Prognostic performance of the IABP-SHOCK II Risk Score among cardiogenic shock subtypes in the critical care cardiology trials network registry

Abstract

Background

Risk stratification has potential to guide triage and decision-making in cardiogenic shock (CS). We assessed the prognostic performance of the IABP-SHOCK II score, derived in Europe for acute myocardial infarct-related CS (AMI-CS), in a contemporary North American cohort, including different CS phenotypes.

Methods

The critical care cardiology trials network (CCCTN) coordinated by the TIMI study group is a multicenter network of cardiac intensive care units (CICU). Participating centers annually contribute ≥2 months of consecutive medical CICU admissions. The IABP-SHOCK II risk score includes age > 73 years, prior stroke, admission glucose > 191 mg/dl, creatinine > 1.5 mg/dl, lactate > 5 mmol/l, and post-PCI TIMI flow grade < 3. We assessed the risk score across various CS etiologies.

Results

Of 17,852 medical CICU admissions 5,340 patients across 35 sites were admitted with CS. In patients with AMI-CS (n = 912), the IABP-SHOCK II score predicted a >3-fold gradient in in-hospital mortality (low risk = 26.5%, intermediate risk = 52.2%, high risk = 77.5%, P < .0001; c-statistic = 0.67; Hosmer-Lemeshow P = .79). The score showed a similar gradient of in-hospital mortality in patients with non-AMI-related CS (n = 2,517, P < .0001) and mixed shock (n = 923, P < .001), as well as in left ventricular (<0.0001), right ventricular (P = .0163) or biventricular (<0.0001) CS. The correlation between the IABP-SHOCK II score and SOFA was moderate (r² = 0.17) and the IABP-SHOCK II score revealed a significant risk gradient within each SCAI stage.

Conclusions

In an unselected international multicenter registry of patients admitted with CS, the IABP- SHOCK II score only moderately predicted in-hospital mortality in a broad population of CS regardless of etiology or irrespective of right, left, or bi-ventricular involvement.

Background

Patients with cardiogenic shock (CS) are at substantial risk of adverse outcomes,^1,2 as reflected by short-term mortality rates from 30% to 40% despite contemporary therapy.^2–5 Early identification as well as timely triage and selection of therapies in patients with CS based on risk for adverse outcomes, including mortality and/or clinical deterioration, has the potential of assisting in therapeutic decisions and in communicating with such risk to families and other healthcare providers, but a role in improving outcomes has not been established.⁶ Determination of risk using practical risk scores may help triage and guide clinical decision-making.

Several risk scores have been developed and validated in broad general critical care populations.^7–9 These scores demonstrate good predictive performance across conditions encountered in the general medical and surgical intensive care units (ICUs). However, the populations used in the development of these risk scores did not include patients with CS, nor include destination therapies such as temporary or durable mechanical circulatory support (MCS), and such scores do not perform well in such patients with CS.^10–13 Risk scores focusing on patients with CS have also been developed. However, they are less established in clinical practice and have mostly focused on acute myocardial infarction-related CS (AMI-CS) with limited assessment in patients presenting with CS other than AMI-CS,^12,14–17 which represent only about 30% of the patients with CS admitted to the contemporary CICU.⁵ The IABP-SHOCK II risk score is a pragmatic tool combining clinical and laboratory parameters that was derived in the largest randomized clinical trial of mechanical circulatory support (MCS) in AMI-CS and validated in an observational European cohort.^16–19 We consider important to examine the real world performance of the IABP SHOCK II score across common etiologies and hemodynamic profiles of CS beyond AMI-CS, which was the original cohort used to derive this score. Such analysis is aimed at providing relevant data for the contemporary management of CS, in which AMI-CS represents less than 50% of all CS.¹²

Accordingly, given the diverse and heterogenous phenotype of CS (eg, mixed, vasodilatory, bi-ventricular, etc), we assessed the prognostic performance of the IABP-SHOCK II score for predicting in-hospital mortality in a prospective North American cohort of patients with CS of varied etiology, including AMICS, non-AMICS and mixed shock, as well as according to right, left or biventricular involvement.

