A predictive model for the early identification of patients at risk for a prolonged intensive care unit length of stay

Outcomes and factors associated with prolonged stays among patients admitted to adult intensive care unit in a resource-limited setting: a multicenter chart review

The length of stay in an intensive care unit is used as a benchmark for measuring resource consumption and quality of care and predicts a higher risk of readmission. The study aimed to assess the outcome and factors associated with prolonged intensive care unit stays among those admitted to adult intensive care units of selected public hospitals in Addis Ababa from January 1, 2022, to December 31, 2022. A multicenter retrospective chart review was conducted involving 409 adult patients. Binary logistic regression was used to assess factors associated with a prolonged stay and chi-square tests were used to assess associations and differences in outcomes for prolonged stays. The study, involving 409 of 421 individuals, revealed a predominantly male (55.0%) and the median age of study participants was 38, with an interquartile range (27, 55). Approximately 16.9% experienced prolonged stays, resulting in a 43.5% mortality rate. After adjustments for confounders, there were significant associations with prolonged stays for sedative/hypnotics, readmission, and complications. The study revealed that for every six patients admitted to the intensive care unit, one patient stayed longer, with nearly half experiencing mortality, demanding increased attention. The study emphasized the critical need for improvement in addressing associations between sedative/hypnotics, readmissions, complications, and prolonged stays.

Prevalence and Risk Factors of Healthcare-Associated Infections among Hospitalized Pediatric Patients: Point Prevalence Survey in Thailand 2021

BACKGROUND

Healthcare-associated infections (HAIs) pose a grave threat to patient safety, morbidity, and mortality, contributing to antimicrobial resistance. Thus, we estimated the point prevalence, risk factors, types, and pathogens of HAIs in hospitalized pediatric patients.

METHODS

A point prevalence survey (PPS) of HAIs in hospitalized pediatric patients < 18 years old was conducted from March to May 2021. Outcomes, risk factors, and types of HAIs associated with HAIs in 41 hospitals across Thailand were collected.

RESULTS

The prevalence of HAIs was 3.9% (95% CI 2.9-5.0%) (56/1443). By ages < 1 month, 1 month-2 years, 2-12 years, and 12-18 years, the prevalence of HAIs was 4.2%, 3.3%, 4.1%, and 3.0%, respectively (p = 0.80). Significant independent risk factors were extended hospital length of stay (LOS) and central venous catheter (CVC) use. Compared to an LOS of <4 days, LOSs of 4-7 days, 8-14 days, and >14 days had adjusted odds ratios (aORs) of 2.65 (95% CI 1.05, 6.68), 5.19 (95% CI 2.00, 13.4), and 9.03 (95% CI 3.97, 20.5), respectively. The use of a CVC had an aOR of 2.45 (95% CI 1.06-5.66). Lower respiratory tract infection (LRTI) was the most common HAI type (46.4%: 26/56). The highest prevalence of HAIs was predominantly observed in LRTI diagnoses, with the highest among these in the <1 month age category at 2.3% (17/738).

CONCLUSION

The prevalence of HAIs in hospitalized pediatric patients was 3.9%. Extended LOS and use of CVC were HAI risk factors. A strategy for reducing LOS and reviewing insertion indications or the early planned removal of a CVC was implemented. The surveillance of HAIs stands as a cornerstone and fundamental component of IPC, offering invaluable insights that enhance hospital IPC interventions aimed at preventing HAIs.

Prediction of prolonged length of stay on the intensive care unit in severely injured patients-a registry-based multivariable analysis

Purpose

Mortality is the primary outcome measure in severely injured trauma victims. However, quality indicators for survivors are rare. We aimed to develop and validate an outcome measure based on length of stay on the intensive care unit (ICU).

Methods

The TraumaRegister DGU of the German Trauma Society (DGU) was used to identify 108,178 surviving patients with serious injuries who required treatment on ICU (2014-2018). In a first step, need for prolonged ICU stay, defined as 8 or more days, was predicted. In a second step, length of stay was estimated in patients with a prolonged stay. Data from the same trauma registry (2019-2022, n = 72,062) were used to validate the models derived with logistic and linear regression analysis.

Results

The mean age was 50 years, 70% were males, and the average Injury Severity Score was 16.2 points. Average/median length of stay on ICU was 6.3/2 days, where 78% were discharged from ICU within the first 7 days. Prediction of need for a prolonged ICU stay revealed 15 predictors among which injury severity (worst Abbreviated Injury Scale severity level), need for intubation, and pre-trauma condition were the most important ones. The area under the receiver operating characteristic curve was 0.903 (95% confidence interval 0.900-0.905). Length of stay prediction in those with a prolonged ICU stay identified the need for ventilation and the number of injuries as the most important factors. Pearson's correlation of observed and predicted length of stay was 0.613. Validation results were satisfactory for both estimates.

Conclusion

Length of stay on ICU is a suitable outcome measure in surviving patients after severe trauma if adjusted for severity. The risk of needing prolonged ICU care could be calculated in all patients, and observed vs. predicted rates could be used in quality assessment similar to mortality prediction. Length of stay prediction in those who require a prolonged stay is feasible and allows for further benchmarking.

Nutrition considerations for patients with persistent critical illness: A narrative review

Critically ill patients experience high rates of malnutrition and significant muscle loss during their intensive care unit (ICU) admission, impacting recovery. Nutrition is likely to play an important role in mitigating the development and progression of malnutrition and muscle loss observed in ICU, yet definitive clinical trials of nutrition interventions in ICU have failed to show benefit. As improvements in the quality of medical care mean that sicker patients are able to survive the initial insult, combined with an aging and increasingly comorbid population, it is anticipated that ICU length of stay will continue to increase. This review aims to discuss nutrition considerations unique to critically ill patients who have persistent critical illness, defined as an ICU stay of >10 days. A discussion of nutrition concepts relevant to patients with persistent critical illness will include energy and protein metabolism, prescription, and delivery; monitoring of nutrition at the bedside; and the role of the healthcare team in optimizing nutrition support.

Surgical scheduling via optimization and machine learning with long-tailed data : Health care management science, in press

Using data from cardiovascular surgery patients with long and highly variable post-surgical lengths of stay (LOS), we develop a modeling framework to reduce recovery unit congestion. We estimate the LOS and its probability distribution using machine learning models, schedule procedures on a rolling basis using a variety of optimization models, and estimate performance with simulation. The machine learning models achieved only modest LOS prediction accuracy, despite access to a very rich set of patient characteristics. Compared to the current paper-based system used in the hospital, most optimization models failed to reduce congestion without increasing wait times for surgery. A conservative stochastic optimization with sufficient sampling to capture the long tail of the LOS distribution outperformed the current manual process and other stochastic and robust optimization approaches. These results highlight the perils of using oversimplified distributional models of LOS for scheduling procedures and the importance of using optimization methods well-suited to dealing with long-tailed behavior.

Forecasting ICU Census by Combining Time Series and Survival Models

Capacity planning of ICUs is essential for effective management of health safety, quality of patient care, and the allocation of ICU resources. Whereas ICU length of stay (LOS) may be estimated using patient information such as severity of illness scoring systems, ICU census is impacted by both patient LOS and arrival patterns. We set out to develop and evaluate an ICU census forecasting algorithm using the Multiple Organ Dysfunction Score (MODS) and the Nine Equivalents of Nursing Manpower Use Score (NEMS) for capacity planning purposes.

