT, ventricular tachycardia

Data are presented as %, mean ± SD, or median (IQR).

3.3. Impact of DI on in‐hospital mortality

Patient‐ and admission‐related variables were matched as shown in Tables 2 and 3. Absolute standardized differences for all matched characteristics were < 20% (mostly <5%), denoting significant reduction in bias. Table 4 represents the differences between study groups post‐matching for both primary and secondary study outcomes. In the first comparison groups, patients with DI were compared to those with neither DI nor AMI. Within the prematched sample, overall mortality was significantly higher among patients with DI (28.5% vs 17.7%; P < 0.001). This relationship disappeared following matching of these groups. Post–propensity matching revealed no difference in hospital‐mortality between patients with DI and those without DI or AMI (26.9% vs 27.0%; adjusted odds ratio [aOR]: 0.99, 95% confidence interval [CI]: 0.92–1.07, P = 0.87).

Table 2.

Characteristics of patients with sepsis with and without DI, before and after propensity matching

Characteristics	Before Propensity‐Score Matching			After Propensity‐Score Matchinga
	Sepsis Without DI, n = 631  328	Sepsis With DI, n = 5537	P Value	Sepsis Without DI, n = 10  339	Sepsis With DI, n = 5269	P Value
Mean age, y	63.2 ±18.0	71.6 ±14.4	<0.001	71.4 ±14.6	71.3 ±14.5	0.71
Female sex	51.0	48.9	0.002	49.5	48.9	0.49
Race			<0.001			0.36
Caucasian	65.9	67.3		68.8	67.6
African American	12.5	10.0		9.9	10.3
Hispanic	10.0	7.6		8.2	7.7
Other or missing	11.7	15.2		13.6	14.3
Weekend admission	26.1	27.5	0.02	27.0	27.2	0.73
Year			<0.001			0.48
2011	21.6	8.4		8.3	8.8
2012	22.2	16.4		16.9	17.0
2013	25.7	29.1		30.2	29.2
2014	30.5	46.1		44.6	45.0
HTN with and without complications	55.4	63.6	<0.001	63.3	63.3	0.96
DM with and without complications	33.0	34.8	0.004	35.4	34.7	0.43
Dyslipidemia	28.5	35.8	<0.001	35.9	35.3	0.44
Chronic pulmonary disease	26.4	29.4	<0.001	28.9	29.4	0.51
Current or past smoker	27.3	24.9	<0.001	25.3	24.9	0.63
History of stroke or TIA	5.3	6.1	0.004	6.4	6.1	0.60
History of MI	4.5	8.8	<0.001	7.4	8.3	0.05
Obesity	14.2	13.2	0.03	13.9	13.2	0.21
Renal failure	21.1	30.0	<0.001	28.3	29.3	0.21
Liver disease	6.5	5.6	0.01	5.8	5.7	0.87
Vasopressor use	2.2	7.8	<0.001	4.2	5.3	0.001
Ventilator use	13.6	24.5	<0.001	21.4	22.6	0.07
Shock	20.6	37.3	<0.001	33.2	34.7	0.07

Abbreviations: DI, demand ischemia; DM, diabetes mellitus; HTN, hypertension; MI, myocardial infarction; SD, standard deviation; TIA, transient ischemic attack

Data are presented as % or mean ± SD.

^aNo covariate exhibits large imbalance defined as │d│ > 0.2.

Table 3.

Characteristics of patients with sepsis and AMI vs DI, before and after propensity matching

Characteristics	Before Propensity‐Score Matching			After Propensity‐Score Matchinga
	Sepsis With AMI, n = 29  289	Sepsis With DI, n = 5537	P Value	Sepsis With AMI, n = 10  121	Sepsis With DI, n = 5360	P Value
Mean age, y	72.5 ± 13.7	71.6 ± 14.4	<0.001	72.1 ± 13.9	72.0 ±14.1	0.62
Female sex	48.8	48.9	0.96	48.7	49.0	0.66
Race			<0.001			0.76
Caucasian	68.9	67.3		69.1	68.2
African American	10.4	10.0		9.6	9.8
Hispanic	8.4	7.6		7.5	7.7
Other or missing	12.3	15.2		13.8	14.3
Weekend admission	27.5	27.5	0.97	27.6	27.3	0.67
Year			<0.001			0.16
2011	24.0	8.4		8.8	8.7
2012	23.0	16.4		17.9	16.9
2013	25.4	29.1		30.6	29.8
2014	27.6	46.1		42.7	44.5
HTN with and without complications	63.1	63.6	0.48	64.0	63.9	0.88
DM with and without complications	38.6	34.8	<0.001	35.9	35.6	0.75
Dyslipidemia	33.6	35.8	<0.002	35.4	35.9	0.57
Chronic pulmonary disease	29.4	29.4	0.99	29.7	29.5	0.83
Current or past smoker	21.9	24.9	<0.001	24.2	24.5	0.69
History of stroke or TIA	6.2	6.1	0.78	6.4	6.3	0.64
History of MI	7.5	8.8	<0.001	8.7	8.7	0.92
Obesity	11.4	13.2	<0.001	12.4	12.7	0.64
Renal failure	32.9	30.0	<0.001	30.6	30.5	0.94
Liver disease	5.1	5.6	0.14	5.2	5.5	0.49
Vasopressor use	5.8	7.8	<0.001	6.9	7.2	0.46
Ventilator use	39.2	24.5	<0.001	27.3	25.2	0.006
Shock	48.7	37.3	<0.001	39.4	37.8	0.054