Methods

Study population

The critical care cardiology trials network (CCCTN) is an investigator-initiated collaborative research network of American Heart Association level 1 cardiac intensive care units (CICUs)²⁰ in the United States and Canada. The CCCTN is scientifically overseen by its academic Executive and Steering Committees, and the data are coordinated by the TIMI study group (Boston, MA). The composition, structure, and data capture of the CCCTN registry have been previously reported,²¹ and the general epidemiology of shock in this cohort has been described.⁵ Participating centers annually contribute at least a 2-month “snapshot” of consecutive medical admissions to the CICU. Patients in this analysis were admitted anytime between September 2017 and September 2021 across 35 CICUs in North America. The CCCTN registry protocol and waiver of informed consent were approved by the institutional review board at each of the participating institutions and no personal identifying health information was collected in the database.

Patients were included in the present analysis if they met criteria for shock (both CS and non-CS), as defined by sustained hemodynamic instability (ie, systolic blood pressure < 90 mm Hg or the need for inotropic or vasopressor agents) with evidence of end-organ hypoperfusion (eg, altered mental status, oliguria, acute kidney injury, hepatic injury, metabolic acidosis, or elevated serum lactate [>2 mmol/L]). The etiology of shock was further classified by the individual site investigators as described previously, including patients with and without CS.⁵ CS was further categorized as either (1) AMI-CS or (2) CS not precipitated by acute coronary syndrome (eg, nonischemic cardiomyopathies, ischemic cardiomyopathies without ACS, or other cardiac causes of shock with the exception of postcardiotomy CS). CS was also characterized based on underlying ventricular dysfunction (ie, left ventricular failure, right ventricular failure, biventricular failure, or other, including those due to arrhythmia, valvular heart disease, or left ventricular outflow track obstruction) as defined by imaging and/or hemodynamic assessment. The category of mixed shock included CS patients for whom more than one type of shock was determined to contribute (eg, both CS and distributive shock).

Risk score calculation and outcome

The IABP-SHOCK II risk score was derived to assess the risk of 30-day mortality after presentation with AMI-CS and includes the following six variables: age > 73 years, prior stroke, admission glucose > 191 mg/dl, creatinine > 1.5 mg/dl, lactate > 5 mmol/l, and postcoronary intervention TIMI flow grade < 3.¹⁸ Each variable is assigned either 1 or 2 points (Table 1) and the patients are classified into 3 a priori risk categories according to total score: low risk for 0-2 points, intermediate for 3-4 points, and high for 5-9 points. Laboratory elements of the score were assessed based on the values at admission to the participating CICU. The only variable with missing values was lactate (n = 707 out of total 5340, 13%) with missing data contributing a value of zero towards the patient’s total score per usual convention. For patients without AMI-CS and those who did not receive coronary intervention, the variable for TIMI flow grade was assigned zero points in the calculation yielding a maximum possible score of 7 instead of 9.

Table 1.

Variables and risk scores included in the IABP-SHOCK II score, SCAI shock stages, and SOFA score.

IABP shock II score variables definition		SCAI shock stages definition in CCCTN
Variable	Points	SCAI A	Acute myocardial infarction or heart failure without shock and normal renal function* and normal lactate
Age > 73	1	SCAI B	Hypotension without shock, no more than minimal renal impairment,* and normal lactate†
History of stroke	2
Glucose > 10.6 mmol/L (191 mg/dl)	1	SCAI C	Site-reported CS, and use of inotropes/vasopressors or MCS, and abnormal lactate† or abnormal renal function* or increased liver function tests‡
Creatinine > 132.6 umol/L (1.5 mg/dl)	1
Arterial lactate > 5 mmol/L	2	SCAI D	Site-reported CS meeting stage C above, and lactate↑ ≥ 50%, or use of multiple inotropes/vasopressors, or MCS (new or > 1 device) initiated > 24 h following admission
TIMI flow < 3 after PCI	2
Total possible score	9	SCAI E	Site-reported cardiogenic shock, and (worst) pH ≤ 7.2 or (worst) lactate ≥ 5.0 mmol/L

