Effectiveness of SAPS III to predict hospital mortality for post-cardiac arrest patients

Prospective comparison of prognostic scores for prediction of outcome after out-of-hospital cardiac arrest: results of the AfterROSC1 multicentric study

BACKGROUND

Out-of-hospital cardiac arrest (OHCA) is a heterogeneous entity with multiple origins and prognoses. An early, reliable assessment of the prognosis is useful to adapt therapeutic strategy, tailor intensity of care, and inform relatives. We aimed primarily to undertake a prospective multicentric study to evaluate predictive performance of the Cardiac Arrest Prognosis (CAHP) Score as compare to historical dataset systematically collected after OHCA (Utstein style criteria). Our secondary aim was to evaluate other dedicated scores for predicting outcome after OHCA and to compare them to Utstein style criteria.

METHODS

We prospectively collected data from 24 French and Belgium Intensive Care Units (ICUs) between August 2020 and June 2022. All cases of non-traumatic OHCA (cardiac and non-cardiac causes) patients with stable return of spontaneous circulation (ROSC) and comatose at ICU admission (defined by Glasgow coma score ≤ 8) on ICU admission were included. The primary outcome was the modified Rankin scale (mRS) at day 90 after cardiac arrest, assessed by phone interviews. A wide range of developed scores (CAHP, OHCA, CREST, C-Graph, TTM, CAST, NULL-PLEASE, and MIRACLE2) were included, and their accuracies in predicting poor outcome at 90 days after OHCA (defined as mRS ≥ 4) were determined using the area under the receiving operating characteristic curve (AUROC) and the calibration belt.

RESULTS

During the study period, 907 patients were screened, and 658 were included in the study. Patients were predominantly male (72%), with a mean age of 61 ± 15, most having collapsed from a supposed cardiac cause (64%). The mortality rate at day 90 was 63% and unfavorable neurological outcomes were observed in 66%. The performance (AUROC) of Utstein criteria for poor outcome prediction was moderate at 0.79 [0.76-0.83], whereas AUROCs from other scores varied from 0.79 [0.75-0.83] to 0.88 [0.86-0.91]. For each score, the proportion of patients for whom individual values could not be calculated varied from 1.4% to 17.4%.

CONCLUSIONS

In patients admitted to ICUs after a successfully resuscitated OHCA, most of the scores available for the evaluation of the subsequent prognosis are more efficient than the usual Utstein criteria but calibration is unacceptable for some of them. Our results show that some scores (CAHP, sCAHP, mCAHP, OHCA, rCAST) have superior performance, and that their ease and speed of determination should encourage their use. Trial registration https://clinicaltrials.gov/ct2/show/NCT04167891.

Prediction model of in-hospital mortality in intensive care unit patients with cardiac arrest: a retrospective analysis of MIMIC -IV database based on machine learning

BACKGROUND

Both in-hospital cardiac arrest (IHCA) and out-of-hospital cardiac arrest (OHCA) have higher incidence and lower survival rates. Predictors of in-hospital mortality for intensive care unit (ICU) admitted cardiac arrest (CA) patients remain unclear.

METHODS

The Medical Information Mart for Intensive Care IV (MIMIC-IV) database was used to perform a retrospective study. Patients meeting the inclusion criteria were identified from the MIMIC-IV database and randomly divided into training set (n = 1206, 70%) and validation set (n = 516, 30%). Candidate predictors consisted of the demographics, comorbidity, vital signs, laboratory test results, scoring systems, and treatment information on the first day of ICU admission. Independent risk factors for in-hospital mortality were screened using the least absolute shrinkage and selection operator (LASSO) regression model and the extreme gradient boosting (XGBoost) in the training set. Multivariate logistic regression analysis was used to build prediction models in training set, and then validated in validation set. Discrimination, calibration and clinical utility of these models were compared using the area under the curve (AUC) of the receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA). After pairwise comparison, the best performing model was chosen to build a nomogram.

RESULTS

Among the 1722 patients, in-hospital mortality was 53.95%. In both sets, the LASSO, XGBoost,the logistic regression(LR) model and the National Early Warning Score 2 (NEWS 2) models showed acceptable discrimination. In pairwise comparison, the prediction effectiveness was higher with the LASSO,XGBoost and LR models than the NEWS 2 model (p < 0.001). The LASSO,XGBoost and LR models also showed good calibration. The LASSO model was chosen as our final model for its higher net benefit and wider threshold range. And the LASSO model was presented as the nomogram.

