Variable selection with stepwise and best subset approaches

While purposeful selection is performed partly by software and partly by hand, the stepwise and best subset approaches are automatically performed by software. Two R functions stepAIC() and bestglm() are well designed for stepwise and best subset regression, respectively. The stepAIC() function begins with a full or null model, and methods for stepwise regression can be specified in the direction argument with character values "forward", "backward" and "both". The bestglm() function begins with a data frame containing explanatory variables and response variables. The response variable should be in the last column. Varieties of goodness-of-fit criteria can be specified in the IC argument. The Bayesian information criterion (BIC) usually results in more parsimonious model than the Akaike information criterion.

Model building strategy for logistic regression: purposeful selection

Logistic regression is one of the most commonly used models to account for confounders in medical literature. The article introduces how to perform purposeful selection model building strategy with R. I stress on the use of likelihood ratio test to see whether deleting a variable will have significant impact on model fit. A deleted variable should also be checked for whether it is an important adjustment of remaining covariates. Interaction should be checked to disentangle complex relationship between covariates and their synergistic effect on response variable. Model should be checked for the goodness-of-fit (GOF). In other words, how the fitted model reflects the real data. Hosmer-Lemeshow GOF test is the most widely used for logistic regression model.

Introduction to machine learning: k-nearest neighbors

Machine learning techniques have been widely used in many scientific fields, but its use in medical literature is limited partly because of technical difficulties. k-nearest neighbors (kNN) is a simple method of machine learning. The article introduces some basic ideas underlying the kNN algorithm, and then focuses on how to perform kNN modeling with R. The dataset should be prepared before running the knn() function in R. After prediction of outcome with kNN algorithm, the diagnostic performance of the model should be checked. Average accuracy is the mostly widely used statistic to reflect the kNN algorithm. Factors such as k value, distance calculation and choice of appropriate predictors all have significant impact on the model performance.

Multiple imputation with multivariate imputation by chained equation (MICE) package

Multiple imputation (MI) is an advanced technique for handing missing values. It is superior to single imputation in that it takes into account uncertainty in missing value imputation. However, MI is underutilized in medical literature due to lack of familiarity and computational challenges. The article provides a step-by-step approach to perform MI by using R multivariate imputation by chained equation (MICE) package. The procedure firstly imputed m sets of complete dataset by calling mice() function. Then statistical analysis such as univariate analysis and regression model can be performed within each dataset by calling with() function. This function sets the environment for statistical analysis. Lastly, the results obtained from each analysis are combined by using pool() function.

Small studies may overestimate the effect sizes in critical care meta-analyses: a meta-epidemiological study

INTRODUCTION

Small-study effects refer to the fact that trials with limited sample sizes are more likely to report larger beneficial effects than large trials. However, this has never been investigated in critical care medicine. Thus, the present study aimed to examine the presence and extent of small-study effects in critical care medicine.

METHODS

Critical care meta-analyses involving randomized controlled trials and reported mortality as an outcome measure were considered eligible for the study. Component trials were classified as large (≥100 patients per arm) and small (<100 patients per arm) according to their sample sizes. Ratio of odds ratio (ROR) was calculated for each meta-analysis and then RORs were combined using a meta-analytic approach. ROR<1 indicated larger beneficial effect in small trials. Small and large trials were compared in methodological qualities including sequence generating, blinding, allocation concealment, intention to treat and sample size calculation.

RESULTS

A total of 27 critical care meta-analyses involving 317 trials were included. Of them, five meta-analyses showed statistically significant RORs <1, and other meta-analyses did not reach a statistical significance. Overall, the pooled ROR was 0.60 (95% CI: 0.53 to 0.68); the heterogeneity was moderate with an I2 of 50.3% (chi-squared = 52.30; P = 0.002). Large trials showed significantly better reporting quality than small trials in terms of sequence generating, allocation concealment, blinding, intention to treat, sample size calculation and incomplete follow-up data.

