The association between sodium fluctuations and mortality in surgical patients requiring intensive care

Effect of long-term serum sodium levels on the prognosis of patients on maintenance hemodialysis

Abnormal serum Na (SNa) levels are common in patients with chronic kidney disease (CKD) which is associated with increased morbidity and mortality. There are relatively few studies on the effect of SNa indicators on the prognosis of patients undergoing maintenance hemodialysis (MHD). We aim to investigate the effect of long-term SNa levels on the survival and prognosis of patients undergoing hemodialysis (HD). Newly entered HD patients in the registration system of Zhejiang Provincial Dialysis Quality Control Center between January 1, 2010 and December 31, 2019 were included and followed up until December 31, 2020. Multiple sodium levels were collected from patients, defining long-term SNa as the mean of multiple SNa, according to which patients were grouped, with the prognostic differences between subgroups compared by Kaplan-Meier modeling and multifactorial Cox regression modeling. Finally, a total of 21,701 patients were included in this study and Cox regression showed that decreased SNa levels (Na < 135 mmol/L, HR = 1.704, 95% CI 1.408-2.063, p < 0.001; 135≦Na≦137.5 mmol/L, HR = 1.127,95% CI 1.016-1.250, p = 0.024) and elevated SNa levels (142.5 < Na≦145mmol/L, HR = 1.198, 95% CI 1.063-1.350, p = 0.003; Na > 145mmol/L, HR = 2.150, 95% CI 1.615-2.863, p < 0.001) were all independent risk factors for all-cause mortality in MHD patients.

The association between nutritional status measured by body mass index and outcomes in the pediatric intensive care unit

Aim/Introduction

The relationship between nutritional status upon admission to a pediatric intensive care unit (PICU) and clinical outcomes remains unclear. We examined the relationship between nutrition status, as indicated by body mass index-for-age (BMI-for-age), and clinical outcomes in the PICU.

Method

In this retrospective study at a tertiary care center, records of 1,015 critically ill children and adolescents aged one month to 18 years old with available anthropometric parameters were included. The nutritional status upon admission was determined by calculating the BMI-for-age z-score using the WHO growth charts as the reference. The participants were categorized as underweight (BMI-for-age z-score < -2), normal weight (-2 ≤ BMI-for-age z-score ≤ +1), and overweight/obese (BMI-for-age z-score > +1). Multi-variate odds ratios (OR) with 95% confidence intervals (CI) were used to investigate the association between malnutrition (being underweight and overweight/obese) and odds of Prolonged PICU stay (≥7 days) and PICU mortality after controlling for descriptive characteristics, Glasgow Coma Scale score status, fluctuations in serum sodium, and acute kidney injury confounders.

Results

The proportions of patients in underweight, normal weight, and overweight/obese categories were 34.2%, 45.8%, and 20%, respectively. During the study period, 21.5% of patients had prolonged PICU stay, and 5.6% of patients in PICU died. Compared to normal-weight patients, underweight patients had higher odds of prolonged PICU stay (OR: 1.52; 95% CI: 1.05-2.22) and PICU mortality (OR: 2.12; 95% CI: 1.22-4.01). Age- and gender-stratified full-adjusted analysis showed that the increased odds of prolonged PICU stay remained significant among underweight boys and underweight individuals aged 5-19 years old. Furthermore, the increased odds of PICU mortality remained significant among underweight individuals aged 2-5 years old. However, being overweight or obese during PICU admission did not demonstrate a significant association with our outcomes in the total sample or subgroup analysis.

Conclusion

Our findings showed that PICU patients who were underweight had higher odds of prolonged PICU stay and PICU mortality than their normal-weight counterparts. This underscores the importance of closely monitoring underweight patients in the PICU upon admission in order to improve clinical outcomes.

Development and validation of potential phenotypes of serum electrolyte disturbances in critically ill patients and a Web-based application

BACKGROUND

Electrolyte disturbances are highly heterogeneous and severely affect the prognosis of critically ill patients. Our study was to determine whether data-driven phenotypes of seven electrolytes have prognostic relevance in critically ill patients.

METHODS

We extracted patient information from three large independent public databases, and clustered the electrolyte distribution of ICU patients based on the extreme value, median value and coefficient of variation of electrolytes. Three plausible clinical phenotypes were calculated using K-means clustering algorithm as the basic clustering method. MIMIC-IV was considered a training set, and two others have been designated as verification set. The robustness of the model was then validated from different angles, providing dynamic and interactive visual charts for more detailed characterization of phenotypes.

