Original Research

Journal of Intensive Care Medicine

1-7
Comparative Prognostic Accuracy of Risk ª The Author(s) 2019
Article reuse guidelines:
sagepub.com/journals-permissions
Prediction Models for Cardiogenic Shock DOI: 10.1177/0885066619878125
journals.sagepub.com/home/jic

Robert J. H. Miller, MD1 , Danielle Southern, MSc2,

Stephen B. Wilton, MD1,2, Matthew T. James, MD2,3,
Bryan Har, MD1, Greg Schnell, MD1, Sean van Diepen, MD, Msc4,
and Andrew D. M. Grant, MD1

Abstract
Objectives: Despite advances in medical therapy, reperfusion, and mechanical support, cardiogenic shock remains associated
with excess morbidity and mortality. Accurate risk stratification may improve patient management. We compared the accuracy of
established risk scores for cardiogenic shock. Methods: Patients admitted to tertiary care center cardiac care units in the
province of Alberta in 2015 were assessed for cardiogenic shock. The Acute Physiology and Chronic Health Evaluation-II
(APACHE-II), CardShock, intra-aortic balloon pump (IABP) Shock II, and sepsis-related organ failure assessment (SOFA) risk
scores were compared. Receiver operating characteristic curves were used to assess discrimination of in-hospital mortality and
compared using DeLong’s method. Calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test. Results: The
study included 3021 patients, among whom 510 (16.9%) had cardiogenic shock. Patients with cardiogenic shock had longer median
hospital stays (median 11.0 vs 4.1 days, P < .001) and were more likely to die (29.0% vs 2.5%, P < .001). All risk scores were
adequately calibrated for predicting hospital morality except for the APACHE-II score (Hosmer-Lemeshow P < .001). Dis-
crimination of in-hospital mortality with the APACHE-II (area under the curve [AUC]: 0.72, 95% confidence interval [CI]: 0.66-
0.76) and IABP-Shock II (AUC: 0.73, 95% CI: 0.68-0.77) scores were similar, while the CardShock (AUC: 0.76, 95% CI: 0.72-0.81)
and SOFA (AUC: 0.76, 95%CI: 0.72-0.81) scores had better discrimination for predicting in-hospital mortality. Conclusions: In a
real-world population of patients with cardiogenic shock, existing risk scores had modest prognostic accuracy, with no clear
superior score. Further investigation is required to improve the discriminative abilities of existing models or establish novel
methods.

