Evaluating Prognostic Bias of Critical Illness Severity Scores Based on Age, Sex, and Primary Language in the United States: A Retrospective Multicenter Study

OBJECTIVES

Although illness severity scoring systems are widely used to support clinical decision-making and assess ICU performance, their potential bias across different age, sex, and primary language groups has not been well-studied.

DESIGN SETTING AND PATIENTS

We aimed to identify potential bias of Sequential Organ Failure Assessment (SOFA) and Acute Physiology and Chronic Health Evaluation (APACHE) IVa scores via large ICU databases.

SETTING/PATIENTS

This multicenter, retrospective study was conducted using data from the Medical Information Mart for Intensive Care (MIMIC) and eICU Collaborative Research Database. SOFA and APACHE IVa scores were obtained from ICU admission. Hospital mortality was the primary outcome. Discrimination (area under receiver operating characteristic [AUROC] curve) and calibration (standardized mortality ratio [SMR]) were assessed for all subgroups.

INTERVENTIONS

Not applicable.

MEASUREMENTS AND MAIN RESULTS

A total of 196,310 patient encounters were studied. Discrimination for both scores was worse in older patients compared with younger patients and female patients rather than male patients. In MIMIC, discrimination of SOFA in non-English primary language speakers patients was worse than that of English speakers (AUROC 0.726 vs. 0.783, p < 0.0001). Evaluating calibration via SMR showed statistically significant underestimations of mortality when compared with overall cohort in the oldest patients for both SOFA and APACHE IVa, female patients (1.09) for SOFA, and non-English primary language patients (1.38) for SOFA in MIMIC.

CONCLUSIONS

Differences in discrimination and calibration of two scores across varying age, sex, and primary language groups suggest illness severity scores are prone to bias in mortality predictions. Caution must be taken when using them for quality benchmarking and decision-making among diverse real-world populations.

Illness severity assessment of older adults in critical illness using machine learning (ELDER-ICU): an international multicentre study with subgroup bias evaluation

BACKGROUND

Comorbidity, frailty, and decreased cognitive function lead to a higher risk of death in elderly patients (more than 65 years of age) during acute medical events. Early and accurate illness severity assessment can support appropriate decision making for clinicians caring for these patients. We aimed to develop ELDER-ICU, a machine learning model to assess the illness severity of older adults admitted to the intensive care unit (ICU) with cohort-specific calibration and evaluation for potential model bias.

METHODS

In this retrospective, international multicentre study, the ELDER-ICU model was developed using data from 14 US hospitals, and validated in 171 hospitals from the USA and Netherlands. Data were extracted from the Medical Information Mart for Intensive Care database, electronic ICU Collaborative Research Database, and Amsterdam University Medical Centers Database. We used six categories of data as predictors, including demographics and comorbidities, physical frailty, laboratory tests, vital signs, treatments, and urine output. Patient data from the first day of ICU stay were used to predict in-hospital mortality. We used the eXtreme Gradient Boosting algorithm (XGBoost) to develop models and the SHapley Additive exPlanations method to explain model prediction. The trained model was calibrated before internal, external, and temporal validation. The final XGBoost model was compared against three other machine learning algorithms and five clinical scores. We performed subgroup analysis based on age, sex, and race. We assessed the discrimination and calibration of models using the area under receiver operating characteristic (AUROC) and standardised mortality ratio (SMR) with 95% CIs.

FINDINGS

Using the development dataset (n=50 366) and predictive model building process, the XGBoost algorithm performed the best in all types of validations compared with other machine learning algorithms and clinical scores (internal validation with 5037 patients from 14 US hospitals, AUROC=0·866 [95% CI 0·851-0·880]; external validation in the US population with 20 541 patients from 169 hospitals, AUROC=0·838 [0·829-0·847]; external validation in European population with 2411 patients from one hospital, AUROC=0·833 [0·812-0·853]; temporal validation with 4311 patients from one hospital, AUROC=0·884 [0·869-0·897]). In the external validation set (US population), the median AUROCs of bias evaluations covering eight subgroups were above 0·81, and the overall SMR was 0·99 (0·96-1·03). The top ten risk predictors were the minimum Glasgow Coma Scale score, total urine output, average respiratory rate, mechanical ventilation use, best state of activity, Charlson Comorbidity Index score, geriatric nutritional risk index, code status, age, and maximum blood urea nitrogen. A simplified model containing only the top 20 features (ELDER-ICU-20) had similar predictive performance to the full model.