RESULTS

15,340, 12,445 and 2147 ICU patients with electrolyte records during early ICU stay in MIMIC-IV, eICU-CRD and AmsterdamUMCdb were enrolled. After clustering, three reasonable and interpretable phenotypes are defined as α, β and γ according to the order of clusters. The α and γ phenotype, with significant differences in electrolyte distribution and clinical variables, higher 28-day mortality and longer length of ICU stay (p < 0.001), was further demonstrated by robustness analysis. The α phenotype has significant kidney injury, while the β phenotype has the best prognosis. In addition, the assignment methods of the three phenotypes were developed into a web-based tool for further verification and application.

CONCLUSIONS

Three different clinical phenotypes were identified that correlated with electrolyte distribution and clinical outcomes. Further validation and characterization of these phenotypes is warranted.

Plasma sodium during the recovery of renal function in critically ill adult patients: Multicenter prospective cohort study

BACKGROUND

Sodium increases during acute kidney injury (AKI) recovery. Both hypernatremia and positive fluid balances are associated with increased mortality. We aimed to evaluate the association between daily fluid balance and daily plasma sodium during the recovery from AKI among critical patients.

METHODS

Adult patients with AKI were enrolled in four ICUs and followed up for four days or until ICU discharge or hemodialysis initiation. Day zero was the peak day of creatinine. The primary outcome was daily plasma sodium; the main exposure was daily fluid balance.

RESULTS

93 patients were included. The median age was 66 years; 68% were male. Plasma sodium increased in 79 patients (85%), and 52% presented hypernatremia. We found no effect of daily fluid balance on plasma sodium (β -0.26, IC95%: -0.63-0.13; p = 0.19). A higher total sodium variation was observed in patients with lower initial plasma sodium (β -0.40, IC95%: -0.53 to -0.27; p < 0.01), higher initial urea (β 0.07, IC95%: 0.04-0.01; p < 0.01), and higher net sodium balance (β 0.002, IC95%: 0.0001-0.01; p = 0.05).

CONCLUSIONS

The increase in plasma sodium is common during AKI recovery and can only partially be attributed to the water and electrolyte balances. The incidence of hypernatremia in this population of patients is higher than in the general critically ill patient population.

Effect of fosfomycin-induced hypernatremia on patients' hospital stay length and survival

BACKGROUND AND OBJECTIVE

Hypernatremia is a possible side effect of intravenous fosfomycin. The aim of this study was to investigate the effects of changes in sodium (Na) levels on hospital stay and survival in patients hospitalized in the intensive care unit receiving fosfomycin.

SUBJECTS AND METHODS

This study was conducted retrospectively on the files of patients over the age of 60, who were admitted to the Internal Medicine Intensive Care Unit. Plasma sodium levels were observed and documented over a period of 14 days. The patients were divided into two groups (Hypernatremia group Na > 145 mEq/L vs normonatremia group 135-145 mEq/L). In addition, daily sodium changes were noted for 14 days in patients.

RESULTS

The mean age of the patients was 75 years. Hospitalization days were longer for hypernatremia patients (31.5 days vs 41 days, p = 0.003). Patients with hypernatremia had an extended duration of stay in the intensive care unit. (21 days vs 31 days p = 0.002). The 1-month survival rate was 61.4% in patients with hypernatremia and 24.9% in patients without hypernatremia (p = 0.004). The absence of hypernatremia increases mortality by 2.09 times (95% CI 1.35-3.23). When discharge and mortality rates were analyzed according to sodium fluctuation, discharged patients exhibited a lower sodium fluctuation (4 min/max (-10/19) vs 6 min/max (-16/32) p < 0.001).

CONCLUSION

In conclusion, the strength of our study is that it specifically focuses on the consequences of the sodium fluctuation on patient management and provides results.

Association between serum sodium trajectory and mortality in patients with acute kidney injury: a retrospective cohort study

INTRODUCTION

Dysnatremia is strongly associated with poor prognosis in acute kidney injury (AKI); however, the impact of sodium trajectories on the prognosis of patients with AKI has not yet been well elucidated. This study aimed to assess the association between sodium trajectories in patients with AKI and mortality at 30-day and 1-year follow-up.

METHODS

This retrospective cohort study used data from Medical Information Mart for Intensive Care (MIMIC)-IV database, and patients diagnosed with AKI within 48 h after admission were enrolled. Group-based trajectory models (GBTM) were applied to map the developmental course of the serum sodium fluctuations. Kaplan-Meier survival curve was used to compare differences in mortality in AKI patients with distinct serum sodium trajectories. Hazard ratios (HRs) were calculated to determine the association between trajectories and prognosis using Cox proportional hazard models.