CONCLUSIONS

The LASSO model enabled good prediction of in-hospital mortality in ICU admission CA patients, which may be widely used in clinical decision-making.

External validation of the PROLOGUE score to predict neurological outcome in adult patients after cardiac arrest: a prospective cohort study

BACKGROUND

The PROLOGUE score (PROgnostication using LOGistic regression model for Unselected adult cardiac arrest patients in the Early stages) is a novel prognostic model for the prediction of neurological outcome after cardiac arrest, which showed exceptional performance in the internal validation. The aim of this study is to validate the PROLOGUE score in an independent cohort of unselected adult cardiac arrest patients and to compare it to the thoroughly validated Out-of-Hospital Cardiac Arrest (OHCA) and Cardiac Arrest Hospital Prognosis (CAHP) scores.

METHODS

This study included consecutive adult cardiac arrest patients admitted to the intensive care unit (ICU) of a Swiss tertiary teaching hospital between October 2012 and July 2022. The primary endpoint was poor neurological outcome at hospital discharge, defined as a Cerebral Performance Category (CPC) score of 3 to 5 including death.

RESULTS

Of 687 patients included in the analysis, 321 (46.7%) survived to hospital discharge with good neurological outcome, 68 (9.9%) survived with poor neurological outcome and 298 (43.4%) died. The PROLOGUE score showed an area under the receiver operating characteristic curve (AUROC) of 0.83 (95% CI 0.80 to 0.86) and good calibration for the prediction of the primary outcome. The OHCA and CAHP score showed similar performance (AUROC 0.83 and 0.84 respectively), the differences between the three scores were not significant (p = 0.495). In a subgroup analysis, the PROLOGUE score performed equally in out-of-hospital and in-hospital cardiac arrest patients whereas the OHCA and CAHP score performed significantly better in OHCA patients.

CONCLUSION

The PROLOGUE score showed good prognostic accuracy for the early prediction of neurological outcome in adult cardiac arrest survivors in our cohort and might support early goals-of-care discussions in the ICU. Trial registration Not applicable.

Interpretable machine learning models for predicting in-hospital death in patients in the intensive care unit with cerebral infarction

BACKGROUND

Research on patients with cerebral infarction in the Intensive Care Unit (ICU) is still lacking. Our study aims to develop and validate multiple machine-learning (ML) models using two large ICU databases-Medical Information Mart for Intensive Care version III (MIMIC-III) and eICU Research Institute Database (eRI)-to guide clinical practice.

METHODS

We collected clinical data from patients with cerebral infarction in the MIMIC-III and eRI databases within 24 h of admission. The opinion of neurologists and the Least Absolute Shrinkage and Selection Operator regression was used to screen for relevant clinical features. Using eRI as the training set and MIMIC-III as the test set, we developed and validated six ML models. Based on the results of the model validation, we select the best model and perform the interpretability analysis on it.

RESULTS

A total of 4,338 patients were included in the study (eRI:3002, MIMIC-III:1336), resulting in a total of 18 clinical characteristics through screening. Model validation results showed that random forest (RF) was the best model, with AUC and F1 scores of 0.799 and 0.417 in internal validation and 0.733 and 0.498 in external validation, respectively; moreover, its sensitivity and recall were the highest of the six algorithms for both the internal and external validation. The explanatory analysis of the model showed that the three most important variables in the RF model were Acute Physiology Score-III, Glasgow Coma Scale score, and heart rate, and that the influence of each variable on the judgement of the model was consistent with medical knowledge.

CONCLUSION

Based on a large sample of patients and advanced algorithms, our study bridges the limitations of studies on this area. With our model, physicians can use the admission information of cerebral infarction patients in the ICU to identify high-risk groups among them who are prone to in-hospital death, so that they could be more alert to this group of patients and upgrade medical measures early to minimize the mortality of patients.