Keywords
cardiogenic shock, risk prediction, shock, acute coronary syndrome

Introduction
Cardiogenic shock is the most common cause of death in
patients following acute myocardial infarction (MI).1,2 Despite
major advances in the treatment of MI and heart failure, the
prognosis of patients who develop cardiogenic shock remains
poor.3 Prompt revascularization and hemodynamic support
may stabilize and facilitate recovery in some patients with
cardiogenic shock4; however, there is a subset of patients who
will require prolonged use of inotropic medications, mechani-
cal circulatory support (MCS), or even heart transplantation.5
These more aggressive treatments are resource-intensive and
have an increased risk of complications.3 Accurately identify-
ing patients in cardiogenic shock at the highest risk may help
target aggressive therapies, potentially improving outcomes
while minimizing costs and complications associated with
unnecessary interventions.6
Multiple risk prediction tools for patients with shock have
been developed. The Acute Physiology and Chronic Health
1
Department of Cardiac Sciences, Libin Cardiovascular Institute of Alberta,
Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada
2
Department of Community Health Sciences, O’Brien Institute for Public
Health, Cumming School of Medicine, University of Calgary, Calgary, Alberta,
Canada
3
Department of Medicine, Libin Cardiovascular Institute of Alberta, Cumming
School of Medicine, University of Calgary, Calgary, Alberta, Canada
4
Division of Cardiology, Department of Critical Care, Faculty of Medicine,
University of Alberta, Edmonton, Alberta, Canada
Received May 29, 2019. Received revised July 10, 2019.
Accepted September 04, 2019.
Corresponding Author:
Andrew D. M. Grant, Department of Cardiac Sciences, Libin Cardiovascular
Institute of Alberta, Cumming School of Medicine, University of Calgary, Room
C-835, 1403–29th Street NW, Calgary, Alberta, Canada T2N 2Y8.
Email: andrew.grant@albertahealthservices.ca
2 Journal of Intensive Care Medicine XX(X)
Evaluation-II (APACHE-II) and Sequential Organ Failure
Assessment (SOFA) scores were originally derived for use in
a broader group of critically ill general medical and surgical
patients.7,8 Initial studies investigating their utility in patients
with cardiogenic shock showed only modest predictive perfor-
mance.9 Subsequently, the CardShock score was derived in a
cohort of patients with all causes of cardiogenic shock and
reported improved discrimination compared to the APACHE-
II score with area under the curve (AUC) of 0.85 and 0.76,
respectively.10 Importantly, the authors noted that the Card-
Shock score did not perform as well in a validation cohort with
an AUC of 0.71.10 More recently, data from the intra-aortic
balloon pump (IABP) in cardiogenic Shock II (IABP-ShockII)
was used to develop a risk score which may have improved
discriminatory value.11 External validation of these scores in an
unselected tertiary care population of patients presenting with
cardiogenic shock is lacking and their comparative prognostic
accuracy remains unclear.12
This study was a retrospective cohort study of patients pre-
senting with cardiogenic shock to tertiary care coronary care
units (CCUs) in Alberta, Canada. The study aim was to exter-
nally validate the performance of 4 existing risk models for
discriminating all-cause mortality in patients with cardiogenic
shock and to compare their prognostic performance.
Materials and Methods
Patient Population
Patients admitted to a tertiary care CCU in Alberta between
January 1, 2015, and December 31, 2015, were identified. Two
centers (Foothills Medical Center and Mazankowski Heart
Institute) were included, which provide advanced cardiology
coverage to over 4 million people through a centralized referral
system. Patient demographics, admission date, admission diag-
nosis, and discharge date were retrieved for the TRACER data-
base which prospectively collects information on patients
admitted to tertiary care CCUs in Alberta. Data automatically
collected in the TRACER database has shown high agreement
with manually collected data, including the APACHE-II and
SOFA scores.13 Clinical information including vital signs, use
of inotropic or vasopressor support, MCS, and laboratory
investigations throughout admission were used to identify
patients with evidence of cardiogenic shock. Patients were
classified as cardiogenic shock if they met a modified defini-
tion of the Should We Emergently Revascularize Occluded
Coronaries for Cardiogenic Shock (SHOCK) trial clinical cri-
teria for enrollment. This included sustained systolic blood
pressure (SBP) <90 mm Hg for at least 30 minutes or required
inotropic or MCS to maintain a SBP >90 mm Hg with evidence
of end-organ hypoperfusion.4 Markers of end-organ dysfunc-
tion included elevated lactate (>2.0 mmol/L),4 reduced urine
output (defined as <0.5 mL/kg/h for at least 6 hours), or acute
kidney injury based on Kidney Disease Improving Global Out-
comes definition.14 Admission diagnoses were reviewed and
patients with a primary or secondary diagnosis including acute
coronary syndrome (ACS), cardiomyopathy, or dysrhythmias
were included in the analysis. Early revascularization was
defined as percutaneous coronary intervention or coronary
artery bypass grafting occurring within 12 hours of admission
to CCU.4
Medical comorbidities including cardiovascular risk factors
(hypertension, diabetes, dyslipidemia, family history, and
smoking history) and other comorbidities (cerebrovascular dis-
ease, peripheral vascular disease, chronic lung disease, chronic
renal failure, chronic liver disease, malignancy, and deep vein
thrombosis/pulmonary embolism) were determined through
linkage with the Alberta Provincial Project for Outcome
Assessment in Coronary Heart Disease (APPROACH) data-
base. The APPROACH database prospectively collects data
on all patients undergoing cardiac catheterization in Alberta15
and has been successfully linked with several other adminis-
trative databases previously for clinical research.16 Trained
physicians or technicians enter demographics, medical history,
and angiographic characteristics.17 Use of inotropic or vaso-
pressor support was captured within the TRACER database.
Use of an inotrope for less than 60 consecutive minutes was
not included. Milrinone, dobutamine, and dopamine were con-
sidered inotropic supports and norepinephrine, vasopressin,
and epinephrine were considered vasopressor supports. All
patients who were managed with IABP, Impella (Impella CP;
Abiomed, Aachen, Germany), extracorporeal membrane oxy-
genation (ECMO), or surgical left ventricular assist device
(LVAD) implantation during the index hospitalization were
classified as MCS recipients.
The APPROACH database to determine the primary out-
comes of all-cause in-CCU and in-hospital mortality identified
through a quarterly linkage with the Alberta Bureau of Vital
Statistics. APACHE-II and SOFA scores are recorded in the
TRACER database by the initial attending physician.7 Existing
clinical information was used to determine the CardShock and
IABP-ShockII scores.8 Components included in the risk scores
are summarized in Supplemental Table 1. Lactate and glucose
were not collected in all patients with cardiogenic shock (miss-
ing in 10 and 18, respectively) and were assumed to be normal
for the purpose of risk score calculation. In patients presenting
with a non-ACS diagnosis without coronary angiography,
thrombolysis in myocardial infarction (TIMI) flow grade was
assumed to be 3. Patients presenting with ACS who died prior
to coronary angiography (n ¼ 10) were assumed to have TIMI
flow grade <3.
Statistical Methods
Our primary analysis was performed on the subset of patients
with cardiogenic shock. Categorical variables were summar-
ized as number (proportion) and were compared with a Fisher
exact test. Continuous variables had a nonparametric distribu-
tion and therefore were summarized as median (interquartile
range) and compared with a Wilcoxon rank-sum test. Receiver-
operating characteristic (ROC) curves were constructed for
each of the risk scores, applying them as continuous variables.
Miller et al 3

