Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data

Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data Nicolò Ghionzoli , View ORCID Profile Andrea Stefanini , Geza Halasz , Carlotta Sorini Dini , Carlotta Sciaccaluga , Francesca Maria Righini , Valerio Guarrasi , Arianna Francesconi , Paolo Soda , View ORCID Profile Maria Concetta Pastore , Hatem Soliman Aboumarie , Vito Piazza , Marco Marini , View ORCID Profile Domenico Gabrielli , Matteo Cameli , Serafina Valente doi: https://doi.org/10.1101/2025.04.28.25326615 Nicolò Ghionzoli a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andrea Stefanini a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Andrea Stefanini For correspondence: astefanini94{at}gmail.com Geza Halasz b Department of Cardio-Thoraco-Vascular Sciences, A.O. San Camillo-Forlanini , Rome, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Carlotta Sorini Dini a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Carlotta Sciaccaluga a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Francesca Maria Righini a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Valerio Guarrasi c Unit of Computer Systems and Bioinformatics, Department of Engineering, Università Campus Bio-Medico di Roma , Rome, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Arianna Francesconi c Unit of Computer Systems and Bioinformatics, Department of Engineering, Università Campus Bio-Medico di Roma , Rome, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Paolo Soda c Unit of Computer Systems and Bioinformatics, Department of Engineering, Università Campus Bio-Medico di Roma , Rome, Italy d Department of Diagnostic and Intervention, Radiation Physics, Biomedical Engineering, Umeå University , Umeå, Sweden PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Maria Concetta Pastore a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Maria Concetta Pastore Hatem Soliman Aboumarie e Department of Anaesthetics and Critical Care, Harefield Hospital, Royal Brompton and Harefield Hospitals , Hill End Road, Harefield UB9 6JH, UK f School of Cardiovascular, Metabolic Sciences and Medicine, King’s College London , UK MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Vito Piazza b Department of Cardio-Thoraco-Vascular Sciences, A.O. San Camillo-Forlanini , Rome, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Marco Marini g Department of Cardiovascular Sciences, Clinic of Cardiology , Ospedali Riuniti, Ancona, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Domenico Gabrielli b Department of Cardio-Thoraco-Vascular Sciences, A.O. San Camillo-Forlanini , Rome, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Domenico Gabrielli Matteo Cameli a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Serafina Valente a Department of Medical Biotechnologies, Division of Cardiology, University of Siena , Siena, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract Background Sub-phenotyping cardiogenic shock (CS) patients using non-traditional clustering methods represents a step toward precision medicine, potentially improving outcomes in this heterogeneous and high-mortality condition. This study aimed to apply an unsupervised machine learning approach to integrate clinical and advanced echocardiographic data, identifying CS sub-phenotypes associated with different outcomes and features, beyond etiology. Methods This multicenter observational study prospectively analyzed 172 patients admitted to Cardiac Intensive Care Units with overt CS, from 2021. An exploratory statistical analysis preceded patient clustering using the Elbow Method and K-Means algorithm, based on clinical presentation. Dimensionality reduction was performed with Principal Component Analysis. Phenotypes were further stratified according to the Society for Cardiovascular Angiography and Interventions (SCAI) stages. Results Five distinct Phenotypes (labeled from I to V) were identified, showing progressively increasing in-hospital mortality rates: 25% (I), 32% (II), 39% (III), 41% (IV), and 60% (V). Kaplan-Meier analysis demonstrated a stepwise increase in mortality risk. Phenotypes IV and V had significantly higher mortality than Phenotype I (HR: 2.78 [95% CI, 1.07-7.19] and HR: 2.80 [95% CI, 1.10-7.14]; P < 0.05). Mortality prediction remained independent after adjustment for confounding factors, and independently of SCAI stage. Phenotype I had the lowest mortality, with higher arterial pressure and moderate left ventricular (LV) dysfunction, whereas Phenotype II exhibited marked LV failure. Oppositely, Phenotypes IV and V had severe congestion despite only mild LV impairment. Conclusions Machine learning, newly integrating echocardiographic data, identified five distinct CS Phenotypes, each with unique clinical/echocardiographic features and mortality risks. These insights could support personalized treatment strategies in CS patients, pending further validation. WHAT IS NEW? This machine learning-based analysis, newly integrating echocardiographic data, identified 5 distinct CS Phenotypes. These Phenotypes have been shown to be easily feasible at the bedside, each with unique mortality risks and features, including varying patterns of cardiac dysfunction and dilatation. WHAT ARE THE CLINICAL IMPLICATIONS? The 5 CS Phenotypes lay the foundation for personalized treatment strategies and improve risk stratification beyond SCAI staging in CS patients. Further external validation of the 5 Phenotypes to other independent CS cohorts, and additional investigations that account for trajectories of patients across Phenotypes, could establish their role in enhancing clinical outcomes. INTRODUCTION