Predicting neurological outcome in adult patients with cardiac arrest: systematic review and meta-analysis of prediction model performance

This work aims to assess the performance of two post-arrest (out-of-hospital cardiac arrest, OHCA, and cardiac arrest hospital prognosis, CAHP) and one pre-arrest (good outcome following attempted resuscitation, GO-FAR) prediction model for the prognostication of neurological outcome after cardiac arrest in a systematic review and meta-analysis. A systematic search was conducted in Embase, Medline, and Web of Science Core Collection from November 2006 to December 2021, and by forward citation tracking of key score publications. The search identified 1'021 records, of which 25 studies with a total of 124'168 patients were included in the review. A random-effects meta-analysis of C-statistics and overall calibration (total observed vs. expected [O:E] ratio) was conducted. Discriminatory performance was good for the OHCA (summary C-statistic: 0.83 [95% CI 0.81-0.85], 16 cohorts) and CAHP score (summary C-statistic: 0.84 [95% CI 0.82-0.87], 14 cohorts) and acceptable for the GO-FAR score (summary C-statistic: 0.78 [95% CI 0.72-0.84], five cohorts). Overall calibration was good for the OHCA (total O:E ratio: 0.78 [95% CI 0.67-0.92], nine cohorts) and the CAHP score (total O:E ratio: 0.78 [95% CI 0.72-0.84], nine cohorts) with an overestimation of poor outcome. Overall calibration of the GO-FAR score was poor with an underestimation of good outcome (total O:E ratio: 1.62 [95% CI 1.28-2.04], five cohorts). Two post-arrest scores showed good prognostic accuracy for predicting neurological outcome after cardiac arrest and may support early discussions about goals-of-care and therapeutic planning on the intensive care unit. A pre-arrest score showed acceptable prognostic accuracy and may support code status discussions.

Comparison of different clinical risk scores to predict long-term survival and neurological outcome in adults after cardiac arrest: results from a prospective cohort study

Background

Several scoring systems have been used to predict short-term outcome in patients with out-of-hospital cardiac arrest (OHCA), including the disease-specific OHCA and CAHP (Cardiac Arrest Hospital Prognosis) scores, as well as the general severity-of-illness scores Acute Physiology and Chronic Health Evaluation II (APACHE II) and Simplified Acute Physiology Score II (SAPS II). This study aimed to assess the prognostic performance of these four scores to predict long-term outcomes (≥ 2 years) in adult cardiac arrest patients.

Methods

This is a prospective single-centre cohort study including consecutive cardiac arrest patients admitted to intensive care in a Swiss tertiary academic medical centre. The primary endpoint was 2-year mortality. Secondary endpoints were neurological outcome at 2 years post-arrest assessed by Cerebral Performance Category with CPC 1–2 defined as good and CPC 3–5 as poor neurological outcome, and 6-year mortality.

Results

In 415 patients admitted to intensive care, the 2-year mortality was 58.1%, with 96.7% of survivors showing good neurological outcome. The 6-year mortality was 82.5%. All four scores showed good discriminatory performance for 2-year mortality, with areas under the receiver operating characteristics curve (AUROC) of 0.82, 0.87, 0.83 and 0.81 for the OHCA, CAHP, APACHE II and SAPS II scores. The results were similar for poor neurological outcome at 2 years and 6-year mortality.

Conclusion

This study suggests that two established cardiac arrest-specific scores and two severity-of-illness scores provide good prognostic value to predict long-term outcome after cardiac arrest and thus may help in early goals-of-care discussions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13613-022-01048-y.

Development and validation of an interpretable prehospital return of spontaneous circulation (P-ROSC) score for patients with out-of-hospital cardiac arrest using machine learning: A retrospective study

BACKGROUND

Return of spontaneous circulation (ROSC) before arrival at the emergency department is an early indicator of successful resuscitation in out-of-hospital cardiac arrest (OHCA). Several ROSC prediction scores have been developed with European cohorts, with unclear applicability in Asian settings. We aimed to develop an interpretable prehospital ROSC (P-ROSC) score for ROSC prediction based on patients with OHCA in Asia.