Table 1. Overall Patient Characteristics.a

All Patients, n ¼ 3021 Cardiogenic Shock, n ¼ 510 No Cardiogenic Shock, N ¼ 2511 P Value

Age (years), median (IQR) 64.1 (54.5-74.9) 65.8 (55.8-75.8) 63.5 (54.2-74.6) .009
Male gender, n (%) 2172 (71.9%) 362 (70.9%) 1810 (72.1%) .627
CCU length of stay (days), median (IQR) 1.9 (1.2-3.0) 3.2 (1.6-5.9) 1.8 (1.2-2.7) <.001
Hospitalization length (days), median (IQR) 4.7 (2.9-9.7) 11.0 (5.0-24.0) 4.1 (2.8-7.8) <.001
ACS presentation, n (%) 2224 (73.6%) 292 (57.2%) 1932 (76.9%) <.001
Early revascularization, n (%) 2004 (90.1%) 246 (84.2%) 1758 (91.0%) <.001
Hypertension, n (%) 1026 (34.0%) 102 (20.0%) 924 (36.8%) <.001
Diabetes, n (%) 421 (13.9%) 47 (9.2%) 374 (14.9%) .001
Dyslipidemia, n (%) 721 (23.9%) 51 (10.0%) 670 (26.7%) <.001
History of CAD, n (%) 1552 (51.4%) 129 (25.3%) 1423 (56.7%) <.001
History of CHF, n (%) 261 (8.6%) 73 (14.3%) 188 (7.5%) <.001
APACHE-II score, median (IQR) 10 (7-12) 14 (10-20) 9 (7-11) <.001
CardShock score, median (IQR) 1 (1-2) 2 (1-3) 1 (1-2) <.001
IABP-Shock II score, median (IQR) 0 (0-1) 1 (0-2) 0 (0-1) <.001
SOFA score, median (IQR) 2 (2-4) 6 (3-8) 2 (2-3) <.001
Abbreviations: ACS, acute coronary syndrome; APACHE-II, Acute Physiology and Chronic Health Evaluation-II; CAD, coronary artery disease; CCU, coronary
care unit; CHF, congestive heart failure; IQR, interquartile range; SOFA, Sequential Organ Failure Assessment.
a
Continuous variables are shown as median (interquartile range). Categorical variables are shown as number (proportion).