CONCLUSIONS

Small trials are more likely to report larger beneficial effects than large trials in critical care medicine, which could be partly explained by the lower methodological quality in small trials. Caution should be practiced in the interpretation of meta-analyses involving small trials.

Machine learning for the prediction of volume responsiveness in patients with oliguric acute kidney injury in critical care

BACKGROUND AND OBJECTIVES

Excess fluid balance in acute kidney injury (AKI) may be harmful, and conversely, some patients may respond to fluid challenges. This study aimed to develop a prediction model that can be used to differentiate between volume-responsive (VR) and volume-unresponsive (VU) AKI.

METHODS

AKI patients with urine output < 0.5 ml/kg/h for the first 6 h after ICU admission and fluid intake > 5 l in the following 6 h in the US-based critical care database (Medical Information Mart for Intensive Care (MIMIC-III)) were considered. Patients who received diuretics and renal replacement on day 1 were excluded. Two predictive models, using either machine learning extreme gradient boosting (XGBoost) or logistic regression, were developed to predict urine output > 0.65 ml/kg/h during 18 h succeeding the initial 6 h for assessing oliguria. Established models were assessed by using out-of-sample validation. The whole sample was split into training and testing samples by the ratio of 3:1.

MAIN RESULTS

Of the 6682 patients included in the analysis, 2456 (36.8%) patients were volume responsive with an increase in urine output after receiving > 5 l fluid. Urinary creatinine, blood urea nitrogen (BUN), age, and albumin were the important predictors of VR. The machine learning XGBoost model outperformed the traditional logistic regression model in differentiating between the VR and VU groups (AU-ROC, 0.860; 95% CI, 0.842 to 0.878 vs. 0.728; 95% CI 0.703 to 0.753, respectively).

CONCLUSIONS

The XGBoost model was able to differentiate between patients who would and would not respond to fluid intake in urine output better than a traditional logistic regression model. This result suggests that machine learning techniques have the potential to improve the development and validation of predictive modeling in critical care research.

Cystatin C in prediction of acute kidney injury: a systemic review and meta-analysis

BACKGROUND

Cystatin C (CysC) has been proposed as a filtration marker for the early detection of acute kidney injury (AKI); however, a wide range of its predictive accuracy has been reported.

STUDY DESIGN

Meta-analysis of diagnostic test studies.

SETTING & POPULATION

Various clinical settings of AKI, including patients after cardiac surgery, pediatric patients, and critically ill patients.

SELECTION CRITERIA

Computerized search of PubMed, Current Contents, CINAHL, and EMBASE from inception until November 15, 2010, was performed to identify potentially relevant articles. Inclusion criteria were studies investigating the diagnostic accuracy of CysC level to predict AKI. There were no language restrictions in the search.

INDEX TESTS

Increasing or increased serum CysC level or urinary CysC excretion.

REFERENCE TESTS

The outcome was the development of AKI, primarily based on serum creatinine level (definition varied across studies).

RESULTS

We analyzed data from 19 studies and 11 countries involving 3,336 patients. Of these studies, 13 could be included in the meta-analysis. Across all settings, the diagnostic OR for serum CysC level to predict AKI was 27.7 (95% CI, 12.8-59.8), with sensitivity and specificity of 0.86 and 0.82, respectively. The area under the receiver operating characteristic curve (AUROC) of serum CysC levelto predict AKI was 0.87 (95% CI, 0.81-0.93). In an analysis excluding studies that did not clearly define the measurement time point, early serum CysC (within 24 hours after renal insult or intensive care unit admission) remained of diagnostic value. For the diagnostic value of urinary CysC excretion, the diagnostic OR was 3.10 (95% CI, 2.00-4.81), with sensitivity and specificity of0.61 and 0.67, respectively. TheAUROC of urinary CysC excretion to predict AKI was 0.67 (95% CI, 0.63-0.71) [corrected].

LIMITATIONS

Variation in criteria for definitions of index and reference tests, absence of measured glomerular filtration rate in most studies.

