Exploring relationships between in-hospital mortality and hospital case volume using random forest: results of a cohort study based on a nationwide sample of German hospitals, 2016–2018

Abstract Background Relationships between in-hospital mortality and case volume were investigated for various patient groups in many empirical studies with mixed results. Typically, those studies relied on (semi-)parametric statistical models like logistic regression. Those models impose strong assumptions on the functional form of the relationship between outcome and case volume. The aim of this study was to determine associations between in-hospital mortality and hospital case volume using random forest as a flexible, nonparametric machine learning method. Methods We analyzed a sample of 753,895 hospital cases with stroke, myocardial infarction, ventilation > 24 h, COPD, pneumonia, and colorectal cancer undergoing colorectal resection treated in 233 German hospitals over the period 2016–2018. We derived partial dependence functions from random forest estimates capturing the relationship between the patient-specific probability of in-hospital death and hospital case volume for each of the six considered patient groups. Results Across all patient groups, the smallest hospital volumes were consistently related to the highest predicted probabilities of in-hospital death. We found strong relationships between in-hospital mortality and hospital case volume for hospitals treating a (very) small number of cases. Slightly higher case volumes were associated with substantially lower mortality. The estimated relationships between in-hospital mortality and case volume were nonlinear and nonmonotonic. Conclusion Our analysis revealed strong relationships between in-hospital mortality and hospital case volume in hospitals treating a small number of cases. The nonlinearity and nonmonotonicity of the estimated relationships indicate that studies applying conventional statistical approaches like logistic regression should consider these relationships adequately.

Download Full-text

Higher programmatic volume in paediatric heart surgery is associated with better early outcomes

Cardiology in the Young ◽

10.1017/s1047951115002073 ◽

2015 ◽

Vol 25 (8) ◽

pp. 1572-1578 ◽

Cited By ~ 6

Author(s):

Andrzej Kansy ◽

Tjark Ebels ◽

Christian Schreiber ◽

Jeffrey P. Jacobs ◽

Zdzislaw Tobota ◽

...

Keyword(s):

Risk Factors ◽

Logistic Regression ◽

Hospital Mortality ◽

Heart Surgery ◽

Major Complication ◽

High Volume ◽

Patient Specific ◽

Hospital Death ◽

Data Set ◽

Multivariable Logistic Regression

AbstractObjectivePrevious analyses have suggested an association between centre volume and in-hospital mortality, post-operative complications, and mortality in those patients who suffer from a complication. We sought to determine the nature of this association using a multicentre cohort.MethodsAll the patients, aged 18 years or younger, undergoing heart surgery at centres participating in the European Congenital Heart Surgeons Database (2003–2013) were included. Programmes were grouped as follows: small <150; medium 150–250; large 251–349; very large >350. Multivariable logistic regression was used to identify the differences between groups with the adjusted in-hospital mortality, onset of any and/or major complication, and in-hospital mortality in those patients with any and/or major complication. The outcomes were adjusted for patient specific risk factors and surgical risk factors.ResultsThe data set consisted of 119,345 procedures performed in 99 centres. Overall, in-hospital mortality was 4.63%; complications occurred in 23.4% of the patients. In-hospital mortality in patients with complications was 13.82%. Multivariable logistic regression showed that the risk of in-hospital death was higher in low- and medium-volume centres (p<0.001). The rate of the occurrence of any post-operative complication in small, medium, and large programmes was lower compared with very large centres (p<0.001). Low- and medium-volume centres were associated with significantly higher mortality in patients with any complication (p<0.001).ConclusionsOur analysis showed that the risk of in-hospital mortality was lower in higher-volume centres. Although the risk of complications is higher in high-volume centres, the mortality associated with complications that occurred in these centres was lower.

