Using an Artificial Neural Networks (ANNs) Model for Prediction of Intensive Care Unit (ICU) Outcome and Length of Stay at Hospital in Traumatic Patients

Hematocrit and lactate trends help predict outcomes in trauma independent of CT and other clinical parameters

Background

Hematocrit and lactate have an established role in trauma as indicators of bleeding and cell death, respectively. The wide availability of CT imaging and clinical data poses the question of how these can be used in combination to predict outcomes.

Purpose

To assess the utility of hematocrit or lactate trends in predicting intensive care unit (ICU) admission and hospital length of stay (LOS) in patients with torso trauma combined with clinical parameters and injury findings on CT.

Materials and Methods

This was a single-center retrospective study of adults with torso trauma in one year. Trends were defined as a unit change per hour. CT findings and clinical parameters were explanatory variables. Outcomes were ICU admission and hospital LOS. Multivariate logistic and negative binomial regression models were used to calculate the odds ratio (OR) and incident rate ratio (IRR).

Results

Among 840 patients, 561 (72% males, age 39 ± 18) were included, and 168 patients (30%) were admitted to the ICU. Decreasing hematocrit trend [OR 2.54 (1.41-4.58), p = 0.002] and increasing lactate trend [OR 3.85 (1.35-11.01), p = 0.012] were associated with increased odds of ICU admission. LOS median was 2 (IQR: 1-5) days. Decreasing hematocrit trend [IRR 1.37 (1.13-1.66), p = 0.002] and increasing lactate trend [2.02 (1.43-2.85), p < 0.001] were associated with longer hospital LOS.

Conclusion

Hematocrit and lactate trends may be helpful in predicting ICU admission and LOS in torso trauma independent of organ injuries on CT, age, or admission clinical parameters.

A Machine Learning-Based Approach to Predict Prognosis and Length of Hospital Stay in Adults and Children With Traumatic Brain Injury: Retrospective Cohort Study

BACKGROUND

The treatment and care of adults and children with traumatic brain injury (TBI) constitute an intractable global health problem. Predicting the prognosis and length of hospital stay of patients with TBI may improve therapeutic effects and significantly reduce societal health care burden. Applying novel machine learning methods to the field of TBI may be valuable for determining the prognosis and cost-effectiveness of clinical treatment.

OBJECTIVE

We aimed to combine multiple machine learning approaches to build hybrid models for predicting the prognosis and length of hospital stay for adults and children with TBI.

METHODS

We collected relevant clinical information from patients treated at the Neurosurgery Center of the Second Affiliated Hospital of Anhui Medical University between May 2017 and May 2022, of which 80% was used for training the model and 20% for testing via screening and data splitting. We trained and tested the machine learning models using 5 cross-validations to avoid overfitting. In the machine learning models, 11 types of independent variables were used as input variables and Glasgow Outcome Scale score, used to evaluate patients' prognosis, and patient length of stay were used as output variables. Once the models were trained, we obtained and compared the errors of each machine learning model from 5 rounds of cross-validation to select the best predictive model. The model was then externally tested using clinical data of patients treated at the First Affiliated Hospital of Anhui Medical University from June 2021 to February 2022.

RESULTS

The final convolutional neural network-support vector machine (CNN-SVM) model predicted Glasgow Outcome Scale score with an accuracy of 93% and 93.69% in the test and external validation sets, respectively, and an area under the curve of 94.68% and 94.32% in the test and external validation sets, respectively. The mean absolute percentage error of the final built convolutional neural network-support vector regression (CNN-SVR) model predicting inpatient time in the test set and external validation set was 10.72% and 10.44%, respectively. The coefficient of determination (R²) was 0.93 and 0.92 in the test set and external validation set, respectively. Compared with back-propagation neural network, CNN, and SVM models built separately, our hybrid model was identified to be optimal and had high confidence.

CONCLUSIONS

This study demonstrates the clinical utility of 2 hybrid models built by combining multiple machine learning approaches to accurately predict the prognosis and length of stay in hospital for adults and children with TBI. Application of these models may reduce the burden on physicians when assessing TBI and assist clinicians in the medical decision-making process.

