131–136, 2018
Received 31 May 2017; first review completed 19 Jun 2017; accepted in final form 10 Jul 2017
Downloaded from https://journals.lww.com/shockjournal by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3tjcLwhL8g9a8Clf3LIPrl59qreEzz126Vo41IgDj4hfuAyrr6GljjQ== on 02/08/2019
ABSTRACT—Objective: Identify predictors of cardiogenic etiology among emergency department (ED) patients with
hypotension, and use these predictors to create a clinical tool to discern cardiogenic etiology of hypotension. Methods: This
secondary analysis evaluated a prospective cohort of consecutive patients with hypotension in an urban, academic, tertiary
care ED from November 2012 to September 2013. We included adults with hypotension, defined as a new vasopressor
requirement, systolic blood pressure (SBP) < 90 mm Hg after at least 1 L of crystalloid or 2 units packed red blood cells, or
SBP < 90 mm Hg and fluids withheld due to concern for fluid overload. The primary outcome was cardiogenic etiology,
adjudicated by two physician chart review, with 25% paired chart review (kappa ¼ 0.92). We used multivariable logistic
regression to predict cardiogenic etiology, utilizing clinical data abstracted from the electronic medical record. We created a
prediction score from significant covariates and calculated its test characteristics for cardiogenic hypotension. Results: Of
700 patients with hypotension, 107 (15.3%, 95% CI: 12.6%–18.0%) had cardiogenic etiology. Independent predictors of
cardiogenic etiology were shortness of breath (OR 4.1, 95% CI: 2.5–6.7), troponin > 0.1 ng/mL (37.5, 7.1–198.2),
electrocardiographic ischemia (8.9, 4.0–19.8), history of heart failure (2.0, 1.1–3.3), and absence of fever (4.5,
2.3–8.7) (area under the curve [AUC] ¼ 0.83). The prediction score created from these predictors yielded 78% sensitivity
and 77% specificity for cardiogenic etiology (AUC ¼ 0.827). Conclusions: Clinical predictors offer reasonable ED screening
sensitivity for cardiogenic hypotension, while demonstrating sufficient specificity to facilitate early cardiac interventions.
KEYWORDS—Cardiogenic, emergency department, shock
131
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.
132 SHOCK VOL. 49, No. 2 HENNING ET AL.
known heart failure or hemodialysis dependent), or a vasopressor requirement. Sample size justification
We excluded patients with hypotension due to atrial fibrillation with rapid
ventricular response or supraventricular tachycardia who were discharged once This study sought to identify predictors of cardiogenic hypotension using
rate control was achieved, intoxication, and hypotension due to transient multivariable logistic regression. We estimated a priori that 15% of patients had
medication effect. We also excluded patients who had a documented baseline cardiogenic etiology, allowing up to 10 predictors to be included in the logistic
systolic blood pressure < 90 mm Hg, unless a 10 mm Hg decrease in systolic regression model using the n/10 rule to prevent over fitting.
blood pressure occurred.
ED screening and enrollment occurred using an alert preprogrammed into
clinical information technology systems. All patients with hypotension noted RESULTS
at triage, in nursing notes, or on the bedside monitors (two readings more
than 5 min apart), or who required vasopressor medications at any point Of 700 patients enrolled in the parent study, 107 of 700
during their ED stay, were prospectively identified for possible inclusion in the (15.3%) had a cardiogenic cause as the most likely etiology of
study. Eligible patients then underwent a confirmatory chart review to affirm
the presence of shock and absence of exclusion criteria. This confirmatory hypotension in the ED. Seventy-six of 700 (10.9%) met criteria
review and subsequent data abstraction typically occurred 1 month after ED for cardiogenic shock, and 68 of 700 (9.7%) were cardiogenic
arrival and was performed without subsequent knowledge of the hospital patients who died in-hospital. Cardiogenic patients had a higher
course.
prevalence of comorbidities generally associated with cardiac
Data collection disease, including coronary artery disease, heart failure, myo-
cardial infarction, and chronic obstructive pulmonary disease,
Elements of the history of present illness, initial vital signs, physical
examination, past medical history, and medications were abstracted from the while non-cardiogenic hypotension was more frequently asso-
hospital record into an a priori standardized database for each enrolled patient. ciated with alcoholism and end-stage liver disease (Table 1).