RESULTS

A total of 9,314 AKI patients were enrolled. Three distinct sodium trajectories were identified including: (i) stable group (ST, in which the serum sodium levels remained relatively stable, n = 4,935; 53.0%), (ii) descending group (DS, in which the serum sodium levels declined, n = 2,994; 32.15%) and (iii) ascending group (AS, in which the serum sodium levels were elevated, n = 1,383; 14.85%). There was no significant difference in age and gender distribution among the groups. The 30-day mortality rates were 7.9% in ST, 9.5% in DS and 16.6% in AS (p < 0.001). The results of 1-year mortality rates were similar (p < 0.001). In adjusted analysis, patients in the DS (HR = 1.22, 95% confidence interval [CI], 1.04-1.43, p = 0.015) and AS (HR = 1.68, 95% CI, 1.42-2.01, p = 0.013) groups had higher risks of 30-day mortality compared to those in the ST group.

CONCLUSION

In patients with AKI, the serum sodium trajectories were independently associated with 30-day and 1-year mortality. Association between serum sodium level trajectories and prognosis in patients with AKI deserve further study.

CARDS, a Novel Prognostic Index for Risk Stratification and In-Hospital Monitoring

Background: Sodium fluctuation is independently associated with clinical deterioration. We developed and validated a prognostic index based on sodium fluctuation for risk stratification and in-hospital monitoring. Methods: This study included 33,323 adult patients hospitalized at a tertiary care hospital in 2014. The first 28,279 hospitalizations were analyzed to develop the model and then the validity of the model was tested using data from 5044 subsequent hospitalizations. We predict in-hospital mortality using age, comorbidity, range of sodium fluctuation, and duration of sodium fluctuation, abbreviated as CARDS. Results: In-hospital mortality was similar in the derivation (0.6%) and validation (0.4%) cohorts. In the derivation cohort, four independent risk factors for mortality were identified using logistic regression: age (66-75, 2 points; >75, 3 points); Charlson comorbidity index (>2, 5 points); range of sodium fluctuation (7-10, 4 points; >10, 10 points); and duration of fluctuation (≤3, 3 points). The AUC was 0.907 (95% CI: 0.885-0.928) in the derivation cohort and 0.932 (95% CI: 0.895-0.970) in the validation cohort. In the derivation cohort, in-hospital mortality was 0.106% in the low-risk group (0-7 points), 1.076% in the intermediate-risk group (8-14 points), and 8.463% in the high-risk group (15-21 points). In the validation cohort, in-hospital mortality was 0.049% in the low-risk group, 1.064% in the intermediate-risk group, and 8.403% in the high-risk group. Conclusions: These results suggest that patients at low, intermediate, and high risk for in-hospital mortality may be identified by CARDS mainly based on sodium fluctuation.

A Retrospective Cohort Study: Predicting 90-Day Mortality for ICU Trauma Patients with a Machine Learning Algorithm Using XGBoost Using MIMIC-III Database

Objective

The aim of this study was to develop and validate a machine learning-based predictive model that predicts 90-day mortality in ICU trauma patients.

Methods

Data of patients with severe trauma were extracted from the Medical Information Mart for Intensive Care III (MIMIC-III) database. The performances of mortality prediction models generated using nine machine learning extreme gradient boosting (XGBoost), logistic regression, random forest, AdaBoost, multilayer perceptron (MLP) neural networks, support vector machine (SVM), light gradient boosting machine (GBM), k nearest neighbors (KNN) and gaussian naive bayes (GNB). The performance of the model was evaluated in terms of discrimination, calibration and clinical application.

Results

We found that the accuracy, sensitivity, specificity, PPV, NPV and F1 score of our proposed XGBoost model were 82.8%, 79.7%, 77.6%, 51.2%, 91.5% and 0.624, respectively. Among the nine models, the XGBoost model performed best. Compared with traditional logistic regression, the calibration curves of the XGBoost model and decision curve analysis (DCA) performed well.

Conclusion

Our study shows that the XGBoost model outperforms other machine learning models in predicting 90-day mortality in trauma patients. It can be used to assist clinicians in the early identification of mortality risk factors and early intervention to reduce mortality.

Risk stratification and predictive value of serum sodium fluctuation for adverse prognosis in acute coronary syndrome patients

BACKGROUND

Serum sodium fluctuation (SF) as an indicator of the extent of changes in serum sodium is associated with increased mortality in hospitalized patients. However, there is no consensus on diagnostic criteria for SF, and its impact on the outcome of patients with acute coronary syndrome (ACS) remains uncertain. We defined SF and assessed its association with adverse prognosis in hospitalized ACS patients.

METHODS

Patients diagnosed with ACS were consecutively recruited. The serum SF rate (SFR) was defined as the ratio of the difference between the highest and lowest serum sodium levels during hospitalization to the initial serum sodium level on admission. The Cox proportional hazards model was performed to evaluate the association between SFR and mortality. The dose-response relationships of SFR with mortality was characterized by restricted cubic splines (RCS) model. The predictive performance of SF for mortality was assessed by the area under the receiver operating characteristic curves (AUCs).