Machine-learning prediction for hospital length of stay using a French medico-administrative database

INTRODUCTION

Prolonged Hospital Length of Stay (PLOS) is an indicator of deteriorated efficiency in Quality of Care. One goal of public health management is to reduce PLOS by identifying its most relevant predictors. The objective of this study is to explore Machine Learning (ML) models that best predict PLOS.

METHODS

Our dataset was collected from the French Medico-Administrative database (PMSI) as a retrospective cohort study of all discharges in the year 2015 from a large university hospital in France (APHM). The study outcomes were LOS transformed into a binary variable (long vs. short LOS) according to the 90^th percentile (14 days). Logistic regression (LR), classification and regression trees (CART), random forest (RF), gradient boosting (GB) and neural networks (NN) were applied to the collected data. The predictive performance of the models was evaluated using the area under the ROC curve (AUC).

RESULTS

Our analysis included 73,182 hospitalizations, of which 7,341 (10.0%) led to PLOS. The GB classifier was the most performant model with the highest AUC (0.810), superior to all the other models (all p-values <0.0001). The performance of the RF, GB and NN models (AUC ranged from 0.808 to 0.810) was superior to that of the LR model (AUC = 0.795); all p-values <0.0001. In contrast, LR was superior to CART (AUC = 0.786), p < 0.0001. The variable most predictive of the PLOS was the destination of the patient after hospitalization to other institutions. The typical clinical profile of these patients (17.5% of the sample) was the elderly patient, admitted in emergency, for a trauma, a neurological or a cardiovascular pathology, more often institutionalized, with more comorbidities notably mental health problems, dementia and hemiplegia.

DISCUSSION

The integration of ML, particularly the GB algorithm, may be useful for health-care professionals and bed managers to better identify patients at risk of PLOS. These findings underscore the need to strengthen hospitals through targeted allocation to meet the needs of an aging population.

Decision support by machine learning systems for acute management of severely injured patients: A systematic review

Introduction

Treating severely injured patients requires numerous critical decisions within short intervals in a highly complex situation. The coordination of a trauma team in this setting has been shown to be associated with multiple procedural errors, even of experienced care teams. Machine learning (ML) is an approach that estimates outcomes based on past experiences and data patterns using a computer-generated algorithm. This systematic review aimed to summarize the existing literature on the value of ML for the initial management of severely injured patients.

Methods

We conducted a systematic review of the literature with the goal of finding all articles describing the use of ML systems in the context of acute management of severely injured patients. MESH search of Pubmed/Medline and Web of Science was conducted. Studies including fewer than 10 patients were excluded. Studies were divided into the following main prediction groups: (1) injury pattern, (2) hemorrhage/need for transfusion, (3) emergency intervention, (4) ICU/length of hospital stay, and (5) mortality.

Results

Thirty-six articles met the inclusion criteria; among these were two prospective and thirty-four retrospective case series. Publication dates ranged from 2000 to 2020 and included 32 different first authors. A total of 18,586,929 patients were included in the prediction models. Mortality was the most represented main prediction group (n = 19). ML models used were artificial neural network ( n = 15), singular vector machine (n = 3), Bayesian network (n = 7), random forest (n = 6), natural language processing (n = 2), stacked ensemble classifier [SuperLearner (SL), n = 3], k-nearest neighbor (n = 1), belief system (n = 1), and sequential minimal optimization (n = 2) models. Thirty articles assessed results as positive, five showed moderate results, and one article described negative results to their implementation of the respective prediction model.

Conclusions

While the majority of articles show a generally positive result with high accuracy and precision, there are several requirements that need to be met to make the implementation of such models in daily clinical work possible. Furthermore, experience in dealing with on-site implementation and more clinical trials are necessary before the implementation of ML techniques in clinical care can become a reality.