The history and physical examination came exclusively from the ED attending Cardiogenic patients had higher mean troponin concentrations,
and resident charts. Basic demographics, length of stay, disposition data, and all
laboratory values were obtained from the electronic health record. Electrocar- whereas non-cardiogenic patients demonstrated higher initial
diogram ischemia was defined as any electrocardiographic findings that temperatures and percentage of immature granulocytes
suggested ischemia according to the final cardiologist interpretation. (Table 2).
A study investigator adjudicated each patient’s underlying cause of insta-
bility using all available information (e.g., diagnostic testing results, physician The cohort had an overall mortality of 19.6% (95% CI:
notes, etc.) at the end of the hospital stay. Underlying causes were initially 16.8%–22.7%) before discharge. Patients with cardiogenic
classified as septic, cardiogenic, hemorrhagic, hypovolemic, anaphylactic,
neurogenic, or other based on previously published methods and definitions
(10, 11, 15), and then reclassified as a binary outcome: cardiogenic or non- TABLE 1. Demographics of patients with cardiogenic and
cardiogenic. Patients were systematically screened for evidence to determine a non-cardiogenic etiologies of hypotension
cause of shock, and each patient was assigned a single diagnosis that Non- P
represented the most likely cause of shock in the ED. To determine inter-
Cardiogenic cardiogenic value
rater reliability, a second physician reviewer adjudicated a 25% sample,
and the agreement between the two reviewers was found to be sufficient to n 107 593
proceed with a single adjudication for each patient (kappa ¼ 0.92, 95% CI: Age in years (16) 68.3 (18.5) 65.2 (16.6) 0.08
0.83 –1.0). Male (%) 58 (54.2) 334 (56.3) 0.68
Race (%)
Outcomes African American 15 (14.0) 61 (10.3) 0.25
Asian 5 (4.7) 23 (3.9) 0.70
The primary outcome was the adjudicated diagnosis: cardiogenic or non-
cardiogenic etiology of hypotension. Secondary outcomes included cardiogenic Hispanic 3 (2.8) 26 (4.3) 0.47
shock in the ED (patients meeting any one criteria for end-organ damage: White 66 (61.7) 410 (69.1) 0.13
altered mental status, lactate 2 mmol/L, or vasopressor requirement) (10, 16, Other 18 (16.8) 73 (12.3) 0.20
17) and in-hospital mortality. Code status in ED
Do not intubate 16 (15.0) 76 (12.8) 0.55
Data analysis Do not resuscitate 16 (15.0) 80 (13.5) 0.05
Comorbidities
Data analysis was performed using SAS v9.3 statistical software (SAS Alcoholism 2 (1.9) 46 (7.8) 0.03
Institute Inc, Cary, NC). A binary outcome variable was created for the primary Diabetes 36 (33.6) 154 (26.0) 0.10
outcome of cardiogenic etiology (i.e., cardiogenic or non-cardiogenic). A
Foley catheter 1 (0.9) 21 (3.5) 0.15
univariate assessment of the association between diagnosis and binary clinical
covariates was performed using chi-square. For continuous covariates, Student t Vascular catheter 3 (2.8) 59 (10.0) 0.02
test or Wilcoxon rank sum was used as appropriate. A multivariate logistic Heart failure 32 (29.9) 102 (17.2) <0.01
regression model was to predict cardiogenic etiology of hypotension. All Coronary artery disease 41 (38.3) 128 (21.6) <0.01
covariates from the initial univariate analysis with P < 0.1 were included in Chronic obstructive pulmonary 20 (18.7) 68 (11.5) 0.04
the logistic regression model selection process. A stepwise model selection disease
process was used, with P < 0.05 required for entry and P < 0.1 required to stay Dementia 6 (5.6) 41 (6.9) 0.62