INTERPRETATION

The ELDER-ICU model reliably predicts the risk of in-hospital mortality using routinely collected clinical features. The predictions could inform clinicians about patients who are at elevated risk of deterioration. Prospective validation of this model in clinical practice and a process for continuous performance monitoring and model recalibration are needed.

FUNDING

National Institutes of Health, National Natural Science Foundation of China, National Special Health Science Program, Health Science and Technology Plan of Zhejiang Province, Fundamental Research Funds for the Central Universities, Drug Clinical Evaluate Research of Chinese Pharmaceutical Association, and National Key R&D Program of China.

The Global Open Source Severity of Illness Score (GOSSIS)

OBJECTIVES

To develop and demonstrate the feasibility of a Global Open Source Severity of Illness Score (GOSSIS)-1 for critical care patients, which generalizes across healthcare systems and countries.

DESIGN

A merger of several critical care multicenter cohorts derived from registry and electronic health record data. Data were split into training (70%) and test (30%) sets, using each set exclusively for development and evaluation, respectively. Missing data were imputed when not available.

SETTING/PATIENTS

Two large multicenter datasets from Australia and New Zealand (Australian and New Zealand Intensive Care Society Adult Patient Database [ANZICS-APD]) and the United States (eICU Collaborative Research Database [eICU-CRD]) representing 249,229 and 131,051 patients, respectively. ANZICS-APD and eICU-CRD contributed data from 162 and 204 hospitals, respectively. The cohort included all ICU admissions discharged in 2014-2015, excluding patients less than 16 years old, admissions less than 6 hours, and those with a previous ICU stay.

INTERVENTIONS

Not applicable.

MEASUREMENTS AND MAIN RESULTS

GOSSIS-1 uses data collected during the ICU stay's first 24 hours, including extrema values for vital signs and laboratory results, admission diagnosis, the Glasgow Coma Scale, chronic comorbidities, and admission/demographic variables. The datasets showed significant variation in admission-related variables, case-mix, and average physiologic state. Despite this heterogeneity, test set discrimination of GOSSIS-1 was high (area under the receiver operator characteristic curve [AUROC], 0.918; 95% CI, 0.915-0.921) and calibration was excellent (standardized mortality ratio [SMR], 0.986; 95% CI, 0.966-1.005; Brier score, 0.050). Performance was held within ANZICS-APD (AUROC, 0.925; SMR, 0.982; Brier score, 0.047) and eICU-CRD (AUROC, 0.904; SMR, 0.992; Brier score, 0.055). Compared with GOSSIS-1, Acute Physiology and Chronic Health Evaluation (APACHE)-IIIj (ANZICS-APD) and APACHE-IVa (eICU-CRD), had worse discrimination with AUROCs of 0.904 and 0.869, and poorer calibration with SMRs of 0.594 and 0.770, and Brier scores of 0.059 and 0.063, respectively.

CONCLUSIONS

GOSSIS-1 is a modern, free, open-source inhospital mortality prediction algorithm for critical care patients, achieving excellent discrimination and calibration across three countries.

Clinically Interpretable Machine Learning Models for Early Prediction of Mortality in Older Patients with Multiple Organ Dysfunction Syndrome (MODS): An International Multicenter Retrospective Study

BACKGROUND

Multiple organ dysfunction syndrome (MODS) is associated with a high risk of mortality among older patients. Current severity scores are limited in their ability to assist clinicians with triage and management decisions. We aim to develop mortality prediction models for older patients with MODS admitted to the ICU.

METHODS

The study analyzed older patients from 197 hospitals in the US and one hospital in the Netherlands. The cohort was divided into the young-old (65-80 years) and old-old (≥80 years), which were separately used to develop and evaluate models including internal, external and temporal validation. Demographic characteristics, comorbidities, vital signs, laboratory measurements, and treatments were used as predictors. We used the XGBoost algorithm to train models, and the SHAP method to interpret predictions.