DESIGN

Retrospective observational study.

SETTING

We developed the algorithm using data from the Medical-Surgical ICU (MSICU) at University Hospital, London, Canada and validated using data from the Critical Care Trauma Centre (CCTC) at Victoria Hospital, London, Canada.

PATIENTS

Adult patient admissions (7,434) to the MSICU and (9,075) to the CCTC from 2015 to 2021.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We developed an Autoregressive integrated moving average time series model that forecasts patients arriving in the ICU and a survival model using MODS, NEMS, and other factors to estimate patient LOS. The models were combined to create an algorithm that forecasts ICU census for planning horizons ranging from 1 to 7 days. We evaluated the algorithm quality using several fit metrics. The root mean squared error ranged from 2.055 to 2.890 beds/d and the mean absolute percentage error from 9.4% to 13.2%. We show that this forecasting algorithm provides a better fit when compared with a moving average or a time series model that directly forecasts ICU census. Additionally, we evaluated the performance of the algorithm using data during the global COVID-19 pandemic and found that the error of the forecasts increased proportionally with the number of COVID-19 patients in the ICU.

CONCLUSIONS

It is possible to develop accurate tools to forecast ICU census. This type of algorithm may be important to clinicians and managers when planning ICU capacity as well as staffing and surgical demand planning over a short time horizon.

Data-driven methodology to predict the ICU length of stay: A multicentre study of 99,492 admissions in 109 Brazilian units

PURPOSE

The length of stay (LoS) is one of the most used metrics for resource use in Intensive Care Units (ICU). We propose a structured data-driven methodology to predict the ICU length of stay and the risk of prolonged stay, and its application in a large multicenter Brazilian ICU database.

METHODS

Demographic data, comorbidities, complications, laboratory data, and primary and secondary diagnosis were prospectively collected and retrospectively analysed by a data-driven methodology, which includes eight different machine learning models and a stacking model. The study setting included 109 mixed-type ICUs from 38 Brazilian hospitals and the external validation was performed by 93 medical-surgical ICUs of 55 hospitals in Brazil.

RESULTS

A cohort of 99,492 adult ICU admissions were included from the 01^st of January to the 31^st of December 2019. The stacking model combining Random Forests and Linear Regression presented the best results to predict ICU length of stay (RMSE = 3.82; MAE = 2.52; R² = 0.36). The prediction model for the risk of long stay were accurate to early identify prolonged stay patients (Brier Score = 0.04, AUC = 0.87, PPV = 0.83, NPV = 0.95).

CONCLUSION

The data-driven methodology to predict ICU length of stay and the risk of long-stay proved accurate in a large multicentre cohort of general ICU patients. The proposed models are helpful to predict the individual length of stay and to early identify patients with high risk of prolonged stay.

Predict models for prolonged ICU stay using APACHE II, APACHE III and SAPS II scores: A Japanese multicenter retrospective cohort study

Prolonged ICU stays are associated with high costs and increased mortality. Thus, early prediction of such stays would help clinicians to plan initial interventions, which could lead to efficient utilization of ICU resources. The aim of this study was to develop models for predicting prolonged stays in Japanese ICUs using APACHE II, APACHE III and SAPS II scores. In this multicenter retrospective cohort study, we analyzed the cases of 85,558 patients registered in the Japanese Intensive care Patient Database between 2015 and 2019. Prolonged ICU stay was defined as an ICU stay of >14 days. Multivariable logistic regression analyses were performed to develop three predictive models for prolonged ICU stay using APACHE II, APACHE III and SAPS II scores, respectively. After exclusions, 79,620 patients were analyzed, 2,364 of whom (2.97%) experienced prolonged ICU stays. Multivariable logistic regression analyses showed that severity scores, BMI, MET/RRT, postresuscitation, readmission, length of stay before ICU admission, and diagnosis at ICU admission were significantly associated with higher risk of prolonged ICU stay in all models. The present study developed predictive models for prolonged ICU stay using severity scores. These models may be helpful for efficient utilization of ICU resources.

Machine Learning-based Prediction of Prolonged Intensive Care Unit Stay for Critical Patients with Spinal Cord Injury

STUDY DESIGN

A retrospective cohort study.

OBJECTIVE

The objective of the study was to develop machine-learning (ML) classifiers for predicting prolonged intensive care unit (ICU)-stay and prolonged hospital-stay for critical patients with spinal cord injury (SCI).

SUMMARY OF BACKGROUND DATA

Critical patients with SCI in ICU need more attention. SCI patients with prolonged stay in ICU usually occupy vast medical resources and hinder the rehabilitation deployment.

METHODS

A total of 1599 critical patients with SCI were included in the study and labeled with prolonged stay or normal stay. All data were extracted from the eICU Collaborative Research Database and the Medical Information Mart for Intensive Care III-IV Database. The extracted data were randomly divided into training, validation and testing (6:2:2) subdatasets. A total of 91 initial ML classifiers were developed, and the top three initial classifiers with the best performance were further stacked into an ensemble classifier with logistic regressor. The area under the curve (AUC) was the main indicator to assess the prediction performance of all classifiers. The primary predicting outcome was prolonged ICU-stay, while the secondary predicting outcome was prolonged hospital-stay.

RESULTS

In predicting prolonged ICU-stay, the AUC of the ensemble classifier was 0.864 ± 0.021 in the three-time five-fold cross-validation and 0.802 in the independent testing. In predicting prolonged hospital-stay, the AUC of the ensemble classifier was 0.815 ± 0.037 in the three-time five-fold cross-validation and 0.799 in the independent testing. Decision curve analysis showed the merits of the ensemble classifiers, as the curves of the top three initial classifiers varied a lot in either predicting prolonged ICU-stay or discriminating prolonged hospital-stay.

CONCLUSION

The ensemble classifiers successfully predict the prolonged ICU-stay and the prolonged hospital-stay, which showed a high potential of assisting physicians in managing SCI patients in ICU and make full use of medical resources.Level of Evidence: 3.

Risk factors for prolonged intensive care unit stays in patients after cardiac surgery with cardiopulmonary bypass: A retrospective observational study

Objectives

Patients after cardiac surgery with cardiopulmonary bypass (CPB) require a stay in the ICU postoperatively. This study aimed to investigate the incidence of prolonged length of stay (LOS) in the ICU after cardiac surgery with CPB and identify associated risk factors.

Methods

The current investigation was an observational, retrospective study that included 395 ICU patients who underwent cardiac surgery with CPB at a tertiary hospital in Guangzhou from June 2015 to June 2017. Data were obtained from the hospital database. Binary logistic regression modeling was used to analyze risk factors for prolonged ICU LOS.

Results

Of 395 patients, 137 (34.7%) had a prolonged ICU LOS (>72.0 h), and the median ICU LOS was 50.9 h. Several variables were found associated with prolonged ICU LOS: duration of CPB, prolonged mechanical ventilation and non-invasive assisted ventilation use, PaO₂/FiO₂ ratios within 6 h after surgery, type of surgery, red blood cell infusion during surgery, postoperative atrial arrhythmia, postoperative ventricular arrhythmia (all P < 0.05).

Conclusions

These findings are clinically relevant for identifying patients with an estimated prolonged ICU LOS, enabling clinicians to facilitate earlier intervention to reduce the risk and prevent resulting delayed recovery.