CONCLUSION

Serum CysC appears to be a good biomarker in the prediction of AKI, whereas urinary CysC excretion has only moderate diagnostic value.

Lactate clearance is a useful biomarker for the prediction of all-cause mortality in critically ill patients: a systematic review and meta-analysis*

OBJECTIVES

Lactate clearance has been widely investigated for its prognostic value in critically ill patients. However, the results are conflicting. The present study aimed to explore the diagnostic accuracy of lactate clearance in predicting mortality in critically or acutely ill patients.

DATA SOURCES

Databases of Medline, Embase, Scopus, and Web of Knowledge were searched from inception to June 2013.

STUDY SELECTION

Studies investigating the prognostic value of lactate clearance were defined as eligible. The searched item consisted of terms related to critically ill patients and terms related to lactate clearance.

DATA EXTRACTION

The following data were extracted: the name of the first author, publication year, subjects and setting, mean age of study population, sample size, male percentage, mortality of study cohort, definition of clearance, and the initial lactate level. Relative risk was reported to estimate the predictive value of lactate clearance on mortality rate, with relative risk less than 1 indicating that lactate clearance was a protective factor. Meta-analysis of diagnostic accuracy of lactate clearance in predicting mortality was performed by using hierarchical summary receiver operating characteristic model.

DATA SYNTHESIS

A total of 15 original articles were included in the study. Because of the significant heterogeneity across studies (I = 61.4%), random-effects model was used to pool relative risks. The pooled relative risk for mortality was 0.38 (95% CI, 0.29-0.50). The overall sensitivity and specificity for lactate clearance to predict mortality were 0.75 (95% CI, 0.58-0.87) and 0.72 (95% CI, 0.61-0.80), respectively. The diagnostic performance improved slightly when meta-analysis was restricted to ICU patients, with sensitivity and specificity of 0.83 (95% CI, 0.67-0.92) and 0.67 (95% CI, 0.59-0.75), respectively.

CONCLUSION

Our study demonstrates that lactate clearance is predictive of lower mortality rate in critically ill patients, and its diagnostic performance is optimal for clinical utility.

Drawing Nomograms with R: applications to categorical outcome and survival data

Outcome prediction is a major task in clinical medicine. The standard approach to this work is to collect a variety of predictors and build a model of appropriate type. The model is a mathematical equation that connects the outcome of interest with the predictors. A new patient with given clinical characteristics can be predicted for outcome with this model. However, the equation describing the relationship between predictors and outcome is often complex and the computation requires software for practical use. There is another method called nomogram which is a graphical calculating device allowing an approximate graphical computation of a mathematical function. In this article, we describe how to draw nomograms for various outcomes with nomogram() function. Binary outcome is fit by logistic regression model and the outcome of interest is the probability of the event of interest. Ordinal outcome variable is also discussed. Survival analysis can be fit with parametric model to fully describe the distributions of survival time. Statistics such as the median survival time, survival probability up to a specific time point are taken as the outcome of interest.

Ultrasonographic measurement of the respiratory variation in the inferior vena cava diameter is predictive of fluid responsiveness in critically ill patients: systematic review and meta-analysis

Respiratory variation in the inferior vena cava (ΔIVC) has been extensively studied with respect to its value in predicting fluid responsiveness, but the results are conflicting. This systematic review was aimed at investigating the diagnostic accuracy of ΔIVC in predicting fluid responsiveness. Databases including Medline, Embase, Scopus and Web of Knowledge were searched from inception to May 2013. Studies exploring the diagnostic performance of ΔIVC in predicting fluid responsiveness were included. To allow for more between- and within-study variance, a hierarchical summary receiver operating characteristic model was used to pool the results. Subgroup analyses were performed for patients on mechanical ventilation, spontaneously breathing patients and those challenged with colloids and crystalloids. A total of 8 studies involving 235 patients were eligible for analysis. Cutoff values of ΔIVC varied across studies, ranging from 12% to 40%. The pooled sensitivity and specificity in the overall population were 0.76 (95% confidence interval [CI]: 0.61-0.86) and 0.86 (95% CI: 0.69-0.95), respectively. The pooled diagnostic odds ratio (DOR) was 20.2 (95% CI: 6.1-67.1). The diagnostic performance of ΔIVC appeared to be better in patients on mechanical ventilation than in spontaneously breathing patients (DOR: 30.8 vs. 13.2). The pooled area under the receiver operating characteristic curve was 0.84 (95% CI: 0.79-0.89). Our study indicates that ΔIVC measured with point-of-care ultrasonography is of great value in predicting fluid responsiveness, particularly in patients on controlled mechanical ventilation and those resuscitated with colloids.