RESULTS

The study retrospectively enrolled 1856 ACS patients, of which 36 (1.94%) patients dead within 1 year. Multivariate Cox analysis showed that SFR was independently associated with higher risk of 1-year mortality (HR = 1.17, 95% CI: 1.111-1.244, P < 0.001). RCS analysis showed the optimal threshold for SFR was 5%, and the 1-year cumulative mortality was higher in the abnormal SF group (SFR ≥ 5%) compared with the normal SF group (SFR < 5%, P < 0.01). The AUCs of SF for predicting mortality within 1 month, 6 months, and 1 year were 0.842 (95% CI: 0.781-0.904), 0.830 (95% CI:0.736-0.926), 0.703 (95% CI:0.595--0.811), respectively. Even in patients with normal baseline serum sodium, abnormal SF group demonstrated a significantly higher 1-year mortality compared to normal SF group (HR = 4.955, 95% CI: 1.919-12.795).

CONCLUSION

The SFR during hospitalization is an adequate predictor of adverse outcomes in ACS patients, independent of serum sodium level at admission. Additional research is warranted to ascertain whether interventions targeting SF confer measurable clinical benefits.

Euvolaemic hyponatraemia as a rare first presentation of chronic hypopituitarism

A man in his late 40s with no significant medical history presented with 2 weeks of lethargy, nausea and dizziness, alongside worsening headaches. Initial assessment revealed severe hyponatraemia and secondary hypothyroidism; urgent MRI pituitary was requested with a clinical suspicion of pituitary apoplexy. This demonstrated a likely cystic pituitary adenoma, with further testing revealing pituitary gland suppression, leading to a diagnosis of chronic secondary hypopituitarism. Initiating hormone replacement allowed substantial reported improvements in this patient's quality of life.A review of the patient's work-up revealed areas in which best practice was not followed. Cortisol measurements and paired urinary and serum osmolalities were initially not sent, nor results appropriately chased. A subsequent literature review identified that conformation with national and local guidelines on hyponatraemia management is poor. This patient's case, when combined with the literature review, provides evidence to support methods to increase educational awareness of an appropriate work-up of hyponatraemia among clinicians.

Nomogram Prediction Model of Hypernatremia on Mortality in Critically Ill Patients

Objective

To investigate the value of hypernatremia in the intensive care unit (ICU) for the risk prediction of mortality in severe patients.

Methods

Clinical data of critically ill patients admitted to the ICU of Beijing Friendship Hospital, were collected for retrospective analysis. Univariate and multivariate logistic regression analyses were employed to analyze the influencing factors. Nomograms predicting the mortality were constructed with R software and validated with repeated sampling.

Results

A total of 442 cases were eligible for this study. Hypernatremia within 48 hours of ICU admission, change in sodium concentration (CNa+) within 48 hours, septic shock, APACHE II score, hyperlactatemia within 48 hours, use of continuous renal replacement therapy (CRRT) within 48 hours, and the use of mechanical ventilation (MV) within 48 hours of ICU admission were all identified as independent risk factors for death within 28 days of ICU admission. These predictors were included in a nomogram of 28-day mortality in severe patients, which was constructed using R software.

Conclusion

The nomogram could predict the individualized risk of 28-day mortality based on the above factors. The model has better discrimination and accuracy and has high clinical application value.

Association between serum sodium and in-hospital mortality among critically ill patients with spontaneous subarachnoid hemorrhage

Objective

The aim of this study was to retrospectively explore the relationship between serum sodium and in-hospital mortality and related factors in critically ill patients with spontaneous subarachnoid hemorrhage (SAH).

Methods

Data were collected from the Medical Information Mart for Intensive Care IV database. Restricted cubic splines were used to explore the relationship between serum sodium and in-hospital mortality. Receiver operating characteristic analysis was used to calculate the optimal cutoff value of sodium fluctuation, and decision curve analysis was plotted to show the net benefit of different models containing serum sodium.

Results

A total of 295 patients with spontaneous SAH were included in the retrospective analysis. The level of sodium on ICU admission and minimum sodium in the ICU had a statistically significant non-linear relationship with in-hospital mortality (non-linear P-value < 0.05, total P-value < 0.001). Serum sodium on ICU admission, minimum serum sodium during ICU, and sodium fluctuation were independently associated with in-hospital mortality with odds ratios being 1.23 (95% confidence interval (CI): 1.04-1.45, P = 0.013), 1.35 (95% CI: 1.18-1.55, P < 0.001), and 1.07 (95% CI: 1.00-1.14, P = 0.047), respectively. The optimal cutoff point was 8.5 mmol/L to identify in-hospital death of patients with spontaneous SAH with sodium fluctuation, with an AUC of 0.659 (95% CI 0.573-0.744).