METHODS

This retrospective study examined patients who suffered from OHCA between Jan 1, 2009 and Jun 17, 2018 using data recorded in the Pan-Asian Resuscitation Outcomes Study (PAROS) registry. AutoScore, an interpretable machine learning framework, was used to develop P-ROSC. On the same cohort, the P-ROSC was compared with two clinical scores, the RACA and the UB-ROSC. The predictive power was evaluated using the area under the curve (AUC) in the receiver operating characteristic analysis.

FINDINGS

170,678 cases were included, of which 14,104 (8.26%) attained prehospital ROSC. The P-ROSC score identified a new variable, prehospital drug administration, which was not included in the RACA score or the UB-ROSC score. Using only five variables, the P-ROSC score achieved an AUC of 0.806 (95% confidence interval [CI] 0.799-0.814), outperforming both RACA and UB-ROSC with AUCs of 0.773 (95% CI 0.765-0.782) and 0.728 (95% CI 0.718-0.738), respectively.

INTERPRETATION

The P-ROSC score is a practical and easily interpreted tool for predicting the probability of prehospital ROSC.

FUNDING

This research received funding from SingHealth Duke-NUS ACP Programme Funding (15/FY2020/P2/06-A79).

Performance of Sequential Organ Failure Assessment and Simplified Acute Physiology Score II for Post-Cardiac Surgery Patients in Intensive Care Unit

Background: The aim of this study is to assess the performance of Sequential Organ Failure Assessment (SOFA) score and Simplified Acute Physiology Score (SAPS II) on outcomes of patients with cardiac surgery and identify the cutoff values to provide a reference for early intervention.

Methods: All data were extracted from MIMIC-III (Medical Information Mart for Intensive Care-III) database. Cutoff values were calculated by the receiver-operating characteristic curve and Youden indexes. Patients were grouped, respectively, according to the cutoff values of SOFA and SAPS II. A non-adjusted model and adjusted model were established to evaluate the prediction of risk. Comparison of clinical efficacy between two scoring systems was made by decision curve analysis (DCA). The primary outcomes of this study were in-hospital mortality, 28-day mortality, 90-day mortality, and 1-year mortality after cardiac surgery. The secondary outcomes included length of hospital stay and intensive care unit (ICU) stay and the incidence of acute kidney injury (AKI) within 7 days after ICU admission.

Results: A total of 6,122 patients were collected and divided into the H-SOFA group (SOFA ≥ 7) and L-SOFA group (SOFA < 7) or H-SAPS II group (SAPS II ≥ 43) and L-SAPS II group (SAPS II < 43). In-hospital mortality, 28-day mortality, 90-day mortality, and 1-year mortality were higher, the length of hospital and ICU stay were longer in the H-SOFA group than in the L-SOFA group (p < 0.05), while the incidence of AKI was not significantly different. In-hospital mortality, 28-day mortality, 90-day mortality, 1-year mortality, and the incidence of AKI were all significantly higher in the H-SAPS II group than in the L-SAPS II group (p < 0.05). Hospital stay and ICU stay were longer in the H-SAPS II group than in the L-SAPS II group (p < 0.05). According to DCA, the SAPS II scoring system had more net benefits on assessing the long-term mortality compared with the SOFA scoring system.

Conclusion: Exceeding the cutoff values of SOFA and SAPS II scores could lead to increased mortality and extended length of ICU and hospital stay. The SAPS II scoring system had a better discriminative performance of 90-day mortality and 1-year mortality in post-cardiac surgery patients than the SOFA scoring system. Emphasizing the critical value of the scoring system is of significance for timely treatment.

Severity indexes of blunt trauma victims in intensive therapy: prediction capacity for mortality

OBJECTIVE

To identify the predictive capacity for mortality of the indexes Revised Trauma Score, Rapid Emergency Medicine Score, modified Rapid Emergency Medicine Score, and Simplified Acute Physiology Score III in blunt trauma victims hospitalized in an intensive care unit and compare their performance.

METHOD

Retrospective cohort of patients with blunt trauma in an intensive care unit from medical records. Receiver Operating Characteristic and a 95% confidence interval of the area under the curve were analyzed to compare results.

RESULTS

Out of 165 analyzed patients, 66.7% have received surgical treatment. The mortality in the intensive care unit and in the hospital was 17.6% and 20.6%, respectively. For the mortality in the intensive care unit, the area under the curve varied from 0.672 to 0.738; however, better results have been observed in surgical patients (0.747 to 0.811). Similar results have been observed for in-hospital mortality. In all analyses, the areas under the curve of the indexes presented no significant difference.