in the model. Model discrimination was assessed using area under the curve Hypertension 56 (52.3) 250 (42.2) 0.05
(AUC). The Hosmer–Lemeshow test was used to assess model calibration. The Intravenous drug use 1 (0.9) 8 (1.4) 0.73
model was internally validated using bootstrapping techniques. The logistic
End-stage liver disease 2 (1.9) 57 (9.6) <0.01
regression analysis was repeated on 1,000 different samples of the same size
drawn with replacement from the original dataset. Median b-estimates and odds Peripheral vascular disease 11 (10.3) 39 (6.6) 0.17
ratios with 95% confidence intervals were recalculated for each model covariate Chronic renal insufficiency 13 (12.2) 38 (6.4) 0.04
using these repeated samples. Hemodialysis 7 (6.5) 54 (9.1) 0.39
b-coefficients from the initial logistic regression analysis were used to Stroke 11 (10.3) 43 (7.3) 0.28
assign weight to each independent predictor. The test characteristics for Myocardial infarction 14 (13.1) 34 (5.7) <0.01
different score thresholds were calculated to optimize sensitivity and specificity Human immunodeficiency virus 0 (0) 11 (1.9) 0.16
when predicting cardiogenic etiology for hypotension. As sensitivity analyses, Hematologic malignancy 3 (2.8) 22 (3.7) 0.64
the prediction score test characteristics were calculated in two subpopulations Solid malignancy 19 (17.8) 151 (25.5) 0.09
of cardiogenic hypotension to assess the efficacy in identifying the highest risk
Solid organ transplant 0 (0) 18 (3.0) 0.07
cardiogenic population most likely to benefit from early interventions: cardio-
genic shock in the ED and cardiogenic patients who died in hospital. ED indicates emergency department.
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.
SHOCK FEBRUARY 2018 CARDIOGENIC HYPOTENSION IN THE EMERGENCY DEPARTMENT 133
TABLE 2. Clinical results associated with cardiogenic and TABLE 4. Prediction scores assigned for each covariate to predict
non-cardiogenic etiologies of hypotension cardiogenic hypotension
Non- P Covariate Score
Cardiogenic cardiogenic value
Troponin 0.1 ng/mL 3
Initial vital signs (SD) Electrocardiogram ischemia 2
Temperature (8F) 97.2 (6.5) 100.2 (6.3) <0.01 Reported shortness of breath 1.5
Heart rate (beats/min) 88 (28.6) 94 (24.6) 0.02 Absence of reported or measured fever 1.5
Respiratory rate (breaths/min 21 (5.6) 20 (6.1) 0.34 History of heart failure 0.5
Systolic blood pressure (mm Hg) 92 (23.3) 87 (16.2) 0.01
A total score 2 suggests cardiogenic hypotension.
Diastolic blood pressure (mm Hg) 56 (15.7) 51 (12.6) <0.01
Oxygen saturation (%) 97 (4.3) 97 (4.5) 0.55
Laboratory results (SD) (Table 5). In the sensitivity analyses for higher-risk cardiogenic
White blood cell count (cells/mm3) 12.6 (7.3) 12.8 (9.3) 0.79
patients, a score 2 had 82% sensitivity and 75% specificity
Immature granulocytes (cells/mm3) 1.1 (1.1) 7.0 (9.2) <0.01
Hematocrit (%) 36.0 (7.7) 33.8 (7.6) <0.01 identifying cardiogenic shock in the ED and 79% sensitivity
Platelets (1,000 cells/mm3) 253 (116.2) 236 (146.1) 0.25 and 71% specificity for patients who died in-hospital. In the
Bicarbonate (mEq/L) 21.9 (21.9) 22.0 (5.6) 0.86 overall study population, 28.6% of patients with a score 2 had
Creatinine (mg/dL) 2.1 (1.8) 2.0 (1.9) 0.55 cardiogenic shock in the ED, and 10.4% of patients with a
International normalized ratio 1.8 (1.7) 1.8 (1.5) 0.83
score 2 had cardiogenic hypotension and died before dis-
Glucose (mg/dL) 159 (92.1) 148 (93.3) 0.31
Lactate (mmol/L) 3.3 (3.0) 2.8 (2.3) 0.09 charge (Fig. 1).