Conclusion

Among patients with spontaneous SAH, we found a J-shaped association between serum sodium on ICU admission and minimum sodium values during ICU with in-hospital mortality. Sodium fluctuation above 8.5 mmol/L was independently associated with in-hospital mortality. These results require being tested in prospective trials.

The Prognostic Importance of Serum Sodium for Mortality among Critically Ill Patients Requiring Continuous Renal Replacement Therapy

BACKGROUND

Serum sodium derangement is common in critically ill patients requiring continuous renal replacement therapy (CRRT). We aimed to assess the association between serum sodium before and during CRRT with mortality.

METHODS

This is a historical cohort study of 1,520 critically ill patients receiving CRRT from December 2006 through November 2015 in a tertiary hospital in the United States. Using logistic regression analysis, we used serum sodium before CRRT, mean serum sodium, and serum sodium changes during CRRT to predict 90-day mortality after CRRT initiation.

RESULTS

Compared with the normal serum sodium levels, the odds ratio (OR) of 90-day mortality in patients with serum sodium before CRRT of 143-147 and ≥148 mmol/L were 1.45 (95% CI 1.03-2.05) and 2.24 (95% CI 1.33-3.87), respectively. There was no significant increase in 90-day mortality in serum sodium of ≤137 mmol/L. During CRRT, the mean serum sodium levels of ≤137 (OR 1.41; 95% CI 1.01-1.98) and ≥143 mmol/L (OR 1.52; 95% CI 1.14-2.03) were associated with higher 90-day mortality. The greater serum sodium changes during CRRT were associated with higher 90-mortality (OR 1.35; 95% CI 1.21-1.51 per 5-mmol/L increase).

CONCLUSION

Before CRRT initiation, hypernatremia and during CRRT, hypo- and hypernatremia were associated with increased mortality.

Serum sodium variability and acute kidney injury: a retrospective observational cohort study on a hospitalized population

Aim of our study was to analyze the association between serum sodium (Na) variability and acute kidney injury (AKI) development. We performed a retrospective observational cohort study on the inpatient population admitted to Fondazione Policlinico Universitario A. Gemelli IRCCS between January 1, 2010 and December 31, 2014 with inclusion of adult patients with ≥ 2 Na and ≥ 2 serum creatinine measurements. We included only patients with ≥ 2 Na measurements before AKI development. The outcome of interest was AKI. The exposures of interest were hyponatremia, hypernatremia and Na fluctuations before AKI development. Na variability was evaluated using the coefficient of variation (CV). Multivariable Cox proportional hazards and logistic regression models were fitted to obtain hazard ratios (HRs), odds ratios (ORs) and 95% confidence intervals (CIs) for the association between the exposures of interest and AKI. Overall, 56,961 patients met our inclusion criteria. During 1541 person-years of follow-up AKI occurred in 1450 patients. In multivariable hazard models, patients with pre-existent dysnatremia and those who developed dysnatremia had a higher risk of AKI compared with patients with normonatremia. Logistic models suggested a higher risk for AKI in the 3rd (OR 1.41, 95% CI 1.18, 1.70, p < 0.001) and 4th (OR 1.53, 95% CI 1.24, 1.91, p < 0.001) highest quartiles of Na CV with a significant linear trend across quartiles (p trend < 0.001). This association was also independent from Na highest and lowest peak value. Dysnatremia is a common condition and is positive associated with AKI development. Furthermore, high Na variability might be considered an independent early indicator for kidney injury development.

Increased mortality risk associated with serum sodium variations and borderline hypo- and hypernatremia in hospitalized adults

Background

This study aimed to evaluate short-term and long-term mortalities in a cohort of unselected hospitalized patients with serum sodium concentration ([Na⁺]) variations within and outside of reference range.

Methods

All adult patients admitted to the Mayo Clinic, Rochester, MN, USA from January 2011 to December 2013 (n = 147358) were retrospectively screened. Unique patients admitted during the study period were examined. The main exposure was serum [Na⁺] variation. Outcome measures were hospital and 1-year all-cause mortalities.

Results

A total of 60944 patients, mean age 63 ± 17 years, were studied. On admission, 17% (n = 10066) and 1.4% (n = 852) had hypo- and hypernatremia, respectively. During the hospital stay, 11044 and 4128 developed hypo- and hypernatremia, respectively, accounting for 52.3 and 82.9% of the total hypo- and hypernatremic patients. Serum [Na⁺] variations of ≥6 mEq/L occurred in 40.6% (n = 24 740) of the 60 944 patients and were significantly associated with hospital and 1-year mortalities after adjusting potential confounders (including demographics, comorbidities, estimated glomerular filtration rate, admission serum [Na⁺], number of [Na⁺] measurements and length of hospital stay). Adjusted odds ratios for hospital and 1-year mortalities increased with increasing [Na⁺] variations in a dose-dependent manner, from 1.47 to 5.48 (all 95% confidence intervals >1.0). Moreover, in fully adjusted models, [Na⁺] variations (≥6 mEq/L) within the reference range (135–145 mEq/L) or borderline hypo- or hypernatremia (133–137 and 143–147 mEq/L, respectively) compared with 138–142 mEq/L were associated with increased hospital and 1-year mortalities.