Univariate description and bivariate statistical inference: the first step delving into data

In observational studies, the first step is usually to explore data distribution and the baseline differences between groups. Data description includes their central tendency (e.g., mean, median, and mode) and dispersion (e.g., standard deviation, range, interquartile range). There are varieties of bivariate statistical inference methods such as Student's t-test, Mann-Whitney U test and Chi-square test, for normal, skews and categorical data, respectively. The article shows how to perform these analyses with R codes. Furthermore, I believe that the automation of the whole workflow is of paramount importance in that (I) it allows for others to repeat your results; (II) you can easily find out how you performed analysis during revision; (III) it spares data input by hand and is less error-prone; and (IV) when you correct your original dataset, the final result can be automatically corrected by executing the codes. Therefore, the process of making a publication quality table incorporating all abovementioned statistics and P values is provided, allowing readers to customize these codes to their own needs.

Prevalence and impact of acute renal impairment on COVID-19: a systematic review and meta-analysis

BACKGROUND

The aim of this study is to assess the prevalence of abnormal urine analysis and kidney dysfunction in COVID-19 patients and to determine the association of acute kidney injury (AKI) with the severity and prognosis of COVID-19 patients.

METHODS

The electronic database of Embase and PubMed were searched for relevant studies. A meta-analysis of eligible studies that reported the prevalence of abnormal urine analysis and kidney dysfunction in COVID-19 was performed. The incidences of AKI were compared between severe versus non-severe patients and survivors versus non-survivors.

RESULTS

A total of 24 studies involving 4963 confirmed COVID-19 patients were included. The proportions of patients with elevation of sCr and BUN levels were 9.6% (95% CI 5.7-13.5%) and 13.7% (95% CI 5.5-21.9%), respectively. Of all patients, 57.2% (95% CI 40.6-73.8%) had proteinuria, 38.8% (95% CI 26.3-51.3%) had proteinuria +, and 10.6% (95% CI 7.9-13.3%) had proteinuria ++ or +++. The overall incidence of AKI in all COVID-19 patients was 4.5% (95% CI 3.0-6.0%), while the incidence of AKI was 1.3% (95% CI 0.2-2.4%), 2.8% (95% CI 1.4-4.2%), and 36.4% (95% CI 14.6-58.3%) in mild or moderate cases, severe cases, and critical cases, respectively. Meanwhile, the incidence of AKI was 52.9%(95% CI 34.5-71.4%), 0.7% (95% CI - 0.3-1.8%) in non-survivors and survivors, respectively. Continuous renal replacement therapy (CRRT) was required in 5.6% (95% CI 2.6-8.6%) severe patients, 0.1% (95% CI - 0.1-0.2%) non-severe patients and 15.6% (95% CI 10.8-20.5%) non-survivors and 0.4% (95% CI - 0.2-1.0%) survivors, respectively.

CONCLUSION

The incidence of abnormal urine analysis and kidney dysfunction in COVID-19 was high and AKI is closely associated with the severity and prognosis of COVID-19 patients. Therefore, it is important to increase awareness of kidney dysfunction in COVID-19 patients.