Cardiogenic shock (CS) is a subtype of circulatory shock in which end-organ perfusion has inadequate cardiac output as primum movens . It is characterized by the need for vasoactive support and clinical signs of tissue hypoperfusion, together with increased serum lactate levels, resulting in a wide spectrum of phenotypes that reflect diverse etiologies, pathogenetic mechanisms, stages of severity and patient-specific conditions [ 1 – 4 ]. Despite advancements in management, including mechanical circulatory supports, CS remains associated with an extremely high in-hospital mortality rate, ranging from 30% to 60% [ 1 , 2 , 5 , 6 ]. As the public health burden of CS is expected to rise over time, several prognostic scores and classifications have been formulated, in order to best allocate available resources and appropriate therapeutic strategies [ 7 – 12 ]. Specifically, the Society for Cardiovascular Angiography and Interventions (SCAI) proposed a five-stage classification of CS, validated in retrospective studies and recently updated, based on clinical assessment of hypoperfusion, lactate levels, and hemodynamic evaluation [ 12 – 14 ]. Additionally, the sub-phenotyping of patients with CS continues to evolve to disentangle heterogeneity, and the use of an impartial machine learning (ML)-based clustering algorithm – which simplifies routine clinical data into reproducible CS sub-phenotypes – may yield new insights that could support more personalized care [ 15 ]. This approach has been effectively applied for the phenotyping of several clinical syndromes and disorders [ 16 , 17 ], including CS, with the identification of distinct phenotypes, each with specific and reproducible correlations to mortality [ 18 , 19 ]. All these classifications include both hemodynamic variables and indices of metabolic derangement, but lack echocardiographic data, which is the promptest and safest diagnostic modality [ 2 , 20 ], with relevant prognostic role [ 21 – 26 ]. An integrative approach combining clinical variables and echocardiographic parameters could help identify differences in underlying disease mechanisms, offering insights with prognostic and therapeutic implications. We primary aimed to apply an unsupervised ML approach integrating clinical and imaging data (including advanced echocardiography), to identify clinically relevant CS sub-phenotypes associated with different outcomes and treatment responsiveness, as a promising step forward in precision medicine, beyond etiology. As for the secondary objectives, we aimed to characterize the clinical profiles and outcomes associated with identified sub-phenotypes and their reproducibility. Finally, we sought to improve risk stratification of CS patients, in particular by defining subsets of mortality risk within the SCAI staging system. METHODS STUDY DESIGN AND POPULATION In this prospective, observational, multicenter, non-profit study, patients diagnosed with CS at admission to the three Italian cardiac intensive care units were enrolled from November 1st, 2021 to September 30th, 2024. CS diagnosis was physician adjudicated at each site based on the European Society of Cardiology guidelines [ 3 ]. In particular CS was diagnosed based on the presence of clinical signs of hypoperfusion (altered mental status, oliguria, cold extremities, narrow pulse pressure, dizziness) associated with biochemical markers of hypoperfusion (elevated creatinine, metabolic acidosis and elevated serum lactate), with or without hypotension (defined as systolic blood pressure 90 mmHg) [ 3 ]. Thus, patients underwent clinical, biohumoral, and echocardiographic evaluation at admission, within 3 hours from CS identification. Then, all the following clinical events were registered during the course of patients’ hospital stay: in-hospital mortality, major arrhythmias, use of temporary mechanical circulatory supports, heart transplantation and/or left ventricular (LV) assist devices implantation. Only adult patients (aged ≥ 18 years) were considered for analysis, and patients with incomplete data for key variables (in particular missing outcome data) were excluded. All patients provided written informed consent and the study was conducted in accordance with the Declaration of Helsinki. The study was approved by the Tuscany Region Local Ethic Committee for Clinical Trial – section south-west extended area ( approval number: 20438 ). The corresponding author had full access to all the study data and takes responsibility for its integrity and the data analysis. CLINICAL, LABORATORISTIC AND ECHOCARDIOGRAPHIC PARAMETERS For each patient the following data were collected and extracted from electronic health records: baseline demographic characteristic, cardiovascular risk factors, number of vessels affected and culprit lesion (if applicable) if acute coronary syndrome (ACS) etiology, vital signs, cardiac arrest as onset, principal drugs used during hospitalization, blood tests at admission (including blood count, C-reactive protein, kidney function, liver function, high-sensitivity troponin, N-terminal pro B-type natriuretic peptide, arterial lactate levels, arterial oxygen saturation, central venous oxygen saturation), and the use of mechanical circulatory supports. Echocardiography was performed by an expert operator using a fully equipped machine (Vivid E9, GE, Horthern, Norway). All parameters were measured according to the European Association of Cardiovascular Imaging/American Society of Echocardiography guidelines [ 27 ] [ Supplemental methods ]. Speckle tracking echocardiography was performed offline using the dedicated 2D strain softwares (Echopac, GE), by