Troponin (ng/mL) 0.76 (1.8) 0.17 (0.3) <0.01
Cardiac specific (%)
Troponin 0.1 ng/mL 13 (12.2) 2 (0.3) <0.01 DISCUSSION
Shortness of breath 48 (44.9) 109 (18.4) <0.01
Among patients with persistent hypotension in the ED, the
Electrocardiogram ischemia 25 (23.4) 14 (2.4) <0.01
Absence of reported or 95 (88.8) 366 (61.7) <0.01 likelihood of a cardiogenic etiology can be assessed using
Measured fever simple history, physical, and laboratory information. A clinical
SD indicates standard deviation. prediction score of 2 has reasonable sensitivity and specifici-
ty for cardiogenic cause of hypotension, increasing the likeli-
hood of cardiogenic hypotension from 15.3% in the overall
hypotension had mortality of 26.2% (95% CI: 18.7%–35.3%) hypotensive population to 38.3%. Only 5% of patients with < 2
which increased to 34.2% (95% CI: 24.5%–45.4%) if evidence points had cardiogenic hypotension.
of cardiogenic shock (altered mental status, lactate 2 mmol/ The clinical prediction score identified patients within the
L, or vasopressor requirement) was observed in the ED. two higher-risk sub-populations of cardiogenic hypotension
Patients determined to have cardiogenic hypotension, but not with similar efficacy. While prospective validation and rule
meeting criteria for shock, had 6.5% (95% CI: 0.8%–21.8%) refinement to include echocardiography are necessary, this
mortality (Supplemental Table 1, http://links.lww.com/SHK/ novel prediction score demonstrated adequate sensitivity for
A618). high-risk cardiogenic patients to justify clinical implementa-
The covariates reported shortness of breath (4.1, 95% CI: tion as an ED screening tool. The specificity of 75% signifi-
2.5–6.7), history of heart failure (2.0, 95% CI: 1.1–3.3), cantly narrows the hypotensive population to one where the
absence of measured or reported fever (4.5, 95% CI: 2.3– assessment for early MCS would not overwhelm cardiology or
8.7), troponin 0.1 ng/mL (OR 37.5, 95% CI: 7.1–198.2), and system resources. Based on the low incidence of cardiogenic
ischemia on electrocardiogram (8.9, 95% CI: 4.0–19.8) (AUC shock across an already infrequent group of patients with
0.827) were independently predicted cardiogenic hypotension persistent hypotension, only one in four patients screened in
(Table 3). The bootstrapping internal validation provided simi- by the score would have cardiogenic shock. Yet, using the study
lar estimates (Supplemental Table 2, http://links.lww.com/ institution as a model, a patient would meet score threshold less
SHK/A619). than once daily. In exchange for the opportunity to initiate early
The prediction score assigned points based on the covariate cardiac interventions, the resources required to implement this
b-coefficients from the logistic regression model (Table 4). A screening may be inexpensive to the health system.
composite score of 2 or more had 78% sensitivity and 77% The logistic regression model and prediction score give
specificity for cardiogenic hypotension in this population weights to the various findings typically integrated into an
TABLE 3. Logistic regression model predicting cardiogenic hypotension among all ED patients with persistent hypotension
Covariate AOR 95% CI b-coefficient Std. error P value
Troponin 0.1 (ng/mL) 37.5 7.1–198.2 3.62 0.85 <0.01
Electrocardiogram ischemia 8.9 4.0–19.8 2.18 0.41 <0.01
Report shortness of breath 4.1 2.5–6.7 1.41 0.41 <0.01
History of heart failure 2.0 1.1–3.3 0.67 0.27 <0.01
Absence of reported or measured fever 4.5 2.3–8.7 1.50 0.34 <0.01
AUC ¼ 0.827 Hosmer–Lemeshow ¼ 0.09
AUC indicates area under the curve.
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.
134 SHOCK VOL. 49, No. 2 HENNING ET AL.
TABLE 5. Test characteristics of a prediction score 2 for cardiogenic hypotension, cardiogenic shock in the ED, and cardiogenic patients
who died in-hospital among overall population with persistent hypotension
Sens (%), Spec (%), PPV (%), NPV (%),
Population n Score 2 (95% CI) (95% CI) (95% CI) (95% CI)
Cardiogenic hypotension 107 83 78 (70–90) 77 (70–82) 38 (32–45) 95 (93–97)
Cardiogenic shock in the ED 76 62 82 (73–90) 75 (72–79) 29 (23–35) 97 (96–99)
Cardiogenic hypotension with mortality 28 22 79 (63–94) 71 (68–74) 26 (20–32) 98 (96–99)
ED indicates emergency department; NPV, negative predictive value; PPV, positive predictive value; Sens, sensitivity; Spec, specificity.