Propensity score method: a non-parametric technique to reduce model dependence

Propensity score analysis (PSA) is a powerful technique that it balances pretreatment covariates, making the causal effect inference from observational data as reliable as possible. The use of PSA in medical literature has increased exponentially in recent years, and the trend continue to rise. The article introduces rationales behind PSA, followed by illustrating how to perform PSA in R with MatchIt package. There are a variety of methods available for PS matching such as nearest neighbors, full matching, exact matching and genetic matching. The task can be easily done by simply assigning a string value to the method argument in the matchit() function. The generic summary() and plot() functions can be applied to an object of class matchit to check covariate balance after matching. Furthermore, there is a useful package PSAgraphics that contains several graphical functions to check covariate balance between treatment groups across strata. If covariate balance is not achieved, one can modify model specifications or use other techniques such as random forest and recursive partitioning to better represent the underlying structure between pretreatment covariates and treatment assignment. The process can be repeated until the desirable covariate balance is achieved.

Accuracy of stroke volume variation in predicting fluid responsiveness: a systematic review and meta-analysis

PURPOSE

Stroke volume variation (SVV) appears to be a good predictor of fluid responsiveness in critically ill patients. However, a wide range of its predictive values has been reported in recent years. We therefore undertook a systematic review and meta-analysis of clinical trials that investigated the diagnostic value of SVV in predicting fluid responsiveness.

METHODS

Clinical investigations were identified from several sources, including MEDLINE, EMBASE, WANFANG, and CENTRAL. Original articles investigating the diagnostic value of SVV in predicting fluid responsiveness were considered to be eligible. Participants included critically ill patients in the intensive care unit (ICU) or operating room (OR) who require hemodynamic monitoring.

RESULTS

A total of 568 patients from 23 studies were included in our final analysis. Baseline SVV was correlated to fluid responsiveness with a pooled correlation coefficient of 0.718. Across all settings, we found a diagnostic odds ratio of 18.4 for SVV to predict fluid responsiveness at a sensitivity of 0.81 and specificity of 0.80. The SVV was of diagnostic value for fluid responsiveness in OR or ICU patients monitored with the PiCCO or the FloTrac/Vigileo system, and in patients ventilated with tidal volume greater than 8 ml/kg.

CONCLUSIONS

SVV is of diagnostic value in predicting fluid responsiveness in various settings.

Efficacy and safety of regional citrate anticoagulation in critically ill patients undergoing continuous renal replacement therapy

PURPOSE

Regional citrate anticoagulation (RCA) is an attractive anticoagulation mode in continuous renal replacement therapy (CRRT) because it restricts the anticoagulatory effect to the extracorporeal circuit. In recent years, several randomized controlled trials have been conducted to investigate its superiority over other anticoagulation modes. Thus, we performed a systematic review of available evidence on the efficacy and safety of RCA.

METHODS

A systematic review of randomized controlled trials investigating the efficacy and safety of RCA was performed. PubMed, Current Contents, CINAHL, and EMBASE databases were searched to identify relevance articles. Data on circuit life span, bleeding events, metabolic derangement, and mortality were abstracted. Mean difference was used for continuous variables, and risk ratio was used for binomial variables. The random effects or fixed effect model was used to combine these data according to heterogeneity. The software Review Manager 5.1 was used for the meta-analysis.

RESULTS

Six studies met our inclusion criteria, which involved a total of 658 circuits. In these six studies patients with liver failure or a high risk of bleeding were excluded. The circuit life span in the RCA group was significantly longer than that in the control group, with a mean difference of 23.03 h (95% CI 0.45-45.61 h). RCA was able to reduce the risk of bleeding, with a risk ratio of 0.28 (95% CI 0.15-0.50). Metabolic stability (electrolyte and acid-base stabilities) in performing RCA was comparable to that in other anticoagulation modes, and metabolic derangements (hypernatremia, metabolic alkalosis, and hypocalcemia) could be easily controlled without significant clinical consequences. Two studies compared mortality rate between RCA and control groups, with one reported similar mortality rate and the other reported superiority of RCA over the control group (hazards ratio 0.7).