A Robust Design-Based Expert System for Feature Selection and COVID-19 Pandemic Prediction in Japan

Expert systems are frequently used to make predictions in various areas. However, the practical robustness of expert systems is not as good as expected, mainly due to the fact that finding an ideal system configuration from a specific dataset is a challenging task. Therefore, how to optimize an expert system has become an important issue of research. In this paper, a new method called the robust design-based expert system is proposed to bridge this gap. The technical process of this system consists of data initialization, configuration generation, a genetic algorithm (GA) framework for feature selection, and a robust mechanism that helps the system find a configuration with the highest robustness. The system will finally obtain a set of features, which can be used to predict a pandemic based on given data. The robust mechanism can increase the efficiency of the system. The configuration for training is optimized by means of a genetic algorithm (GA) and the Taguchi method. The effectiveness of the proposed system in predicting epidemic trends is examined using a real COVID-19 dataset from Japan. For this dataset, the average prediction accuracy was 60%. Additionally, 10 representative features were also selected, resulting in a selection rate of 67% with a reduction rate of 33%. The critical features for predicting the epidemic trend of COVID-19 were also obtained, including new confirmed cases, ICU patients, people vaccinated, population, population density, hospital beds per thousand, middle age, aged 70 or older, and GDP per capital. The main contribution of this paper is two-fold: Firstly, this paper has bridged the gap between the pandemic research and expert systems with robust predictive performance. Secondly, this paper proposes a feature selection method for extracting representative variables and predicting the epidemic trend of a pandemic disease. The prediction results indicate that the system is valuable to healthcare authorities and can help governments get hold of the epidemic trend and strategize their use of healthcare resources.

Machine learning combining CT findings and clinical parameters improves prediction of length of stay and ICU admission in torso trauma

OBJECTIVE

To develop machine learning (ML) models capable of predicting ICU admission and extended length of stay (LOS) after torso (chest, abdomen, or pelvis) trauma, by using clinical and/or imaging data.

MATERIALS AND METHODS

This was a retrospective study of 840 adult patients admitted to a level 1 trauma center after injury to the torso over the course of 1 year. Clinical parameters included age, sex, vital signs, clinical scores, and laboratory values. Imaging data consisted of any injury present on CT. The two outcomes of interest were ICU admission and extended LOS, defined as more than the median LOS in the dataset. We developed and tested artificial neural network (ANN) and support vector machine (SVM) models, and predictive performance was evaluated by area under the receiver operating characteristic (ROC) curve (AUC).

RESULTS

The AUCs of SVM and ANN models to predict ICU admission were up to 0.87 ± 0.03 and 0.78 ± 0.02, respectively. The AUCs of SVM and ANN models to predict extended LOS were up to 0.80 ± 0.04 and 0.81 ± 0.05, respectively. Predictions based on imaging alone or imaging with clinical parameters were consistently more accurate than those based solely on clinical parameters.

CONCLUSIONS

The best performing models incorporated imaging findings and outperformed those with clinical findings alone. ML models have the potential to help predict outcomes in trauma by integrating clinical and imaging findings, although further research may be needed to optimize their performance.

KEY POINTS

• Artificial neural network and support vector machine-based models were used to predict the intensive care unit admission and extended length of stay after trauma to the torso. • Our input data consisted of clinical parameters and CT imaging findings derived from radiology reports, and we found that combining the two significantly enhanced the prediction of both outcomes with either model. • The highest accuracy (83%) and highest area under the receiver operating characteristic curve (0.87) were obtained for artificial neural networks and support vector machines, respectively, by combining clinical and imaging features in the prediction of intensive care unit admission.

Machine learning techniques for mortality prediction in critical traumatic patients: anatomic and physiologic variables from the RETRAUCI study

Background

Interest in models for calculating the risk of death in traumatic patients admitted to ICUs remains high. These models use variables derived from the deviation of physiological parameters and/or the severity of anatomical lesions with respect to the affected body areas. Our objective is to create different predictive models of the mortality of critically traumatic patients using machine learning techniques.

