Machine learning model to predict hypotension after starting continuous renal replacement therapy

Hypotension after starting continuous renal replacement therapy (CRRT) is associated with worse outcomes compared with normotension, but it is difficult to predict because several factors have interactive and complex effects on the risk. The present study applied machine learning algorithms to develop models to predict hypotension after initiating CRRT. Among 2349 adult patients who started CRRT due to acute kidney injury, 70% and 30% were randomly assigned into the training and testing sets, respectively. Hypotension was defined as a reduction in mean arterial pressure (MAP) ≥ 20 mmHg from the initial value within 6 h. The area under the receiver operating characteristic curves (AUROCs) in machine learning models, such as support vector machine (SVM), deep neural network (DNN), light gradient boosting machine (LGBM), and extreme gradient boosting machine (XGB) were compared with those in disease-severity scores such as the Sequential Organ Failure Assessment and Acute Physiology and Chronic Health Evaluation II. The XGB model showed the highest AUROC (0.828 [0.796–0.861]), and the DNN and LGBM models followed with AUROCs of 0.822 (0.789–0.856) and 0.813 (0.780–0.847), respectively; all machine learning AUROC values were higher than those obtained from disease-severity scores (AUROCs < 0.6). Although other definitions of hypotension were used such as a reduction of MAP ≥ 30 mmHg or a reduction occurring within 1 h, the AUROCs of machine learning models were higher than those of disease-severity scores. Machine learning models successfully predict hypotension after starting CRRT and can serve as the basis of systems to predict hypotension before starting CRRT.

Machine Learning Prediction Models for Mechanically Ventilated Patients: Analyses of the MIMIC-III Database

Background: Mechanically ventilated patients in the intensive care unit (ICU) have high mortality rates. There are multiple prediction scores, such as the Simplified Acute Physiology Score II (SAPS II), Oxford Acute Severity of Illness Score (OASIS), and Sequential Organ Failure Assessment (SOFA), widely used in the general ICU population. We aimed to establish prediction scores on mechanically ventilated patients with the combination of these disease severity scores and other features available on the first day of admission.Methods: A retrospective administrative database study from the Medical Information Mart for Intensive Care (MIMIC-III) database was conducted. The exposures of interest consisted of the demographics, pre-ICU comorbidity, ICU diagnosis, disease severity scores, vital signs, and laboratory test results on the first day of ICU admission. Hospital mortality was used as the outcome. We used the machine learning methods of k-nearest neighbors (KNN), logistic regression, bagging, decision tree, random forest, Extreme Gradient Boosting (XGBoost), and neural network for model establishment. A sample of 70% of the cohort was used for the training set; the remaining 30% was applied for testing. Areas under the receiver operating characteristic curves (AUCs) and calibration plots would be constructed for the evaluation and comparison of the models' performance. The significance of the risk factors was identified through models and the top factors were reported.Results: A total of 28,530 subjects were enrolled through the screening of the MIMIC-III database. After data preprocessing, 25,659 adult patients with 66 predictors were included in the model analyses. With the training set, the models of KNN, logistic regression, decision tree, random forest, neural network, bagging, and XGBoost were established and the testing set obtained AUCs of 0.806, 0.818, 0.743, 0.819, 0.780, 0.803, and 0.821, respectively. The calibration curves of all the models, except for the neural network, performed well. The XGBoost model performed best among the seven models. The top five predictors were age, respiratory dysfunction, SAPS II score, maximum hemoglobin, and minimum lactate.Conclusion: The current study indicates that models with the risk of factors on the first day could be successfully established for predicting mortality in ventilated patients. The XGBoost model performs best among the seven machine learning models.