CONCLUSIONS

RCA is effective in maintaining circuit patency and reducing the risk of bleeding, and thus can be recommended for CRRT if and when metabolic monitoring is adequate and the protocol is followed. However, the safety of citrate in patients with liver failure cannot be concluded from current analysis. The metabolic stability can be easily controlled during RCA. Survival benefit from RCA is still controversial due to limited evidence.

Survival analysis in the presence of competing risks

Survival analysis in the presence of competing risks imposes additional challenges for clinical investigators in that hazard function (the rate) has no one-to-one link to the cumulative incidence function (CIF, the risk). CIF is of particular interest and can be estimated non-parametrically with the use cuminc() function. This function also allows for group comparison and visualization of estimated CIF. The effect of covariates on cause-specific hazard can be explored using conventional Cox proportional hazard model by treating competing events as censoring. However, the effect on hazard cannot be directly linked to the effect on CIF because there is no one-to-one correspondence between hazard and cumulative incidence. Fine-Gray model directly models the covariate effect on CIF and it reports subdistribution hazard ratio (SHR). However, SHR only provide information on the ordering of CIF curves at different levels of covariates, it has no practical interpretation as HR in the absence of competing risks. Fine-Gray model can be fit with crr() function shipped with the cmprsk package. Time-varying covariates are allowed in the crr() function, which is specified by cov2 and tf arguments. Predictions and visualization of CIF for subjects with given covariate values are allowed for crr object. Alternatively, competing risk models can be fit with riskRegression package by employing different link functions between covariates and outcomes. The assumption of proportionality can be checked by testing statistical significance of interaction terms involving failure time. Schoenfeld residuals provide another way to check model assumption.

Missing data imputation: focusing on single imputation

Complete case analysis is widely used for handling missing data, and it is the default method in many statistical packages. However, this method may introduce bias and some useful information will be omitted from analysis. Therefore, many imputation methods are developed to make gap end. The present article focuses on single imputation. Imputations with mean, median and mode are simple but, like complete case analysis, can introduce bias on mean and deviation. Furthermore, they ignore relationship with other variables. Regression imputation can preserve relationship between missing values and other variables. There are many sophisticated methods exist to handle missing values in longitudinal data. This article focuses primarily on how to implement R code to perform single imputation, while avoiding complex mathematical calculations.

Impact of delirium on clinical outcome in critically ill patients: a meta-analysis

CONTEXT

Delirium is prevalent in the intensive care unit (ICU) and has been associated with negative clinical outcomes. However, a quantitative and systematic assessment of published studies has not been conducted.

OBJECTIVE

Meta-analysis of clinical observational studies was performed to investigate the association between delirium and clinical outcomes.

DATA SOURCES AND STUDY SELECTION

Relevant studies were identified by investigators from databases including Medline, Embase, OVID and EBSCO from inception to May 2012. Studies that reported the association of delirium with clinical outcomes in critical care setting were included.

DATA EXTRACTION

Data were extracted independently by reviewers and summary effects were obtained using random effects model.

DATA SYNTHESIS

Of the 16 studies included, 14 studies involving 5891 patients reported data on mortality, and delirious patients had higher mortality rate than non-delirious patients (odds ratio [OR]: 3.22; 95% confidence interval [CI]: 2.30-4.52). Delirious patients had higher rate of complications (OR: 6.5; 95% CI: 2.7-15.6), and were more likely to be discharged to skilled placement (OR: 2.59; 95% CI: 1.59-4.21). Furthermore, patients with delirium had longer length of stay in both ICU (weighted mean difference [WMD]: 7.32 days; 95% CI: 4.63-10.01) and hospital (WMD: 6.53 days; 95% CI: 3.03-10.03), and they spent more time on mechanical ventilation (WMD: 7.22 days; 95% CI: 5.15-9.29).

CONCLUSION

Delirium in critically ill patients is associated with higher mortality rate, more complications, longer duration of mechanical ventilation, and longer length of stay in ICU and hospital.