Methods

We used 9625 records from the RETRAUCI database (National Trauma Registry of 52 Spanish ICUs in the period of 2015–2019). Hospital mortality was 12.6%. Data on demographic variables, affected anatomical areas and physiological repercussions were used. The Weka Platform was used, along with a ten-fold cross-validation for the construction of nine supervised algorithms: logistic regression binary (LR), neural network (NN), sequential minimal optimization (SMO), classification rules (JRip), classification trees (CT), Bayesian networks (BN), adaptive boosting (ADABOOST), bootstrap aggregating (BAGGING) and random forest (RFOREST). The performance of the models was evaluated by accuracy, specificity, precision, recall, F-measure, and AUC.

Results

In all algorithms, the most important factors are those associated with traumatic brain injury (TBI) and organic failures. The LR finds thorax and limb injuries as independent protective factors of mortality. The CT generates 24 decision rules and uses those related to TBI as the first variables (range 2.0–81.6%). The JRip detects the eight rules with the highest risk of mortality (65.0–94.1%). The NN model uses a hidden layer of ten nodes, which requires 200 weights for its interpretation. The BN find the relationships between the different factors that identify different patient profiles. Models with the ensemble methodology (ADABOOST, BAGGING and RandomForest) do not have greater performance. All models obtain high values ​​in accuracy, specificity, and AUC, but obtain lower values ​​in recall. The greatest precision is achieved by the SMO model, and the BN obtains the best recall, F-measure, and AUC.

Conclusion

Machine learning techniques are useful for creating mortality classification models in critically traumatic patients. With clinical interpretation, the algorithms establish different patient profiles according to the relationship between the variables used, determine groups of patients with different evolutions, and alert clinicians to the presence of rules that indicate the greatest severity.

Artificial neural networks can predict trauma volume and acuity regardless of center size and geography: A multicenter study

BACKGROUND

Trauma has long been considered unpredictable. Artificial neural networks (ANN) have recently shown the ability to predict admission volume, acuity, and operative needs at a single trauma center with very high reliability. This model has not been tested in a multicenter model with differing climate and geography. We hypothesize that an ANN can accurately predict trauma admission volume, penetrating trauma admissions, and mean Injury Severity Score (ISS) with a high degree of reliability across multiple trauma centers.

METHODS

Three years of admission data were collected from five geographically distinct US Level I trauma centers. Patients with incomplete data, pediatric patients, and primary thermal injuries were excluded. Daily number of traumas, number of penetrating cases, and mean ISS were tabulated from each center along with National Oceanic and Atmospheric Administration data from local airports. We trained a single two-layer feed-forward ANN on a random majority (70%) partitioning of data from all centers using Bayesian Regularization and minimizing mean squared error. Pearson's product-moment correlation coefficient was calculated for each partition, each trauma center, and for high- and low-volume days (>1 standard deviation above or below mean total number of traumas).

RESULTS

There were 5,410 days included. There were 43,380 traumas, including 4,982 penetrating traumas. The mean ISS was 11.78 (SD = 6.12). On the training partition, we achieved R = 0.8733. On the testing partition (new data to the model), we achieved R = 0.8732, with a combined R = 0.8732. For high- and low-volume days, we achieved R = 0.8934 and R = 0.7963, respectively.

CONCLUSION

An ANN successfully predicted trauma volumes and acuity across multiple trauma centers with very high levels of reliability. The correlation was highest during periods of peak volume. This can potentially provide a framework for determining resource allocation at both the trauma system level and the individual hospital level.

LEVEL OF EVIDENCE

Care Management, level IV.

DICOM Metadata Analysis for Population Studies

This article reports an experimental study to determine how to use the stored Digital Imaging and Communication in Medicine (DICOM) metadata to perform population studies. As a case study, it was considered three types of medical imaging studies (i.e. routine head computed tomography, thorax computed radiography and thorax digital radiography) stored in the picture archiving and communication systems (PACS) of three healthcare institutions. The final sample consisted of DICOM metadata belonging to 1370360 images, corresponding to 109160 medical imaging studies performed on 72716 patients. The study followed a methodological approach that allows the identification of the number of patients with performed studies by age group and gender, as well as the average number of studies by patient, age group and gender in each one of the three healthcare institutions. The results show the relevance of the aggregation and analyses of DICOM metadata stored in heterogeneous PACS facilities.