a single independent operator, for each center who analyzed all the images acquired by a second experienced operator in the same center [ Supplemental methods ]. VARIABLE HANDLING AND CLUSTER ANALYSIS An initial statistical analysis focused on candidate variables and in-hospital mortality correlations, using Pearson’s correlation coefficient, was applied. Results were visualized through heatmaps, highlighting the strength and direction of these relationships. Hence, a careful selection of included variables for the subsequent clustering was based not only on their associations with in-hospital mortality, but also on clinical judgement, reflecting underlying disease and pathophysiological processes, in order to avoid the identification of clusters based on outcomes rather than pathophysiology. A sensitivity analysis was conducted by testing clustering stability with and without specific variables (LV global longitudinal strain (GLS), lactate, right ventricular fractional area change). Patients with high proportions of missing data were excluded, to ensure that overall missingness did not exceed 20% for imputation. Remaining missing values were imputed using the K-Nearest Neighbors Imputer (K = 5). The number of variables was established according to the simple size (considering the Formann formula in which clustering analysis should include 2 n individuals, where n is the number of variables) [ 15 ]. Following the exploratory analysis, clustering was conducted by an external expert using the K-Means algorithm. The number of clusters was optimized using the Elbow Method by evaluating the sum of squared distances and further validated with the Silhouette Score, which measures clustering quality. Each cluster’s distinctiveness was examined in relation to in-hospital mortality as the outcome variable using Chi-square test. Dimensionality reduction was performed using Principal Component Analysis to facilitate the visualization of clusters in a two-dimensional space. The Principal Component Analysis plot highlighted the distribution of clusters and their alignment with the target variable. CONVENTIONAL/TRADITIONAL STATISTICAL ANALYSIS All continuous variables were tested for normality using the Kolmogorov-Smirnov test. Continuous variables were expressed as mean ± standard deviation or median and interquartile ranges (IQR), and depending if two-group or five-group comparison, data were compared using the Student’s t-test/Analysis of Variance (ANOVA) or Mann-Whitney/Kruskal-Wallis test, as appropriate. Categorical data were expressed as counts and percentages and were compared using the Chi-square test. All features within each Phenotype were described and compared to summarize each cluster’s key characteristics and all phenotypes were further stratified according to the SCAI shock stage at admission. Univariate and multivariate Cox regression analysis was also used to assess independent predictors of in-hospital mortality, with adjustment for confounding factors (age, sex, CS etiology and mechanical circulatory supports). To examine the impact of phenotype membership on in-hospital mortality, survival curves were generated by the Kaplan–Meier estimator and compared using log-rank test. Analyses were performed by an external expert using the Statistical Package for Social Sciences software, release 30.0 (SPSS). A P value < 0.05 was considered statistically significant. RESULTS STUDY POPULATION Overall, 172 CS patients were enrolled (mean age 65 ± 14 years, 78% (135 patients) male, and BMI 26.4 ± 4.5 kg/m 2 ). Nearly half of patients had a history of hypertension (54%) and dyslipidemia (56%) and the use of at least one mechanical circulatory support was frequent: 52% (89 patients) with intra-aortic balloon pump, 15% (26) with Impella and 14% (23) with extracorporeal membrane oxygenation. Overall in-hospital mortality was 34%. The most common cause of CS was ACS (56%, 97 patients), with non-ACS etiologies accounting for the remaining 44% (75 patients). As for echocardiographic parameters, the population showed enlarged left ventricle (LV end-diastolic diameter (EDD): 57 ± 11 mm) and left atrium (left atrial volume: 76 ± 17 ml), and severe LV dysfunction as measured by LV ejection fraction (EF) of 22 (15-30) %. Right ventricular global function was nearly normal (S’ tissue Doppler imaging (TDI): 0.09 (0.07-0.12) m/s, right ventricular fractional area change: 32 (25-40) %). Regarding STE parameters, patients showed a reduction in LV strain, with LV GLS -6.1 ± 3.5%. The general characteristics of the overall study population are reported in Table 1 . No sex-based differences were present. The overall missingness of the data was 18%; LV GLS was the feature with the highest missingness (50%), reflecting daily practice challenges in cardiac intensive care unit imaging. View this table: View inline View popup Table 1. Characteristics and outcomes of the overall study population and by Phenotypes at the time of admission or during the Hospital stay. PHENOTYPES OF CARDIOGENIC SHOCK After calculating Pearson’s correlation coefficient, age, creatinine, lactate level, procalcitonin and C-reactive protein showed the greatest association with in-hospital mortality ( Figure S1 ). Then, following a careful selection to avoid the identification of clusters based on outcomes rather than pathophysiology, 8 final variables were included in the clustering analysis, including echocardiographic parameters: mean arterial pressure (MAP), central venous pressure, lactate, creatinine, LV EDD, LV EF, S’ TDI and LV GLS. An optimal cluster number of 5 was determined ( Figure S2 ). K-means algorithm