Download Full-text

Risk factors and clinical outcomes associated with blood culture contamination

Infection Control and Hospital Epidemiology ◽

10.1017/ice.2021.111 ◽

2021 ◽

pp. 1-7

Author(s):

Justin M. Klucher ◽

Kevin Davis ◽

Mrinmayee Lakkad ◽

Jacob T. Painter ◽

Ryan K. Dare

Keyword(s):

Risk Factors ◽

Logistic Regression ◽

Blood Culture ◽

Hospital Mortality ◽

Clinical Outcomes ◽

Tertiary Care ◽

Blood Cultures ◽

Hospital Charges ◽

Patient Specific ◽

Contaminated Blood

Abstract Objective: To determine patient-specific risk factors and clinical outcomes associated with contaminated blood cultures. Design: A single-center, retrospective case-control risk factor and clinical outcome analysis performed on inpatients with blood cultures collected in the emergency department, 2014–2018. Patients with contaminated blood cultures (cases) were compared to patients with negative blood cultures (controls). Setting: A 509-bed tertiary-care university hospital. Methods: Risk factors independently associated with blood-culture contamination were determined using multivariable logistic regression. The impacts of contamination on clinical outcomes were assessed using linear regression, logistic regression, and generalized linear model with γ log link. Results: Of 13,782 blood cultures, 1,504 (10.9%) true positives were excluded, leaving 1,012 (7.3%) cases and 11,266 (81.7%) controls. The following factors were independently associated with blood-culture contamination: increasing age (adjusted odds ratio [aOR], 1.01; 95% confidence interval [CI], 1.01–1.01), black race (aOR, 1.32; 95% CI, 1.15–1.51), increased body mass index (BMI; aOR, 1.01; 95% CI, 1.00–1.02), chronic obstructive pulmonary disease (aOR, 1.16; 95% CI, 1.02–1.33), paralysis (aOR 1.64; 95% CI, 1.26–2.14) and sepsis plus shock (aOR, 1.26; 95% CI, 1.07–1.49). After controlling for age, race, BMI, and sepsis, blood-culture contamination increased length of stay (LOS; β = 1.24 ± 0.24; P < .0001), length of antibiotic treatment (LOT; β = 1.01 ± 0.20; P < .001), hospital charges (β = 0.22 ± 0.03; P < .0001), acute kidney injury (AKI; aOR, 1.60; 95% CI, 1.40–1.83), echocardiogram orders (aOR, 1.51; 95% CI, 1.30–1.75) and in-hospital mortality (aOR, 1.69; 95% CI, 1.31–2.16). Conclusions: These unique risk factors identify high-risk individuals for blood-culture contamination. After controlling for confounders, contamination significantly increased LOS, LOT, hospital charges, AKI, echocardiograms, and in-hospital mortality.

Download Full-text

Preoperative predictions of in-hospital mortality using electronic medical record data

10.1101/329813 ◽

2018 ◽

Cited By ~ 1

Author(s):

Brian Hill ◽

Robert Brown ◽

Eilon Gabel ◽

Christine Lee ◽

Maxime Cannesson ◽

...

Keyword(s):