Abbreviations: AMI, acute myocardial infarction; DI, demand ischemia; DM, diabetes mellitus; HTN, hypertension; MI, myocardial infarction; SD, standard deviation; TIA, transient ischemic attack

Data are presented as % or mean ± SD.

^aNo covariate exhibits large imbalance defined as │d│ > 0.2.

Table 4.

Primary and secondary outcomes among propensity‐matched groups

Endpoint	Sepsis Without DI, n = 10  339	Sepsis With DI, n = 5269	OR (95% CI)	P Value	Sepsis With AMI, n = 10  121	Sepsis With DI, n = 5360	OR (95% CI)	P Value
Primary endpoint
In‐hospital mortality	27.0	26.9	0.99 (0.92–1.07)	0.87	41.1	29.1	0.58 (0.54–0.63)	<0.001
Secondary endpoints
AF/flutter	24.0	30.6	1.39 (1.29–1.50)	<0.001	27.3	31.5	1.22 (1.14–1.31)	<0.001
VT/VF	2.5	3.8	1.52 (1.26–1.83)	<0.001	6.2	3.9	0.61 (0.52–0.72)	<0.001
Cost of stay, USD	64  850	74  216	—	<0.001	87  197	76  052	—	<0.001
Mean LOS, d	6.6	7.1	—	<0.001	7.7	7.2	—	0.001

Abbreviations: AF, atrial fibrillation; AMI, acute myocardial infarction; CI, confidence interval; DI, demand ischemia; LOS, length of stay; OR, odds ratio; USD, United States dollars; VF, ventricular fibrillation; VT, ventricular tachycardia

Data are presented as % unless otherwise noted.

Within the second set of comparison groups, patients with DI were compared to those with AMI. Patients with DI experienced significantly lower hospital mortality than those with AMI prior to matching (28.5% vs 48.3%; P < 0.001), which remained persistent post‐matching (29.1% vs 41.1%; aOR: 0.58, 95% CI: 0.54–0.63, P < 0.001).

3.4. DI and secondary outcomes

Patients with DI had significantly higher rates of atrial fibrillation (AF)/flutter (31.2% vs 16.6%; P < 0.001) and ventricular tachycardia/fibrillation (3.9% vs 1.7%; P < 0.001) pre‐ and post‐matching (aOR: 1.39, 95% CI: 1.29–1.50 for AF/flutter and aOR: 1.52, 95% CI: 1.26–1.83 for ventricular tachycardia/fibrillation) compared to patients without DI or AMI. On the other hand, DI had higher rates of AF/flutter but lower rates of ventricular tachycardia/fibrillation pre‐ and post‐matching compared to those with AMI.

Cost of stay (US$74  216 vs 64  850; P < 0.001) and LOS (7.1 vs 6.6 days; P < 0.001) were significantly higher in the group of patients with DI post‐matching compared to those without DI. However, when compared to patients with AMI, cost of stay (US$76  052 vs 87  197; P < 0.001) and LOS (7.2 vs 7.7 days; P < 0.001) were lower in the DI arm.

4. DISCUSSION

Our findings demonstrate that the incidence of patients diagnosed with DI among patients admitted with sepsis increased steadily, whereas the incidence of AMI decreased from 2011 to 2014. Increasing awareness among clinicians on non‐ACS related rise in Tns among septic patients and strong recommendations from guidelines on recognizing non–MI‐related Tn elevations as a separate category may have contributed to these trends.7 Ramirez and colleagues reported a 5.8% rate of AMI in a 2008 study among patients with community‐acquired pneumonia.12 However, the study identified <30 cases with AMI and included mainly elderly males. In addition, they defined patients with ST‐segment changes and elevated Tns as AMI, which may have overestimated rate of MI in their study. In a 2016 study of 2.6 million sepsis patients, rate of AMI diagnosis was 4.5% and invasive revascularization strategy predicted >50% risk reduction in hospital mortality.13 Although the extensive inflammatory cascade in sepsis may predispose patients to AMI, the incidence, management, and outcomes for AMI in sepsis have not been studied well. Another finding in our study was the relative flat hospital mortality rate among DI patients during the 4 years. Several other studies have also reported decreasing mortality rates with sepsis over time within the United States and outside, mainly attributable to improved management strategies.14