identified 5 CS Phenotypes labeled from I to V ( Figure 1 ) based on their unique clinical and echocardiographic characteristics at presentation ( Figure 2 , Table 2 ). The Principal Component Analysis plot highlighted the distribution of the 5 clusters and their alignment with the target variable ( Figure S3 ). Radar plot was used to display the deviation of clinical and echocardiographic values from the mean value ( Figure 3 ). Baseline key characteristics and outcomes across Phenotypes are summarized in Table 1 . Phenotype I, of intermediate age (66 ± 14 years) and lowest in-hospital mortality, had a preserved MAP (85 ± 17 mmHg), nearly normal kidney function (creatinine: 1.20 (IQR 0.87-1.72) mg/dl) and a slight increase in LV EDD (53 ± 10 mm, P = 0.006) with a moderate reduction in LV EF and LV GLS (respectively 25 (IQR: 20-30) % and -6.6 ± 2.5%; P < 0.001) (with LV GLS intraclass correlation coefficient: 0.99 (IQR: 0.99-0.99). Phenotype II exhibited similar metabolic and hemodynamic features relative to Phenotype I, but, interestingly, marked LV dilation (EDD: 72 (IQR: 66-76) mm; P < 0.001) and dysfunction (LV EF: 15 (IQR: 12-20) %; P < 0.001) and preserved right ventricular function (S’ TDI: 0.09 (IQR 0.07-0,11) m/s). Phenotype III, of intermediate age (65 ± 15 years) and in-hospital mortality, primarily showed RV dysfunction (right ventricular fractional area change: 25 (IQR 23-30) %, P < 0.001; S’ TDI: 0.08 (IQR: 0.07-0.09) m/s, P < 0.001; free wall right ventricular longitudinal strain: -13.5 ± 5.9%, P = 0.003) and a higher grade of lactate level (5.9 ± 5.5 mmol/l), systemic congestion (central venous pressure: 18 ± 4 mmHg, P < 0.001) and renal dysfunction (creatinine: 1.66 (1.29-2.40) mg/dl). Conversely, Phenotype IV, of older age (74 ± 11 years, P = 0.003), and Phenotype V showed the highest in-hospital mortality associated with elevated lactate, relevant systemic congestion (central venous pressure: 13.0 ± 7 mmHg and 15.0 ± 4 mmHg, respectively) but only a mild reduction in LV EF (50 (IQR 43-60) % and 31 (IQR 25-40) %, for each), with preserved LV dimension (EDD: 49 ± 4 mm and 52 ± 5 mm, respectively) and right ventricular function. In addition, Phenotype V exhibited severe hypotension (MAP: 57 ± 11 mmHg; P < 0.001) associated with severe renal dysfunction (creatinine: 6.26 (IQR 5.40-8.81) mg/dl; P < 0.001). Download figure Open in new tab Figure 1. Machine learning and Echo Data for Cardiogenic Shock Phenotypes - Machine Learning-based integration of clinical and echocardiographic multicenter data to identify 5 cardiogenic shock Phenotypes, each with unique mortality risks and features. LV, left ventricular; MAP, mean arterial pressure; RV, right ventricular. Download figure Open in new tab Figure 2. The 5 distinct cardiogenic shock Phenotypes - Outstanding characteristics at presentation of the 5 cardiogenic shock Phenotypes. CVP, central venous pressure; EDD, end-diastolic diameter; EF, ejection fraction; GLS, global longitudinal strain; LAC, lactate; LV, left ventricular; MAP, mean arterial pressure; S’ TDI, s’ wave tissue doppler imaging. Download figure Open in new tab Figure 3. Phenotypes’ clinical and echocardiographic trends – This radar plot illustrates the distribution for each Phenotype of clinical and echocardiographic variables. Data were normalized across all Phenotypes. The dashed gray line marks the mean values. Values above the mean are drawn outside the dashed line, whereas values below are drawn inside. The creatinine value scale for Phenotype V has been reduced (as indicated by the double lines on the purple line) to better adjust the relationships between the other Phenotypes and the mean. CVP, central venous pressure; EDD, end-diastolic diameter; EF, ejection fraction; GLS, global longitudinal strain; LAC, lactate; MAP, mean arterial pressure; S’ TDI, s’ wave tissue doppler imaging. View this table: View inline View popup Download powerpoint Table 2. Peculiar and distinguishing features of each Phenotype compared to the others. As for CS etiology, the majority of patients in Phenotype I, IV and V had ACS-related CS, accounting for 72%, 71%, and 80% of cases, respectively. In contrast, 82% of patients with Phenotype II and 51% with Phenotype III had non-ACS-related CS. No statistically significant difference regarding the etiology was observed between Phenotype I and Phenotype IV (P = 0.842) or Phenotype V (P = 0.640). Finally, no statistically significant differences were found in the use of MCS across all the Phenotypes (P = 0.332; specifically: P = 0.149 for IABP; P = 0.653 for Impella; P = 0.199 for ECMO; P = 0.767 for ECMO + IABP; P = 0.306 for ECMO + Impella). ASSOCIATION OF PHENOTYPES WITH OUTCOME In-hospital mortality rates progressively increased across the phenotypes, with rates of 25%, 32%, 39%, 41% and 60% for Phenotype I through V, respectively ( Figure S4 ). Univariate Cox regression analysis revealed a statistically significant rise in mortality corresponding to higher phenotype classification. Cox regression analysis also revealed that, relative to Phenotype I, patients with Phenotype V and IV were at highest risk of mortality (hazard ratio (HR): 2.78 [95% confidence interval (CI), 1.07-7.19] and HR: 2.80 [95% CI, 1.10-7.14] respectively). At multivariate analysis, Phenotypes predicted mortality also after adjusting for confounding factors (age, sex, ACS-related etiology, MCS utilization; adjusted HR 1.29, [95% CI, 1.07-1.57, P = 0.009) ( Table 3 ). View this table: View inline View popup Download powerpoint Table 3. Multivariate and Bivariate Cox analysis. Survival analysis by Kaplan-Meier showed a stepwise increase