Conclusion

In hospitalized adults, [Na⁺] fluctuation (≥6 mEq/L) irrespective of admission [Na⁺] and borderline hypo- or hypernatremia are independent predictors of progressively increasing short- and long-term mortality burdens.

A statistically rigorous deep neural network approach to predict mortality in trauma patients admitted to the intensive care unit

BACKGROUND

Trauma patients admitted to critical care are at high risk of mortality because of their injuries. Our aim was to develop a machine learning-based model to predict mortality using Fahad-Liaqat-Ahmad Intensive Machine (FLAIM) framework. We hypothesized machine learning could be applied to critically ill patients and would outperform currently used mortality scores.

METHODS

The current Deep-FLAIM model evaluates the statistically significant risk factors and then supply these risk factors to deep neural network to predict mortality in trauma patients admitted to the intensive care unit (ICU). We analyzed adult patients (≥18 years) admitted to the trauma ICU in the publicly available database Medical Information Mart for Intensive Care III version 1.4. The first phase selection of risk factor was done using Cox-regression univariate and multivariate analyses. In the second phase, we applied deep neural network and other traditional machine learning models like Linear Discriminant Analysis, Gaussian Naïve Bayes, Decision Tree Model, and k-nearest neighbor models.

RESULTS

We identified a total of 3,041 trauma patients admitted to the trauma surgery ICU. We observed that several clinical and laboratory-based variables were statistically significant for both univariate and multivariate analyses while others were not. With most significant being serum anion gap (hazard ratio [HR], 2.46; 95% confidence interval [CI], 1.94-3.11), sodium (HR, 2.11; 95% CI, 1.61-2.77), and chloride (HR, 2.11; 95% CI, 1.69-2.64) abnormalities on laboratories, while clinical variables included the diagnosis of sepsis (HR, 2.03; 95% CI, 1.23-3.37), Quick Sequential Organ Failure Assessment score (HR, 1.52; 95% CI, 1.32-3.76). And Systemic Inflammatory Response Syndrome criteria (HR. 1.41; 95% CI, 1.24-1.26). After we used these clinically significant variables and applied various machine learning models to the data, we found out that our proposed DNN outperformed all the other methods with test set accuracy of 92.25%, sensitivity of 79.13%, and specificity of 94.16%; positive predictive value, 66.42%; negative predictive value, 96.87%; and area under the curve of the receiver-operator curve of 0.91 (1.45-1.29).

CONCLUSION

Our novel Deep-FLAIM model outperformed all other machine learning models. The model is easy to implement, user friendly and with high accuracy.

LEVEL OF EVIDENCE

Prognostic study, level II.

A deep learning backcasting approach to the electrolyte, metabolite, and acid-base parameters that predict risk in ICU patients

BACKGROUND

A powerful risk model allows clinicians, at the bedside, to ensure the early identification of and decision-making for patients showing signs of developing physiological instability during treatment. The aim of this study was to enhance the identification of patients at risk for deterioration through an accurate model using electrolyte, metabolite, and acid-base parameters near the end of patients' intensive care unit (ICU) stays.

METHODS

This retrospective study included 5157 adult patients during the last 72 hours of their ICU stays. The patients from the MIMIC-III database who had serum lactate, pH, bicarbonate, potassium, calcium, glucose, chloride, and sodium values available, along with the times at which those data were recorded, were selected. Survivor data from the last 24 hours before discharge and four sets of nonsurvivor data from 48-72, 24-48, 8-24, and 0-8 hours before death were analyzed. Deep learning (DL), random forest (RF) and generalized linear model (GLM) analyses were applied for model construction and compared in terms of performance according to the area under the receiver operating characteristic curve (AUC). A DL backcasting approach was used to assess predictors of death vs. discharge up to 72 hours in advance.

RESULTS

The DL, RF and GLM models achieved the highest performance for nonsurvivors 0-8 hours before death versus survivors compared with nonsurvivors 8-24, 24-48 and 48-72 hours before death versus survivors. The DL assessment outperformed the RF and GLM assessments and achieved discrimination, with an AUC of 0.982, specificity of 0.947, and sensitivity of 0.935. The DL backcasting approach achieved discrimination with an AUC of 0.898 compared with the DL native model of nonsurvivors from 8-24 hours before death versus survivors with an AUC of 0.894. The DL backcasting approach achieved discrimination with an AUC of 0.871 compared with the DL native model of nonsurvivors from 48-72 hours before death versus survivors with an AUC of 0.846.