RESULTS

34,497 young-old (11.3% mortality) and 21,330 old-old (15.7% mortality) patients were analyzed. Discrimination AUROC of internal validation models in 9,046 U.S. patients was as follows: 0.87 and 0.82, respectively; Discrimination of external validation models in 1,905 EUR patients was as follows: 0.86 and 0.85, respectively; and of temporal validation models in 8,690 U.S. patients: 0.85 and 0.78, respectively. These models outperformed standard clinical scores like SOFA and APSIII. The GCS, Charlson Comorbidity Index, and Code Status emerged as top predictors of mortality.

CONCLUSIONS

Our models integrate data spanning physiologic and geriatric-relevant variables that outperform existing scores used in older adults with MODS, which represents a proof of concept of how machine learning can streamline data analysis for busy ICU clinicians to potentially optimize prognostication and decision making.

A contrastive learning approach for ICU false arrhythmia alarm reduction

The high rate of false arrhythmia alarms in Intensive Care Units (ICUs) can lead to disruption of care, negatively impacting patients’ health through noise disturbances, and slow staff response time due to alarm fatigue. Prior false-alarm reduction approaches are often rule-based and require hand-crafted features from physiological waveforms as inputs to machine learning classifiers. Despite considerable prior efforts to address the problem, false alarms are a continuing problem in the ICUs. In this work, we present a deep learning framework to automatically learn feature representations of physiological waveforms using convolutional neural networks (CNNs) to discriminate between true vs. false arrhythmia alarms. We use Contrastive Learning to simultaneously minimize a binary cross entropy classification loss and a proposed similarity loss from pair-wise comparisons of waveform segments over time as a discriminative constraint. Furthermore, we augment our deep models with learned embeddings from a rule-based method to leverage prior domain knowledge for each alarm type. We evaluate our method using the dataset from the 2015 PhysioNet Computing in Cardiology Challenge. Ablation analysis demonstrates that Contrastive Learning significantly improves the performance of a combined deep learning and rule-based-embedding approach. Our results indicate that the final proposed deep learning framework achieves superior performance in comparison to the winning entries of the Challenge.

A novel artificial intelligence based intensive care unit monitoring system: using physiological waveforms to identify sepsis

A massive amount of multimodal data are continuously collected in the intensive care unit (ICU) along each patient stay, offering a great opportunity for the development of smart monitoring devices based on artificial intelligence (AI). The two main sources of relevant information collected in the ICU are the electronic health records (EHRs) and vital sign waveforms continuously recorded at the bedside. While EHRs are already widely processed by AI algorithms for prompt diagnosis and prognosis, AI-based assessments of the patients' pathophysiological state using waveforms are less developed, and their use is still limited to real-time monitoring for basic visual vital sign feedback at the bedside. This study uses data from the MIMIC-III database (PhysioNet) to propose a novel AI approach in ICU patient monitoring that incorporates features estimated by a closed-loop cardiovascular model, with the specific goal of identifying sepsis within the first hour of admission. Our top benchmark results (AUROC = 0.92, AUPRC = 0.90) suggest that features derived by cardiovascular control models may play a key role in identifying sepsis, by continuous monitoring performed through advanced multivariate modelling of vital sign waveforms. This work lays foundations for a deeper data integration paradigm which will help clinicians in their decision-making processes. This article is part of the theme issue 'Advanced computation in cardiovascular physiology: new challenges and opportunities'.

Delaying initiation of diuretics in critically ill patients with recent vasopressor use and high positive fluid balance

BACKGROUND

Fluid overload is associated with poor outcomes. Clinicians might be reluctant to initiate diuretic therapy for patients with recent vasopressor use. We estimated the effect on 30-day mortality of withholding or delaying diuretics after vasopressor use in patients with probable fluid overload.

METHODS

This was a retrospective cohort study of adults admitted to ICUs of an academic medical centre between 2008 and 2012. Using a database of time-stamped patient records, we followed individuals from the time they first required vasopressor support and had >5 L cumulative positive fluid balance (plus additional inclusion/exclusion criteria). We compared mortality under usual care (the mix of care actually delivered in the cohort) and treatment strategies restricting diuretic initiation during and for various durations after vasopressor use. We adjusted for baseline and time-varying confounding via inverse probability weighting.

RESULTS

The study included 1501 patients, and the observed 30-day mortality rate was 11%. After adjusting for observed confounders, withholding diuretics for at least 24 h after stopping most recent vasopressor use was estimated to increase 30-day mortality rate by 2.2% (95% confidence interval [CI], 0.9-3.6%) compared with usual care. Data were consistent with moderate harm or slight benefit from withholding diuretic initiation only during concomitant vasopressor use; the estimated mortality rate increased by 0.5% (95% CI, -0.2% to 1.1%).

