Association of Acidemia With Short-Term Mortality of Acute Myocardial Infarction: A Retrospective Study Base on MIMIC-III Database

Application of an interpretable machine learning method to predict the risk of death during hospitalization in patients with acute myocardial infarction combined with diabetes mellitus

BACKGROUND

Predicting the prognosis of patients with acute myocardial infarction (AMI) combined with diabetes mellitus (DM) is crucial due to high in-hospital mortality rates. This study aims to develop and validate a mortality risk prediction model for these patients by interpretable machine learning (ML) methods.

METHODS

Data were sourced from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2). Predictors were selected by Least absolute shrinkage and selection operator (LASSO) regression and checked for multicollinearity with Spearman's correlation. Patients were randomly assigned to training and validation sets in an 8:2 ratio. Seven ML algorithms were used to construct models in the training set. Model performance was evaluated in the validation set using metrics such as area under the curve (AUC) with 95% confidence interval (CI), calibration curves, precision, recall, F1 score, accuracy, negative predictive value (NPV), and positive predictive value (PPV). The significance of differences in predictive performance among models was assessed utilising the permutation test, and 10-fold cross-validation further validated the model's performance. SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) were applied to interpret the models.

RESULTS

The study included 2,828 patients with AMI combined with DM. Nineteen predictors were identified through LASSO regression and Spearman's correlation. The Random Forest (RF) model was demonstrated the best performance, with an AUC of 0.823 (95% CI: 0.774-0.872), high precision (0.867), accuracy (0.873), and PPV (0.867). The RF model showed significant differences (p < 0.05) compared to the K-Nearest Neighbours and Decision Tree models. Calibration curves indicated that the RF model's predicted risk aligned well with actual outcomes. 10-fold cross-validation confirmed the superior performance of RF model, with an average AUC of 0.828 (95% CI: 0.800-0.842). Significant Variables in RF model indicated that the top eight significant predictors were urine output, maximum anion gap, maximum urea nitrogen, age, minimum pH, maximum international normalised ratio (INR), mean respiratory rate, and mean systolic blood pressure.

CONCLUSION

This study demonstrates the potential of ML methods, particularly the RF model, in predicting in-hospital mortality risk for AMI patients with DM. The SHAP and LIME methods enhance the interpretability of ML models.

Risk Factors and An Interpretability Tool of In-hospital Mortality in Critically Ill Patients with Acute Myocardial Infarction

PURPOSE

We aim to develop and validate an interpretable machine-learning model that can provide critical information for the clinical treatment of critically ill patients with acute myocardial infarction (AMI).

METHODS

All data was extracted from the multi-centre database (training and internal validation cohorts: MIMIC-III/-IV, external validation cohort: eICU). After comparing different machine-learning models and several unbalanced data processing methods, the model with the best performance was selected. Lasso regression was used to build a compact model. Seven evaluation methods, PR, and ROC curves were used to assess the model. The SHapley Additive exPlanations (SHAP) values were calculated to evaluate the feature's importance. The SHAP plots were adopted to explain and interpret the results. A web-based tool was developed to help application.

RESULTS

A total of 12,170 critically ill patients with AMI were included. The balance random forest (BRF) model had the best performance in predicting in-hospital mortality. The compact model did not differ from the full variable model in performance (AUC: 0.891 vs 0.885, P = 0.06). The external validation results also demonstrated the stability of the model (AUC: 0.784). All SHAP plots have shown the contribution ranking of all variables in the model, the relationship trend between variables and outcomes, and the interaction mode between variables. A web-based tool is constructed that can provide individualized risk stratification probabilities (https://github.com/huangtao36/BRF-web-tool) .

CONCLUSION

We built the BRF model and the web-based tool by the model algorithm. The model effect has been verified externally. The tool can help clinical decision-making.

Causal relationship between immune cells and risk of myocardial infarction: evidence from a Mendelian randomization study

Background

Atherosclerotic plaque rupture is a major cause of heart attack. Previous studies have shown that immune cells are involved in the development of atherosclerosis, but different immune cells play different roles. The aim of this study was to investigate the causal relationship between immunological traits and myocardial infarction (MI).