Comparative Safety Profiles of Sedatives Commonly Used in Clinical Practice: A 10-Year Nationwide Pharmacovigilance Study in Korea

This study aims to compare the prevalence and seriousness of adverse events (AEs) among sedatives used in critically ill patients or patients undergoing invasive procedures and to identify factors associated with serious AEs. Retrospective cross-sectional analysis of sedative-related AEs voluntarily reported to the Korea Adverse Event Reporting System from 2008 to 2017 was performed. All AEs were grouped using preferred terms and System Organ Classes per the World Health Organization-Adverse Reaction Terminology. Logistic regression was performed to identify factors associated with serious events. Among 95,188 AEs, including 3132 (3.3%) serious events, the most common etiologic sedative was fentanyl (58.8%), followed by pethidine (25.9%). Gastrointestinal disorders (54.2%) were the most frequent AEs. The most common serious AE was heart rate/rhythm disorders (33.1%). Serious AEs were significantly associated with male sex; pediatrics; etiologic sedative with etomidate at the highest risk, followed by dexmedetomidine, ketamine, and propofol; polypharmacy; combined sedative use; and concurrent use of corticosteroids, aspirin, neuromuscular blockers, and antihistamines (reporting odds ratio > 1, p < 0.001 for all). Sedative-induced AEs are most frequently reported with fentanyl, primarily manifesting as gastrointestinal disorders. Etomidate is associated with the highest risk of serious AEs, with the most common serious events being heart rate/rhythm disorders.

Early Prediction of Mortality, Severity, and Length of Stay in the Intensive Care Unit of Sepsis Patients Based on Sepsis 3.0 by Machine Learning Models

Background: Early prediction of the clinical outcome of patients with sepsis is of great significance and can guide treatment and reduce the mortality of patients. However, it is clinically difficult for clinicians. Methods: A total of 2,224 patients with sepsis were involved over a 3-year period (2016-2018) in the intensive care unit (ICU) of Peking Union Medical College Hospital. With all the key medical data from the first 6 h in the ICU, three machine learning models, logistic regression, random forest, and XGBoost, were used to predict mortality, severity (sepsis/septic shock), and length of ICU stay (LOS) (>6 days, ≤ 6 days). Missing data imputation and oversampling were completed on the dataset before introduction into the models. Results: Compared to the mortality and LOS predictions, the severity prediction achieved the best classification results, based on the area under the operating receiver characteristics (AUC), with the random forest classifier (sensitivity = 0.65, specificity = 0.73, F1 score = 0.72, AUC = 0.79). The random forest model also showed the best overall performance (mortality prediction: sensitivity = 0.50, specificity = 0.84, F1 score = 0.66, AUC = 0.74; LOS prediction: sensitivity = 0.79, specificity = 0.66, F1 score = 0.69, AUC = 0.76) among the three models. The predictive ability of the SOFA score itself was inferior to that of the above three models. Conclusions: Using the random forest classifier in the first 6 h of ICU admission can provide a comprehensive early warning of sepsis, which will contribute to the formulation and management of clinical decisions and the allocation and management of resources.

A deep attention model to forecast the Length Of Stay and the in-hospital mortality right on admission from ICD codes and demographic data

Leveraging the Electronic Health Records (EHR) longitudinal data to produce actionable clinical insights has always been a critical issue for recent studies. Non-forecasted extended hospitalizations account for a disproportionate amount of resource use, the mediocre quality of inpatient care, and avoidable fatalities. The capability to predict the Length of Stay (LoS) and mortality in the early stages of the admission provides opportunities to improve care and prevent many preventable losses. Forecasting the in-hospital mortality is important in providing clinicians with enough insights to make decisions and hospitals to allocate resources, hence predicting the LoS and mortality within the first day of admission is a difficult but a paramount endeavor. The biggest challenge is that few data are available by this time, thus the prediction has to bring in the previous admissions history and free text diagnosis that are recorded immediately on admission. We propose a model that uses the multi-modal EHR structured medical codes and key demographic information to classify the LoS in 3 classes; Short Los (LoS⩽10 days), Medium LoS (10<LoS⩽30 days) and Long LoS (LoS>30 days) as well as mortality as a binary classification of a patient's death during current admission. The prediction has to use data available only within 24 h of admission. The key predictors include previous ICD9 diagnosis codes, ICD9 procedures, key demographic data, and free text diagnosis of the current admission recorded right on admission. We propose a Hierarchical Attention Network (HAN-LoS and HAN-Mor) model and train it to a dataset of over 45321 admissions recorded in the de-identified MIMIC-III dataset. For improved prediction, our attention mechanisms can focus on the most influential past admissions and most influential codes in these admissions. For fair performance evaluation, we implemented and compared the HAN model with previous approaches. With dataset balancing techniques HAN-LoS achieved an AUROC of over 0.82 and a Micro-F1 score of 0.24 and HAN-Mor achieved AUC-ROC of 0.87 hence outperforming the existing baselines that use structured medical codes as well as clinical time series for LoS and Mortality forecasting. By predicting mortality and LoS using the same model, we show that with little tuning the proposed model can be used for other clinical predictive tasks like phenotyping, decompensation,re-admission prediction, and survival analysis.

Intra-abdominal hypertension and hypoxic respiratory failure together predict adverse outcome - A sub-analysis of a prospective cohort

PURPOSE

To assess whether the combination of intra-abdominal hypertension (IAH, intra-abdominal pressure ≥ 12 mmHg) and hypoxic respiratory failure (HRF, PaO2/FiO2 ratio < 300 mmHg) in patients receiving invasive ventilation is an independent risk factor for 90- and 28-day mortality as well as ICU- and ventilation-free days.

METHODS

Mechanically ventilated patients who had blood gas analyses performed and intra-abdominal pressure measured, were included from a prospective cohort. Subgroups were defined by the absence (Group 1) or the presence of either IAH (Group 2) or HRF (Group 3) or both (Group 4). Mixed-effects regression analysis was performed.

RESULTS

Ninety-day mortality increased from 16% (Group 1, n = 50) to 30% (Group 2, n = 20) and 27% (Group 3, n = 100) to 49% (Group 4, n = 142), log-rank test p < 0.001. The combination of IAH and HRF was associated with increased 90- and 28-day mortality as well as with fewer ICU- and ventilation-free days. The association with 90-day mortality was no longer present after adjustment for independent variables. However, the association with 28-day mortality, ICU- and ventilation-free days persisted after adjusting for independent variables.

CONCLUSIONS

In our sub-analysis, the combination of IAH and HRF was not independently associated with 90-day mortality but independently increased the odds of 28-day mortality, and reduced the number of ICU- and ventilation-free days.

A Simple Scoring Tool to Predict Medical Intensive Care Unit Readmissions Based on Both Patient and Process Factors

BACKGROUND

Although many predictive models have been developed to risk assess medical intensive care unit (MICU) readmissions, they tend to be cumbersome with complex calculations that are not efficient for a clinician planning a MICU discharge.

OBJECTIVE

To develop a simple scoring tool that comprehensively takes into account not only patient factors but also system and process factors in a single model to predict MICU readmissions.

DESIGN

Retrospective chart review.

PARTICIPANTS

We included all patients admitted to the MICU of Robert Wood Johnson University Hospital, a tertiary care center, between June 2016 and May 2017 except those who were < 18 years of age, pregnant, or planned for hospice care at discharge.