CONCLUSIONS

Withholding diuretic initiation after vasopressor use in patients with high cumulative positive balance (>5 L) was estimated to increase 30-day mortality. These findings are hypothesis generating and should be tested in a clinical trial.

Quantitative Electroencephalogram Trends Predict Recovery in Hypoxic-Ischemic Encephalopathy

OBJECTIVES

Electroencephalogram features predict neurologic recovery following cardiac arrest. Recent work has shown that prognostic implications of some key electroencephalogram features change over time. We explore whether time dependence exists for an expanded selection of quantitative electroencephalogram features and whether accounting for this time dependence enables better prognostic predictions.

DESIGN

Retrospective.

SETTING

ICUs at four academic medical centers in the United States.

PATIENTS

Comatose patients with acute hypoxic-ischemic encephalopathy.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We analyzed 12,397 hours of electroencephalogram from 438 subjects. From the electroencephalogram, we extracted 52 features that quantify signal complexity, category, and connectivity. We modeled associations between dichotomized neurologic outcome (good vs poor) and quantitative electroencephalogram features in 12-hour intervals using sequential logistic regression with Elastic Net regularization. We compared a predictive model using time-varying features to a model using time-invariant features and to models based on two prior published approaches. Models were evaluated for their ability to predict binary outcomes using area under the receiver operator curve, model calibration (how closely the predicted probability of good outcomes matches the observed proportion of good outcomes), and sensitivity at several common specificity thresholds of interest. A model using time-dependent features outperformed (area under the receiver operator curve, 0.83 ± 0.08) one trained with time-invariant features (0.79 ± 0.07; p < 0.05) and a random forest approach (0.74 ± 0.13; p < 0.05). The time-sensitive model was also the best-calibrated.

CONCLUSIONS

The statistical association between quantitative electroencephalogram features and neurologic outcome changed over time, and accounting for these changes improved prognostication performance.

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports

Chest radiography is an extremely powerful imaging modality, allowing for a detailed inspection of a patient's chest, but requires specialized training for proper interpretation. With the advent of high performance general purpose computer vision algorithms, the accurate automated analysis of chest radiographs is becoming increasingly of interest to researchers. Here we describe MIMIC-CXR, a large dataset of 227,835 imaging studies for 65,379 patients presenting to the Beth Israel Deaconess Medical Center Emergency Department between 2011-2016. Each imaging study can contain one or more images, usually a frontal view and a lateral view. A total of 377,110 images are available in the dataset. Studies are made available with a semi-structured free-text radiology report that describes the radiological findings of the images, written by a practicing radiologist contemporaneously during routine clinical care. All images and reports have been de-identified to protect patient privacy. The dataset is made freely available to facilitate and encourage a wide range of research in computer vision, natural language processing, and clinical data mining.

How is the Doctor Feeling? ICU Provider Sentiment is Associated with Diagnostic Imaging Utilization

The judgment of intensive care unit (ICU) providers is difficult to measure using conventional structured electronic medical record (EMR) data. However, provider sentiment may be a proxy for such judgment. Utilizing 10 years of EMR data, this study evaluates the association between provider sentiment and diagnostic imaging utilization. We extracted daily positive / negative sentiment scores of written provider notes, and used a Poisson regression to estimate sentiment association with the total number of daily imaging reports. After adjusting for confounding factors, we found that (1) negative sentiment was associated with increased imaging utilization $(p < 0.01)$, (2) sentiment's association was most pronounced at the beginning of the ICU stay $(p < 0.01)$, and (3) the presence of any form of sentiment increased diagnostic imaging utilization up to a critical threshold $(p < 0.01)$. Our results indicate that provider sentiment may clarify currently unexplained variance in resource utilization and clinical practice.

The association between autoimmune disease and 30-day mortality among sepsis ICU patients: a cohort study

Introduction

Sepsis results from a dysregulated host response to an infection that is associated with an imbalance between pro- and anti-inflammatory cytokines. This imbalance is hypothesized to be a driver of patient mortality. Certain autoimmune diseases modulate the expression of cytokines involved in the pathophysiology of sepsis. However, the outcomes of patients with autoimmune disease who develop sepsis have not been studied in detail. The objective of this study is to determine whether patients with autoimmune diseases have different sepsis outcomes than patients without these comorbidities.