The AUC for each ROC curve was compared to the values
obtained for each alternate score using the method described
by DeLong et al18 The calibration of the risk scores was
assessed across risk deciles using the Hosmer-Lemeshow
goodness-of-fit (GOF) test. Separate subgroup analyses were
performed stratified by the presence of MCS and ACS. All
statistical tests were 2-sided and P values of < .05 were con-
sidered statistically significant. All analyses were performed
using Stata/IC version 13.1 (StataCorp, College Station,
Texas). This study was approved by the institutional review
board at the University of Calgary.
Results
Patient Population
A total of 3021 patients were admitted to the CCU during the
study period, among whom 510 (16.9%) had cardiogenic
shock. Baseline population characteristics for all patients are
shown in Table 1. Patients with cardiogenic shock were more
frequently older, had fewer preadmission comorbidities, longer
median hospital stays (median: 11.0 vs 4.1 days, P < .001), and
were more likely to die (29.0% vs 2.5%, P < .001). Baseline
population characteristics for patients with cardiogenic shock
stratified by survival status are presented in Table 2. Among
the 510 patients with cardiogenic shock, 148 (29.0%) died in-
hospital, with 103 (20.2%) dying in CCU. In unadjusted anal-
yses, patients who died in hospital were more frequently older
and more likely to require inotropic support (56.1% vs 36.5%,
P < .001), vasopressor support (87.8% vs 66.0%, P < .001), or
mechanical ventilation (62.8% vs 34.8%, P < .001). Mechan-
ical circulatory support use was not more common in patients
who survived (35.1% vs 27.7%, P ¼ .120), although LVAD
implantation was (8.3% vs 2.0%, P < .001). There was a trend
toward more frequent early revascularization in patients with
ACS who survived to discharge (87.1% vs 77.8%, P ¼ .055).
Risk Score Comparison
The individual risk score ROC curves for in-hospital mortality
are shown in Figure 1 and Table 3. The APACHE-II score
demonstrated modest discrimination and poor calibration for
in-hospital mortality (AUC: 0.71, 95% confidence interval
[CI]: 0.66-0.76, GOF P < .001). The discrimination of the
CardShock (AUC: 0.76, 95% CI: 0.72-0.81, GOF P ¼ .628),
IABP-ShockII (AUC: 0.73, 95% CI: 0.68-0.77, GOF P ¼ .222),
and SOFA score (AUC: 0.76, 95%CI: 0.72-0.81, GOF
P ¼ .637) were similar and adequately calibrated. Calibration
graphs for in-hospital mortality are shown in Figure 2. In all
models, there was a trend toward overestimating risk in the
higher risk categories.
The ROC curves for in-CCU mortality are shown in Figure 3.
The calibration of the APACHE-II score for CCU mortality was
poor (AUC: 0.75, 95% CI: 0.69-0.81, GOF P < .001) and
discrimination of the CardShock (AUC: 0.75, 95%
CI: 0.70-0.81, GOF P ¼ .851), IABP-ShockII (AUC: 0.72,
95% CI: 0.67-0.78, GOF P ¼ .240), and SOFA scores
(AUC: 0.79, 95%CI: 0.74-0.84, GOF P ¼ .805) were similar
and adequately calibrated.
Subgroup Analyses
Performance of the risk scores was also assessed in strata
defined by the use of MCS (Supplemental Table 2). There were
no significant differences between AUC of scores in patients
with or without MCS. The prognostic accuracy of all models
was also examined in patients with and without ACS (Supple-
mental Table 3). Model performance was not significantly dif-
ferent in patients with or without an underlying diagnosis of
ACS. However, there was a trend toward higher c-statistic in
patients with ACS.
4 Journal of Intensive Care Medicine XX(X)