SOFA score variable definition
Score	Glasgow coma score	PaO 2/FiO2, mmHg	Mean arterial pressure (mmHg) or Vasopressors administered (μg/kg/min)	Platelets x 103/mm3	Bilirubin, mg/dl(μmol/L)	Creatinine, mg/dl(μmol/L) or urine output (mL/day)
0	15	</ = 400	MAP >/ = 70	>/ = 150	< 1.2 (< 20)	< 1.2 (< 110)
1	13-14	< 400	MAP < 70	< 150	1.2-1.9 (20-32)	1.2-1.9 (110-170)
2	10-12	< 300	Dopamine </ = 5 or dobutamine (any dose)	< 100	2.0-5.9 (33-101)	2.0-3.4 (171-299)
3	6-9	< 200 with ventilatory support	Dopamine > 5 or epinephrine </ = 0.1 or norepinephrine </ = 0.1	< 50	6.0-11.9 (102-204)	3.5-4.9 (300-440) or UO < 500 mL/day
4	< 6	< 100 with ventilatory support	Dopamine > 15 or epinephrine > 0.1 or norepinephrine > 0.1	< s	> 12.0 (> 204)	> 5.0 (> 440) UO < 500 mL/day

CS, cardiogenic shock; MCS, mechanical circulatory support.

^*Normal renal function is defined as estimated glomerular filtration rate (eGFR) > 90 mL/min/1.73 m²; minimal renal impairment is defined as eGFR 60-90 mL/min/1.73 m²; abnormal renal function defined as eGFR < 45 mL/min/1.73 m².

^†Normal lactate is defined as worst lactate < 2 mmol/L or missing; abnormal lactate ≥ 2 mmol/L.

^‡Increased liver function tests: alanine transaminase or aspartate aminotransferase > 150 IU/L or 3× the upper limit of normal.

The primary outcome of this analysis was in-hospital death. We also examined metrics of resource utilization, including use of invasive hemodynamic monitoring (pulmonary artery catheter, central line, or arterial line), mechanical ventilation (MV), renal replacement therapy, maximum number of vasoactive medications, length of stay as well as use and timing of MCS.

Statistical analysis

Data are reported as counts and percentages for categorical variables, and as medians with 25th and 75th percentiles for continuous data. In-hospital mortality and resource utilization were stratified by risk category according to the IABP-SHOCK II score. In-hospital mortality rates were compared using the Pearson’s chi-square test. Discriminatory performance is reported using the c-statistic. In-hospital mortality rates were also compared across categories both of the sequential organ failure assessment (SOFA) score, given its validation in critical care populations with previously defined cutoffs,⁵ and the Society for Cardiovascular Angiography and Intervention (SCAI) shock stages (Table 1).²² Correlation between the IABP-SHOCK II score and SOFA score was assessed by calculating the Spearman correlation coefficient classifying SOFA stages as a continuous variable. Assessment of score calibration was performed using the Hosmer-Lemeshow goodness-of-fit test, as well as by plotting the observed vs expected risk based on the original predicted categorical risk. A sensitivity analysis of IABP-SHOCK II score performance without the variable for TIMI flow grade was performed among all patients, including those with AMI-CS, as well as for all patients with nonmissing lactate. Multivariable assessment of each individual variable comprising the score was performed for different shock etiologies. Multivariate logistic regression was also performed including IABP SHOCK II score, SOFA score, and SCAI shock stages. A nominal α level of <0.05 was considered statistically significant. All reported P values were 2-sided. Statistical analysis was performed using SAS System V9.4 (SAS Institute Inc, Cary, NC).

Results

Distribution of risk score

Of the 17,852 medical CICU admissions, a total of 17,486 subjects had IABP-SHOCK II scores available, among whom 5,340 (30.5%) patients had shock (AMI-CS = 912, CS without MI = 2,517, Mixed shock = 923 and 988 with other shock) and were included in this analysis. Table 2 displays the clinical and demographic characteristics of patients with shock according to their IABP-SHOCK II score risk categories. Patients in the high-risk category were older, more likely to be women, and had higher rates of prior cerebrovascular disease, coronary disease, hypertension, diabetes, chronic kidney disease, and atrial fibrillation, which is expected for some of these variables which are part of the IABP SHOCK II score itself (Table 2). Supplementary Tables S1–S3 describe the clinical and demographic characteristics of patients across IABP-SHOCK II risk score categories stratified by subtype of shock (AMI-CS, non-AMI CS, and mixed shock). Consistent with the derivation population from the IABP-SHOCK II trial,¹⁸ the majority of patients fell into the low-risk category (Figure 1). Patients with AMI-CS had the highest percentage of patients in the high-risk category (7.8%).

Table 2.

Clinical and demographic characteristics in patients with shock according to IABP SHOCK II risk category.