CONCLUSIONS

The DL backcasting approach could be used to simultaneously monitor changes in the electrolyte, metabolite, and acid-base parameters of patients who develop physiological instability during ICU treatment and predict the risk of death over a period of hours to days.

Analysis of adult disease characteristics and mortality on MIMIC-III

Purpose

To deeply analyze the basic information and disease information of adult patients in the MIMIC-III (Medical Information Mart for Intensive Care III) database, and provide data reference for clinicians and researchers.

Materials and methods

Tableau2019.1.0 and Navicat12.0.29 were used for data analysis and extraction of disease distribution of adult patients in the MIMIC-III database.

Result

A total of 38,163 adult patients were included in the MIMIC-III database. Only 38,156 patients with the first diagnosis were selected. Among them, 21,598 were males accounting for 56.6% the median age was 66 years (Q1-Q3: 53–78), the median length of a hospital stay was 7 days (Q1-Q3: 4–12), and the median length of an ICU stay was 2.1 days (Q1-Q3: 1.2–4.1). Septicemia was the disease with the highest mortality rate among patients and the total mortality rate was 48.9%. The disease with the largest number of patients at the last time was other forms of chronic ischemic heart disease.

Conclusion

By analyzing the patients’ basic information, the admission spectrum and the disease morbidity and mortality can help more researchers understand the MIMIC-III database and facilitate further research.

Trajectories of Serum Sodium on In-Hospital and 1-Year Survival among Hospitalized Patients

BACKGROUND AND OBJECTIVES

This study aimed to investigate the association between in-hospital trajectories of serum sodium and risk of in-hospital and 1-year mortality in patients in hospital.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

This is a single-center cohort study. All adult patients who were hospitalized from years 2011 through 2013 who had available admission serum sodium and at least three serum sodium measurements during hospitalization were included. The trend of serum sodium during hospitalization was analyzed using group-based trajectory modeling; the five main trajectories were grouped as follows: (1) stable normonatremia, (2) uncorrected hyponatremia, (3) borderline high serum sodium, (4) corrected hyponatremia, and (5) fluctuating serum sodium. The outcome of interest was in-hospital mortality and 1-year mortality. Stable normonatremia was used as the reference group for outcome comparison.

RESULTS

A total of 43,539 patients were analyzed. Of these, 47% had stable normonatremia, 15% had uncorrected hyponatremia, 31% had borderline high serum sodium, 3% had corrected hyponatremia, and 5% had fluctuating serum sodium trajectory. In adjusted analysis, there was a higher in-hospital mortality among those with uncorrected hyponatremia (odds ratio [OR], 1.33; 95% CI, 1.06 to 1.67), borderline high serum sodium (OR, 1.66; 95% CI, 1.38 to 2.00), corrected hyponatremia (OR, 1.50; 95% CI, 1.02 to 2.20), and fluctuating serum sodium (OR, 4.61; 95% CI, 3.61 to 5.88), compared with those with the normonatremia trajectory. One-year mortality was higher among those with uncorrected hyponatremia (hazard ratio [HR], 1.28; 95% CI, 1.19 to 1.38), borderline high serum sodium (HR, 1.18; 95% CI, 1.11 to 1.26), corrected hyponatremia (HR, 1.24; 95% CI, 1.08 to 1.42), and fluctuating serum sodium (HR, 2.10; 95% CI, 1.89 to 2.33) compared with those with the normonatremia trajectory.

CONCLUSIONS

More than half of patients who had been hospitalized had an abnormal serum sodium trajectory during hospitalization. This study demonstrated that not only the absolute serum sodium levels but also their in-hospital trajectories were significantly associated with in-hospital and 1-year mortality. The highest in-hospital and 1-year mortality risk was associated with the fluctuating serum sodium trajectory.

PODCAST

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2020_03_25_CJN.12281019.mp3.