Logistic Regression ◽

Random Forest ◽

Hospital Mortality ◽

Risk Score ◽

Surgical Mortality ◽

Physical Status ◽

Preoperative Risk ◽

Physical Status Classification ◽

Postoperative Risk ◽

Status Classification

AbstractBackgroundPredicting preoperative in-hospital mortality using readily-available electronic medical record (EMR) data can aid clinicians in accurately and rapidly determining surgical risk. While previous work has shown that the American Society of Anesthesiologists (ASA) Physical Status Classification is a useful, though subjective, feature for predicting surgical outcomes, obtaining this classification requires a clinician to review the patient’s medical records. Our goal here is to create an improved risk score using electronic medical records and demonstrate its utility in predicting in-hospital mortality without requiring clinician-derived ASA scores.MethodsData from 49,513 surgical patients were used to train logistic regression, random forest, and gradient boosted tree classifiers for predicting in-hospital mortality. The features used are readily available before surgery from EMR databases. A gradient boosted tree regression model was trained to impute the ASA Physical Status Classification, and this new, imputed score was included as an additional feature to preoperatively predict in-hospital post-surgical mortality. The preoperative risk prediction was then used as an input feature to a deep neural network (DNN), along with intraoperative features, to predict postoperative in-hospital mortality risk. Performance was measured using the area under the receiver operating characteristic (ROC) curve (AUC).ResultsWe found that the random forest classifier (AUC 0.921, 95%CI 0.908-0.934) outperforms logistic regression (AUC 0.871, 95%CI 0.841-0.900) and gradient boosted trees (AUC 0.897, 95%CI 0.881-0.912) in predicting in-hospital post-surgical mortality. Using logistic regression, the ASA Physical Status Classification score alone had an AUC of 0.865 (95%CI 0.848-0.882). Adding preoperative features to the ASA Physical Status Classification improved the random forest AUC to 0.929 (95%CI 0.915-0.943). Using only automatically obtained preoperative features with no clinician intervention, we found that the random forest model achieved an AUC of 0.921 (95%CI 0.908-0.934). Integrating the preoperative risk prediction into the DNN for postoperative risk prediction results in an AUC of 0.924 (95%CI 0.905-0.941), and with both a preoperative and postoperative risk score for each patient, we were able to show that the mortality risk changes over time.ConclusionsFeatures easily extracted from EMR data can be used to preoperatively predict the risk of in-hospital post-surgical mortality in a fully automated fashion, with accuracy comparable to models trained on features that require clinical expertise. This preoperative risk score can then be compared to the postoperative risk score to show that the risk changes, and therefore should be monitored longitudinally over time.Author summaryRapid, preoperative identification of those patients at highest risk for medical complications is necessary to ensure that limited infrastructure and human resources are directed towards those most likely to benefit. Existing risk scores either lack specificity at the patient level, or utilize the American Society of Anesthesiologists (ASA) physical status classification, which requires a clinician to review the chart. In this manuscript we report on using machine-learning algorithms, specifically random forest, to create a fully automated score that predicts preoperative in-hospital mortality based solely on structured data available at the time of surgery. This score has a higher AUC than both the ASA physical status score and the Charlson comorbidity score. Additionally, we integrate this score with a previously published postoperative score to demonstrate the extent to which patient risk changes during the perioperative period.

Download Full-text

Machine Learning-based in-hospital Mortality Prediction Models for Patients With Acute Coronary Syndrome

10.21203/rs.3.rs-134944/v1 ◽

2020 ◽

Author(s):

Jun Ke ◽

Yiwei Chen ◽

Xiaoping Wang ◽

Zhiyong Wu ◽

qiongyao Zhang ◽

...

Keyword(s):