Methods

Using the Multiparameter Intelligent Monitoring in Intensive Care III database (v. 1.4) which contains retrospective clinical data for over 50,000 adult ICU stays, we compared 30-day mortality risk for sepsis patients with and without autoimmune disease. We used logistic regression models to control for known confounders, including demographics, disease severity, and immunomodulation medications. We used mediation analysis to evaluate how the chronic use of immunomodulation medications affects the relationship between autoimmune disease and 30-day mortality.

Results

Our study found a statistically significant 27.00% reduction in the 30-day mortality risk associated with autoimmune disease presence. This association was found to be the strongest (OR 0.71, 95% CI 0.54–0.93, P = 0.014) among patients with septic shock. The autoimmune disease-30-day mortality association was not mediated through the chronic use of immunomodulation medications (indirect effect OR 1.07, 95% CI 1.01–1.13, P = 0.020).

Conclusions

We demonstrated that autoimmune diseases are associated with a lower 30-day mortality risk in sepsis. Our findings suggest that autoimmune diseases affect 30-day mortality through a mechanism unrelated to the chronic use of immunomodulation medications. Since this study was conducted within a single study center, research using data from other medical centers will provide further validation.

Electronic supplementary material

The online version of this article (10.1186/s13054-019-2357-1) contains supplementary material, which is available to authorized users.

The eICU Collaborative Research Database, a freely available multi-center database for critical care research

Critical care patients are monitored closely through the course of their illness. As a result of this monitoring, large amounts of data are routinely collected for these patients. Philips Healthcare has developed a telehealth system, the eICU Program, which leverages these data to support management of critically ill patients. Here we describe the eICU Collaborative Research Database, a multi-center intensive care unit (ICU)database with high granularity data for over 200,000 admissions to ICUs monitored by eICU Programs across the United States. The database is deidentified, and includes vital sign measurements, care plan documentation, severity of illness measures, diagnosis information, treatment information, and more. Data are publicly available after registration, including completion of a training course in research with human subjects and signing of a data use agreement mandating responsible handling of the data and adhering to the principle of collaborative research. The freely available nature of the data will support a number of applications including the development of machine learning algorithms, decision support tools, and clinical research.

You Snooze, You Win: the PhysioNet/Computing in Cardiology Challenge 2018

The PhysioNet/Computing in Cardiology Challenge 2018 focused on the use of various physiological signals (EEG, EOG, EMG, ECG, SaO₂) collected during polysomnographic sleep studies to detect sources of arousal (non-apnea) during sleep. A total of 1,983 polysomnographic recordings were made available to the entrants. The arousal labels for 994 of the recordings were made available in a public training set while 989 labels were retained in a hidden test set. Challengers were asked to develop an algorithm that could label the presence of arousals within the hidden test set. The performance metric used to assess entrants was the area under the precision-recall curve. A total of twenty-two independent teams entered the Challenge, deploying a variety of methods from generalized linear models to deep neural networks.

Representation Learning Approaches to Detect False Arrhythmia Alarms from ECG Dynamics

The high rate of intensive care unit false arrhythmia alarms can lead to disruption of care and slow response time due to desensitization of clinical staff. We study the use of machine learning models to detect false ventricular tachycardia (v-tach) alarms using ECG waveform recordings. We propose using a Supervised Denoising Autoencoder (SDAE) to detect false alarms using a low-dimensional representation of ECG dynamics learned by minimizing a combined reconstruction and classification loss. We evaluate our algorithms on the PhysioNet Challenge 2015 dataset, containing over 500 records (over 300 training and 200 testing) with v-tach alarms. Our results indicate that using the SDAE on Fast Fourier Transformed (FFT) ECG at a beat-by-beat level outperforms several competitive baselines on the task of v-tach false alarm classification. We show that it is important to exploit the underlying known physiological structure using beat-by-beat frequency distribution from multiple cardiac cycles of the ECG waveforms to obtain competitive results and improve over previous entries from the 2015 PhysioNet Challenge.