COVID-19 Associated Pulmonary Aspergillosis in Mechanically Ventilated Patients

Background COVID-19 associated pulmonary aspergillosis (CAPA) occurs in critically ill COVID-19 patients. Risks and outcomes remain poorly understood. Methods A retrospective cohort study of adult mechanically ventilated COVID-19 patients admitted to five Johns Hopkins hospitals was conducted between March and August 2020. CAPA was defined using composite clinical criteria. Fine and Gray competing risks regression was used to analyze clinical outcomes and multilevel mixed-effects ordinal logistic regression was used to compare longitudinal disease severity scores. Results Amongst the cohort of 396 people, 39 met criteria for CAPA. Compared to those without, patients with CAPA were more likely to have underlying pulmonary vascular disease (41% vs 21.6%, p=0.01), liver disease (35.9% vs 18.2%, p=0.02), coagulopathy (51.3% vs 33.1%, p=0.03), solid tumors (25.6% vs 10.9%, p=0.017), multiple myeloma (5.1% vs 0.3%, p=0.027), corticosteroid exposure during index admission (66.7% vs 42.6%, p=0.005), and had a lower BMI (median 26.6 vs 29.9, p=0.04). People with CAPA had worse outcomes as measured by ordinal severity of disease scores, requiring longer time to improvement (adjusted odds ratio 1.081.091.1, p&lt;0.001), and advancing in severity almost twice as fast (subhazard ratio, sHR 1.31.82.5, p&lt;0.001). People with CAPA were intubated twice as long as those without (sHR) 0.40.50.6, p&lt;0.001) and had a longer hospital length of stay [median (IQR) 41.1 (20.5, 72.4) vs 18.5 (10.7, 31.8), p&lt;0.001]. Conclusion CAPA is associated with poor outcomes. Attention towards preventative measures (screening and/or prophylaxis) is warranted in people with high risk of developing CAPA.

COMORBID INTERNAL DISEASES IN MILITARY SERVANTS WHO HAVE SENSED THE EXPOSURE OF EXTREME FACTORS OF MILITARY SERVICE

Objective: to analyze comorbid pathology by methods of its quantitative assessment in servicemen exposed to extreme factors of military service (a set of factors of radiation accidents, the influence of modern armed conflicts and modern combat trauma without blood loss). Materials and methods. Studies of comorbid pathology were performed in 613 servicemen who were treated at the clinical base of the Ukrainian Military Medical Academy (UMMA) in National Military Medical Clinical Center «Main Military Clinical Hospital» NMMCC «MMCH» during 1989–2018 years. Soldiers who suffered from acute radiation sickness (ARS) in 1961 (n = 34), participants in the liquidation of the consequences of the Chornobyl catastrophe (PLCChC) 15 years after participating in the elimination of its consequences and in a later period (respectively PLCChC 1st group (n = 59) and the 2nd group (n = 337)). Soldiers are participants in the anti-terrorist operation (ATO)/Combined Forces (CFO) operation who did not receive modern combat injuries with blood loss (participants in the ATO/CFO, n = 183). All servicemen did not have any diseases limiting their fitness capabilities for military service before being exposed to extreme factors of military service. A cumulative CIRS scale was used to comprehensively assess comorbidity. Results. With increasing time after participation in the elimination of the consequences of radiation accidents, the course of arterial hypertension (AH) is aggravated, but the dose-dependence of the severity of AH on the received radiation dose has not been established. The easier course of hypertension in ATO/CFO servicemen compared to ARS remote servicemen and group 2 PLCChC servicemen can be explained by their younger age at the time of the survey and less time after exposure to extreme military service factors. In the military in the remote period after exposure to a complex of factors of radiation accidents, the frequency of diagnosing diseases by individual organs and body system increases comparing to non-irradiated servicemen. The total number of disease severity scores on the cumulative scale of CIRS diseases is also higher. However, a dose-dependent effect of the severity of comorbid pathology was also not found. These data indicate a higher prevalence of comorbid pathology in servicemen affected by a complex of factors of radiation accf5idents, compared with participants in the anti-terrorist operation / environmental protection. However, the lower severity of comorbid pathology in ATO/CFO participants can also be explained by their younger age at the time of the survey and less time after exposure to extreme factors of military service. Conclusions. For servicemen, with increasing time after participation in the elimination of the consequences of radiation accidents, the course of hypertension without its dose dependence becomes more difficult. The total number of disease severity scores on the cumulative scale of CIRS diseases in servicemen in the remote period after participation in the elimination of the consequences of radiation accidents is higher than in non-irradiated servicemen. However, a dose-dependent effect of the severity of comorbid pathology was also not found. Key words: servicemen, participants of liquidation of consequences of the Chornobyl catastrophe, participants of anti-terrorist operation / operation of the Joint Forces, radiation accidents, acute radiation sickness, ionizing radiation, comorbid pathology, chronic diseases.