MAIN MEASURES

Logistic regression models and a scoring tool for MICU readmissions were developed on a training set of 409 patients, and validated in an independent set of 474 patients.

KEY RESULTS

Readmission rate in the training and validation sets were 8.8% and 9.1% respectively. The scoring tool derived from the training dataset included the following variables: MICU admission diagnosis of sepsis, intubation during MICU stay, duration of mechanical ventilation, tracheostomy during MICU stay, non-emergency department admission source to MICU, weekend MICU discharge, and length of stay in the MICU. The area under the curve of the scoring tool on the validation dataset was 0.76 (95% CI, 0.68-0.84), and the model fit the data well (Hosmer-Lemeshow p = 0.644). Readmission rate was 3.95% among cases in the lowest scoring range and 50% in the highest scoring range.

CONCLUSION

We developed a simple seven-variable scoring tool that can be used by clinicians at MICU discharge to efficiently assess a patient's risk of MICU readmission. Additionally, this is one of the first studies to show an association between MICU admission diagnosis of sepsis and MICU readmissions.

Prediction of Prolonged Intensive Care Unit Length of Stay Following Cardiac Surgery

Intensive care unit (ICU) costs comprise a significant proportion of the total inpatient charges for cardiac surgery. No reliable method for predicting intensive care unit length of stay following cardiac surgery exists, making appropriate staffing and resource allocation challenging. We sought to develop a predictive model to anticipate prolonged ICU length of stay (LOS). All patients undergoing coronary artery bypass grafting (CABG) and/or valve surgery with a Society of Thoracic Surgeons (STS) predicted risk score were evaluated from an institutional STS database. Models were developed using 2014-2017 data; validation used 2018-2019 data. Prolonged ICU LOS was defined as requiring ICU care for at least three days postoperatively. Predictive models were created using lasso regression and relative utility compared. A total of 3283 patients were included with 1669 (50.8%) undergoing isolated CABG. Overall, 32% of patients had prolonged ICU LOS. Patients with comorbid conditions including severe COPD (53% vs 29%, P < 0.001), recent pneumonia (46% vs 31%, P < 0.001), dialysis-dependent renal failure (57% vs 31%, P < 0.001) or reoperative status (41% vs 31%, P < 0.001) were more likely to experience prolonged ICU stays. A prediction model utilizing preoperative and intraoperative variables correctly predicted prolonged ICU stay 76% of the time. A preoperative variable-only model exhibited 74% prediction accuracy. Excellent prediction of prolonged ICU stay can be achieved using STS data. Moreover, there is limited loss of predictive ability when restricting models to preoperative variables. This novel model can be applied to aid patient counseling, resource allocation, and staff utilization.

A model of workflow in the hospital during a pandemic to assist management

We present a computational model of workflow in the hospital during a pandemic. The objective is to assist management in anticipating the load of each care unit, such as the ICU, or ordering supplies, such as personal protective equipment, but also to retrieve key parameters that measure the performance of the health system facing a new crisis. The model was fitted with good accuracy to France’s data set that gives information on hospitalized patients and is provided online by the French government. The goal of this work is both practical in offering hospital management a tool to deal with the present crisis of COVID-19 and offering a conceptual illustration of the benefit of computational science during a pandemic.

What factors predict length of stay in the intensive care unit? Systematic review and meta-analysis

PURPOSE

Studies have shown that a small percentage of ICU patients have prolonged length of stay (LoS) and account for a large proportion of resource use. Therefore, the identification of prolonged stay patients can improve unit efficiency. In this study, we performed a systematic review and meta-analysis to understand the risk factors of ICU LoS.

MATERIALS AND METHODS

We searched MEDLINE, Embase and Scopus databases from inception to November 2018. The searching process focused on papers presenting risk factors of ICU LoS. A meta-analysis was performed for studies reporting appropriate statistics.

RESULTS

From 6906 citations, 113 met the eligibility criteria and were reviewed. A meta-analysis was performed for six factors from 28 papers and concluded that patients with mechanical ventilation, hypomagnesemia, delirium, and malnutrition tend to have longer stay, and that age and gender were not significant factors.

CONCLUSIONS

This work suggested a list of risk factors that should be considered in prediction models for ICU LoS, as follows: severity scores, mechanical ventilation, hypomagnesemia, delirium, malnutrition, infection, trauma, red blood cells, and PaO2:FiO2. Our findings can be used by prediction models to improve their predictive capacity of prolonged stay patients, assisting in resource allocation, quality improvement actions, and benchmarking analysis.

Predictors of Trauma High Resource Consumers in a Mature Trauma System

Background

Extended hospital length of stay (LOS) is widely associated with significant healthcare costs. Since LOS is a known surrogate for cost, we sought to evaluate outliers. We hypothesized that particular characteristics are likely predictive of trauma high resource consumers (THRC) and can be used to more effectively manage care of this population.

Methods

The Pennsylvania Trauma Outcome Study database was retrospectively queried from 2003-2017 for all adult (age ≥15) trauma patients admitted to accredited trauma centers in Pennsylvania. THRC were defined as patients with hospital LOS two standard deviations above the population mean or ≥22 days (p<0.05). Patient demographics, comorbid conditions and clinical variables were compared between THRC and non-THRC to identify potential predictor variables. A multilevel mixed-effects logistic regression model controlling for age, gender, injury severity, admission Glasgow coma score, systolic blood pressure, and injury year assessed the adjusted impact of clinical factors in predicting THRC status. The National Trauma Data Bank (NTDB) was retrospectively queried from 2014-2016 for all adult (age ≥15) trauma patients admitted to state-accredited trauma centers and likewise were assessed for factors associated with THRC.

Results

A total of 465,601 patients met inclusion criteria [THRC: 16,818 (3.6%); non-THRC 448,783 (96.4%)]. Compared to non-THRC counterparts, THRC patients were significantly more severely injured (median ISS: 9 vs. 22, p<0.001). In adjusted analysis, gunshot wound (GSW) to the abdomen, undergoing major surgery and reintubation along with injury to the spine, upper or lower extremities were significantly associated with THRC. From the NTDB, 2 323 945 patients met inclusion criteria. In adjusted analysis, GSW to the abdomen was significantly associated with THRC. Penetrating injury overall was associated with decreased risk of being a THRC in the NTDB dataset. Those who had either GSW to abdomen, surgery, or reintubation required significantly longer LOS (p<0.001).

Conclusions

Reintubation, major surgery, gunshot wound to abdomen, along with injury to the spine, upper or lower extremities are all strongly predictive of THRC. Understanding the profile of the THRC will allow clinicians and case management to proactively put processes in place to streamline care and potentially reduce costs and LOS.

An evaluation of time series summary statistics as features for clinical prediction tasks

Background

Clinical prediction tasks such as patient mortality, length of hospital stay, and disease diagnosis are highly important in critical care research. The existing studies for clinical prediction mainly used simple summary statistics to summarize information from physiological time series. However, this lack of statistics leads to a lack of information. In addition, using only maximum and minimum statistics to indicate patient features fails to provide an adequate explanation. Few studies have evaluated which summary statistics best represent physiological time series.

Methods

In this paper, we summarize 14 statistics describing the characteristics of physiological time series, including the central tendency, dispersion tendency, and distribution shape. Then, we evaluate the use of summary statistics of physiological time series as features for three clinical prediction tasks. To find the combinations of statistics that yield the best performances under different tasks, we use a cross-validation-based genetic algorithm to approximate the optimal statistical combination.