CONCLUSION

The accuracy of the severity indexes was moderate, with an improved performance when applied to surgical patients. The four indexes presented a similar prediction for the analyzed outcomes.

Machine learning applied to a Cardiac Surgery Recovery Unit and to a Coronary Care Unit for mortality prediction

Most established severity-of-illness systems used for prediction of intensive care unit (ICU) mortality were developed targeted at the general ICU population, based on logistic regression (LR). To date, no dynamic predictive tool for ICU mortality has been developed targeted at the Cardiac Surgery Recovery Unit (CSRU) and Coronary Care Unit (CCU) using machine learning (ML). CSRU and CCU adult patients from the MIMIC-III critical care database were studied. The ML methods developed extract ICU data during a 5-h window and demographic features to produce mortality predictions and were compared to six established severity-of-illness systems and LR. In a secondary experiment, additional procedure/surgery and ICU features were added to the models. The ML models developed were the Tree Ensemble (TE), Random Forest, XGBoost Tree Ensemble (XGB), Naive Bayes (NB), and Bayesian network. The discrimination, calibration and accuracy statistics were assessed. The AUROC values were superior for the ML models reaching 0.926 and 0.924 for the XGB, and 0.904 and 0.908 for the TE for ICU mortality prediction in the primary and secondary experiments respectively. Among the conventional systems, the serial SOFA obtained the highest AUROC (0.8405). The Brier score was better for the ML models except the NB over the conventional systems. The accuracy statistics less sensitive to unbalanced cohorts were higher for all the ML models. In conclusion, the predictive power of XGB was excellent, substantially outperforming the conventional systems and LR. The ML models developed in this work offer promising results that could benefit CSRU and CCU.

Current Utility of Sequential Organ Failure Assessment Score: A Literature Review and Future Directions

The Sequential Organ Failure Assessment (SOFA) score is commonly used in the Intensive Care Unit (ICU) to evaluate, prognosticate and assess patients. Since its validation, the SOFA score has served in various settings, including medical, trauma, surgical, cardiac, and neurological ICUs. It has been a strong mortality predictor and literature over the years has documented the ability of the SOFA score to accurately distinguish survivors from non-survivors on admission. Over the years, multiple variations have been proposed to the SOFA score, which have led to the evolution of alternate validated scoring models replacing one or more components of the SOFA scoring system. Various SOFA based models have been used to evaluate specific clinical populations, such as patients with cardiac dysfunction, hepatic failure, renal failure, different races and public health illnesses, etc. This study is aimed to conduct a review of modifications in SOFA score in the past several years. We review the literature evaluating various modifications to the SOFA score such as modified SOFA, Modified SOFA, modified Cardiovascular SOFA, Extra-renal SOFA, Chronic Liver Failure SOFA, Mexican SOFA, quick SOFA, Lactic acid quick SOFA (LqSOFA), SOFA in hematological malignancies, SOFA with Richmond Agitation-Sedation scale and Pediatric SOFA. Various organ systems, their relevant scoring and the proposed modifications in each of these systems are presented in detail. There is a need to incorporate the most recent literature into the SOFA scoring system to make it more relevant and accurate in this rapidly evolving critical care environment. For future directions, we plan to put together most if not all updates in SOFA score and probably validate it in a large database a single institution and validate it in multisite data base.

Use of SOFA score in cardiac arrest research: A scoping review

Background

The Sequential Organ Failure Assessment (SOFA) score is a commonly used severity-of-illness score in cardiac arrest research. Due to its nature, the SOFA score often has missing data. How much data is missing and how that missing data is handled is unknown.

Objectives

We conducted a scoping review on cardiac arrest studies using SOFA, focusing on missing data.

Data sources

PubMed, Embase, and Web of Science.

Study selection

All English-language peer-reviewed studies of cardiac arrest with SOFA as an outcome or exposure were included.

Data extraction

For each study, quantity of missing SOFA data, analytic strategy to handle missing SOFA variables, whether/to what degree mortality influenced the amount of missing SOFA scores), SOFA score modifications, and number of SOFA measurements was extracted.