Appropriateness of Using Vitamin K for the Correction of INR Elevation Secondary to Hepatic Disease in Critically ill Patients: An Observational Study

Background Using vitamin K for correction of coagulopathy in critically ill patients is controversial with limited evidence. This study aims to evaluate the efficacy and safety of vitamin K in the correction of international normalized ratio (INR) elevation secondary to liver disease in critically ill patients. Method A retrospective study of critically ill patients with coagulopathy secondary to liver disease. The primary outcome was to evaluate the association between vitamin K administration and the incidence of new bleeding events in critically ill patients with INR elevation; other outcomes were considered secondary. Patients were categorized into two groups based on vitamin K administration to correct INR elevation. The propensity score was generated based on disease severity scores and the use of pharmacological DVT prophylaxis. Results A total of 98 patients were included in the study. Forty-seven patients (48%) received vitamin K during the study period. The odds of the new bleeding event was not statistically different between groups (OR 2.4, 95% CI 0.28-21.67, P = .42). Delta of INR reduction was observed with a median of 0.63 when the first dose is given ( P-value: <.0001). However the INR reduction with other subsequent doses of vitamin K was not statistically significant. Conclusion The administration of vitamin K for INR correction in critically ill patients with coagulopathy secondary to liver disease was not associated with a lower odds of new bleeding events. Further studies are needed to assess the value of vitamin K administration in critically ill patients with liver diseases related coagulopathy.

Machine Learning Model to Predict Hypotension After Starting Continuous Renal Replacement Therapy

Background: Hypotension after starting continuous renal replacement therapy (CRRT) is associated with worse outcome, but it is difficult to predict because several factors have interactive and complex effects on the risk. The present study applied machine learning algorithms to develop models to predict hypotension after initiating CRRT.Methods: Among 2,349 adult patients who started CRRT due to acute kidney injury, 70% and 30% were randomly assigned into the training and testing sets, respectively. Hypotension was defined as a reduction in mean arterial pressure (MAP) ≥20 mmHg from the initial value within 6 hours. The area under the receiver operating characteristic curves (AUROCs) in machine learning models, such as support vector machine (SVM), deep neural network (DNN), and light gradient boosting machine (LGBM), were compared with those in disease-severity scores such as the Sequential Organ Failure Assessment and Acute Physiology and Chronic Health Evaluation II.Results: The DNN model showed the highest AUROC (0.822 [0.789–0.856]), and the LGBM and SVM models followed with AUROCs of 0.810 (0.776–0.845) and 0.807 (0.772–0.842), respectively; all machine learning AUROC values were higher than those obtained from disease-severity scores (AUROCs <0.6). Although different definitions of hypotension were used such as a reduction of MAP ≥30 mmHg or a reduction occurring within 1 hour, the AUROCs of machine learning models were higher than those of disease-severity scores. These machine learning models were well calibrated.Conclusion: Machine learning models successfully predict hypotension after starting CRRT and can serve as the basis of systems to predict hypotension before starting CRRT.

Serum levels of copper, zinc and disease severity scores in sickle cell disease patients in Benin City, Nigeria

Background: Micronutrient deficiency is recognized in sickle cell anaemia (SCA) but it is not known for certain whether changes in zinc, copper and copper-to-zinc ratio are associated with Sickle cell disease severity scores. Objective: To compare serum levels of copper, zinc and copper-to-zinc ratio in SCA subjects with control group and correlate the variables with objective disease severity scores. Methods: Serum copper and zinc were determined in 100 SCA patients and 50 controls using kits supplied by Centronic, Germany. Unpaired Students’t-test was used to compare the variables between SCA patients in steady clinical state, vaso-occlusive crisis and controls, while Spearman correlation coefficient was used to associate the parameters with disease severity scores. Results: Serum copper level was higher (P=0.008) in SCA patients than controls, while serum zinc level was lower (P<0.001) in SCA patients than controls. The copper/zinc ratio was higher (P<0.001) in SCA patients than controls. Significantly higher (P<0.001) copper and lower (P<0.001) zinc levels were observed in patients in vaso-occlusive crisis than in steady clinical state. Zinc correlated inversely (r=-0.2743; P=0.006) while copper-to-zinc ratio correlated positively with disease severity scores. Conclusion: Copper-to-zinc ratio may be an indicator of disease severity in SCA patients.Keywords: Copper/zinc ratio, disease severity score, sickle cell anaemia.