ED diagnosis. Notably, a history of coronary artery disease or reflecting the lower incidence of cardiogenic shock in the
myocardial infarction were not significant independent predic- emergency setting, expert recommendations give the vague
tors of cardiogenic hypotension after controlling for the other guidance that a ‘‘cardiogenic cause of shock should be consid-
model covariates. Similarly, a history of heart failure only met ered for shock in the ED without an obvious cause’’ (19),
score threshold if accompanied by another independent predic- without acknowledging the diagnostic difficulty in identifying
tor. Conversely, 2 of 15 (13.3%) patients with a tropo- those patients with a cardiogenic cause of shock in the absence
nin 0.1 ng/mL and 14 of 39 (35.9%) with ischemia on of an obvious ischemic event. We found that cardiogenic
electrocardiogram had non-cardiogenic hypotension, remind- hypotension accounted for nearly one of seven ED patients
ing clinicians to avoid basing clinical diagnoses on the presence with persistent hypotension. These cardiogenic patients fre-
or absence of a single data point. Along these lines, our recent quently met criteria for cardiogenic shock in the ED, which was
study demonstrated that patients with suspected septic shock associated with a 34.2% mortality rate (20). Furthermore, the
often did not have measured or reported fever (18). That study majority of our patients with cardiogenic hypotension did not
informs the application of the proposed criteria here, reminding have a troponin > 0.1 ng/mL or electrocardiogram ischemia,
clinicians that while hypotensive patients without signs of fever suggesting that myocardial infarction is less frequently the
have a higher likelihood of cardiogenic cause, sepsis may still cause of shock in ED patients. Still, early MCS may benefit
be present and an assessment for infectious causes is still patients with cardiogenic shock, even without myocardial
necessary. Until novel diagnostic approaches improve the infarction, by offloading the heart to promote recovery and
identification of cardiogenic hypotension, this process provides maintaining cardiac output to prevent end-organ damage as a
a reasonable approach. bridge to more definitive treatment such as durable ventricular
Prior studies describe cardiogenic shock most frequently assist devices or cardiac transplant (22, 23). An emphasis in the
occurring as the sequela of myocardial infarction (19, 20), current literature on managing shock in the context of acute
and typically occurring after a patient is admitted (21). Likely myocardial infarction likely reflects the inherent difficulty
FIG. 1. Demonstration of the patient populations created when stratifying ED patients with persistent hypotension using the proposed prediction
score at a threshold of two points. ED indicates emergency department.
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.
SHOCK FEBRUARY 2018 CARDIOGENIC HYPOTENSION IN THE EMERGENCY DEPARTMENT 135
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.
136 SHOCK VOL. 49, No. 2 HENNING ET AL.
21. Babaev A, Frederick PD, Pasta DJ, Every N, Sichrovsky T, Hochman JS: report: 15,000 patients and counting. J Heart Lung Transplant 34(12):1495–
Trends in management and outcomes of patients with acute myocardial 1504, 2015.
infarction complicated by cardiogenic shock. JAMA 294(4):448–454, 2005. 24. Seif D, Perera P, Mailhot T, Riley D, Mandavia D: Bedside ultrasound in
22. Mody KP, Takayama H, Landes E, Yuzefpolskaya M, Colombo PC, Naka Y, Jorde resuscitation and the rapid ultrasound in shock protocol. Crit Care Res Pract
UP, Uriel N: Acute mechanical circulatory support for fulminant myocarditis 2012(3):1–14, 2012.
complicated by cardiogenic shock. J Cardiovasc Transl Res 7(2):156–164, 2014. 25. Hall MK, Taylor RA, Luty S, Allen IE, Moore CL: Impact of point-of-care
23. Kirklin JK, Naftel DC, Pagani FD, Kormos RL, Stevenson LW, Blume ED, ultrasonography on ED time to disposition for patients with nontraumatic shock.
Myers SL, Miller MA, Baldwin JT, Young JB: Seventh INTERMACS annual Am J Emerg Med 34(6):1022–1030, 2016.
Copyright © 2017 by the Shock Society. Unauthorized reproduction of this article is prohibited.