	Low (n = 3,791) score 0-2	Intermediate (n = 1,312) score 3-4	High (n = 237) score 5-9
Age (y, median IQR)	64 (54, 72)	69 (58,77)	76 (71, 81)
Female (%)	34.2	36.3	36.3
Hypertension	57.6	66.9	74.3
Diabetes mellitus	34.4	44.1	52.3
CAD	35.1	39.7	48.5
Cerebrovascular disease	2.2	25.1	53.6
Heart failure	55.4	44.6	44.7
Chronic kidney disease	27.9	32.9	38.0
Peripheral arterial disease	8.7	s	15.6
Atrial fibrillation	30.7	30.6	38.8
Ventricular arrhythmias	9.4	7.0	5.9
Lactate level on admission	2.0 (1.3-3.1)	6.3 (2.9-9.1)	7.7 (5.9-10.5)
eGFR (mL/min/1.732)	50.1 (29.9-73.1)	36.2 (23.1-53.5)	31.6 (20.7-41.3)
SOFA score	7.0 (5.0-10.0)	10.0 (7.0-13.0)	11 (8.0-13.0)
SCAI SHOCK stage
B or C (%)	77.8	19.6	2.6
D (%)	63.8	29.8	6.4
E (%)	34.5	47	18.5

The reported data represent the % in each category unless otherwise stated.

Figure 1.

Distribution of frequency of risk categories according to the presence of cardiogenic shock and etiology. AMI-CS, acute myocardial infarction related cardiogenic shock.

Performance for in-hospital mortality prediction

Evaluation of the individual risk score variables and their association with in-hospital mortality are shown in Table 3 and Table 4. Lactate >5 mmol/L was associated with the highest relative odds of death across all shock etiologies, except for mixed shock where TIMI flow < 3 had the highest OR. Overall, the IABP-SHOCK II risk score identified a stepwise gradient of increasing in-hospital mortality across all shock categories (Figure 2A, [low risk = 26.6%, intermediate risk =48.5%, high risk = 76.4%, P < .001]). Among patients with AMI-CS (n = 912), there was a >3-fold gradient in in-hospital mortality between patients in the highest vs lowest-risk categories (Figure 2B, (low risk = 26.5%, intermediate risk =52.2%, high risk = 77.5%, P < .0001). Discrimination for in-hospital mortality was modest (c-statistic 0.67) and model calibration was very good in the AMI-CS group (Figure 2B dashed lines, Hosmer-Lemeshow P = .79). Similarly, when assessing the IABP SHOCK II score performance in non AMICS, the score demonstrated a strong gradient of in-hospital mortality by risk category in patients with non-AMI CS (n = 2,517, [low risk = 22.0%, intermediate risk = 40.9%, high risk = 71.6%, Figure 2C, P < .0001]), as well as in patients with mixed shock (n = 923, (low risk = 37.5%, intermediate risk =56.5%, high risk = 80.0%, Figure 2D, P < .0001]). C-statistics were 0.61 in both non-AMI CS and mixed shock, respectively, and 0.63 in the entire shock cohort, including distributive and hypovolemic shock. When specifically examining distributive shock, the IABP SHOCK II score performed similarly in predicting in-hospital mortality, with rate of 31.5%, 50%, and 80% in the low, intermediate, and high-risk categories respectively (P < .0001, c-statistics 0.60; Hosmer-Lemshow P = .52). The results of the sensitivity analysis performed in patients with nonmissing lactate were similar to the primary analysis.

Table 3.

Univariable and multivariable association of the IABP-SHOCK II risk score components with mortality in the overall shock population (AMI-CS, non-AMI CS, and mixed shock) population.

Variable	Univariable analysis (OR, 95% CI)	Multivariable analysis (adjusted-OR, 95% CI)
Age > 73 years	1.66 (1.47-1.88)	1.59 (1.40-1.81)
History of stroke	1.41 (1.18-1.69)	1.33 (1.10-1.61)
Glucose > 191 md/dL	1.89 (1.67-2.13)	1.45 (1.27-1.65)
Creatinine > 1.5 mg/dL	1.79 (1.59-2.01)	1.69 (1.49-1.90)
Lactate > 5 mmol/L	3.76 (3.28-4.31)	3.32 (2.88-3.83)
TIMI flow < 3 post PCI	2.64 (1.82-3.83)	2.76 (1.87-4.09)

Table 4.

Multivariate analysis of each individual variable for mortality predictor across different shock etiologies.