in mortality for the five clusters ( Figure 4 ); the log-rank tests between all survival curves of the different Phenotypes showed a P = 0.109, but a statistically significant difference in cumulative survival between Phenotype I and Phenotype V (Chi-square 4.880, P = 0.027) or Phenotype IV (Chi-square 5.652, P = 0.017) was observed. Download figure Open in new tab Figure 4. Association of Phenotypes with outcome – Kaplan-Meier survival curves for the 5 cardiogenic shock Phenotypes. PHENOTYPES AND SCAI STAGES We compared the distribution of each Phenotype with regard to their SCAI shock stage at admission ( Figure 5 ): within each Phenotypes, the SCAI staging further stratified mortality. Moreover, within each SCAI stage (C-E), the 5 Phenotypes further stratified mortality. Eventually, at bivariate Cox regression analysis, Phenotypes predict mortality independently of SCAI stage (HR 1.24, CI 1.01-1.53, P = 0.041) ( Table 3 ). Download figure Open in new tab Figure 5. Phenotypes and SCAI stages - In-hospital Mortality according to the SCAI Shock Classification and cardiogenic shock Phenotype. SCAI, Society for Cardiovascular Angiography and Interventions. DISCUSSION To date, this is the first prospective, multicenter study to identify CS sub-phenotypes through a ML algorithm, using not only clinical variables, but also echocardiographic data, beyond etiology. These Phenotypes differ in hemodynamic and metabolic characteristics, as well as in cardiac structure and function, suggesting the presence of distinct clinical groups that are expected to correlate with in-hospital mortality, providing a foundation for personalized treatment strategies in CS patients (all distinctive characteristics of Phenotypes have significant P values). The widespread availability and rapid execution of echocardiography allow for the seamless integration of this information, making this phenotyping approach easily feasible at bedside. In this direction, our results showed that marked cardiac dysfunction and dilatation were not a major discriminator of worse prognosis across all the Phenotypes, according with limited studies that demonstrated the lack of association between severe LV dysfunction and mortality [ 24 – 26 ]. Considering a total of 172 patients, the population was significantly representative of overt CS (SCAI stage C-E), and 65% patients received at least one mechanical device for cardiac/circulatory support, ranging from intra-aortic balloon pump to extracorporeal membrane oxygenators. Phenotype I, that showed the lowest mortality and predominantly ACS-related etiology, had a higher mean arterial pressure, normal kidney function and a slight increase in LV diameter with a moderate reduction in EF and LV GLS. Phenotype II, also associated with lower mortality, exhibited similar metabolic and hemodynamic features but, interestingly, marked LV dilation and dysfunction and normal right ventricular function, identifying younger patients probably with a history of dilated cardiomyopathy. Phenotype III primarily identified CS patients of intermediate age and in-hospital mortality, with relevant RV dysfunction combined with marked systemic congestion and metabolic and renal impairment. Otherwise, Phenotype IV mainly detected older and ACS-related CS patients associated with the highest mortality, alongside Phenotype V. In addition, Phenotype V was distinguished by a marked state of hypotension associated with a potentially consequent severe renal dysfunction. Many of the phenotypes’ clinical characteristics align with established clinical scores and classifications, reinforcing the concepts that hemodynamic and metabolic parameters are associated with increased mortality [ 8 , 9 , 12 , 28 – 32 ]. Either Phenotypes IV and V showed relevant metabolic impairment and systemic congestion, but only a mild reduction in LV EF with normal LV dimension and right ventricular function, suggesting that cardiac function differs between Phenotypes and metabolic derangements can occur in the absence of a marked reduction in LV EF. In this sense, it is becoming increasingly necessary to consider circulatory shock – thus including CS – as a syndrome that originates from hemodynamic alterations of various causes but progresses toward a common final pathway of metabolic derangement, ultimately evolving into a cellular and mitochondrial pathology that influences outcome. Moreover, K. Berg-Hansen et al. demonstrated in a recent study that LV GLS proved as a superior predictor of short-term mortality in patients with CS compared with LV EF (which was not associated with in-hospital mortality), indicating that regular LV GLS assessment could improve risk stratification in this cohort [ 21 ]. In our study, clustering was also performed using LV GLS as a variable to evaluate functional cardiac alterations across different Phenotypes. The results revealed that LV GLS impairment occurred alongside a decline in LV EF; however, since the higher sensitivity, LV GLS could give an earlier ability to predict overall prognosis. In addition, these subgroups transcend the concept of etiology-based phenotyping, which remains difficult due to the inherently heterogeneous population even within ACS-related CS category. Indeed, the correlation between Phenotypes and in-hospital mortality is independent of the CS etiology. Phenotypes could also provide support to SCAI staging system: in our analysis, within each Phenotypes, the SCAI staging further stratified mortality, even if SCAI stage E in Phenotypes II, III and IV was relatively underrepresented; this may be at least partially explained by the fact that SCAI classification