Data synthesis

We included 66 studies published between 2006–2019. Five studies were randomized controlled trials, 26 were prospective cohort studies, and 25 were retrospective cohort studies. SOFA was used as an outcome in 36 (55%) and a primary outcome in 10 (15%). Nine studies (14%) mentioned the quantity of missing SOFA data, which ranged from 0 to 76% (median: 10% [IQR: 6%, 42%]). Twenty-seven (41%) studies reported a method to handle missing SOFA. The most common method used excluded subjects with missing data (81%). In the 50 studies using serial SOFA scores, 11 (22%) documented mortality prior to SOFA measurement; which ranged from 3% to 76% (median: 12% [IQR: 6%–35%]).

Conclusions

Missing data is common in cardiac arrest research using SOFA scores. Variability exists in reporting and handling missing SOFA variables.

Combining structured and unstructured data for predictive models: a deep learning approach

Background

The broad adoption of electronic health records (EHRs) provides great opportunities to conduct health care research and solve various clinical problems in medicine. With recent advances and success, methods based on machine learning and deep learning have become increasingly popular in medical informatics. However, while many research studies utilize temporal structured data on predictive modeling, they typically neglect potentially valuable information in unstructured clinical notes. Integrating heterogeneous data types across EHRs through deep learning techniques may help improve the performance of prediction models.

Methods

In this research, we proposed 2 general-purpose multi-modal neural network architectures to enhance patient representation learning by combining sequential unstructured notes with structured data. The proposed fusion models leverage document embeddings for the representation of long clinical note documents and either convolutional neural network or long short-term memory networks to model the sequential clinical notes and temporal signals, and one-hot encoding for static information representation. The concatenated representation is the final patient representation which is used to make predictions.

Results

We evaluate the performance of proposed models on 3 risk prediction tasks (i.e. in-hospital mortality, 30-day hospital readmission, and long length of stay prediction) using derived data from the publicly available Medical Information Mart for Intensive Care III dataset. Our results show that by combining unstructured clinical notes with structured data, the proposed models outperform other models that utilize either unstructured notes or structured data only.

Conclusions

The proposed fusion models learn better patient representation by combining structured and unstructured data. Integrating heterogeneous data types across EHRs helps improve the performance of prediction models and reduce errors.

Artificial neural networks improve early outcome prediction and risk classification in out-of-hospital cardiac arrest patients admitted to intensive care

Background

Pre-hospital circumstances, cardiac arrest characteristics, comorbidities and clinical status on admission are strongly associated with outcome after out-of-hospital cardiac arrest (OHCA). Early prediction of outcome may inform prognosis, tailor therapy and help in interpreting the intervention effect in heterogenous clinical trials. This study aimed to create a model for early prediction of outcome by artificial neural networks (ANN) and use this model to investigate intervention effects on classes of illness severity in cardiac arrest patients treated with targeted temperature management (TTM).

Methods

Using the cohort of the TTM trial, we performed a post hoc analysis of 932 unconscious patients from 36 centres with OHCA of a presumed cardiac cause. The patient outcome was the functional outcome, including survival at 180 days follow-up using a dichotomised Cerebral Performance Category (CPC) scale with good functional outcome defined as CPC 1–2 and poor functional outcome defined as CPC 3–5. Outcome prediction and severity class assignment were performed using a supervised machine learning model based on ANN.

Results

The outcome was predicted with an area under the receiver operating characteristic curve (AUC) of 0.891 using 54 clinical variables available on admission to hospital, categorised as background, pre-hospital and admission data. Corresponding models using background, pre-hospital or admission variables separately had inferior prediction performance. When comparing the ANN model with a logistic regression-based model on the same cohort, the ANN model performed significantly better (p = 0.029). A simplified ANN model showed promising performance with an AUC above 0.852 when using three variables only: age, time to ROSC and first monitored rhythm. The ANN-stratified analyses showed similar intervention effect of TTM to 33 °C or 36 °C in predefined classes with different risk of a poor outcome.

Conclusion

A supervised machine learning model using ANN predicted neurological recovery, including survival excellently, and outperformed a conventional model based on logistic regression. Among the data available at the time of hospitalisation, factors related to the pre-hospital setting carried most information. ANN may be used to stratify a heterogenous trial population in risk classes and help determine intervention effects across subgroups.