The use of increasingly sensitive TnI assays has facilitated the identification of myocardial necrosis in sepsis, which may be indicative of early myocardial dysfunction or strain, reduced microvascular circulation, and direct cardiotoxicity.15, 16 Demand–supply mismatch remains the most popular explanation for elevated Tns in septic patients with or without CAD, and the imbalance in myocardial oxygen demand and supply is thought to be related to hypoxia, hypovolemia, hypotension, tachycardia, and/or anemia.17, 18 Conflicting theories from studies on coronary blood flow and myocardial metabolism in humans with septic shock have been postulated.19, 20 Current research in myocardial dysfunction in sepsis suggests that coronary blood flow actually increases in sepsis.20, 21 Microvascular dysfunction, stress‐related release of cytosolic Tns, direct bacterial myocarditis, micronecrosis, effect of cytokines, caspase activation, free radicals, nitric oxide, and endotoxin‐related myocardial damage may also play a role in elevation of cardiac Tns in the setting of sepsis.22 Increased cardiac filling pressures, increased wall stress, duration of hypotension, and use of inotropes can also lead to elevation in Tns.23 Clinical studies evaluating cardiac function of patients with sepsis by echocardiography also show a reduced ejection fraction, followed by both systolic and diastolic dysfunction.24 Mehta et al. showed that serum TnI concentration correlated with degree of myocardial dysfunction in septic shock, with higher levels increased in those with greater severity of sepsis and higher mortality.25 At baseline, our findings showed that patients with MI had the highest proportion of patients with hospital mortality, higher risk for mortality and severity of illness, shock, acute cardiorespiratory failure, ventricular arrhythmias, and cardiac arrest followed by those with DI and then those with neither DI nor AMI. Although cardiogenic shock was more common among patients with an AMI (≈1 out of 6 shock patients), it comprised <5% of the total study population. Similar patterns were observed for LOS or cost of stay, with AMI patients having the longest LOS and the highest cost of stay among all groups. We are unable to use our database to identify quantitative changes in cardiac enzymes and their correlation with outcomes.

Early recognition and treatment of sepsis has gained momentum nationally after strong backing from societal guidelines.26 The goal of such a strategy is to prevent the multiple organ failure and possible death that may ensue after a systemic inflammatory response. Universally validated prediction scores for mortality in sepsis include various parameters of end‐organ damage such as a low partial pressure of arterial oxygen (PaO₂)/fraction of inspired oxygen (FiO₂) ratio, thrombocytopenia, high serum bilirubin and creatinine, low Glasgow coma score, and hypotension with degree of vasopressor requirement.27 Considering that end‐organ damage in the form of myocardial cellular injury in sepsis may be reflected in the form of elevated Tns, there is a general expectation for such a rise to predict worse outcomes in sepsis. In a retrospective study by Relos et al. conducted in the surgical intensive care unit, moderate elevations in serum TnI (below the threshold for MI diagnosis) were associated with higher mortality and longer hospital and intensive care unit (ICU) LOS.28 In another study of 217 consecutive patients admitted to the ICU, Quenot et al. observed that the elevation in Tn was independently associated with higher mortality.29 Several other studies have shown similar results with higher mortality in the patient subgroup with high Tn levels, including additional biomarkers such as atrial or brain natriuretic peptides.5, 6, 30, 31, 32 However, Kolief et al conducted a study including 260 adult medical ICU patients and argued that left ventricular dysfunction, instead of Tn levels, predicted mortality on multivariate analysis.33 The authors considered cardiac dysfunction only in the presence of cardiac arrest, congestive heart failure, or ACS and evaluated a much higher‐risk population. Right ventricular dysfunction and diastolic dysfunction have been shown to have predictive value in determining outcomes in sepsis.34

The generalizability of the existing data to practice is challenging, considering that the studies were mostly small, single‐center, conducted several years ago, had variable designs in study enrollment or analysis, and used different assays for Tn measurement. The rate of diagnosis of DI was significantly lower in our study when compared with existing literature. We hypothesize that the administrative nature of the database may have been responsible for this variation; however, we also propose that patients with more significant elevations in Tns may have been identified over others leading to inclusion of DI among the diagnostic codes.