Machine Learning ◽

Logistic Regression ◽

Random Forest ◽

Hospital Mortality ◽

Operating Characteristic ◽

Prediction Models ◽

Characteristic Curve ◽

Multivariate Logistic Regression Analysis ◽

Hdl Cholesterol ◽

Coronary Syndrome

Abstract BackgroundThe purpose of this study is to identify the risk factors of in-hospital mortality in patients with acute coronary syndrome (ACS) and to evaluate the performance of traditional regression and machine learning prediction models.MethodsThe data of ACS patients who entered the emergency department of Fujian Provincial Hospital from January 1, 2017 to March 31, 2020 for chest pain were retrospectively collected. The study used univariate and multivariate logistic regression analysis to identify risk factors for in-hospital mortality of ACS patients. The traditional regression and machine learning algorithms were used to develop predictive models, and the sensitivity, specificity, and receiver operating characteristic curve were used to evaluate the performance of each model.ResultsA total of 7810 ACS patients were included in the study, and the in-hospital mortality rate was 1.75%. Multivariate logistic regression analysis found that age and levels of D-dimer, cardiac troponin I, N-terminal pro-B-type natriuretic peptide (NT-proBNP), lactate dehydrogenase (LDH), high-density lipoprotein (HDL) cholesterol, and calcium channel blockers were independent predictors of in-hospital mortality. The study found that the area under the receiver operating characteristic curve of the models developed by logistic regression, gradient boosting decision tree (GBDT), random forest, and support vector machine (SVM) for predicting the risk of in-hospital mortality were 0.963, 0.960, 0.963, and 0.959, respectively. Feature importance evaluation found that NT-proBNP, LDH, and HDL cholesterol were top three variables that contribute the most to the prediction performance of the GBDT model and random forest model.ConclusionsThe predictive model developed using logistic regression, GBDT, random forest, and SVM algorithms can be used to predict the risk of in-hospital death of ACS patients. Based on our findings, we recommend that clinicians focus on monitoring the changes of NT-proBNP, LDH, and HDL cholesterol, as this may improve the clinical outcomes of ACS patients.

Download Full-text

Revisiting the Relationship Between Hospital Case Volume and Outcomes in Abdominally Based Free Flap Breast Reconstruction

Annals of Plastic Surgery ◽

10.1097/sap.0000000000002308 ◽

2020 ◽

Vol 85 (4) ◽

pp. 397-401

Author(s):

Anmol Chattha ◽

Austin D. Chen ◽

Justin Muste ◽

Justin B. Cohen ◽

Bernard T. Lee ◽

...

Keyword(s):

Breast Reconstruction ◽

Free Flap ◽

Case Volume ◽

Hospital Case ◽

The Relationship

Download Full-text

Real-world clopidogrel utilization in acute coronary syndromes: patients selection and outcomes in a single-center experience

Therapeutic Advances in Cardiovascular Disease ◽

10.1177/1753944717738911 ◽

2017 ◽

Vol 11 (12) ◽

pp. 323-331 ◽

Cited By ~ 9

Author(s):

Diego Castini ◽

Simone Persampieri ◽

Sara Cazzaniga ◽

Giulia Ferrante ◽

Marco Centola ◽

...

Keyword(s):

Logistic Regression ◽

Regression Analysis ◽

Adverse Events ◽

Hospital Mortality ◽

Logistic Regression Analysis ◽

Left Ventricular ◽

Hospital Death ◽

Multivariable Logistic Regression Analysis ◽

Killip Class ◽

Female Sex

Background: With this study, we sought to identify patient characteristics associated with clopidogrel prescription and its relationship with in-hospital adverse events in an unselected cohort of ACSs patients. Materials and Methods: We studied all consecutive patients admitted at our institution for ACSs from 2012 to 2014. Patients were divided into two groups based on clopidogrel or novel P2Y12 inhibitors (prasugrel or ticagrelor) prescription and the relationship between clopidogrel use and patient clinical characteristics and in-hospital adverse events was evaluated using logistic regression analysis. Results: The population median age was 68 years (57–77 year) and clopidogrel was prescribed in 230 patients (46%). Patients characteristics associated with clopidogrel prescription were older age, female sex, non-ST-elevation ACS diagnosis, the presence of diabetes mellitus and anemia, worse renal and left ventricular functions and a higher Killip class. Patients on clopidogrel demonstrated a significantly higher incidence of in-hospital mortality (4.8%) than prasugrel and ticagrelor-treated patients (0.4%), while a nonstatistically significant trend emerged considering bleeding events. However, on multivariable logistic regression analysis female sex, the presence of anemia and Killip class were the only variables independently associated with in-hospital death. Conclusion: Patients treated with clopidogrel showed a higher in-hospital mortality. However, clinical variables associated with its use identify a population at high risk for adverse events and this seems to play a major role for the higher in-hospital mortality observed in clopidogrel-treated patients.