The Effectiveness of Scoring Systems in the Prediction of Diagnosis-Based Mortality

Scoring systems are used for mortality and morbidity rating in intensive care conditions, prognosis prediction, standardization of scientific data and the monitoring of clinical quality. The aim of this study was to retrospectively analyze the efficacy of APACHE II (Acute Physiology and Chronic Health Evaluation), APACHE IV and SAPS (Simplified Acute Physiology Score) III prognostic scorings in the prediction of mortality and disease severity of patients admitted to the Anesthesia and Reanimation Clinic Intensive Care Unit (ICU) in Bakırköy Dr. Sadi Konuk Training and Research Hospital according to general and specific diagnoses. A total of 1896 patient files were included in the study. With the exception of single system or head trauma patient groups, a statistically significant difference was found in the mortality prediction rates in all other diagnosis groups (P < 0.05). The discrimination calculated with AUROC fields was sufficient in all groups, and calibration was evaluated as good except for the neurological and neurosurgical patient group. In respect of standard mortality prediction, APACHE II and IV were good in cases of sepsis, and SAPS III made almost exact predictions for cardiovascular diseases, APACHE II for neurological diseases, and APACHE IV for gastrointestinal system diseases. From the results of this study, it was seen that different scoring systems vary in predictions according to the diagnoses, therefore, it can be recommended that the diagnosis should be taken into account more when applying scoring systems.

Predictive modeling in urgent care: a comparative study of machine learning approaches

OBJECTIVE

The growing availability of rich clinical data such as patients' electronic health records provide great opportunities to address a broad range of real-world questions in medicine. At the same time, artificial intelligence and machine learning (ML)-based approaches have shown great premise on extracting insights from those data and helping with various clinical problems. The goal of this study is to conduct a systematic comparative study of different ML algorithms for several predictive modeling problems in urgent care.

DESIGN

We assess the performance of 4 benchmark prediction tasks (eg mortality and prediction, differential diagnostics, and disease marker discovery) using medical histories, physiological time-series, and demographics data from the Medical Information Mart for Intensive Care (MIMIC-III) database.

MEASUREMENTS

For each given task, performance was estimated using standard measures including the area under the receiver operating characteristic (AUC) curve, F-1 score, sensitivity, and specificity. Microaveraged AUC was used for multiclass classification models.

RESULTS AND DISCUSSION

Our results suggest that recurrent neural networks show the most promise in mortality prediction where temporal patterns in physiologic features alone can capture in-hospital mortality risk (AUC > 0.90). Temporal models did not provide additional benefit compared to deep models in differential diagnostics. When comparing the training-testing behaviors of readmission and mortality models, we illustrate that readmission risk may be independent of patient stability at discharge. We also introduce a multiclass prediction scheme for length of stay which preserves sensitivity and AUC with outliers of increasing duration despite decrease in sample size.

Performance on the APACHE II, SAPS II, SOFA and the OHCA score of post-cardiac arrest patients treated with therapeutic hypothermia

Objective

This study assessed the ability of the Acute Physiologic and Chronic Health Evaluation (APACHE) II score, Simplified Acute Physiology Score (SAPS) II, Sequential Organ Failure Assessment (SOFA) score, and out-of-hospital cardiac arrest (OHCA) score to predict the outcome of OHCA patients who underwent therapeutic hypothermia (TH).

Methods

This study included OHCA patients treated with TH between January 2010 and December 2013. The APACHE II score, SAPS II, and SOFA score were calculated at the time of admission and 24 h and 48 h after intensive care unit admission. The OHCA score was calculated at the time of admission. The area under the curve (AUC) of the receiver operating characteristic curve and logistic regression analysis were used to evaluate outcome predictability.

Results

Data from a total of 173 patients were included in the analysis. The APACHE II score at 0 h and 48 h, SAPS II at 48 h, and OHCA score had moderate discrimination for mortality (AUC: 0.715, 0.750, 0.720, 0.740). For neurologic outcomes, the APACHE II score at 0 h and 48 h, SAPS II at 0 h and 48 h, and OHCA score showed moderate discrimination (AUC: 0.752, 0.738, 0.771, 0.771, 0.764). The APACHE II score, SAPS II and SOFA score at various time points, in addition to the OHCA score, were independent predictors of mortality and a poor neurologic outcome.