Results

By experiments using the EHRs of 6,927 patients, we obtained prediction results based on both single statistics and commonly used combinations of statistics under three clinical prediction tasks. Based on the results of an embedded cross-validation genetic algorithm, we obtained 25 optimal sets of statistical combinations and then tested their prediction results. By comparing the performances of prediction with single statistics and commonly used combinations of statistics with quantitative analyses of the optimal statistical combinations, we found that some statistics play central roles in patient representation and different prediction tasks have certain commonalities.

Conclusion

Through an in-depth analysis of the results, we found many practical reference points that can provide guidance for subsequent related research. Statistics that indicate dispersion tendency, such as min, max, and range, are more suitable for length of stay prediction tasks, and they also provide information for short-term mortality prediction. Mean and quantiles that reflect the central tendency of physiological time series are more suitable for mortality and disease prediction. Skewness and kurtosis perform poorly when used separately for prediction but can be used as supplementary statistics to improve the overall prediction effect.

Long-stay pediatric patients in Japanese intensive care units: their significant presence and a newly developed, simple predictive score

Background

The length of stay (LOS) in intensive care units (ICUs) has been used as a good indicator not only for resource consumption but also for health outcomes of patients. However, data regarding pediatric LOS in Japanese ICUs are limited. The primary aim of this study was to characterize the Japanese pediatric ICU patients based on their LOS. Second, we aimed to develop a simple scoring system to predict long-stay pediatric ICU patients on admission.

Methods

We performed a retrospective cohort study using consecutive pediatric data (aged < 16 years) registered in the Japanese Registry of Pediatric Acute Care (JaRPAC) from October 2013 to September 2016, which consisted of descriptive and diagnostic information. The factors for long-stay patients (LSPs; LOS > 14 days) were identified using multiple regression analysis, and subsequently, a simple predictive scoring system was developed based on the results. The validity of the score was prospectively tested using data from the JaRPAC registration from October 2016 to September 2017.

Results

Overall, 4107 patients were included. Although LSPs were few (8.0% [n = 330]), they consumed 38.0% of ICU bed days (9750 for LSPs versus 25,659 overall). Mortality was seven times higher in LSPs than in short-stay patients (9.1% versus 1.3%). An 11-variable simple predictive scoring system was constructed, including Pediatric Index of Mortality 2 ≥ 1 (2 points), liver dysfunction (non-post operation) (2 points), post-cardiopulmonary resuscitation (1 point), circulatory disorder (1 point), post-operative management of liver transplantation (1 point), encephalitis/encephalopathy (1 point), myocarditis/cardiomyopathy (1 point), congenital heart disease (non-post operation) (1 point), lung tissue disease (1 point), Pediatric Cerebral Performance Category scores ≥ 2 (1 point), and age < 2 years (1 point). A score of ≥ 3 points yielded an area under the receiver operating characteristic curve (AUC) of 0.79, sensitivity of 87.0%, and specificity of 59.4% in the original dataset. Reproducibility was confirmed with the internal validation dataset (AUC 0.80, sensitivity 92.6%, and specificity 60.2%).

Conclusions

Pediatric LSPs possess a significant presence in Japanese ICUs with high rates of bed utilization and mortality. The newly developed predictive scoring system may identify pediatric LSPs on admission.

Electronic supplementary material

The online version of this article (10.1186/s40560-019-0392-2) contains supplementary material, which is available to authorized users.

Bedside-measurement of serum cholinesterase activity predicts patient morbidity and length of the intensive care unit stay following major traumatic injury

Major traumatic injury (MTI), a life-threatening condition requiring prompt medical intervention, is associated with an extensive inflammatory response often resulting in multiple organ dysfunction. Early stratification of trauma severity and the corresponding inflammation may help optimize resources at the intensive care unit (ICU). The cholinergic system counters inflammation by quickly modulating the immune response. Serum cholinesterase (butyrylcholinesterase, BChE) is an enzyme that hydrolyses acetylcholine. We tested whether a change in the BChE activity correlates with the morbidity and the length of ICU stay. Blood samples from 10 healthy volunteers and 44 patients with MTI were gathered at hospital admission, followed by measurements 12, 24 and 48 hours later. Point-of-care approach was used to determine the BChE activity. Disease severity was assessed by clinical scoring performed within 24 hours following hospital admission. BChE activity, measured at hospital admission, showed a significant and sustained reduction and correlated with disease severity scores obtained 24 hours following admission. BChE activity, obtained at hospital admission, correlated with the length of ICU stay. Bedside measurement of BChE activity, as a complementary addition to established procedures, might prove useful in the primary assessment of the disease severity and might therefore optimize therapy in the ICU.

Relationship between availability of physiotherapy services and ICU costs

OBJECTIVE

To determine whether 24-h availability of physiotherapy services decreases ICU costs in comparison with the standard 12 h/day availability among patients admitted to the ICU for the first time.

METHODS

This was an observational prevalence study involving 815 patients ≥ 18 years of age who had been on invasive mechanical ventilation (IMV) for ≥ 24 h and were discharged from an ICU to a ward at a tertiary teaching hospital in Brazil. The patients were divided into two groups according to h/day availability of physiotherapy services in the ICU: 24 h (PT-24; n = 332); and 12 h (PT-12; n = 483). The data collected included the reasons for hospital and ICU admissions; Acute Physiology and Chronic Health Evaluation II (APACHE II) score; IMV duration, ICU length of stay (ICU-LOS); and Omega score.

RESULTS

The severity of illness was similar in both groups. Round-the-clock availability of physiotherapy services was associated with shorter IMV durations and ICU-LOS, as well as with lower total, medical, and staff costs, in comparison with the standard 12 h/day availability.

CONCLUSIONS

In the population studied, total costs and staff costs were lower in the PT-24 group than in the PT-12 group. The h/day availability of physiotherapy services was found to be a significant predictor of ICU costs.

Pre-kidney transplant lower extremity impairment and transplant length of stay: a time-to-discharge analysis of a prospective cohort study

BACKGROUND

Few objective tests can be performed at admission for kidney transplantation [KT] to discern risk of increased length of stay [LOS], which is important for patient counseling and is associated with increased costs and mortality. The short physical performance battery [SPPB] is an easily administered, potentially modifiable, 3-part test of lower extremity function. SPPB score is associated with longer hospital LOS in older adults, and may provide similar utility in KT recipients given that ESRD is a disease of accelerated aging. The aim of this study was to characterize the association between SPPB-derived lower extremity function and LOS.

METHODS

The SPPB was administered at KT admission in a prospective cohort of 595 recipients (8/2009-6/2016). The independent association between SPPB impairment (score ≤ 10) and LOS was tested with an adjusted conventional generalized gamma parametric survival model.

RESULTS

Impaired recipients experienced longer LOS (median: 10 vs. 8 days; P <  0.001) with the greatest difference in percent discharged on day 10 (impaired: 54.5%, unimpaired: 73.3%). Discharge typically took 13% longer in the impaired group (relative time = 1.13; 95%CI: 1.05, 1.21, P = 0.001). Discharge for impaired recipients compared to unimpaired was least likely at day 5 (hazard ratio = 0.71; 95% CI:0.68, 0.74, P <  0.001). No differences in the SPPB impairment-LOS relationship were found by age (interaction P = 0.74).