Download Full-text

Gender and Regional Differences in In-Hospital Mortality in Patients with Sickle Cell Anemia (SCA) in the United States: Results from the Nationwide Inpatient Sample (NIS) 1993-2003.

Blood ◽

10.1182/blood.v108.11.461.461 ◽

2006 ◽

Vol 108 (11) ◽

pp. 461-461

Author(s):

Carlton Haywood ◽

Sophie Lanzkron

Keyword(s):

Logistic Regression ◽

Hospital Mortality ◽

Hospital Death ◽

Age At Death ◽

Patient Age ◽

Time Period ◽

Patient Âgé ◽

The Mean ◽

Over Time ◽

Age Of Death

Abstract Background: The purpose of this study was to use the NIS to describe hospital utilization and in-hospital mortality among adults with SCA in the US between 1993-2003. Methods: The NIS is designed to approximate a 20% stratified sample of U.S. community hospitals. We restricted our analyses to discharge records with ICD-9-CM diagnosis codes 28261 or 28262 (SCA without/with crisis), and where the age was listed as 18 or older. Analyses were conducted using tests of linear combinations of coefficients, χ2, and linear and logistic regression. Results: There were an estimated 705,080 hospitalizations over the time period (mean of 64,098 hospitalizations/year). 54% of all hospitalizations were for females. 50% of the hospitalizations were expected to be paid for by Medicaid. The mean patient age over the time period was 31.3 yrs. The mean patient age increased from 30.3 in 1993 to 32.1 in 2003 (p < 0.001). Mean age over time increased even after adjusting for the gender makeup and hospital region (β=0.162, p < 0.001). There were no gender differences in the median age (30) of patients. Mean length of stay (LOS) was 6.5 days for the time period. LOS decreased from 7.5 days in 1993 to 6.4 days in 2003 (p=0.001). Adult women experienced longer LOS than adult men (6.8 days vs. 6.3 days, p <0.001). This difference remained significant even after controlling for age, time, insurance status, and hospital region (β = 0.49, p<0.001). Mean charges/discharge increased from $16,799 in 1993 to $22,281 in 2003, even after adjusting for inflation (p < 0.001). There were an estimated 4497 in-hospital deaths during the time period (0.64% of hospitalizations). The median age at death was 38. The median age at death increased from 35 in 1993 to 42 in 2003 (p = 0.0061). This was due to an increase in age of death (39) for women (p=0.0052). In men the median age of death (37) did not change over time(p=0.4352). In bivariate analyses of median age at death, women were older than men (39 vs. 37 p=0.0056). A simple logistic regression of deaths over time found no significant trends in the odds of an in-hospital death over the time period. In a multivariate model of death over time patients in the South and the West experienced higher odds of an in-hospital death than patients in the Northeast and Midwest. Conclusions: Our analysis shows that women with SCA have longer in-hospital LOS than men, and are older in age at death than men. While the median age at death among persons hospitalized with SCA has been increasing since 1993, this increase is seen exclusively in women. There has been no change in longevity in men hospitalized with SCA over the time period studied.

Download Full-text

Frailty, delirium and hospital mortality of older adults admitted to intensive care, a mediation analysis: the Delirium (Deli) in ICU study.

10.21203/rs.3.rs-52423/v1 ◽

2020 ◽

Author(s):

David Sanchez ◽

Kathleen Brennan ◽

Masar El Safye ◽

Sharon-Ann Shunker ◽

Tony Bogdanoski ◽

...