Table 2. Baseline Patient Characteristics for Patients With Cardiogenic Shock.a

All Patients, Survived to Discharge, In-hospital Mortality,

N ¼ 510 n ¼ 362 (71.0%) n ¼ 148 (29.0%) P Value

Age (years), median (IQR) 65.8 (55.8-75.8) 64.7 (54.1-74.0) 69.4 (60.7-79.1) <.001
Male, n (%) 362 (70.9%) 266 (73.5%) 96 (65.0%) .054
CCU length of stay (days), median (IQR) 3.2 (1.6-5.9) 3.6 (1.9-6.0) 2.7 (0.9-5.4) <.001
Hospitalization length (days), median (IQR) 11.0 (5.0-24.0) 12.8 (7.6-29.9) 4.2 (1.5-12.6) <.001
ACS presentation, n (%) 292 (57.3%) 202 (55.8%) 90 (60.8%) .325
Early revascularization, n (%) 246 (84.2%) 176 (87.1%) 70 (77.8%) .055
Inotropic support, n (%) 215 (42.2%) 132 (36.5%) 83 (56.1%) <.001
Vasopressor support, n (%) 369 (72.4%) 239 (66.0%) 130 (87.8%) <.001
Any MCS, n (%) 168 (32.9%) 127 (35.1%) 41 (27.7%) .120
IABP support, n (%) 129 (25.3%) 96 (26.5%) 33 (22.3%) .369
Impella support, n (%) 14 (2.8%) 6 (4.1%) 8 (2.2%) .246
LVAD, n (%) 33 (6.5%) 30 (8.3%) 3 (2.0%) <.009
ECMO, n (%) 6 (1.2%) 3 (0.8%) 3 (2.0%) .363
Ventilation, n (%) 219 (42.9%) 126 (34.8%) 93 (62.8%) <.001
Therapeutic hypothermia, n (%) 42 (8.2%) 24 (6.6%) 18 (12.2%) .050
Hypertension, n (%) 102 (20.0%) 80 (22.1%) 22 (14.9%) .068
Diabetes, n (%) 52 (9.2%) 39 (9.4%) 13 (8.8%) 1.000
Dyslipidemia, n (%) 51 (10.0%) 40 (11.1%) 11 (7.4%) .256
History of CAD, n (%) 129 (25.3%) 97 (26.8%) 32 (21.6%) .262
History of CHF, n (%) 73 (14.3%) 58 (16.0%) 15 (10.1%) .095
APACHE-II score, median (IQR) 14 (10-20) 13 (9-17) 19 (11-27) <.001
CardShock score, median (IQR) 2 (1-3) 2 (1-3) 3 (2-4) <.001
IABP-Shock II score, median (IQR) 1 (0-2) 1 (0-1) 2 (1-3) <.001
SOFA score, median (IQR) 6 (3-8) 3 (5-7) 8 (6-11) <.001
Abbreviations: ACS, acute coronary syndrome; APACHE-II, Acute Physiology and Chronic Health Evaluation-II; CAD, coronary artery disease; CCU, coronary
care unit; CHF, congestive heart failure; ECMO, extracorporeal membrane oxygenation; IABP, intra-aortic balloon pump; IQR, interquartile range; LVAD, left
ventricular assist device; MCS, mechanical circulatory support; SOFA, Sequential Organ Failure Assessment.
a
Continuous variables are shown as median (interquartile range). Categorical variables are shown as number (proportion).