CONCLUSIONS

Pre-KT SPPB impairment was independently associated with longer LOS regardless of age, indicating that it is a useful, objective tool for pre-KT risk assessment in younger and older recipients that may help inform discharge planning.

Superusers: Drivers of health care resource utilization in the national trauma population

BACKGROUND

Health care spending is driven by a very small percentage of Americans, many of whom are patients with prolonged durations of stay. The objective of this study was to characterize superusers in the trauma population.

METHODS

The National Trauma Data Bank for 2008-2012 was queried. Superusers were defined as those with a duration of stay in the top 0.06% of the population and were compared with the remainder of the population to determine differences in demographic characteristics, comorbidities, prehospital factors, and outcomes. Multivariate analysis was used to determine independent predictors of being classified as a superuser.

RESULTS

A total of 3,617,261 patients met inclusion criteria, with 34,728 qualifying as superusers. Mean duration of stay for superusers was 58.7 days compared with the average 4.6 days (P < .001). Superusers were more likely to be male, black, Medicaid insured, and have a higher Injury Severity Score and lower Glasgow Coma Scale score. The hospital course of superusers was likely to be complicated by pneumonia, acute respiratory distress syndrome, decubitus ulcer, and acute kidney injury.

CONCLUSION

Age, sex, race, and insurance were associated with prolonged use of inpatient care in the trauma patient population. Specific comorbidities and complications are associated with being a superuser. This subset of the trauma population confers a disproportionate burden on the health care system and can serve as a potential target for intervention.

Inclusion of Unstructured Clinical Text Improves Early Prediction of Death or Prolonged ICU Stay

OBJECTIVES

Early prediction of undesired outcomes among newly hospitalized patients could improve patient triage and prompt conversations about patients' goals of care. We evaluated the performance of logistic regression, gradient boosting machine, random forest, and elastic net regression models, with and without unstructured clinical text data, to predict a binary composite outcome of in-hospital death or ICU length of stay greater than or equal to 7 days using data from the first 48 hours of hospitalization.

DESIGN

Retrospective cohort study with split sampling for model training and testing.

SETTING

A single urban academic hospital.

PATIENTS

All hospitalized patients who required ICU care at the Beth Israel Deaconess Medical Center in Boston, MA, from 2001 to 2012.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Among eligible 25,947 hospital admissions, we observed 5,504 (21.2%) in which patients died or had ICU length of stay greater than or equal to 7 days. The gradient boosting machine model had the highest discrimination without (area under the receiver operating characteristic curve, 0.83; 95% CI, 0.81-0.84) and with (area under the receiver operating characteristic curve, 0.89; 95% CI, 0.88-0.90) text-derived variables. Both gradient boosting machines and random forests outperformed logistic regression without text data (p < 0.001), whereas all models outperformed logistic regression with text data (p < 0.02). The inclusion of text data increased the discrimination of all four model types (p < 0.001). Among those models using text data, the increasing presence of terms "intubated" and "poor prognosis" were positively associated with mortality and ICU length of stay, whereas the term "extubated" was inversely associated with them.

CONCLUSIONS

Variables extracted from unstructured clinical text from the first 48 hours of hospital admission using natural language processing techniques significantly improved the abilities of logistic regression and other machine learning models to predict which patients died or had long ICU stays. Learning health systems may adapt such models using open-source approaches to capture local variation in care patterns.

[Cost analysis as a tool for assessing the efficacy of intensive care units]

BACKGROUND

The German "Hospital Structure Act" intends to align the state hospital planning on quality criteria. Within this process cost-utility analyses (CUAs) shall be used to assess the efficacy of medical care. To be objective, CUAs of intensive care units (ICUs) require standardization (adjustment) of costs. The present study analyzed the extent to which treatment costs are related to patient-specific baseline variables (such as type and severity of the primary disease).

METHODS

From 2000-2004, a bottom-up procedure was used to quantify total costs on 14 ICUs in nine German university hospitals. Results were combined with demographic data, and data indicating type (ICD-10 codes) and severity (ICU scoring systems) of the primary disease at ICU admission. Various statistical models were tested to identify that which best described the associations between baseline variables and costs.

RESULTS

In all, 3803 critically ill patients could be examined. The median of treatment costs per patient was 3199 € (IQR 1768-6659 €). No model allowed an acceptably precise adjustment of costs; the estimated mean absolute prognostic error was at least 3860 € (mean relative prognostic error 66%), when we tested an Extreme Gradient Boosting Model.

CONCLUSION

Instruments which are currently available (cost adjustment based on patient-specific baseline variables) do not allow a standardization of costs, and an objective CUA of ICUs. Factors unknown at baseline may cause a large portion of treatment costs.

Variations in Case-Mix-Adjusted Duration of Mechanical Ventilation Among ICUs

OBJECTIVES

To develop a model that predicts the duration of mechanical ventilation and then to use this model to compare observed versus expected duration of mechanical ventilation across ICUs.

DESIGN

Retrospective cohort analysis.

SETTING

Eighty-six eligible ICUs at 48 U.S. hospitals.

PATIENTS

ICU patients receiving mechanical ventilation on day 1 (n = 56,336) admitted from January 2013 to September 2014.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We developed and validated a multivariable logistic regression model for predicting duration of mechanical ventilation using ICU day 1 patient characteristics. Mean observed minus expected duration of mechanical ventilation was then obtained across patients and for each ICU. The accuracy of the model was assessed using R. We defined better performing units as ICUs that had an observed minus expected duration of mechanical ventilation less than -0.5 days and a p value of less than 0.01; and poorer performing units as ICUs with an observed minus expected duration of mechanical ventilation greater than +0.5 days and a p value of less than 0.01. The factors accounting for the majority of the model's explanatory power were diagnosis (71%) and physiologic abnormalities (24%). For individual patients, the difference between observed and mean predicted duration of mechanical ventilation was 3.3 hours (95% CI, 2.8-3.9) with R equal to 21.6%. The mean observed minus expected duration of mechanical ventilation across ICUs was 3.8 hours (95% CI, 2.1-5.5), with R equal to 69.9%. Among the 86 ICUs, 66 (76.7%) had an observed mean mechanical ventilation duration that was within 0.5 days of predicted. Five ICUs had significantly (p < 0.01) poorer performance (observed minus expected duration of mechanical ventilation, > 0.5 d) and 14 ICUs significantly (p < 0.01) better performance (observed minus expected duration of mechanical ventilation, < -0.5 d).

CONCLUSIONS

Comparison of observed and case-mix-adjusted predicted duration of mechanical ventilation can accurately assess and compare duration of mechanical ventilation across ICUs, but cannot accurately predict an individual patient's mechanical ventilation duration. There are substantial differences in duration of mechanical ventilation across ICU and their association with unit practices and processes of care warrants examination.

Which Models Can I Use to Predict Adult ICU Length of Stay? A Systematic Review

OBJECTIVE

We systematically reviewed models to predict adult ICU length of stay.

DATA SOURCES

We searched the Ovid EMBASE and MEDLINE databases for studies on the development or validation of ICU length of stay prediction models.

STUDY SELECTION

We identified 11 studies describing the development of 31 prediction models and three describing external validation of one of these models.