Sodium Fluctuations and Mortality in a General Hospitalized Population

Background/Aims: Aim of our study was to describe the association between natremia (Na) fluctuation and hospital mortality in a general population admitted to a tertiary medical center. Methods: We performed a retrospective observational cohort study on the patient population admitted to the Fondazione Policlinico A. Gemelli IRCCS Hospital between January 2010 and December 2014 with inclusion of adult patients with at least 2 Na values available and with a normonatremic condition at hospital admission. Patients were categorized according to all Na values recorded during hospital stay in the following groups: normonatremia, hyponatremia, hypernatremia, and mixed dysnatremia. The difference between the highest or the lowest Na value reached during hospital stay and the Na value read at hospital admission was used to identify the maximum Na fluctuation. Cox proportional hazards models were used to estimate hazard ratios (HRs) for in-hospital death in the groups with dysnatremias and across quartiles of Na fluctuation. Covariates assessed were age, sex, highest and lowest Na level, Charlson/Deyo score, cardiovascular diseases, cerebrovascular diseases, dementia, congestive heart failure, severe kidney disease, estimated glomerular filtration rate, and number of Na measurements during hospital stay. Results: 46,634 admissions matched inclusion criteria. Incident dysnatremia was independently associated with in-hospital mortality (hyponatremia: HR 3.11, 95% CI 2.53, 3.84, p < 0.001; hypernatremia: HR 5.12, 95% CI 3.94, 6.65, p < 0.001; mixed-dysnatremia: HR 4.94, 95% CI 3.08, 7.92, p < 0.001). We found a higher risk of in-hospital death by linear increase of quartile of Na fluctuation (p trend <0.001) irrespective of severity of dysnatremia (HR 2.34, 95% CI 1.55, 3.54, p < 0.001, for the highest quartile of Na fluctuation compared with the lowest). Conclusions: Incident dysnatremia is associated with higher hospital mortality. Fluctuation of Na during hospital stay is a prognostic marker for hospital death independent of dysnatremia severity.

Critical Care, Critical Data

As big data, machine learning, and artificial intelligence continue to penetrate into and transform many facets of our lives, we are witnessing the emergence of these powerful technologies within health care. The use and growth of these technologies has been contingent on the availability of reliable and usable data, a particularly robust resource in critical care medicine where continuous monitoring forms a key component of the infrastructure of care. The response to this opportunity has included the development of open databases for research and other purposes; the development of a collaborative form of clinical data science intended to fully leverage these data resources, and the creation of data-driven applications for purposes such as clinical decision support. Most recently, data levels have reached the thresholds required for the development of robust artificial intelligence features for clinical purposes. The systematic capture and analysis of clinical data in both individuals and populations allows us to begin to move toward precision medicine in the intensive care unit (ICU). In this perspective review, we examine the fundamental role of data as we present the current progress that has been made toward an artificial intelligence (AI)-supported, data-driven precision critical care medicine.

Multivariable analysis of the association between electrolyte disturbances and mortality in cats

OBJECTIVES

Electrolyte disorders have been individually associated with mortality in small populations of cats with specific conditions, but the associations and interactions between electrolyte disturbances and outcome have not been evaluated in a large, heterogeneous population. It was hypothesized that abnormalities of sodium, chloride, potassium and calcium concentrations would be independently and proportionately associated with death from natural causes and with all-cause mortality in cats.

METHODS

An electronic database containing 7064 electrolyte profiles was constructed to assess the association between disorders of sodium, potassium, corrected-chloride and ionized calcium concentrations with non-survival by multivariable modelling. A second database containing 2388 records was used to validate the models constructed from the first database.

RESULTS

All four electrolytes assessed had non-linear U-shaped associations with case fatality rates, wherein concentrations clustered around the reference interval had the lowest case fatality rates, while progressively abnormal concentrations were associated with proportionately increased risk of non-survival (area under the receiver operator characteristic curve [AUROC] 0.689) or death (AUROC 0.750).

CONCLUSIONS AND RELEVANCE

Multivariable modelling suggested that these electrolyte disturbances were associated with non-survival and with death from natural causes independent of each other. The present study suggests that measurement of electrolyte concentrations is an important component of the assessment of cats in emergency rooms or intensive care units. Future studies should focus on confirming these associations in a prospective manner accounting for disease severity.

Electrolyte Disturbances Are Associated with Non-Survival in Dogs-A Multivariable Analysis

Electrolyte disorders have been individually associated with mortality in small populations of dogs and cats with specific conditions, but the associations and interactions between electrolyte disturbances and outcome have not been evaluated in a large, heterogeneous population. It was hypothesized that abnormalities of sodium, chloride, potassium, and calcium concentrations would be independently and proportionately associated with death from natural causes and with all-cause mortality in dogs. An electronic database containing 33,117 electrolyte profiles was constructed to retrospectively assess the association between disorders of sodium, potassium, corrected chloride, and ionized calcium concentrations with non-survival and with death excluding euthanasia by multivariable modeling. A second database containing 11,249 records was used to validate the models constructed from the first database. All four electrolytes assessed had non-linear U-shaped associations with case fatality rates, wherein concentrations clustered around the reference interval had the lowest case fatality rates, while progressively abnormal concentrations were associated with proportionately increased risk of non-survival (AUROC 0.624) or death (AUROC 0.678). Multivariable modeling suggested that these electrolyte disturbances were associated with non-survival and with death from natural causes independent of each other. This study suggests that measurement of electrolyte concentrations is an important component of the assessment of dogs in emergency rooms or intensive care units. Future studies should focus on confirming these associations in a prospective manner accounting for disease severity.