our study, there was no clearly superior score with respect to

Figure 1. Receiver-operating characteristic curves for predicting
in-hospital mortality. APACHE-II indicates Acute Physiology and
Chronic Health Evaluation II; IABP-II, intra-aortic balloon pump II;
SOFA, Sequential Organ Failure Assessment.
Discussion
We used a real-world population of patients with cardiogenic
shock to externally validate 4 established risk scores. In-
hospital mortality in patients with cardiogenic shock was high,
similar to a large multicenter report where cardiogenic shock
was associated with more than 30% in-hospital mortality.19 In
discriminating in-CCU or in-hospital mortality. However, the
APACHE-II score was poorly calibrated suggesting that it may
not be appropriate to apply in patients with cardiogenic shock.
Our results highlight the importance of external validation for
risk scores, ideally in the population in which it will be applied,
before implementation into clinical care.12
There have been several attempts at generating cardiogenic
shock risk scores and it is not clear why existing risk scores
have performed poorly to date. We found no clear superior risk
score despite different derivation cohorts and model selection
methods. Cardiogenic shock populations likely have high lev-
els of heterogeneity, so it may be that no single derivation
cohort is entirely reflective of an external population. This is
evidenced by the relatively high proportion of ACS in our
population compared with other multicenter studies.19 In our
study, the SOFA score had numerically higher AUC values for
both in-CCU mortality and in-hospital mortality. However,
such small differences are unlikely to have a clinical impact
and ease of use based on number and complexity of variables
may be a more practical feature that impacts selection and
uptake of risk scores into practice between scores. In addition,
risk scores trended toward better performance in patients pre-
senting with ACS. This is not surprising since 2 scores were
derived in ACS populations. However, it does suggest that
Miller et al 5

Table 3. Comparison of AUC Values for Risk Scores.a

In-Hospital Mortality, CCU Mortality,

Risk Score AUC (95% CI), n ¼ 148 P Value vs APACHE II AUC (95% CI), n ¼ 103 P Value vs IABP Shock-II

APACHE II 0.71 (0.66-0.76) – 0.75 (0.69-0.81) .404

CardShock 0.76 (0.72-0.81) .022 0.75 (0.70-0.81) .157
IABP-Shock II 0.73 (0.68-0.77) .574 0.72 (0.67-0.78) –
SOFA 0.76 (0.72-0.81) .034 0.79 (0.74-0.84) .036

Abbreviations: AUC, area under the curve; APACHE-II, Acute Physiology and Chronic Health Evaluation-II; CCU, coronary care unit; CI, confidence interval;
IABP, intra-aortic balloon pump; SOFA, Sequential Organ Failure Assessment.
a
Comparison of AUC values for risk scores for discrimination of in-hospital mortality (n ¼ 148) and mortality in CCU (n ¼ 103).

Figure 2. Calibration graphs for in-hospital mortality showing observed versus predicted risk for each decline. Reference line indicates perfect
calibration (intercept 0, slope 1). APACHE-II indicates Acute Physiology and Chronic Health Evaluation II; IABP-II, intra-aortic balloon pump II;
SOFA, Sequential Organ Failure Assessment.