DATA EXTRACTION

Clinicians use ICU length of stay predictions for planning ICU capacity, identifying unexpectedly long ICU length of stay, and benchmarking ICUs. We required the model variables to have been published and for the models to be free of organizational characteristics and to produce accurate predictions, as assessed by R across patients for planning and identifying unexpectedly long ICU length of stay and across ICUs for benchmarking, with low calibration bias. We assessed the reporting quality using the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies.

DATA SYNTHESIS

The number of admissions ranged from 253 to 178,503. Median ICU length of stay was between 2 and 6.9 days. Two studies had not published model variables and three included organizational characteristics. None of the models produced predictions with low bias. The R was 0.05-0.28 across patients and 0.01-0.64 across ICUs. The reporting scores ranged from 49 of 78 to 60 of 78 and the methodologic scores from 12 of 22 to 16 of 22.

CONCLUSION

No models completely satisfy our requirements for planning, identifying unexpectedly long ICU length of stay, or for benchmarking purposes. Physicians using these models to predict ICU length of stay should interpret them with reservation.

Predicting Prolonged Stay in the ICU Attributable to Bleeding in Patients Offered Plasma Transfusion

In blood transfusion studies, plasma transfusion (PPT) and bleeding are known to be associated with risk of prolonged ICU length of stay (ICU-LOS). However, as patients can show significant heterogeneity in response to a treatment, there might exists subgroups with differential effects. The existence and characteristics of these subpopulations in blood transfusion has not been well-studied. Further, the impact of bleeding in patients offered PPT on prolonged ICU-LOS is not known. This study presents a causal and predictive framework to examine these problems. The two-step approach first estimates the effect of bleeding in PPT patients on prolonged ICU-LOS and then estimates risks of bleeding and prolonged ICU-LOS. The framework integrates a classification model for risks prediction and a regression model to predict actual LOS. Results showed that the effect of bleeding in PPT patients significantly increases risk of prolonged ICU-LOS (55%, p=0.00) while no bleeding significantly reduces ICU-LOS (4%, p=0.046).

Random Survival Forests for Predicting the Bed Occupancy in the Intensive Care Unit

Predicting the bed occupancy of an intensive care unit (ICU) is a daunting task. The uncertainty associated with the prognosis of critically ill patients and the random arrival of new patients can lead to capacity problems and the need for reactive measures. In this paper, we work towards a predictive model based on Random Survival Forests which can assist physicians in estimating the bed occupancy. As input data, we make use of the Sequential Organ Failure Assessment (SOFA) score collected and calculated from 4098 patients at two ICU units of Ghent University Hospital over a time period of four years. We compare the performance of our system with a baseline performance and a standard Random Forest regression approach. Our results indicate that Random Survival Forests can effectively be used to assist in the occupancy prediction problem. Furthermore, we show that a group based approach, such as Random Survival Forests, performs better compared to a setting in which the length of stay of a patient is individually assessed.

The role of ascitic fluid viscosity in differentiating the nature of ascites and in the prediction of renal impairment and duration of ICU stay

BACKGROUND AND AIM

Serum-ascites albumin gradient (SAAG) has been used in the classification of ascites for the last 20 years but it has some drawbacks. This study searches for possible correlations between ascitic fluid viscosity and the etiology of ascites, renal impairment, and length of ICU stay.

MATERIALS AND METHODS

The study was conducted in Zagazig University Hospital, Egypt. It included 240 patients with ascites due to various causes. The patients were divided into two groups: the cirrhotic ascites group, which included 120 patients, and the noncirrhotic ascites group, which included 120 patients. Ascitic patients on medical management with diuretics, antibiotics, paracentesis, and infusion of plasma or albumin were excluded.The laboratory analysis included routine investigations to detect the cause of ascites as well as specific investigations such as ascitic fluid viscosity using a falling ball viscosimeter (microviscosimeter) at 37°C.

RESULTS

The mean ascitic viscosity of patients with SAAG at least 1.1 was 1.16±0.56, which was associated with serum creatinine 1.35±0.52 mg/dl and ICU stay of 3.3±1.2 days. In patients with SAAG less than 1.1 g/dl, the mean ascitic viscosity was 2.98±0.87, with serum creatinine 2.1±0.56 mg/dl and ICU stay of 7.1±1.3 days. Ascitic viscosity can discriminate ascites due to portal hypertension from those associated with nonportal hypertension at a cut-off value of 1.65; it can predict renal impairment in hepatic patients at a cut-off of 1.35 and long ICU stay at a cut-off of 1.995 using receiver operating characteristic analysis.

CONCLUSION

Ascitic viscosity measurement is rapid, inexpensive, and requires small sample volumes. Ascitic viscosity can discriminate ascites due to portal hypertension from those associated with nonportal hypertension at a cut-off value of 1.65. It can predict renal impairment in hepatic patients at a cut-off of 1.35 and long ICU stay at a cut-off of 1.995.

Timing of onset and burden of persistent critical illness in Australia and New Zealand: a retrospective, population-based, observational study

BACKGROUND

Critical care physicians recognise persistent critical illness as a specific syndrome, yet few data exist for the timing of the transition from acute to persistent critical illness. Defining the onset of persistent critical illness as the time at which diagnosis and illness severity at intensive care unit (ICU) arrival no longer predict outcome better than do simple pre-ICU patient characteristics, we measured the timing of this onset at a population level in Australia and New Zealand, and the variation therein, and assessed the characteristics, burden of care, and hospital outcomes of patients with persistent critical illness.

METHODS

In this retrospective, population-based, observational study, we used data for ICU admission in Australia and New Zealand from the Australian and New Zealand Intensive Care Society Adult Patient Database. We included all patients older than 16 years of age admitted to a participating ICU. We excluded patients transferred from another hospital and those admitted to an ICU for palliative care or awaiting organ donation. The primary outcome was in-hospital mortality. Using statistical methods in evenly split development and validation samples for risk score development, we examined the ability of characteristics to predict in-hospital mortality.

FINDINGS

Between Jan, 2000, and Dec, 2014, we studied 1 028 235 critically ill patients from 182 ICUs across Australia and New Zealand. Among patients still in an ICU, admission diagnosis and physiological derangements, which accurately predicted outcome on admission (area under the receiver operating characteristics curve 0·898 [95% CI 0·897-0·899] in the validation cohort), progressively lost their predictive ability and no longer predicted outcome more accurately than did simple antecedent patient characteristics (eg, age, sex, or chronic health status) after 10 days in the ICU, thus empirically defining the onset of persistent critical illness. This transition occurred between day 7 and day 22 across diagnosis-based subgroups and between day 6 and day 15 across risk-of-death-based subgroups. Cases of persistent critical illness accounted for only 51 509 (5·0%) of the 1 028 235 patients admitted to an ICU, but for 1 029 345 (32·8%) of 3 138 432 ICU bed-days and 2 197 108 (14·7%) of 14 961 693 hospital bed-days. Overall, 12 625 (24·5%) of 51 509 patients with persistent critical illness died and only 23 968 (46·5%) of 51 509 were discharged home.

INTERPRETATION

Onset of persistent critical illness can be empirically measured at a population level. Patients with this condition consume vast resources, have high mortality, have much less chance of returning home than do typical ICU patients, and require dedicated future research. ICU clinicians should be aware that the risk of in-hospital mortality can change quickly over the first 2 weeks of an ICU course and be sure to incorporate such changes in their decision making and prognostication.

FUNDING

None.