The Role of Baroreflex Sensitivity in Acute Hypotensive Episodes Prediction in the Intensive Care Unit

A life threatening condition in Intensive Care Unit (ICU) is the Acute Hypotensive Episode (AHE). Patients experiencing an AHE may suffer from irreversible organ damage associated with increased mortality. Predicting the onset of AHE could be of pivotal importance to establish appropriate and timely interventions. We propose a method that, using waveforms widely acquired in ICU, like Arterial Blood Pressure (ABP) and Electrocardiogram (ECG), will extract features relative to the cardiac system to predict whether or not a patient will experience a hypotensive episode. Specifically, we want to assess if there are hidden patterns in the dynamics of baroreflex able to improve the prediction of AHEs. We will investigate the predictive power of features related to the baroreflex by performing classifications with and without them. Results are obtained using 17 classifiers belonging to different model families: classification trees, Support Vector Machines (SVMs), K-Nearest Neighbors (KNNs) replicated with different set of hyper-parameters and logistic regression. On average, the use of baroreflex features in the AHE prediction process increases the Area Under the Curve (AUC) by 10%.

tableone: An open source Python package for producing summary statistics for research papers

Objectives

In quantitative research, understanding basic parameters of the study population is key for interpretation of the results. As a result, it is typical for the first table (“Table 1”) of a research paper to include summary statistics for the study data. Our objectives are 2-fold. First, we seek to provide a simple, reproducible method for providing summary statistics for research papers in the Python programming language. Second, we seek to use the package to improve the quality of summary statistics reported in research papers.

Materials and Methods

The tableone package is developed following good practice guidelines for scientific computing and all code is made available under a permissive MIT License. A testing framework runs on a continuous integration server, helping to maintain code stability. Issues are tracked openly and public contributions are encouraged.

Results

The tableone software package automatically compiles summary statistics into publishable formats such as CSV, HTML, and LaTeX. An executable Jupyter Notebook demonstrates application of the package to a subset of data from the MIMIC-III database. Tests such as Tukey’s rule for outlier detection and Hartigan’s Dip Test for modality are computed to highlight potential issues in summarizing the data.

Discussion and Conclusion

We present open source software for researchers to facilitate carrying out reproducible studies in Python, an increasingly popular language in scientific research. The toolkit is intended to mature over time with community feedback and input. Development of a common tool for summarizing data may help to promote good practice when used as a supplement to existing guidelines and recommendations. We encourage use of tableone alongside other methods of descriptive statistics and, in particular, visualization to ensure appropriate data handling. We also suggest seeking guidance from a statistician when using tableone for a research study, especially prior to submitting the study for publication.

Real-time mortality prediction in the Intensive Care Unit

Real-time prediction of mortality for intensive care unit patients has the potential to provide physicians with a simple and easily interpretable synthesis of patient acuity. Here we extract data from a random time during each patient's ICU stay. We believe this sampling scheme allows for the application of the model(s) across a future patient's entire ICU stay. The AUROC of a Gradient Boosting model was high (AUROC=0.920), even though no information about diagnosis or comorbid burden was utilized. We also compare models using data from the first 24 hours of a patient's stay against published severity of illness scores, and find the Gradient Boosting model greatly outperformed the frequently used Simplified Acute Physiology Score II (AUROC = 0.927 vs. 0.809). We nuance this performance with comparison to the literature, provide our interpretation, and discuss potential avenues for improvement.

AF Classification from a Short Single Lead ECG Recording: the PhysioNet/Computing in Cardiology Challenge 2017

The PhysioNet/Computing in Cardiology (CinC) Challenge 2017 focused on differentiating AF from noise, normal or other rhythms in short term (from 9-61 s) ECG recordings performed by patients. A total of 12,186 ECGs were used: 8,528 in the public training set and 3,658 in the private hidden test set. Due to the high degree of inter-expert disagreement between a significant fraction of the expert labels we implemented a mid-competition bootstrap approach to expert relabeling of the data, levering the best performing Challenge entrants' algorithms to identify contentious labels. A total of 75 independent teams entered the Challenge using a variety of traditional and novel methods, ranging from random forests to a deep learning approach applied to the raw data in the spectral domain. Four teams won the Challenge with an equal high F1 score (averaged across all classes) of 0.83, although the top 11 algorithms scored within 2% of this. A combination of 45 algorithms identified using LASSO achieved an F1 of 0.87, indicating that a voting approach can boost performance.