latent class analyses may help further refine future risk estima-
tion attempts by identifying important subgroups of patients
with cardiogenic shock.20 Finally, the factors that predict
short-term mortality differ from those that predict intermediate
or long-term mortality.2,4 Therefore, it may not be possible to
generate a single score to predict events at multiple time
horizons. Implementing risk prediction models into electronic
health records may help to alleviate the burden of repeated risk
calculations and allow for risk estimation at different time
intervals. This approach has been shown to be a viable and
potentially superior in some studies21,22 and could allow for
implementation of more complex models.
6 Journal of Intensive Care Medicine XX(X)
Figure 3. Receiver-operating characteristic curves for predicting in-
cardiac care unit (CCU) mortality. APACHE-II indicates Acute Phy-
siology and Chronic Health Evaluation II; IABP-II, intra-aortic balloon
pump II; SOFA, Sequential Organ Failure Assessment.
Appropriate therapy for patients with cardiogenic shock
involves early and selected use of inotropic supports and MCS,
which was common in our population. While the use of ino-
tropic and vasopressor supports were higher in patients who
died in hospital, this is likely a reflection of the sicker patient
population. However, it may partially reflect the known
adverse effects of these drugs.23 In addition, the use of ECMO
or Impella support were relatively uncommon in our cohort
compared to IABP which has not demonstrated efficacy for
shock following acute MI.24 It is not clear how physicians’
decision-making regarding the use of MCS interacts with exist-
ing risk estimation tools or our estimation of their utility. How-
ever, the 4 existing risk scores did not perform significantly
different in patients with or without MCS supports. Prospective
studies would be required to determine if a strategy of early
MCS use in patients identified as high-risk is beneficial.
Limitations
Our study has several important limitations including its retro-
spective nature. The SOFA and APACHE-II scores are routi-
nely collected, but the CardShock and IABP-Shock II scores
were derived from existing clinical information. Since this was
a retrospective analysis of prospectively collected data, there
was variation in collection timing which may have impacted
our results. Only patients who survived to CCU were identified,
and therefore our data do not account for patients who may
have died prior to presentation or admission to CCU. In addi-
tion, the present study reflects a snapshot of clinical practice
over 1 year and changes in clinical practice have likely
occurred over time. We used the APACHE-II score instead
of the APACHE-IV score which has been shown to have super-
ior discriminatory ability in the general intensive care popula-
tion.25 Finally, there may be a treatment bias in that patients at
higher risk were treated more aggressively, therefore minimiz-
ing differences in outcomes in the higher risk population. This
would tend to decrease the predictive capacity of risk scores but
should do so in a uniform fashion.
Conclusions
Cardiogenic shock is present in a significant proportion of
patients admitted to tertiary care CCUs and is associated with
a substantial increase in mortality. Existing cardiogenic shock
risk scores have modest prognostic accuracy in this patient
population. Further investigation is required to improve the
discriminative abilities of existing risk prediction models or
establish novel methods for risk prediction. In addition, pro-
spective studies are needed to assess the potential benefit of
using risk scores to guide early MCS utilization.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to
the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for
the research, authorship, and/or publication of this article: Dr Robert J.
H. Miller is supported by the Arthur J. E. Child Fellowship grant.
ORCID iD
Robert J. H. Miller, MD https://orcid.org/0000-0003-4676-2433
Supplemental Material
Supplemental material for this article is available online.
References
1. Goldberg RJ, Makam RCP, Yarzebski J, McManus DD, Lessard
D, Gore JM. Decade-long trends (2001–2011) in the incidence
and hospital death rates associated with the in-hospital develop-
ment of cardiogenic shock after acute myocardial infarction. Circ
Cardiovasc Qual Out. 2016;9(2):117-125.
2. Reynolds HR, Hochman JS. Cardiogenic shock: current concepts