Our study is the largest to date to assess mortality in patients with DI and includes a study population representative of what is seen in clinical practice. We separated patients with AMI into their own group for the purpose of analysis and matched for several risk factors, including patient demographics, prior history, comorbidities, variability in practice (weekend admissions, year of admission), and high‐risk variables such as vasopressor use, shock, and ventilator use. Our post‐matching analysis revealed that patients with DI did not experience higher in‐hospital mortality compared to patients with neither DI nor AMI. Despite no notable difference in hospital mortality, the DI group had a greater burden of atrial or ventricular arrhythmias, longer LOS, and higher cost of stay associated with it than did patients with neither DI nor AMI. Patients with AMI experienced significantly higher hospital mortality following propensity matching when compared to DI, in addition to having higher LOS, cost of stay, and rate of ventricular arrhythmias. Patients with DI had slightly higher rates of atrial arrhythmias, possibly because increased metabolic stress experienced by the heart under conditions of acidosis and electrolyte abnormalities accompanying sepsis may predispose them to atrial fibrillation or flutter. Another likely contributing factor may have been that DI may have occurred as a consequence of atrial arrhythmias in several patients with sepsis; however, the lack of a timeline associated with diagnoses within the NIS database prevents verification of this phenomenon. Our findings differ from Stein et al., who reported that patients with DI or type 2 MI were older, had a higher‐risk CV risk profile, and had higher 30‐day and 1‐year mortality compared to patients with AMI.20 However, their study had only a fraction of patients with DI (4.5% of total amounting to 127 patients), and that subset had different baseline profiles, was not exclusive to septic patients, and had significantly lower mortality rates compared with our study population. Other studies have also shown higher mortality among those with DI on intermediate‐term to longer‐term follow‐up; however, they were smaller, studied different patient groups, and reported relatively extended follow‐up.35, 36, 37

4.1. Study limitations

The retrospective nature of the study design inherits the possibility for selection bias. Diagnosis of DI and AMI is complicated and depends on several factors, such as symptoms, electrocardiographic changes, and cardiac biomarkers (especially Tn levels [timing of test and trends]), among others, which are unavailable within the NIS database. We are unable to perform an objective verification of the accuracy of these diagnoses, which remains a limitation in using such a study design and its dependence on ICD‐9‐CM coding. Some of these drawbacks are attenuated by the extremely robust size of the database and availability of information from across the nation. In addition, several of the ICD‐9‐CM codes used in our study have been used in previously published data.38, 39, 40 Despite the limitations, our dataset is representative of real‐world practice and outcomes. The database does not provide information beyond hospital discharge, which may result in underestimation of DI's short‐term impact. There is potential for miscoding diagnoses or procedures, and variations exist in coding practices. Relevant information such as degree of Tn elevation, hemodynamics, supportive measures, and temporal sequence of events could not be assessed. The possibility that clinicians monitored patients with DI more closely and attempted earlier corrective measures, including hemodynamic correction and management, cannot be excluded. We attempted to correct for several of the biases by matching for several potential confounders, including markers for severity of disease. Using a clinical database accounting for clinical risk‐scoring systems such as the Sequential Organ Failure Assessment (SOFA) or the Acute Physiology and Chronic Health Evaluation II (APACHE II) score among critically ill patients may be a good approach for future study designs.

5. CONCLUSION

From 2011 to 2014, we observed an increasing trend for DI and a decreasing trend for AMI in sepsis. DI‐related in‐hospital mortality was unchanged during the study period. Highest mortality was observed within the cohort with AMI in sepsis. Patients with DI had more baseline comorbidities, experienced higher rates of arrhythmias, and had longer LOS and higher cost of hospital stay when compared to patients with neither DI nor AMI; however, close matching neutralized the signal for increased hospital mortality with DI. Additional research exploring clarity on the definition of AMI in sepsis, prompt identification of DI vs AMI, and optimal management strategies within this high‐risk patient population is needed.

Author contributions

Mahek Shah and Soumya Patnaik contributed equally to this article.

Conflicts of interest

The authors declare no potential conflicts of interest.

Supporting information

Appendix Table 1 ICD‐9CM codes used for identification of admissions for sepsis.

Appendix Table 2. ICD‐9CM codes for diagnosis and procedures.

Appendix Figure 1. Trends in hospital mortality according to study subgroup between 2011 and 2014 using National Inpatient Sample

Shah M, Patnaik S, Maludum O, et al. Mortality in sepsis: Comparison of outcomes between patients with demand ischemia, acute myocardial infarction, and neither demand ischemia nor acute myocardial infarction. Clin Cardiol. 2018;41:936–944. 10.1002/clc.22978