An enhanced cerebral recovery index for coma prognostication following cardiac arrest

Prognostication of coma outcomes following cardiac arrest is both qualitative and poorly understood in current practice. Existing quantitative metrics are powerful, but lack rigorous approaches to classification. This is due, in part, to a lack of available data on the population of interest. In this paper we describe a novel retrospective data set of 167 cardiac arrest patients (spanning three institutions) who received electroencephalography (EEG) monitoring. We utilized a subset of the collected data to generate features that measured the connectivity, complexity and category of EEG activity. A subset of these features was included in a logistic regression model to estimate a dichotomized cerebral performance category score at discharge. We compared the predictive performance of our method against an established EEG-based alternative, the Cerebral Recovery Index (CRI) and show that our approach more reliably classifies patient outcomes, with an average increase in AUC of 0.27.

Right Ventricular Function, Peripheral Edema, and Acute Kidney Injury in Critical Illness

Introduction

The cardiorenal syndrome generally focuses on left ventricular function, and the importance of the right ventricle as a determinant of renal function is described less frequently. In a cohort of critically ill patients with echocardiographic measurements obtained within 24 hours of admission to the intensive care unit, we examined the association of right ventricular function with acute kidney injury (AKI) and AKI-associated mortality. We also examined whether clinical measurement of volume overload modified the association between ventricular function and AKI in a subpopulation with documented admission physical examinations.

Methods

Among 1879 critically ill patients with echocardiographic ventricular measurements, 43% (n = 807) had ventricular dysfunction—21% (n = 388), 9% (n = 167), and 13% (n = 252) with isolated left ventricular dysfunction, isolated right ventricular dysfunction, and biventricular dysfunction, respectively. Overall, ventricular dysfunction was associated with a 43% higher adjusted risk of AKI (95% confidence interval [CI] 1.14–1.80; P = 0.002) compared with those with normal biventricular function, whereas isolated left ventricular dysfunction, isolated right ventricular dysfunction, and biventricular dysfunction were associated with a 1.34 (95% CI 1.00-1.77, P = 0.05), 1.35 (95% CI 0.90–2.10, P = 0.14) and 1.67 (95% CI 1.23–2.31, P = 0.002) higher adjusted risk. Although an episode of AKI was associated with an approximately 2-fold greater risk of hospital mortality in those with isolated left ventricular dysfunction and biventricular dysfunction, in those with isolated right ventricular dysfunction, AKI was associated with a 7.85-fold greater risk of death (95% CI 2.89–21.3, P < 0.001). Independent of ventricular function, peripheral edema was an important determinant of AKI.

Discussion

Like left ventricular function, right ventricular function is an important determinant of AKI and AKI-associated mortality. Volume overload, independently of ventricular function, is a risk factor for AKI. Whether establishment of euvolemia might mitigate AKI risk will require further study.

Estimating Patient's Health State Using Latent Structure Inferred from Clinical Time Series and Text

Modern intensive care units (ICUs) collect large volumes of data in monitoring critically ill patients. Clinicians in the ICUs face the challenge of interpreting large volumes of high-dimensional data to diagnose and treat patients. In this work, we explore the use of Hierarchical Dirichlet Processes (HDP) as a Bayesian nonparametric framework to infer patients' states of health by combining multiple sources of data. In particular, we employ HDP to combine clinical time series and text from the nursing progress notes in a probabilistic topic modeling framework for patient risk stratification. Given a patient cohort, we use HDP to infer latent "topics" shared across multimodal patient data from the entire cohort. Each topic is modeled as a multinomial distribution over a vocabulary of codewords, defined over heterogeneous data sources. We evaluate the clinical utility of the learned topic structure using the first 24-hour ICU data from over 17,000 adult patients in the MIMIC-II database to estimate patients' risks of in-hospital mortality. Our results demonstrate that our approach provides a viable framework for combining different data modalities to model patient's states of health, and can potentially be used to generate alerts to identify patients at high risk of hospital mortality.