Theoretical bounds and approximation of the probability mass function of future hospital bed demand

Failing to match the supply of resources to the demand for resources in a hospital can cause non-clinical transfers, diversions, safety risks, and expensive under-utilized resource capacity. Forecasting bed demand helps achieve appropriate safety standards and cost management by proactively adjusting staffing levels and patient flow protocols. This paper defines the theoretical bounds on optimal bed demand prediction accuracy and develops a flexible statistical model to approximate the probability mass function of future bed demand. A case study validates the model using blinded data from a mid-sized Massachusetts community hospital. This approach expands upon similar work by forecasting multiple days in advance instead of a single day, providing a probability mass function of demand instead of a point estimate, using the exact surgery schedule instead of assuming a cyclic schedule, and using patient-level duration-varying length-of-stay distributions instead of assuming patient homogeneity and exponential length of stay distributions. The primary results of this work are an accurate and lengthy forecast, which provides managers better information and more time to optimize short-term staffing adaptations to stochastic bed demand, and a derivation of the minimum mean absolute error of an ideal forecast.

Use of photoplethysmography to predict mortality in intensive care units

PURPOSE

The aim of this study was to evaluate and compare the capacity to predict hemodynamic variables obtained with photoplethysmography (PPG) and Acute Physiology and Chronic Health Evaluation (APACHE II) in patients hospitalized in the intensive care unit (ICU).

MATERIALS AND METHODS

A prospective cohort study was conducted in the adult ICU of Hospital Nossa Senhora da Conceição, located in Tubarão, Santa Catarina, Brazil. The data collected included the diagnosis for hospitalization, age, gender, clinical or surgical profile, PPG pulse curve signal, and APACHE II score in the first 24 hours. A bivariate and a multivariate logistic regressions were performed, with death as an outcome. A mortality model using artificial neural networks (ANNs) was proposed.

RESULTS

A total of 190 individuals were evaluated. Most of them were males (6:5), with a median age of 67 (54-75) years, and the main reasons for hospitalization were cardiovascular and neurological causes; half of them were surgical cases. APACHE II median score was 14 (8-19), with a median length of stay of 6 (3-15) days, and 28.4% of the patients died. The following factors were associated with mortality: age (OR=1.023; 95% CI 1.001-1.044; P=0.039), clinical profile (OR=5.481; 95% CI 2.646-11.354; P<0.001), APACHE II (OR=1.168; 95% CI 1.106-1.234; P<0.001), heart rate in the first 24 hours (OR=1.020; 95% CI 1.001-1.039; P=0.036), and time between the systolic and diastolic peak (∆T) intervals obtained with PPG (OR=0.989; 95% CI 0.979-0.998; P=0.015). Compared with the accuracy (area under the receiver-operating characteristic curve) 0.780 of APACHE II (95% CI 0.711-0.849; P<0.001), the multivariate logistic model showed a larger area of 0.858 (95% CI 0.803-0.914; P<0.001). In the model using ANNs, the accuracy was 0.895 (95% CI 0.851-0.940; P<0.001).

CONCLUSION

The mortality models using variables obtained with PPG, with the inclusion of epidemiological parameters, are very accurate and, if associated to APACHE II, improve prognostic accuracy. The use of ANN was even more accurate, indicating that this tool is important to help in the clinical judgment of the intensivist.

Length of Hospital Stay Prediction at the Admission Stage for Cardiology Patients Using Artificial Neural Network

For hospitals' admission management, the ability to predict length of stay (LOS) as early as in the preadmission stage might be helpful to monitor the quality of inpatient care. This study is to develop artificial neural network (ANN) models to predict LOS for inpatients with one of the three primary diagnoses: coronary atherosclerosis (CAS), heart failure (HF), and acute myocardial infarction (AMI) in a cardiovascular unit in a Christian hospital in Taipei, Taiwan. A total of 2,377 cardiology patients discharged between October 1, 2010, and December 31, 2011, were analyzed. Using ANN or linear regression model was able to predict correctly for 88.07% to 89.95% CAS patients at the predischarge stage and for 88.31% to 91.53% at the preadmission stage. For AMI or HF patients, the accuracy ranged from 64.12% to 66.78% at the predischarge stage and 63.69% to 67.47% at the preadmission stage when a tolerance of 2 days was allowed.