The association between the Th-17 immune response and pulmonary complications in a trauma ICU population

BACKGROUND

The overall immunopathology of the T-helper cell (Th)-17 immune response has been implicated in various inflammatory diseases including pulmonary inflammation; however its potential role in acute respiratory distress syndrome (ARDS) is not defined. This study aimed to evaluate the Th-17 response in bronchoalveolar lavage fluid (BALF) and blood and from trauma patients with pulmonary complications.

METHODS

A total of 21 severely injured intensive care unit (ICU) subjects, who were mechanically ventilated and undergoing bronchoscopy, were enrolled. BALF and blood were collected and analyzed for Th-1 (interferon [IFN]γ), Th-2 (interleukin [IL]-4, -10), Th-17 (IL-17A, -17F, -22, 23) and pro-inflammatory (IL-1β, IL-6, tumor necrosis factor [TNF]α) cytokine levels.

RESULTS

Significant levels of the Th-17 cytokines IL-17A, -17F and -21 and IL-6 (which can be classified as a Th-17 cytokine) were observed in the BALF of all subjects. There were no significant differences in Th-17 cytokines between those subjects with ARDS and those without, with the exception of plasma and BALF IL-6, which was markedly greater in ARDS subjects, as compared with controls and non-ARDS subjects.

CONCLUSIONS

Trauma patients with pulmonary complications exhibited a significant Th-17 response in the lung and blood, suggesting that this pro-inflammatory milieu may be a contributing factor to such complications.

Evidence-Based Pediatric Outcome Predictors to Guide the Allocation of Critical Care Resources in a Mass Casualty Event

OBJECTIVE

ICU resources may be overwhelmed by a mass casualty event, triggering a conversion to Crisis Standards of Care in which critical care support is diverted away from patients least likely to benefit, with the goal of improving population survival. We aimed to devise a Crisis Standards of Care triage allocation scheme specifically for children.

DESIGN

A triage scheme is proposed in which patients would be divided into those requiring mechanical ventilation at PICU presentation and those not, and then each group would be evaluated for probability of death and for predicted duration of resource consumption, specifically, duration of PICU length of stay and mechanical ventilation. Children will be excluded from PICU admission if their mortality or resource utilization is predicted to exceed predetermined levels ("high risk"), or if they have a low likelihood of requiring ICU support ("low risk"). Children entered into the Virtual PICU Performance Systems database were employed to develop prediction equations to assign children to the exclusion categories using logistic and linear regression. Machine Learning provided an alternative strategy to develop a triage scheme independent from this process.

SETTING

One hundred ten American PICUs

SUBJECTS

: One hundred fifty thousand records from the Virtual PICU database.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

The prediction equations for probability of death had an area under the receiver operating characteristic curve more than 0.87. The prediction equation for belonging to the low-risk category had lower discrimination. R for the prediction equations for PICU length of stay and days of mechanical ventilation ranged from 0.10 to 0.18. Machine learning recommended initially dividing children into those mechanically ventilated versus those not and had strong predictive power for mortality, thus independently verifying the triage sequence and broadly verifying the algorithm.

CONCLUSION

An evidence-based predictive tool for children is presented to guide resource allocation during Crisis Standards of Care, potentially improving population outcomes by selecting patients likely to benefit from short-duration ICU interventions.

A history of outcome prediction in the ICU

PURPOSE OF REVIEW

There are few first-hand accounts that describe the history of outcome prediction in critical care. This review summarizes the authors' personal perspectives about the development and evolution of Acute Physiology and Chronic Health Evaluation over the past 35 years.

RECENT FINDINGS

We emphasize what we have learned in the past and more recently our perspectives about the current status of outcome prediction, and speculate about the future of outcome prediction.

SUMMARY

There is increasing evidence that superior accuracy in outcome prediction requires complex modeling with detailed adjustment for diagnosis and physiologic abnormalities. Thus, an automated electronic system is recommended for gathering data and generating predictions. Support, either public or private, is required to assist users and to update and improve models. Current outcome prediction models have increasingly focused on benchmarks for resource use, a trend that seems likely to increase in the future.

Ventilação mecânica e a lesão renal aguda em pacientes na unidade de terapia intensiva

Objetivo Verificar o impacto do emprego da ventilação mecânica em pacientes internados na Unidade de Terapia Intensiva e a ocorrência de lesão renal aguda. Métodos Estudo de coorte, prospectivo, quantitativo, desenvolvido com 27 pacientes sob suporte de ventilação mecânica internados na unidade de terapia intensiva em um hospital público. Resultados A maioria (55,6%) dos pacientes foi classificada no estágio de lesão renal, de acordo com a classificação Risk, Injury, Failure, Loss, End-Stage (RIFLE). Dentre os pacientes, 45,8% estavam sob ventilação mecânica com pressão expiratória final positiva entre 5cmH2O e 10cmH2O, os quais evoluíram com lesão renal aguda. Acute Physiology and Chronic Health Disease Classification System II (APACHE II) apresentou associação significativa com disfunção renal (p=0,046). Conclusão O emprego da ventilação mecânica invasiva com pressão expiratória final positiva em pacientes graves pode determinar prejuízos à função renal dos pacientes internados em unidade de terapia intensiva.

Would triage predictors perform better than first-come, first-served in pandemic ventilator allocation?

BACKGROUND

In a pandemic, needs for ventilators might overwhelm the limited supply. Outcome predictors have been proposed to guide ventilator triage allocation decisions. However, pandemic triage predictors have not been validated. This quantitative simulation study evaluated outcomes resulting from allocation strategies varying in their performance for selecting short-stay survivors as favorable candidates for ventilators.

METHODS

A quantitative simulation modeled a pandemic surge. Postulated numbers of potential daily admissions presented randomly from a specified population, with a limited number of available ventilators. Patients were triaged to ventilator care vs palliation or turned away to palliation if no ventilator was available. Simulated triage was conducted according to a set of hypothetical triage tools varying in sensitivity and specificity to select favorable ventilator candidates vs first-come, first-served allocation. Death was assumed for palliation. Survival or death was counted for patients who were ventilated according to the specified characteristic of each randomly selected patient.

RESULTS

Triage predictors with intermediate-quality performance resulted in a median daily mortality of 80%, similar to first-come, first-served allocation. A poor-quality predictor resulted in a worse mortality of 90%. Only a high-quality predictor (sensitivity 90%, specificity 90%) resulted in a substantially lower 60% mortality.

CONCLUSIONS

Performance of unvalidated pandemic ventilator triage predictors is unknown and possibly inferior to first-come, first-served allocation. Poor performance of unvalidated predictors proposed for triage would represent an inadequate plan for stewarding scarce resources and would deprive some patients of fair access to a ventilator, thus falling short of sound ethical foundations.

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration

The TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) Statement includes a 22-item checklist, which aims to improve the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. This explanation and elaboration document describes the rationale; clarifies the meaning of each item; and discusses why transparent reporting is important, with a view to assessing risk of bias and clinical usefulness of the prediction model. Each checklist item of the TRIPOD Statement is explained in detail and accompanied by published examples of good reporting. The document also provides a valuable reference of issues to consider when designing, conducting, and analyzing prediction model studies. To aid the editorial process and help peer reviewers and, ultimately, readers and systematic reviewers of prediction model studies, it is recommended that authors include a completed checklist in their submission. The TRIPOD checklist can also be downloaded from www.tripod-statement.org.