is dynamic and patients may progress to higher stages during hospital stay [ 26 ]. Furthermore, within each SCAI stage (C-E), the 5 Phenotypes further stratified mortality, and Phenotypes predicted mortality independently of SCAI stage, even after adjustment for confounding factors. Hence, the Phenotypes seem to be compatible with the SCAI stage at admission and may provide further support for it in more personalized therapy. Clustering methods have been effectively utilized for phenotyping CS patients in different studies: Soussi S et al. individuated four biomarker-driven CS sub-phenotypes [ 33 ], while Zweck et al. employed K-means algorithm to identify three subgroups based on 6 admission laboratory selected variables. The three clusters individuated were labeled as “non-congested,” “cardiorenal,” and “cardiometabolic” subgroups based on their clinical characteristics [ 18 , 19 , 34 ]. Using K-means clustering to sub-phenotype patients with CS could offer some advantages: one of the main strengths of K-means lies in its simplicity and interpretability, making it an accessible tool for exploratory analysis. The algorithm’s computational efficiency is particularly well-suited to datasets of moderate size, such as in this study. Moreover, as an unsupervised method, K-means does not require labeled data, which is particularly advantageous when the goal is to uncover hidden patterns or identify novel phenotypes based on clinical features. Using this method, we found 5 Phenotypes based not only on their clinical variables (as in previous studies), but also on echocardiographic characteristics at presentation which differ from each other, suggesting the individuation of distinct clinical clusters. STUDY LIMITATIONS Major limitations of this study include the limited number of patients and clinical variables, the inability to assess long-term outcomes and the lack of an external validation to other independent CS cohorts, which is important to assess the reproducibility of the analysis. Further larger prospective studies that account for temporal changes in patients’ parameters are needed, to capture the potential transition between Phenotypes during hospitalization. Moreover, the ML method assumes that clusters are spherical and evenly distributed in the feature space, which may not align with the complex, heterogeneous nature of clinical data. Furthermore, the algorithm is sensitive to outliers, which can disproportionately impact the positioning of cluster centroids, and its limited capacity to handle high-dimensional data effectively without preprocessing. Finally, the dependence of speckle-tracking echocardiography on image quality should be considered, reflecting daily practice challenges in cardiac intensive care unit imaging, though sensitivity analysis confirmed the robustness of clustering results even when LV GLS was excluded. CONCLUSIONS We identified 5 CS Phenotypes through a ML approach, combining echocardiographic data with metabolic and hemodynamic variables. These subgroups, characterized by unique mortality risks and features, including varying patterns of cardiac dysfunction and dilatation, offer a framework for tailoring treatment strategies and improving risk stratification beyond SCAI staging in CS patients. Data Availability The corresponding author had full access to all the study data and takes responsibility for its integrity and the data analysis. Sources of Funding no funding related to this work. Disclosures none. Supplemental Material Supplemental Methods Figure S1-S4 Footnotes ↵ ‡ Co-last authors. Nonstandard Abbreviations and Acronyms ACS acute coronary syndrome CS cardiogenic shock EDD end-diastolic diameter EF ejection fraction GLS global longitudinal strain LV left ventricular MAP mean arterial pressure ML machine learning SCAI Society for Cardiovascular Angiography and Interventions S’ TDI wave s’ tissue Doppler imaging REFERENCES 1. ↵ Lüsebrink E , Binzenhöfer L , Adamo M , Lorusso R , Mebazaa A , Morrow DA , Price S , Jentzer JC , Brodie D , Combes A , et al. Cardiogenic shock . Lancet . 2024 Nov 16 ; 404 ( 10466 ): 2006 – 2020 . doi: 10.1016/S0140-6736(24)01818-X . OpenUrl CrossRef PubMed 2. ↵ Chioncel O , Parissis J , Mebazaa A , Thiele H , Desch S , Bauersachs J , Harjola VP , Antohi EL , Arrigo M , Ben Gal T , et al. Epidemiology, pathophysiology and contemporary management of cardiogenic shock - a position statement from the Heart Failure Association of the European Society of Cardiology . Eur J Heart Fail . 2021 Feb ; 23 ( 2 ): 345 . doi: 10.1002/ejhf.2152 . OpenUrl CrossRef 3. ↵ McDonagh TA , Metra M , Adamo M , Gardner RS , Baumbach A , Böhm M , Burri H , Butler J , Čelutkienė J , Chioncel O , et al ; ESC Scientific Document Group. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure . Eur Heart J . 2021 Dec 21 ; 42 ( 48 ): 4901 . doi: 10.1093/eurheartj/ehab670 . OpenUrl CrossRef PubMed 4. ↵ Lawler PR , Mehra MR . Advancing from a “hemodynamic model” to a “mechanistic disease-modifying model” of cardiogenic shock . J Heart Lung Transplant . 2018 Nov ; 37 ( 11 ): 1285 – 1288 . doi: 10.1016/j.healun.2018.07.009 . OpenUrl CrossRef PubMed 5. ↵ Riccardi M , Pagnesi M , Chioncel O , Mebazaa A , Cotter G , Gustafsson F , Tomasoni D , Latronico N , Adamo M , Metra M . Medical therapy of cardiogenic shock: Contemporary use of inotropes and vasopressors . Eur J Heart Fail . 2024 Feb ; 26 ( 2 ): 411 – 431 . doi: 10.1002/ejhf.3162 . OpenUrl CrossRef PubMed 6. ↵ Sinha SS , Rosner CM , Tehrani BN , Maini A , Truesdell AG , Lee SB , Bagchi P , Cameron J , Damluji AA , Desai M , et al. Cardiogenic Shock From Heart Failure Versus Acute Myocardial Infarction: Clinical Characteristics, Hospital Course, and 1-Year Outcomes . Circ Heart Fail . 