Keyword(s):

Older Adults ◽

Intensive Care ◽

Hospital Mortality ◽

Hospital Stay ◽

Mediation Analysis ◽

Baseline Period ◽

Acute Episode ◽

Hospital Death ◽

Frail Patients ◽

The Relationship

Abstract Background As the population ages clinical frailty among older adults admitted to intensive care has been proposed as an important determinant of patient outcomes. Among this group of frail patients an acute episode of delirium is also common, and both frailty and delirium increase the risk of mortality. However, the complex relationship between frailty, delirium and mortality has not been extensively explored in the intensive care setting. Therefore, the aim of this study was to explore the relationship between clinical frailty, acute delirium and hospital mortality of older adults admitted to intensive care. Methods This study is part of a Delirium in Intensive Care (Deli) study that is being conducted across the South Western Sydney Local Health District, between May 2019 and April 2020. During the initial 6-month baseline period, clinical frailty status on admission to ICU, among adults aged 50-years or more, acute episodes of delirium, and the outcomes of ICU and hospital stay will be described. Mediation analysis was used to assess the relationship between frailty, delirium and risk of hospital death. Results During the 6-month baseline period 997 patients, aged 50-years or more, were included in this study. The average age was 71-years (IQR, 63–79), 55% were male (n = 537). Among these patients 39.2% (95% CI 36.1–42.3%, n = 396) had a Clinical Frailty Score (CFS) of 5 or more, and 13.0% (n = 127) had at least one acute episode of delirium. Frail patients were at greater risk of an episode of delirium (17% versus 10%, adjusted Rate Ratio (adjRR) = 1.61, 95% Confidence Interval (CI) 1.14–2.28, p = 0.007), had a longer hospital stay (2.6 days, 95% CI 1–7 days, p = 0.009), and higher risk of hospital mortality (19% versus 7%, adjRR = 2.43, 95% CI 1.68–3.57, p < 0.001), when compared to non-frail patients. Patients who were frail and experienced an acute episode of delirium in the ICU had a 35% rate of hospital mortality, versus 10% among non-frail patients who also experienced delirium in the ICU (p = 0.034, for interaction between frailty, delirium and hospital mortality). The proportion of the effect of frailty and risk of hospital mortality mediated by an acute episode of delirium in the ICU was estimated to 9.4% (95% CI 2–24%). Conclusion This study has been able to show that clinical frailty on admission increases the risk of delirium by approximately 60%, and both increase the risk of hospital mortality. One in three frail patients who experienced an acute episode of delirium during their stay in the ICU did not survive to hospital discharge. These results suggest the importance of recognising clinical frailty in the ICU setting, not just to improve the prediction of outcomes from critical illness, but to identify patients at the greatest risk of adverse events such as delirium, and institute measures to reduce risk, and, importantly to discuss these issues in an open and empathetic way with the patient and their families.

Download Full-text

Abstract TMP83: Association Between Hospital Procedural Volumes and Clinical Outcomes For Patients With Non-Traumatic Subarachnoid Hemorrhage

Stroke ◽

10.1161/str.51.suppl_1.tmp83 ◽

2020 ◽

Vol 51 (Suppl_1) ◽

Author(s):

Dana Leifer ◽

Gregg Fonarow ◽

Anne Hellkamp ◽

David Baker ◽

Brian Hoh ◽

...

Keyword(s):