and improving outcomes. Circulation. 2008;117(5):686-697.
3. Werdan K, Gielen S, Ebelt H, Hochman JS. Mechanical circula-
tory support in cardiogenic shock. Eur Heart J. 2014;35(3):
156-167.
4. Hochman JS, Sleeper LA, Webb JG, et al. Early revascularization
in acute myocardial infarction complicated by cardiogenic shock.
N Engl J Med. 1999;341(9):625-634.
5. Thiele H, Ohman EM, Desch S, Eitel I, de Waha S. Management
of cardiogenic shock. Eur Heart J. 2015;36(20):1223-1230.
6. van Diepen S, Katz JN, Albert NM, et al. Contemporary manage-
ment of cardiogenic shock: a scientific statement from the Amer-
ican Heart Association. Circulation. 2017;136(6):232-268.
7. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE
II: a severity of disease classification system. Crit Care Med.
1985;13(10):818-829.
8. Jones AE, Trzeciak S, Kline JA. The Sequential Organ Failure
Assessment score for predicting outcome in patients with severe
sepsis and evidence of hypoperfusion at the time of emergency
department presentation. Criti Care Med. 2009;37(5):1649-1654.
9. Kellner P, Prondzinsky R, Pallmann L, et al. Predictive value of
outcome scores in patients suffering from cardiogenic shock com-
plicating AMI: APACHE II, APACHE III, Elebute-stoner, SOFA,
Miller et al
and SAPS II. Med Klin Intensivmed Notfmed. 2013;108(8):
666-674.
10. Harjola VP, Lassus J, Sionis A, et al. Clinical picture and risk
prediction of short-term mortality in cardiogenic shock. Eur J
Heart Fail. 2015;17(5):501-509.
11. Pöss J, Köster J, Fuernau G, et al. Risk stratification for patients in
cardiogenic shock after acute myocardial infarction. J Am Coll
Cardiol. 2017;69(15):1913-1920.
12. Blaha MJ. The critical importance of risk score calibration. Time
for transformative approach to risk score validation? J Am Coll
Cardiol. 2016;67(18):2131-2134.
13. Brundin-Mather R, Soo A, Zuege DJ, et al. Secondary EMR
data for quality improvement and research: a comparison of
manual and electronic data collection from an integrated crit-
ical care electronic medical record system. J Crit Care. 2018;
47:295-301.
14. Kidney Disease: Improving Global Outcomes (KDIGO) Acute
Kidney Injury Work Group. KDIGO clinical practice guideline
for acute kidney injury. Kidney Int Supp. 2012;2(1):6-16.
15. Ghali WA, Knudtson ML. Overview of the Alberta Provincial
Project for outcome assessment in coronary heart disease. Can J
Cardiol. 2000;16(10):1.
16. Southern D, James M, Wilton S, et al. Expanding the impact of a
longstanding Canadian Cardiac registry through data linkage:
challenges and opportunities. Int J Population Data Sci. 2018;
3:3-9.
17. Southern DA, Norris CM, Quan H, et al. An administrative data
merging solution for dealing with missing data in a clinical reg-
istry: adaptation from ICD-9 to ICD-10. BMC Med Res Methodol.
2008;8(1):1.
7
18. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the
areas under two or more correlated receiver operating character-
istic curves: a nonparametric approach. Biometrics. 1988;44(3):
837-845.
19. Berg DD, Bohula EA, Diepen SV, et al. Epidemiology of shock in
contemporary cardiac intensive care units. Circ Cardiovasc Qual
Outcome. 2019;12(3):e005618.
20. Gardlund B, Dmitrieva NO, Pieper CF, Finfer S, Marshall JC,
Taylor Thompson B. Six subphenotypes in septic shock: latent
class analysis of the PROWESS shock study. J Crit Care. 2018;
47:70-79.
21. Amarasingham R, Velasco F, Xie B, et al. Electronic medical
record-based multicondition models to predict the risk of 30 day
readmission or death among adult medicine patients: validation
and comparison to existing models. BMC Med Inform Decis Mak.
2015;15:39.
22. Calvert J, Mao Q, Hoffman JL, et al. Using electronic health
record collected clinical variables to predict medical intensive
care unit mortality. Ann Med Surg. 2016;11:52-57.
23. Hauptman PJ, Mikolajczak P, George A, et al. Chronic inotropic
therapy in end-stage heart failure. Am Heart J. 2006;152(6):1096.
24. Thiele H, Zeymer U, Neumann FJ, et al. Intraaortic balloon sup-
port for myocardial infarction with cardiogenic shock. N Engl J
Med. 2012;367(14):1287-1296.
25. Varghese YE, Kalaiselvan MS, Renuka MK, Arunkumar AS.
Comparison of Acute Physiology and Chronic Health Evaluation
II (APACHE II) and Acute Physiology and Chronic Health Eva-
luation IV (APACHE IV) severity of Illness Scoring Systems, in a
multidisciplinary ICU. J Anaesthesiol Clin Pharmacol. 2017;
33(2):248-253.