A global priority list of the TOp TEn resistant Microorganisms (TOTEM) study at intensive care: a prioritization exercise based on multi-criteria decision analysis

The World Health Organization (WHO) proposed a global priority pathogen list (PPL) of multidrug-resistant (MDR) bacteria. Our current objective was to provide global expert ranking of the most serious MDR bacteria present at intensive care units (ICU) that have become a threat in clinical practice. A proposal addressing a PPL for ICU, arising from the WHO Global PPL, was developed. Based on the supporting data, the pathogens were grouped in three priority tiers: critical, high, and medium. A multi-criteria decision analysis (MCDA) was used to identify the priority tiers. After MCDA, mortality, treatability, and cost of therapy were of highest concern (scores of 19/20, 19/20, and 15/20, respectively) while dealing with PPL, followed by healthcare burden and resistance prevalence. Carbapenem-resistant (CR) Acinetobacter baumannii, carbapenemase-expressing Klebsiella pneumoniae (KPC), and MDR Pseudomonas aeruginosa were identified as critical organisms. High-risk organisms were represented by CR Pseudomonas aeruginosa, methicillin-resistant Staphylococcus aureus, and extended-spectrum beta-lactamase (ESBL) Enterobacteriaceae. Finally, ESBL Serratia marcescens, vancomycin-resistant Enterococci, and TMP-SMX-resistant Stenotrophomonas maltophilia were identified as medium priority. We conclude that education, investigation, funding, and development of new antimicrobials for ICU organisms should focus on carbapenem-resistant Gram-negative organisms.

Causal mediation analysis in the context of clinical research

Clinical researches usually collected numerous intermediate variables besides treatment and outcome. These variables are often incorrectly treated as confounding factors and are thus controlled using a variety of multivariable regression models depending on the types of outcome variable. However, these methods fail to disentangle underlying mediating processes. Causal mediation analysis (CMA) is a method to dissect total effect of a treatment into direct and indirect effect. The indirect effect is transmitted via mediator to the outcome. The mediation package is designed to perform CMA under the assumption of sequential ignorability. It reports average causal mediation effect (ACME), average direct effect (ADE) and total effect. Also, the package provides visualization tool for these estimated effects. Sensitivity analysis is designed to examine whether the results are robust to the violation of the sequential ignorability assumption since the assumption has been criticized to be too strong to be satisfied in research practice.

Mechanical power normalized to predicted body weight as a predictor of mortality in patients with acute respiratory distress syndrome

PURPOSE

Protective mechanical ventilation based on multiple ventilator parameters such as tidal volume, plateau pressure, and driving pressure has been widely used in acute respiratory distress syndrome (ARDS). More recently, mechanical power (MP) was found to be associated with mortality. The study aimed to investigate whether MP normalized to predicted body weight (norMP) was superior to other ventilator variables and to prove that the discrimination power cannot be further improved with a sophisticated machine learning method.

METHODS

The study included individual patient data from eight randomized controlled trials conducted by the ARDSNet. The data was split 3:1 into training and testing subsamples. The discrimination of each ventilator variable was calculated in the testing subsample using the area under receiver operating characteristic curve. The gradient boosting machine was used to examine whether the discrimination could be further improved.

RESULTS

A total of 5159 patients with acute onset ARDS were included for analysis. The discrimination of norMP in predicting mortality was significantly better than the absolute MP (p = 0.011 for DeLong's test). The gradient boosting machine was not able to improve the discrimination as compared to norMP (p = 0.913 for DeLong's test). The multivariable regression model showed a significant interaction between norMP and ARDS severity (p < 0.05). While the norMP was not significantly associated with mortality outcome (OR 0.99; 95% CI 0.91-1.07; p = 0.862) in patients with mild ARDS, it was associated with increased risk of mortality in moderate (OR 1.11; 95% CI 1.02-1.23; p = 0.021) and severe (OR 1.13; 95% CI 1.03-1.24; p < 0.008) ARDS.