Neural Network Prediction of ICU Length of Stay Following Cardiac Surgery Based on Pre-Incision Variables

Background

Advanced predictive analytical techniques are being increasingly applied to clinical risk assessment. This study compared a neural network model to several other models in predicting the length of stay (LOS) in the cardiac surgical intensive care unit (ICU) based on pre-incision patient characteristics.

Methods

Thirty six variables collected from 185 cardiac surgical patients were analyzed for contribution to ICU LOS. The Automatic Linear Modeling (ALM) module of IBM-SPSS software identified 8 factors with statistically significant associations with ICU LOS; these factors were also analyzed with the Artificial Neural Network (ANN) module of the same software. The weighted contributions of each factor (“trained” data) were then applied to data for a “new” patient to predict ICU LOS for that individual.

Results

Factors identified in the ALM model were: use of an intra-aortic balloon pump; O₂ delivery index; age; use of positive cardiac inotropic agents; hematocrit; serum creatinine ≥ 1.3 mg/deciliter; gender; arterial pCO₂. The r² value for ALM prediction of ICU LOS in the initial (training) model was 0.356, p <0.0001. Cross validation in prediction of a “new” patient yielded r² = 0.200, p <0.0001. The same 8 factors analyzed with ANN yielded a training prediction r² of 0.535 (p <0.0001) and a cross validation prediction r² of 0.410, p <0.0001. Two additional predictive algorithms were studied, but they had lower prediction accuracies. Our validated neural network model identified the upper quartile of ICU LOS with an odds ratio of 9.8(p <0.0001).

Conclusions

ANN demonstrated a 2-fold greater accuracy than ALM in prediction of observed ICU LOS. This greater accuracy would be presumed to result from the capacity of ANN to capture nonlinear effects and higher order interactions. Predictive modeling may be of value in early anticipation of risks of post-operative morbidity and utilization of ICU facilities.

Five Years Survival of Patients After Liver Transplantation and Its Effective Factors by Neural Network and Cox Poroportional Hazard Regression Models

BACKGROUND

Transplantation is the only treatment for patients with liver failure. Since the therapy imposes high expenses to the patients and community, identification of effective factors on survival of such patients after transplantation is valuable.

OBJECTIVES

The current study attempted to model the survival of patients (two years old and above) after liver transplantation using neural network and Cox Proportional Hazards (Cox PH) regression models. The event is defined as death due to complications of liver transplantation.

PATIENTS AND METHODS

In a historical cohort study, the clinical findings of 1168 patients who underwent liver transplant surgery (from March 2008 to march 2013) at Shiraz Namazee Hospital Organ Transplantation Center, Shiraz, Southern Iran, were used. To model the one to five years survival of such patients, Cox PH regression model accompanied by three layers feed forward artificial neural network (ANN) method were applied on data separately and their prediction accuracy was compared using the area under the receiver operating characteristic curve (ROC). Furthermore, Kaplan-Meier method was used to estimate the survival probabilities in different years.

RESULTS

The estimated survival probability of one to five years for the patients were 91%, 89%, 85%, 84%, and 83%, respectively. The areas under the ROC were 86.4% and 80.7% for ANN and Cox PH models, respectively. In addition, the accuracy of prediction rate for ANN and Cox PH methods was equally 92.73%.

CONCLUSIONS

The present study detected more accurate results for ANN method compared to those of Cox PH model to analyze the survival of patients with liver transplantation. Furthermore, the order of effective factors in patients' survival after transplantation was clinically more acceptable. The large dataset with a few missing data was the advantage of this study, the fact which makes the results more reliable.