Methods

To assess the causal association of immunological profiles with myocardial infarction based on publicly available genome-wide studies, we used a two-sample mendelian randomization (MR) approach with inverse variance weighted (IVW) as the main analytical method. Sensitivity analyses were used to assess heterogeneity and horizontal pleiotropy.

Results

A two-sample MR analysis was conducted using IVW as the primary method. At a significance level of 0.001, we identified 47 immunophenotypes that have a significant causal relationship with MI. Seven of these were present in B cells, five in cDC, four in T cells at the maturation stage, six in monocytes, five in myeloid cells, 12 in TBNK cells, and eight in Treg cells. Sensitivity analyses were performed to confirm the robustness of the MR results.

Conclusions

Our results provide strong evidence that multiple immune cells have a causal effect on the risk of myocardial infarction. This discovery provides a new avenue for the development of therapeutic treatments for myocardial infarction and a new target for drug development.

Association between red blood cell distribution width-to-albumin ratio and prognosis of patients with acute myocardial infarction

BACKGROUND

Red blood cell distribution width (RDW) and albumin level were considered to be related to the prognosis of patients with acute myocardial infarction (AMI). This study aims to investigate the correlation between RAR and 90-day mortality in AMI patients.

METHODS

Data of AMI patients were obtained from the Medical Information Mart for Intensive Care III (MIMIC-III) database. According to the median, RAR < 4.32 was regarded as low RAR level group, and RAR ≥ 4.32 as high RAR level group; low RDW level group was defined as < 14.00%, and high RDW level group as ≥ 14.00%; albumin < 3.30 g/dL was low level group, and albumin ≥ 3.30 g/dL as high level group. The outcome was the mortality rate within 90 days after admission to ICU. Univariate and multivariate Cox models were performed to determine the relationship between RAR and 90-day mortality in AMI patients with hazard ratio (HR) and 95% confidence interval (CI). Stratification analyses were conducted to explore the effect of RAR on 90-day mortality in different subgroups of age, gender, simplified acute physiology score II (SAPS II), elixhauser comorbidity index (ECI) score, treatment modalities and white blood cell.

RESULTS

Of the total 2081 AMI patients, 543 (26.09%) died within 90-day follow-up duration. The results showed that high RAR (HR = 1.65, 95% CI 1.34-2.03) and high RDW levels (HR = 1.31, 95% CI 1.08-1.61) were associated with an increased risk of death in AMI patients, and that high albumin level was related to a decreased risk of death (HR = 0.77, 95%CI 0.64-0.93). The relationship of RAR level and the mortality of AMI patients was also observed in the subgroup analysis. Additionally, the finding indicated that RAR might be a more effective biomarker for predicting 90-day mortality of AMI patients than albumin, RDW.

CONCLUSION

RAR may be a potential marker for the prognostic assessment of AMI, and a high RAR level was correlated with increased risk of 90-day mortality of AMI patients.

Base excess is associated with the risk of all-cause mortality in critically ill patients with acute myocardial infarction

Background

Base excess (BE) represents an increase or decrease of alkali reserves in plasma to diagnose acid-base disorders, independent of respiratory factors. Current findings about the prognostic value of BE on mortality of patients with acute myocardial infarction (AMI) are still unclear. The purpose of this study was to explore the prognostic significance of BE for short-term all-cause mortality in patients with AMI.

Methods

A total of 2,465 patients diagnosed with AMI in the intensive care unit from the Medical Information Mart for Intensive Care III (MIMIC-III) database were included in our study, and we explored the association of BE with 28-day and 90-day all-cause mortality using Cox regression analysis. We also used restricted cubic splines (RCS) to evaluate the relationship between BE and hazard ratio (HR). The primary outcomes were 28-day and 90-day all-cause mortality.