2022 Jun; 15 ( 6 ): e009279 . doi: 10.1161/CIRCHEARTFAILURE.121.009279 . OpenUrl CrossRef PubMed 7. ↵ Harjola VP , Lassus J , Sionis A , Køber L , Tarvasmäki T , Spinar J , Parissis J , Banaszewski M , Silva-Cardoso J , Carubelli V , et al ; CardShock Study Investigators; GREAT network. Clinical picture and risk prediction of short-term mortality in cardiogenic shock . Eur J Heart Fail . 2015 May ; 17 ( 5 ): 501 – 9 . doi: 10.1002/ejhf.260 . Epub 2015 Mar 28. Erratum in: Eur J Heart Fail. 2015 Sep;17(9):984. doi: 10.1002/ejhf.349 OpenUrl CrossRef PubMed 8. ↵ Kapur NK , Kanwar M , Sinha SS , Thayer KL , Garan AR , Hernandez-Montfort J , Zhang Y , Li B , Baca P , Dieng F , et al. Criteria for Defining Stages of Cardiogenic Shock Severity . J Am Coll Cardiol . 2022 Jul 19 ; 80 ( 3 ): 185 – 198 . doi: 10.1016/j.jacc.2022.04.049 . OpenUrl CrossRef PubMed 9. ↵ Panoulas V , Ilsley C . Rapid Classification and Treatment Algorithm of Cardiogenic Shock Complicating Acute Coronary Syndromes: The SAVE ACS Classification . J Interv Cardiol . 2022 Jan 12 ; 2022 : 9948515 . doi: 10.1155/2022/9948515 . OpenUrl CrossRef PubMed 10. Beer BN , Jentzer JC , Weimann J , Dabboura S , Yan I , Sundermeyer J , Kirchhof P , Blankenberg S , Schrage B , Westermann D . Early risk stratification in patients with cardiogenic shock irrespective of the underlying cause - the Cardiogenic Shock Score . Eur J Heart Fail . 2022 Apr ; 24 ( 4 ): 657 – 667 . doi: 10.1002/ejhf.2449 . OpenUrl CrossRef PubMed 11. Pöss J , Köster J , Fuernau G , Eitel I , de Waha S , Ouarrak T , Lassus J , Harjola VP , Zeymer U , Thiele H , et al. Risk Stratification for Patients in Cardiogenic Shock After Acute Myocardial Infarction . J Am Coll Cardiol . 2017 Apr 18 ; 69 ( 15 ): 1913 – 1920 . OpenUrl FREE Full Text 12. ↵ Jentzer JC , van Diepen S , Barsness GW , Henry TD , Menon V , Rihal CS , Naidu SS , Baran DA . Cardiogenic Shock classification to predict mortality in the cardiac intensive care unit . J Am Coll Cardiol . 2019 ; 74 : 2117 – 2128 . OpenUrl FREE Full Text 13. Naidu SS , Baran DA , Jentzer JC , Hollenberg SM , van Diepen S , Basir MB , Grines CL , Diercks DB , Hall S , Kapur NK , et al. SCAI SHOCK Stage Classification Expert Consensus Update: A Review and Incorporation of Validation Studies: This statement was endorsed by the American College of Cardiology (ACC), American College of Emergency Physicians (ACEP), American Heart Association (AHA), European Society of Cardiology (ESC) Association for Acute Cardiovascular Care (ACVC), International Society for Heart and Lung Transplantation (ISHLT), Society of Critical Care Medicine (SCCM), and Society of Thoracic Surgeons (STS) in December 2021 . J Am Coll Cardiol . 2022 Mar 8 ; 79 ( 9 ): 933 – 946 . doi: 10.1016/j.jacc.2022.01.018 . OpenUrl CrossRef PubMed 14. ↵ Schrage B , Dabboura S , Yan I , Hilal R , Neumann JT , Sörensen NA , Goßling A , Becher PM , Grahn H , Wagner T , Seiffert M , et al. Application of the SCAI classification in a cohort of patients with cardiogenic shock . Catheter Cardiovasc Interv . 2020 Jan 11 . 15. ↵ Jentzer JC , Rayfield C , Soussi S , Berg DD , Kennedy JN , Sinha SS , Baran DA , Brant E , Mebazaa A , Billia F , et al. Machine Learning Approaches for Phenotyping in Cardiogenic Shock and Critical Illness: Part 2 of 2 . JACC Adv . 2022 Oct 28 ; 1 ( 4 ): 100126 . doi: 10.1016/j.jacadv.2022.100126 . OpenUrl CrossRef PubMed 16. ↵ Reddy K , Sinha P , O’Kane CM , Gordon AC , Calfee CS , McAuley DF . Subphenotypes in critical care: translation into clinical practice . Lancet Respir Med . 2020 Jun ; 8 ( 6 ): 631 – 643 . doi: 10.1016/S2213-2600(20)30124-7 . OpenUrl CrossRef PubMed 17. ↵ Raverdy V , Tavaglione F , Chatelain E , Lassailly G , De Vincentis A , Vespasiani-Gentilucci U , Qadri SF , Caiazzo R , Verkindt H , Saponaro C , et al. Data-driven cluster analysis identifies distinct types of metabolic dysfunction-associated steatotic liver disease . Nat Med . 2024 Dec ; 30 ( 12 ): 3624 – 3633 . doi: 10.1038/s41591-024-03283-1 . OpenUrl CrossRef PubMed 18. ↵ Zweck E , Thayer KL , Helgestad OKL , Kanwar M , Ayouty M , Garan AR , Hernandez-Montfort J , Mahr C , Wencker D , Sinha SS , et al. Phenotyping Cardiogenic Shock . J Am Heart Assoc . 2021 Jul 20 ; 10 ( 14 ): e020085 . doi: 10.1161/JAHA.120.020085 . OpenUrl CrossRef PubMed 19. ↵ Jentzer JC , Soussi S , Lawler PR , Kennedy JN , Kashani KB . Validation of cardiogenic shock phenotypes in a mixed cardiac intensive care unit population . Catheter Cardiovasc Interv . 2022 Mar ; 99 ( 4 ): 1006 – 1014 . doi: 10.1002/ccd.30103 . OpenUrl CrossRef PubMed 20. ↵ Pastore MC , Ilardi F , Stefanini A , Mandoli GE , Palermi S , Bandera F , Benfari G , Esposito R , Lisi M , Pasquini A , et al ; Working Group of Echocardiography of the Italian Society of Cardiology. Bedside Ultrasound for Hemodynamic Monitoring in Cardiac Intensive Care Unit . J Clin Med . 2022 Dec 19 ; 11 ( 24 ): 7538 . doi: 10.3390/jcm11247538 . OpenUrl CrossRef PubMed 21. ↵ Berg-Hansen K , Ito S , Oh J , Yang JH , Wiggers H , Jentzer JC . Global Longitudinal Strain is a Predictor of Mortality in Patients with Cardiogenic Shock . Eur Heart J Cardiovasc Imaging . 2024 Dec 7 : jeae316 . doi: 10.1093/ehjci/jeae316 . OpenUrl CrossRef 22. Jentzer JC , Wiley BM , Anavekar NS , Pislaru SV , Mankad SV , Bennett CE , Barsness GW , Hollenberg SM , Holmes DR Jr , Oh JK . Noninvasive Hemodynamic Assessment of Shock Severity and Mortality Risk Prediction in the Cardiac Intensive Care Unit . JACC Cardiovasc Imaging . 