CONCLUSIONS

The study showed that norMP was a good ventilator variable associated with mortality, and its predictive discrimination cannot be further improved with a sophisticated machine learning method. Further experimental trials are needed to investigate whether adjusting ventilator variables according to norMP will significantly improve clinical outcomes.

Hierarchical cluster analysis in clinical research with heterogeneous study population: highlighting its visualization with R

Big data clinical research typically involves thousands of patients and there are numerous variables available. Conventionally, these variables can be handled by multivariable regression modeling. In this article, the hierarchical cluster analysis (HCA) is introduced. This method is used to explore similarity between observations and/or clusters. The result can be visualized using heat maps and dendrograms. Sometimes, it would be interesting to add scatter plot and smooth lines into the panels of the heat map. The inherent R heatmap package does not provide this function. A series of scatter plots can be created using lattice package, and then background color of each panel is mapped to the regression coefficient by using custom-made panel functions. This is the unique feature of the lattice package. Dendrograms and color keys can be added as the legend elements of the lattice system. The latticeExtra package provides some useful functions for the work.

Comparing baseline characteristics between groups: an introduction to the CBCgrps package

A usual practice in observational studies is the comparison of baseline characteristics of participants between study groups. The overall population can be grouped by clinical outcome or exposure status. A combined table reporting baseline characteristics is usually displayed, for the overall population and then separately for each group. The last column usually gives the P value for the comparison between study groups. In the conventional research model, the variables for which data are collected are limited in number. It is thus feasible to calculate descriptive data one by one and to manually create the table. The availability of EHR and big data mining techniques makes it possible to explore a far larger number of variables. However, manual tabulation of big data is particularly error prone; it is exceedingly time-consuming to create and revise such tables manually. In this paper, we introduce an R package called CBCgrps, which is designed to automate and streamline the generation of such tables when working with big data. The package contains two functions, twogrps() and multigrps(), which are used for comparisons between two and multiple groups, respectively.

Identification of subclasses of sepsis that showed different clinical outcomes and responses to amount of fluid resuscitation: a latent profile analysis

BACKGROUND AND OBJECTIVE

Sepsis is a heterogeneous disease and identification of its subclasses may facilitate and optimize clinical management. This study aimed to identify subclasses of sepsis and its responses to different amounts of fluid resuscitation.

METHODS

This was a retrospective study conducted in an intensive care unit at a large tertiary care hospital. The patients fulfilling the diagnostic criteria of sepsis from June 1, 2001 to October 31, 2012 were included. Clinical and laboratory variables were used to perform the latent profile analysis (LPA). A multivariable logistic regression model was used to explore the independent association of fluid input and mortality outcome.

RESULTS

In total, 14,993 patients were included in the study. The LPA identified four subclasses of sepsis: profile 1 was characterized by the lowest mortality rate and having the largest proportion and was considered the baseline type; profile 2 was characterized by respiratory dysfunction; profile 3 was characterized by multiple organ dysfunction (kidney, coagulation, liver, and shock), and profile 4 was characterized by neurological dysfunction. Profile 3 showed the highest mortality rate (45.4%), followed by profile 4 (27.4%), 2 (18.2%), and 1 (16.9%). Overall, the amount of fluid needed for resuscitation was the largest on day 1 (median 5115 mL, interquartile range (IQR) 2662 to 8800 mL) and decreased rapidly on day 2 (median 2140 mL, IQR 900 to 3872 mL). Higher cumulative fluid input in the first 48 h was associated with reduced risk of hospital mortality for profile 3 (odds ratio (OR) 0.89, 95% CI 0.83 to 0.95 for each 1000 mL increase in fluid input) and with increased risk of death for profile 4 (OR 1.20, 95% CI 1.11 to 1.30).

CONCLUSION

The study identified four subphenotypes of sepsis, which showed different mortality outcomes and responses to fluid resuscitation. Prospective trials are needed to validate our findings.