Variable	AMICS (OR, 95% CI)	CS without MI (OR, 95% CI)	Mixed Shock (OR, 95% CI)	No Shock (OR, 95% CI)
Age > 73 years	2.32 (1.69-3.18)	1.70 (1.38-2.08)	1.65 (1.23-2.21)	1.90 (1.59-2.27)
History of stroke	0.90 (0.54-1.49)	1.25 (0.92-1.71)	1.82 (1.20-2.74)	1.22 (0.94-1.59)
Glucose > 191 md/dL	1.32 (0.97-1.78)	1.37 (1.10-1.69)	1.30 (0.97-1.75)	1.39 (1.13-1.71)
Creatinine > 1.5 mg/dL	2.09 (1.55-2.81)	1.55 (1.28-1.87)	1.79 (1.36-2.37)	2.53 (2.12-3.02)
Lactate > 5 mmol/L	4.35 (3.07-6.15)	3.04 (2.44-3.79)	2.22 (1.60-3.08)	7.34 (5.47-9.87)
TIMI flow < 3 post PCI	2.65 (1.65-4.24)	N/A	4.52 (1.20-17.02)	0.73 (0.34-1.58)

Figure 2.

In hospital Mortality prediction by risk category according to the different etiologies of shock. A, All patients with shock: intermediate risk OR 2.6, 95% CI 2.28-2.96, high risk OR 8.93 (6.56-12.15); B, AMICS: intermediate risk OR 3.02, 95% CI 2.23-4.08, high risk OR 9.51 (5.29-17.11); C, Non-AMI CS: intermediate risk OR 2.44, 95% CI 1.99-3.00, high risk OR 8.93 (5.33-14.97); D, Mixed shock: intermediate risk OR 2.16, 95% CI 1.61-2.89, high risk OR 6.65 (3.26-13.57). Doted lines represent the expected in-hospital mortality according to the original IABP SHOCK II score study (derivation cohort). A, AMICS; B, Non-AMI CS; C, Mixed shock. AMICS, acute myocardial infarction related cardiogenic shock.

Among left ventricular, right ventricular or biventricular CS subgroups, there was an association between IABP-SHOCK II risk score and all-cause in-hospital mortality in all shock profiles according to type of ventricle involved (Figure 3). Compared with patients with predominant LV shock, in-hospital mortality in patients with RV shock was higher in each risk category, particularly in the highest risk group. Similarly, the presence of biventricular shock was associated with the highest in-hospital mortality, followed by RV and then LV predominant shock.

Figure 3.

IABP SHOCK II score in-hospital mortality risk prediction according to shock predominance. Reference group for odd ratio, P-value and C-index is the low-risk category group.

The correlation between IABP-SHOCK II score and SOFA was only modest in the overall CS cohort (r² = 0.12). The c-statistic for in-hospital mortality using the SOFA score was 0.68 in all shock patients. The integration of both scores is shown in Supplementary Figure 1A and B. A subgroup of patients (n = 2,469) had SCAI Shock stage data collected starting in 2019. The IABP SHOCK II score demonstrated a stepwise gradient for risk prediction for in-hospital mortality within each SCAI Shock stage C (P = .012), D (P < .0001) E (P =.0009) (Figure 4). IABP-SHOCK II score and SCAI Shock stages are not tightly correlated in the overall CS cohort (r² = 0.06). Table 5 reports the results of multivariate logistic regression for in-hospital mortality according to IABP SHOCK II, SOFA score and SCAI shock stages in the overall CS cohort.

Figure 4.

In-hospital mortality risk stratified by IABP SHOCK II and SOFA scores in all patients with shock. All categories with > = 10 admissions are included.

Table 5.

Logistic regression for IABP SHOCK II, SOFA and SCAI shock staging for all patients with cardiogenic shock.

Score	OR (95% CI)	P-value
IABP SHOCK II score	1.31 (1.21-1.42)	< .0001
SOFA score	1.14 (1.10-1.18)	< .0001
SCAI shock stages (continuous)	2.80 (2.37-3.31)	< .0001

Resource utilization by risk score

Table 6 displays metrics of ICU resource utilization according to risk categories. For example, for patients with all etiologies of shock, the percent usage of mechanical ventilation increased across IABP-SHOCK II risk profile (low risk 47.8%, intermediate risk 64.8%, and high risk 72.2%), however the duration of mechanical ventilation decreased with each risk category. The number of inotropic agents used increased with each risk category, with most of the patients in the lower risk category receiving only 1 inotropic agent and those in the high-risk group having the higher percentages of patients receiving more than 3 inotropic agents (Supplementary Figure 2). Supplementary Table 5 describes resource utilization by risk category according to shock etiology.