2021 Feb ; 14 ( 2 ): 321 – 332 . doi: 10.1016/j.jcmg.2020.05.038 . OpenUrl CrossRef PubMed 23. Singam NSV , Tabi M , Wiley B , Anavekar N , Jentzer J . Echocardiographic findings in cardiogenic shock due to acute myocardial infarction versus heart failure . Int J Cardiol . 2023 Aug 1 ; 384 : 38 – 47 . doi: 10.1016/j.ijcard.2023.04.041 . OpenUrl CrossRef PubMed 24. ↵ B.C Picarra , A.R Santos , J.A Pais , M Carrington , D Bras , K Congo , A.R Rocha , D Neves , R Guerreiro , J Aguiar , Portuguese Registry on Acute Coronary Syndromes, Cardiogenic shock without severe left ventricular dysfunction after acute myocardial infarction: population characterization and impact in prognosis , European Heart Journal , Volume 41 , Issue Supplement_ 2 , November 2020 , ehaa946.1792 , doi: 10.1093/ehjci/ehaa946.1792 OpenUrl CrossRef 25. Sundermeyer J , Kellner C , Beer BN , Besch L , Dettling A , Bertoldi LF , Blankenberg S , Dauw J , Dindane Z , Eckner D , et al. Association between left ventricular ejection fraction, mortality and use of mechanical circulatory support in patients with non-ischaemic cardiogenic shock . Clin Res Cardiol . 2024 Apr ; 113 ( 4 ): 570 – 580 . doi: 10.1007/s00392-023-02332-y . OpenUrl CrossRef PubMed 26. ↵ Morici N , Frea S , Bertaina M , Sacco A , Corrada E , Dini CS , Briani M , Tedeschi M , Saia F , Colombo C , et al. SCAI stage reclassification at 24 h predicts outcome of cardiogenic shock: Insights from the Altshock-2 registry . Catheter Cardiovasc Interv . 2023 Jan ; 101 ( 1 ): 22 – 32 . doi: 10.1002/ccd.30484 . OpenUrl CrossRef PubMed 27. ↵ Lang RM , Badano LP , Mor-Avi V , Afilalo J , Armstrong A , Ernande L , Flachskampf FA , Foster E , Goldstein SA , Kuznetsova T , et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging . J Am Soc Echocardiogr . 2015 Jan ; 28 ( 1 ): 1 – 39 .e14. OpenUrl CrossRef PubMed 28. ↵ Frydland M , Møller JE , Wiberg S , Lindholm MG , Hansen R , Henriques JPS , Møller-Helgestad OK , Bang LE , Frikke-Schmidt R , Goetze JP , et al. Lactate is a Prognostic Factor in Patients Admitted With Suspected ST-Elevation Myocardial Infarction . Shock . 2019 Mar ; 51 ( 3 ): 321 – 327 . doi: 10.1097/SHK.0000000000001191 . OpenUrl CrossRef PubMed 29. Verhaeghe M , Hachimi-Idrissi S . Blood lactate and lactate kinetics as treatment and prognosis markers for tissue hypoperfusion . Acta Clin Belg . 2020 Feb ; 75 ( 1 ): 1 – 8 . OpenUrl PubMed 30. Hayıroğlu Mİ , Keskin M , Uzun AO , Yıldırım Dİ , Kaya A , Çinier G , Bozbeyoğlu E , Yıldırımtürk Ö , Kozan Ö , Pehlivanoğlu S . Predictors of In-Hospital Mortality in Patients With ST-Segment Elevation Myocardial Infarction Complicated With Cardiogenic Shock . Heart Lung Circ . 2019 Feb ; 28 ( 2 ): 237 – 244 . OpenUrl PubMed 31. Galusko V , Wenzl FA , Vandenbriele C , Panoulas V , Lüscher TF , Gorog DA . Current and novel biomarkers in cardiogenic shock . Eur J Heart Fail . 2025 Jan 17 . doi: 10.1002/ejhf.3531 . OpenUrl CrossRef 32. ↵ Bocchino PP , Frea S , Sacco A , Bertaina M , Pappalardo F , Tavazzi G , Morici N , Angelini F , Garatti L , Briani M , et al. Organ perfusion pressure predicts outcomes in cardiogenic shock patients . Eur J Heart Fail . 2025 Feb 16 . doi: 10.1002/ejhf.3627 . OpenUrl CrossRef 33. ↵ Soussi S , Tarvasmäki T , Kimmoun A , Ahmadiankalati M , Azibani F , Dos Santos CC , Duarte K , Gayat E , Jentzer JC , Harjola VP , et al. Identifying biomarker-driven subphenotypes of cardiogenic shock: analysis of prospective cohorts and randomized controlled trials . EClinicalMedicine . 2024 Dec 18 ; 79 : 103013 . doi: 10.1016/j.eclinm.2024.103013 . OpenUrl CrossRef PubMed 34. ↵ Zweck E , Kanwar M , Li S , Sinha SS , Garan AR , Hernandez-Montfort J , Zhang Y , Li B , Baca P , Dieng F , et al. Clinical Course of Patients in Cardiogenic Shock Stratified by Phenotype . JACC Heart Fail . 2023 Oct ; 11 ( 10 ): 1304 – 1315 . doi: 10.1016/j.jchf.2023.05.007 . OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted May 01, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data Nicolò Ghionzoli , Andrea Stefanini , Geza Halasz , Carlotta Sorini Dini , Carlotta Sciaccaluga , Francesca Maria Righini , Valerio Guarrasi , Arianna Francesconi , Paolo Soda , Maria Concetta Pastore , Hatem Soliman Aboumarie , Vito Piazza , Marco Marini , Domenico Gabrielli , Matteo Cameli , Serafina Valente medRxiv 2025.04.28.25326615; doi: https://doi.org/10.1101/2025.04.28.25326615 Share This Article: Copy Citation Tools Identifying Cardiogenic Shock Sub-Phenotypes with Machine Learning: A Multicenter Study Combining Clinical and Echocardiographic Data Nicolò Ghionzoli , Andrea Stefanini , Geza Halasz , Carlotta Sorini Dini , Carlotta Sciaccaluga , Francesca Maria Righini , Valerio Guarrasi , Arianna Francesconi , Paolo Soda , Maria Concetta Pastore , Hatem Soliman Aboumarie , Vito Piazza , Marco Marini , Domenico Gabrielli , Matteo Cameli , Serafina Valente medRxiv 2025.04.28.25326615; doi: https://doi.org/10.1101/2025.04.28.25326615 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cardiovascular Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4425) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15221) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6588) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1368) Health Systems and Quality Improvement (1612) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6588) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5440) Public and Global Health (9219) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (710) Sports Medicine (529) Surgery (710) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffab74ad9b606db',t:'MTc3OTQ0MTMyMg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();