Results

When stratified according to quantiles, low BE levels at admission were strongly associated with higher 28-day and 90-day all-cause mortality. Multivariable Cox proportional hazard models revealed that low BE was an independent risk factor of 28-day all-cause mortality [HR 4.158, 95% CI 3.203–5.398 (low vs. normal BE) and HR 1.354, 95% CI 0.896–2.049 (high vs. normal BE)] and 90-day all-cause mortality [HR 4.078, 95% CI 3.160–5.263 (low vs. normal BE) and HR 1.369, 95% CI 0.917–2.045 (high vs. normal BE)], even after adjustment for significant prognostic covariates. The results were also consistent in subgroup analysis. RCS revealed an “L-type” relationship between BE and 28-day and 90-day all-cause mortality, as well as adjusting for confounding variables. Meanwhile, Kaplan–Meier survival curves were stratified by combining BE with carbon dioxide partial pressure (PaCO2), and patients had the highest mortality in the group which had low BE (&lt; 3.5 mEq/L) and high PaCO2 (&gt; 45 mmHg) compared with other groups.

Conclusion

Our study revealed that low BE was significantly associated with 28-day and 90-day mortality in patients with AMI and indicated the value of stratifying the mortality risk of patients with AMI by BE.

Association of Increased Pulse Wave Velocity With Long-Term Clinical Outcomes in Patients With Preserved Ankle-Brachial Index After Acute Myocardial Infarction

BACKGROUND

Low ankle-brachial index (ABI) is an established risk factor for long-term cardiovascular outcomes in patients with acute myocardial infarction (AMI), and brachial-ankle pulse wave velocity (ba-PWV) may also be a risk factor. However, there is a significant overlap between low ABI and high ba-PWV. The purpose of this retrospective study was to examine whether increased ba-PWV was associated with long-term clinical outcomes in AMI patients with normal ABI.

METHODS

We included 932 AMI patients with normal ABI and divided them into the high PWV group (≥1,400 cm/s; n=646) and the low PWV group (<1400 cm/s; n=286) according to the ba-PWV values measured during the AMI hospitalisation. The primary endpoint was the major adverse cardiovascular events (MACE) defined as the composite of all-cause death, nonfatal myocardial infarction, and hospitalisation for heart failure.

RESULTS

During the median follow-up duration of 541 days (Q1: 215 days-Q3: 1,022 days), a total of 154 MACE were observed. The Kaplan-Meier curves showed that MACE was more frequently observed in the high PWV group than in the low PWV group (p<0.001). The multivariate Cox hazard analysis revealed that high ba-PWV was significantly associated with MACE (hazard ratio [HR] 1.587; 95% CI 1.002-2.513; p=0.049) after controlling multiple confounding factors.

CONCLUSIONS

High ba-PWV was significantly associated with long-term adverse events in AMI patients with normal ABI. Our results suggest the usefulness of PWV as a prognostic marker in AMI with normal ABI.

Body mass index linked to short-term and long-term all-cause mortality in patients with acute myocardial infarction

PURPOSES OF STUDY

This study aimed to elucidate the relationship between obesity and short-term and long-term mortality in patients with acute myocardial infarction (AMI) by analysing the body mass index (BMI).

STUDY DESIGN

A retrospective cohort study was performed on adult intensive care unit (ICU) patients with AMI in the Medical Information Mart for Intensive Care III database. The WHO BMI classification was used in the study. The Kaplan-Meier curve was used to show the likelihood of survival in patients with AMI. The relationships of the BMI classification with short-term and long-term mortality were assessed using Cox proportional hazard regression models.

RESULTS

This study included 1295 ICU patients with AMI, who were divided into four groups according to the WHO BMI classification. Our results suggest that obese patients with AMI tended to be younger (p<0.001), be men (p=0.001) and have higher blood glucose and creatine kinase (p<0.001) compared with normal weight patients. In the adjusted model, compared with normal weight AMI patients, those who were overweight and obese had lower ICU risks of death HR=0.64 (95% CI 0.46 to 0.89) and 0.55 (0.38 to 0.78), respectively, inhospital risks of death (0.77 (0.56 to 1.09) and 0.61 (0.43 to 0.87)) and long-term risks of death (0.78 0.64 to 0.94) and 0.72 (0.59 to 0.89). On the other hand, underweight patients had higher risks of short-term(ICU or inhospital mortality) and long-term mortality compared with normal weight patients (HR=1.39 (95% CI 0.58 to 3.30), 1.46 (0.62 to 3.42) and 1.99 (1.15 to 3.44), respectively).