Table 6.

Resource utilization in the CICU in patients with shock.

IABP risk score	Low 0-2	Intermediate 3-4	High > 5
n = 5339	n = 3791	n = 1312	n = 237
Days of ICU Care* (median, 25th-75th percentiles)	4.8 (2.3-9.3)	3.7 (1.5-7.6)	2.7 (0.7-5.2)
Days of hospital stay* (median, 25th-75th percentiles)	11.1 (5.7-21.1)	8.0 (2.5-16.8)	3.2 (0.8-8.9)
Days of ICU care in survivors (median, 25th-75th percentiles)	4.9 (2.6-9.3)	5.1 (2.7-9.3)	4.6 (3.2-7.8)
Days of hospital stay in survivors* (median, 25th-75th percentiles)	12.9 (7.1-23.6)	13.7 (7.9-23.4)	10.3 (7.8-18.6)
Days of ICU care in nonsurvivors (median, 25th-75th percentiles)*	4.5 (1.8-9.2)	2.2 (0.7-5.5)	1.7 (0.6-4.1)
Days of hospital stay in nonsurvivors (median, 25th-75th percentiles)	5.8 (2.2-13.9)	2.7 (0.9-7.7)	1.8 (0.6-5.0)
Mechanical ventilation*	1811 (47.8%)	850 (64.8%)	171 (72.2%)
Renal replacement therapy*	524 (13.8%)	289 (22.0%)	48 (20.3%)
Invasive hemodynamic monitoring†	2940 (77.6%)	1072 (81.7%)	192 (81.0%)
Pulmonary artery catheter	1509 (39.8%)	472 (36.0%)	83 (35.0%)
Arterial line*	2333 (61.5%)	931 (71.0%)	172 (72.6%)

Data presented are the n (%) unless otherwise specified.

^*P < .0001.

^†0.037

Mechanical circulatory support by risk score

In patients with AMI-CS, MCS was used in 59.8% of the low-risk category, 62.3% of the intermediate-risk category, and 59.2% of the high-risk category. In patients with non-AMI-CS, MCS was used in 30.6% of the low-risk category, 25.1% of the intermediate risk category, and 18.9% of the high-risk category. Patients in the low-risk group were found to have the longest admission to MCS initiation time. (Supplementary Table S6). IABP was the most commonly used mechanical support device in all risk categories and across different etiologies of shock (Supplementary Figure S3). In-hospital mortality prediction by IABP SHOCK II score was similar between patients receiving MCS and those without MCS, however the prediction gradient was not as robust for patients receiving either percutaneous LVAD (eg, Impella) or ECMO when compared to those without MCS or those receiving IABP (Figure 5).

Figure 5.

In-hospital mortality prediction in percentage according to presence and type of mechanical circulatory support in the entire cohort. ECMO, extracorporeal membrane oxygenation; pLVAD, percutaneous left ventricular assist device (eg, Impella).

Discussion

In the present study, we analyzed the ability of the previously published IABP-SHOCK II score to stratify mortality risk in a contemporary, multicenter, North American cohort of patients with shock of different etiologies admitted to tertiary care CICUs. Our findings demonstrate that the IABP-SHOCK II score can stratify CICU and inhospital mortality in a broad population of patients with CS, alone or in conjunction with either the SOFA score or SCAI shock classification. This finding includes patients with AMI-CS, in which it performed better, as well as patients with CS without AMI and patients with shock subtypes. Similarly, as a novel finding, the IABP-SHOCK II score performed consistently across patients with left, right or biventricular shock predominance and for patients who did and did not receive temporary MCS.