Conclusions

The APACHE II score, SAPS II, SOFA score, and OHCA score have different capabilities in discriminating and estimating hospital mortality and neurologic outcomes. The OHCA score, APACHE II score and SAPS II at time zero and 48 h offer moderate predictive accuracy. Other scores at 0 h and 48 h, except for the SOFA score, are independently associated with 30-day mortality and poor cerebral performance.

Early predictors of poor outcome after out-of-hospital cardiac arrest

Background

Early identification of predictors for a poor long-term outcome in patients who survive the initial phase of out-of-hospital cardiac arrest (OHCA) may facilitate future clinical research, the process of care and information provided to relatives. The aim of this study was to determine the association between variables available from the patient’s history and status at intensive care admission with outcome in unconscious survivors of OHCA.

Methods

Using the cohort of the Target Temperature Management trial, we performed a post hoc analysis of 933 unconscious patients with OHCA of presumed cardiac cause who had a complete 6-month follow-up. Outcomes were survival and neurological function as defined by the Cerebral Performance Category (CPC) scale at 6 months after OHCA. After multiple imputations to compensate for missing data, backward stepwise multivariable logistic regression was applied to identify factors independently predictive of a poor outcome (CPC 3–5). On the basis of these factors, a risk score for poor outcome was constructed.

Results

We identified ten independent predictors of a poor outcome: older age, cardiac arrest occurring at home, initial rhythm other than ventricular fibrillation/tachycardia, longer duration of no flow, longer duration of low flow, administration of adrenaline, bilateral absence of corneal and pupillary reflexes, Glasgow Coma Scale motor response 1, lower pH and a partial pressure of carbon dioxide in arterial blood value lower than 4.5 kPa at hospital admission. A risk score based on the impact of each of these variables in the model yielded a median (range) AUC of 0.842 (0.840–0.845) and good calibration. Internal validation of the score using bootstrapping yielded a median (range) AUC corrected for optimism of 0.818 (0.816–0.821).

Conclusions

Among variables available at admission to intensive care, we identified ten independent predictors of a poor outcome at 6 months for initial survivors of OHCA. They reflected pre-hospital circumstances (six variables) and patient status on hospital admission (four variables). By using a simple and easy-to-use risk scoring system based on these variables, patients at high risk for a poor outcome after OHCA may be identified early.

Therapeutic hypothermia after cardiac arrest: outcome predictors

OBJECTIVE

The determination of coma patient prognosis after cardiac arrest has clinical, ethical and social implications. Neurological examination, imaging and biochemical markers are helpful tools accepted as reliable in predicting recovery. With the advent of therapeutic hypothermia, these data need to be reconfirmed. In this study, we attempted to determine the validity of different markers, which can be used in the detection of patients with poor prognosis under hypothermia.

METHODS

Data from adult patients admitted to our intensive care unit for a hypothermia protocol after cardiac arrest were recorded prospectively to generate a descriptive and analytical study analyzing the relationship between clinical, neurophysiological, imaging and biochemical parameters with 6-month outcomes defined according to the Cerebral Performance Categories scale (good 1-2, poor 3-5). Neuron-specific enolase was collected at 72 hours. Imaging and neurophysiologic exams were carried out in the 24 hours after the rewarming period.

RESULTS

Sixty-seven patients were included in the study, of which 12 had good neurological outcomes. Ventricular fibrillation and electroencephalographic theta activity were associated with increased likelihood of survival and improved neurological outcomes. Patients who had more rapid cooling (mean time of 163 versus 312 minutes), hypoxic-ischemic brain injury on magnetic resonance imaging or neuron-specific enolase > 58ng/mL had poor neurological outcomes (p < 0.05).

CONCLUSION

Hypoxic-ischemic brain injury on magnetic resonance imaging and neuron-specific enolase were strong predictors of poor neurological outcomes. Although there is the belief that early achievement of target temperature improves neurological prognoses, in our study, there were increased mortality and worse neurological outcomes with earlier target-temperature achievement.