CONCLUSIONS

Overweight and obesity were protective factors for the short-term and long-term risks of death in patients with AMI.

Deep-Learning-Based Survival Prediction of Patients in Coronary Care Units

BACKGROUND

A survival prediction model based on deep learning has higher accuracy than the CPH model in predicting the survival of CCU patients, and it also has a better discrimination ability. We collected information on patients with various diseases in coronary care units (CCUs) from the Medical Information Mart for Intensive Care III (MIMIC-III) database. The purpose of this study was to use this information to construct a neural-network model based on deep learning to predict the survival probabilities of patients with conditions that are common in CCUs.

METHOD

We collected information on patients in the United States with five common diseases in CCUs from 2001 to 2012. We randomly divided the patients into a training cohort and a testing cohort at a ratio of 7 : 3 and applied a survival prediction method based on deep learning to predict their survival probability. We compared our model with the Cox proportional-hazards regression (CPH) model and used the concordance indexes (C-indexes), receiver operating characteristic (ROC) curve, and calibration plots to evaluate the predictive performance of the model.

RESULTS

The 3,388 CCU patients included in the study were randomly divided into 2,371 in the training cohort and 1,017 in the testing cohort. The stepwise regression results showed that the important factors affecting patient survival were the type of disease, age, race, anion gap, glucose, neutrophils, white blood cells, potassium, creatine kinase, and blood urea nitrogen (P < 0.05). We used the training cohort to construct a deep-learning model, for which the C-index was 0.833, or about 5% higher than that for the CPH model (0.786). The C-index of the deep-learning model for the test cohort was 0.822, which was also higher than that for the CPH model (0.782). The areas under the ROC curve for the 28-day, 90-day, and 1-year survival probabilities were 0.875, 0.865, and 0.874, respectively, in the deep-learning model, respectively, and 0.830, 0.843, and 0.806 in the CPH model. These values indicate that the survival analysis model based on deep learning is better than the traditional CPH model in predicting the survival of CCU patients.

CONCLUSION

A survival prediction model based on deep learning has higher accuracy than the CPH model in predicting the survival of CCU patients, and it also has a better discrimination ability.

The Association of Coagulation Indicators and Coagulant Agents With 30-Day Mortality of Critical Diabetics

Diabetes, regarded as a global health concerned disease, was focused by the World Health Organization (WHO). Patients with diabetes may have a hypercoagulable and hypo-fibrinolysis state. There is lots of research about cardiovascular effects on diabetes patients, but less about the coagulation system. This study is designed to investigate the relationship between coagulation indicators and 30-day mortality of critical diabetes patients. In this retrospective, single-center study, we included adult patients diagnosed with diabetes. Data, including demographic, complication, laboratory tests, scoring system, and anticoagulant treatment, were extracted from Medical Information Mart for Intensive Care (MIMIC-III). The receiver operating characteristic (ROC) curve and Kaplan-Meier curve were applied to predict the association of mortality and coagulation indicators. Cox hazard regression model and subgroup analysis were used to analyze the risk factors associated with 30-day mortality. A total of 4026 patients with diabetes mellitus were included in our study, of whom 3312 survived after admitted to the hospital and 714 died. Cox hazard regression showed anticoagulant therapy might decrease the risk of 30-day mortality after adjusted. In age <70 subgroup analysis, we found that patients with PTT <26.8 s or lightly increased PT may increase odds of 30-day hospital death (HR, 95%CI, 2.044 (1.376, 3.034), 1.562 (1.042, 2.343)). When age >70, lightly increased PTT may reduce the risk of mortality, but PT >16.3 s, a high level of hypo-coagulation state, increase risk of mortality (HR, 95%CI, 0.756 (0.574, 0.996), 1.756 (1.129, 2.729)). Critical diabetes patients may benefit from anticoagulant agents. The abnormal coagulant function is related to the risk of 30-day mortality.