Risk stratification in shock

Risk scores can facilitate the classification of severity of illness and the identification of patients at higher risk of adverse outcomes. While they have not been validated to assist in selecting therapies or escalation of treatment (as opposed to decision-making tools), they can be useful in communicating risk of adverse outcomes to patients, families, and other team members, and have the potential of being used to adjust for severity of illness in clinical registries according to risk of mortality or clinical deterioration . In addition, such risk scores can enable tracking of risk-adjusted ICU performance metrics²³ and potentially assist in patient selection for clinical trials.²⁴ As well, accurate risk stratification can facilitate shared-decision making, especially for end-of-life goals of care.^12,25 While few studies have evaluated the applicability of general ICU scores in patients in the CICU^26,27 and specifically those with CS,^10,13,19 several risk scores have been specifically developed for patients with CS, including the IABP-SHOCK II score, as a practical tool for risk stratification in CS.^10,11,16,28

In patients with AMI-CS, the IABP-SHOCK II score may provide the best prognostic discrimination for early mortality when compared to SAPS II, CardShock, and SCAI shock staging, according to a substudy of the CULPRIT-SHOCK trial that included 1,055 patients.¹⁷ Such findings suggest continued potential value of the IABP-SHOCK II score in the context of contemporary CS staging using the SCAI classification system, particularly as it may be easier to operationalize IABP-SHOCK II scores using electronic medical records, in comparison with SCAI shock stages as their definition may be more loose and may differ between centers, practitioners and cohorts. Our results align with prior published data demonstrating that SCAI classification performs better in highest risk categories in contrast to IABP SHOCK II score, ¹⁷ as well as with previously published data from our cohort, where SCAI shock staging demonstrated a c-statistic of 0.852, while it was 0.761 for the IABP-SHOCK II score in 1,991 CCCTN patients.²⁹ Similarly, in a prior study of 181 patients with AMI-CS undergoing pharmacoinvasive treatment,³⁰ the IABP-SHOCK II score predicted mortality with no apparent interaction between the score and median symptom-to-needle time or fibrinolytic-to-catheterization time. In addition, when the IABP-SHOCK II score was originally developed, it underwent external validation in the subgroup of patients with AMI-CS (n = 137) from the European CardShock registry,¹⁴ demonstrating a higher c-statistic (0.73) than in our cohort; we note the much larger cohort size in our current study, including a multinational group of CICUs less prone to selection biases inherent to single center practices. Lastly, the diminished performance of the IABP-SHOCK II score in patients with non-AMI-CS is expected, particularly insofar as the presence of TIMI flow grade < 3 after PCI is a crucial predictor of in-hospital mortality that is not relevant for those without AMI, although it was a strong predictor for patients with mixed shock (Supplementary Table 4).

In non-AMI CS, some risk scores have been evaluated in this population, including CRASH,³¹ Inova Heart and Vascular Institute,³² the cardiogenic shock score (CSS),¹⁶ and CardShock.¹⁴ However, our study, which has a significantly larger cohort, extends these findings to patients with mixed shock as well as by analyzing performance among important hemodynamic subsets of CS (right, left, biventricular). In addition, our study adds to an analysis that compared the IABP-SHOCK II score with the CardShock score in another real-world population (n = 696) with both AMI-CS and non-AMI CS, in which the overall AUC was 0.752 with lower predictive performance in non-AMI shock patients (n = 262, AUC 0.619).³³ Lastly, our study provides insight into the utilization of CICU therapies, including mechanical ventilation, hemodynamic monitoring, MCS and renal replacement in detail fashion according to shock etiology and severity. This data can provide useful information for practicing clinicians, to understand the impact of CS severity and progression in treatment and resource needs.

Clinical implications

When applying risk scores, it is important to take into consideration their performance in the population of interest, the feasibility of incorporating the scores into daily practice, and the specific patient demographics and clinical settings where they are being applied.³⁴ Accordingly, considering that AMI-CS only represents about one third of patients with CS in the modern CICU, understanding the performance of the IABP-SHOCK II score, as a simple clinical tool, in other CS subtypes is relevant to current practice.⁵ Our study incorporates a substantially broader clinical spectrum of patients than the original score development cohort, making our findings novel and relevant for clinicians managing CS patients in the contemporary CICU.