Logistic Regression ◽

Subarachnoid Hemorrhage ◽

Hospital Mortality ◽

Clinical Outcomes ◽

Sensitivity And Specificity ◽

Case Volume ◽

Logistic Regression Models ◽

Improved Outcomes ◽

Volume Outcome ◽

Traumatic Subarachnoid Hemorrhage

Introduction: Previous studies of patients with non-traumatic subarachnoid hemorrhage (SAH) suggest better outcomes at hospitals with higher case and procedural volumes, but the shape of the volume-outcome curve has not been defined. We sought to establish minimum volume criteria for SAH and aneurysm obliteration procedures that could be used for comprehensive stroke centers (CSC) certification. Methods: The 8,512 SAH discharges in the National Inpatient Sample (NIS) from 2010-11 were analyzed. Logistic regression models were used to evaluate the association between clinical outcomes (in-hospital mortality and the NIS-SAH Outcome Measure (NIS-SOM)) and 3 measures of hospital annual case volume (ACV) (nontraumatic SAH discharges, coiling, and clipping procedures). Sensitivity and specificity analyses for the association of desirable clinical outcomes with different volume thresholds were performed. Results: 28.7% of cases underwent clipping and 20.1% underwent coiling with rates of 21.2%for in-hospital mortality and 38.6% for poor outcome on the NIS-SOM. The mean (range) of SAH ACV, coiling ACV, and clipping ACV were 30.9 (1-195), 8.7 (0-94), and 6.1 (0-69). Logistic regression demonstrated improved outcomes with increasing ACVs of SAH discharges and procedures for aneurysm obliteration, with attenuation of the benefit beyond 35 SAH cases/yr. Sensitivity and specificity analyses with different ACV thresholds confirmed the results. Analysis of previously proposed ACV thresholds, including those used as minimum standards for CSC certification, showed that hospitals with more than 35 SAH cases annually had better outcomes compared to hospitals with fewer cases, but some hospitals below this threshold had similar outcomes to those with more cases. The adjusted odds ratio favoring better outcomes with SAH ACV ≥ 35 compared to SAH 20 to 34 was 0.82 for the NIS-SOM (p=0.0054) and 0.80 (p=0.0055) for in-hospital mortality. Conclusions: Outcomes for SAH patients improve with increasing hospital case volumes and procedure volumes, with consistently better outcomes for hospitals with more than 35 SAH cases per year.

Download Full-text

Venoarterial extracorporeal membrane oxygenation in adult patients: predictors of mortality

Perfusion ◽

10.1177/0267659109348725 ◽

2009 ◽

Vol 24 (4) ◽

pp. 225-230 ◽

Cited By ~ 10

Author(s):

JingwenLi ◽

Cun Long ◽

Song Lou ◽

Feilong Hei ◽

Kun Yu ◽

...

Keyword(s):

Logistic Regression ◽

Extracorporeal Membrane Oxygenation ◽

Hospital Mortality ◽

Adult Patients ◽

Hospital Death ◽

Predictive Values ◽

Predictors Of Mortality ◽

Membrane Oxygenation ◽

Venoarterial Extracorporeal Membrane Oxygenation ◽

Clinical Records

Background: Extracorporeal membrane oxygenation is a cardiopulmonary supportive therapy. In this study, we reviewed our experience with extracorporeal membrane oxygenation support and tried to identify measurable values which might predict in-hospital mortality. Methods: From January 2004 through December 2008, 50 of 21,298 adult patients received venoarterial extracorporeal membrane oxygenation. We retrospectively analyzed clinical records of these 50 consecutive patients. Details of demographics, preoperative measurements, clinical characteristics at the time of extracorporeal membrane oxygenation implantation, extracorporeal membrane oxygenation-related complications and in-hospital mortality were collected. Logistic regression analyses were performed to investigate predictors of mortality. A p-value ≤ 0.05 was accepted as significant. Results: Thirty-eight patients were weaned from extracorporeal membrane oxygenation and 33 patients survived to discharge. The overall survival rate was 66%. In a multiple logistic regression analysis, blood lactate level before initiation of extracorporeal membrane oxygenation was a risk factor associated with in-hospital mortality (OR 1.27 95% CI 1.042-1.542). To evaluate the utility of the lactate in predicting mortality, a conventional receiver operating characteristic curve was produced. Sensitivity and specificity were optimal at a cut-off point of 12.6mmol/L, with an area under the curve of 0.752. The positive and negative predictive values were 73.3% and 83.9%, respectively. Conclusions: Extracorporeal membrane oxygenation is a justifiable alternative treatment for postoperative refractory cardiac and pulmonary dysfunction which could rescue more than sixty percent of otherwise fatal patients. Patients with pre-extracorporeal membrane oxygenation lactate levels above 12.6mmol/L are at higher risks for in-hospital death. Evidence-based therapy for this group of high risk patients is needed.

Download Full-text