Another advantage of our study is the more granular assessment of shock phenotypes according to right, left or bi-ventricular involvement, which has not been previously evaluated systematically using other CS risk scores. This is of pivotal importance as under-recognition of RV involvement in CS, is associated with worse outcomes and has significant therapeutic and prognostic implications. Our study demonstrates the performance of the IABP-SHOCK II score in patients with right, left and biventricular CS, which highlights the role of this score across a spectrum of congestive profiles. Our findings also illustrate the higher in-hospital mortality risk in patients with right ventricular and biventricular shock than those with isolated left ventricular involvement in each risk category, although the type of ventricular involvement did not dramatically alter score performance. The prognostic performance in such patients is relevant to clinical practice given the increased recognition of the importance of RV and biventricular shock phenotypes.^35–38 Moreover, as hemodynamics appear to play a critical role in the early management of CS and have been featured prominently in other risk scores ³⁸ their integration with the IABP-SHOCK II score along with SCAI shock staging may provide strong predictive discrimination.

In patients who received MCS support, we observed a similar prognostic performance for in-hospital mortality prediction of the IABP-SHOCK II score compared to patients managed without MCS, although such findings were less robust for patients receiving advanced MCS (pLVAD or ECMO), who had higher observed in-hospital mortality particularly at lower IABP-SHOCK II scores. Notably, our data show a lack of relationship between inhospital mortality risk and MCS use, with patients in the low and intermediate risk categories receiving the most ventricular support with ECMO, while the overall use of MCS did not significantly differ between risk categories (Supplementary Figure 3). Such findings differ from the SCAI SHOCK staging, that has shown a correlation between risk severity and MCS use,³⁹ demonstrating the difference between a risk-stratification tool (eg, IABP-SHOCK-II score) and a severity of illness classification scheme (eg, SCAI shock stage).²² It is likely also relevant that the level of MCS can drive SCAI staging while IABP-SHOCK II is agnostic of hemodynamic support use, adding potential benefit to its use in clinical practice, by providing a global risk stratification ability to all patients, with our without circulatory support. This discrepancy underscores the important distinction between shock severity itself and mortality risk—while mortality risk is higher in those with greater shock severity, a heavy burden of nonmodifiable risk factors for mortality, not always captured in risk scores, can result in high observed mortality despite modest shock severity. This concept is elaborated in the updated SCAI shock classification statement, which introduced the 3-axis model for understanding risk stratification in CS that incorporates shock severity, clinical phenotype, and nonmodifiable risk factors. The IABP-SHOCK II score includes primarily nonmodifiable risk factors, apart from lactate level which is a surrogate for shock severity and glucose level, as a marker of stress response activation. As such, IABP-SHOCK II score is useful for mortality risk stratification but may be less relevant for selection of therapeutic interventions. Lastly, the ideal risk score should incorporate clinical, biochemical, hemodynamic, etiologic and imaging parameters in order to accurately predict adverse outcomes and, if validated, to assist in selecting therapies, including early transfer to shock centers, escalation of MCS or identification of futility.

Limitations

Limitations of the present study include the small number of patients in the high-risk subgroup, albeit similar to the distribution in the original IABP-SHOCK II score development study. In addition, our cohort did not include patients who died before arrival to CICU, as opposed to the development and validation cohort that may have included patients who died in the cardiac catheterization laboratory. In our cohort and the highest risk scores observed in the AMI-CS cohort might have been due to the additional variable of absence of TIMI 3 flow providing additional points. Of note in our cohort there was no association between prior cerebrovascular disease or stroke and in-hospital mortality, and between hyperglycemia on admission and in-hospital mortality, which was not the case in the derivation cohort, probably suggestive of a less sick population in the original derivation and validation cohorts for IABP SHOCK II score. Such discrepancies could explain why our c-statistic values were lower than expected. The absence of invasive hemodynamic parameters, such as the pulmonary artery pulsatility index, in the IABP SHOCK II and other scores is a potential limitation; however, it also enables them to be applied when invasive hemodynamics are not available.

Conclusions

The IABP-SHOCK II score performs moderately for the purpose of stratifying risk of in hospital mortality in patients with CS with different etiologies and ventricular involvement. The IABP-SHOCK II score may be a useful tool for in-hospital mortality prediction in real world patients with CS of different etiologies, inclusive of LV, RV, and biventricular shock, and may provide a complementary and objective risk assessment when added to SCAI shock staging and hemodynamic assessment. Such risk assessment has the potential to assist the clinician in determining risk of adverse outcomes when discussing treatment decisions with patients, relatives and other colleagues about prognosis and identify who may benefit or be harmed from particular escalation of therapies, according to their specific prognosis with the ability of assisting in how to discern futility and care pathways as indicated. Lastly, they may assist in adjusting for risk categories for clinical trial selection or performance metric assessment.