A Model-Based Machine Learning Approach to Probing Autonomic Regulation From Nonstationary Vital-Sign Time Series

Physiological variables, such as heart rate (HR), blood pressure (BP) and respiration (RESP), are tightly regulated and coupled under healthy conditions, and a break-down in the coupling has been associated with aging and disease. We present an approach that incorporates physiological modeling within a switching linear dynamical systems (SLDS) framework to assess the various functional components of the autonomic regulation through transfer function analysis of nonstationary multivariate time series of vital signs. We validate our proposed SLDS-based transfer function analysis technique in automatically capturing 1) changes in baroreflex gain due to postural changes in a tilt-table study including ten subjects, and 2) the effect of aging on the autonomic control using HR/RESP recordings from 40 healthy adults. Next, using HR/BP time series of more than 450 adult ICU patients, we show that our technique can be used to reveal coupling changes associated with severe sepsis (AUC = 0.74, sensitivity = 0.74, specificity = 0.60). Our findings indicate that reduced HR/BP coupling is significantly associated with severe sepsis even after adjusting for clinical interventions (P  0.001). These results demonstrate the utility of our approach in phenotyping complex vital-sign dynamics, and in providing mechanistic hypotheses in terms of break-down of autoregulatory systems under healthy and disease conditions.

An open access database for the evaluation of heart sound algorithms

In the past few decades, analysis of heart sound signals (i.e. the phonocardiogram or PCG), especially for automated heart sound segmentation and classification, has been widely studied and has been reported to have the potential value to detect pathology accurately in clinical applications. However, comparative analyses of algorithms in the literature have been hindered by the lack of high-quality, rigorously validated, and standardized open databases of heart sound recordings. This paper describes a public heart sound database, assembled for an international competition, the PhysioNet/Computing in Cardiology (CinC) Challenge 2016. The archive comprises nine different heart sound databases sourced from multiple research groups around the world. It includes 2435 heart sound recordings in total collected from 1297 healthy subjects and patients with a variety of conditions, including heart valve disease and coronary artery disease. The recordings were collected from a variety of clinical or nonclinical (such as in-home visits) environments and equipment. The length of recording varied from several seconds to several minutes. This article reports detailed information about the subjects/patients including demographics (number, age, gender), recordings (number, location, state and time length), associated synchronously recorded signals, sampling frequency and sensor type used. We also provide a brief summary of the commonly used heart sound segmentation and classification methods, including open source code provided concurrently for the Challenge. A description of the PhysioNet/CinC Challenge 2016, including the main aims, the training and test sets, the hand corrected annotations for different heart sound states, the scoring mechanism, and associated open source code are provided. In addition, several potential benefits from the public heart sound database are discussed.

Proton Pump Inhibitors Are Not Associated With Acute Kidney Injury in Critical Illness

Recent epidemiologic data linking proton pump inhibitor (PPI) use to acute and chronic kidney dysfunction is yet to be validated in other populations, and mechanisms have not been explored. Using a large, well phenotyped inception cohort of 15 063 critically ill patients, we examined the risk of acute kidney injury (AKI), as defined by the Kidney Disease Improving Global Outcomes criteria guidelines, according to prior use of a PPI, histamine-2 receptor antagonist (H₂ RA), or neither. A total of 3725 (24.7%) patients reported PPI use prior to admission, while 905 (6.0%) patients reported H₂ RA use. AKI occurred in 747 (20.0%) and 163 (18.0%) of PPI and H₂ RA users respectively, compared to 1712 (16.2%) of those not taking acid suppressive medications. In unadjusted analysis, PPI and H₂ RA users had a 28% (95%CI 1.17-1.41, P < .001) and 10% (95%CI 0.91-1.30, P = .31) higher risk of AKI compared to those taking neither class of medication. However, in sequential models that included adjustment for demographics, cardiovascular comorbidities, indications for PPI use, and severity of illness, the effect of PPI on the risk of AKI was attenuated, and in the adjusted analysis, PPI was not associated with AKI (OR 1.02; 95%CI 0.91-1.13, P = .73). The presence of sterile pyuria and hypomagnesemia did not modify the association between PPI use and AKI. In summary, after adjustment for demographics, illness severity, and the indication for PPI use, PPI use prior to admission is not associated with critical illness AKI.

False alarm reduction in critical care

High false alarm rates in the ICU decrease quality of care by slowing staff response times while increasing patient delirium through noise pollution. The 2015 PhysioNet/Computing in Cardiology Challenge provides a set of 1250 multi-parameter ICU data segments associated with critical arrhythmia alarms, and challenges the general research community to address the issue of false alarm suppression using all available signals. Each data segment was 5 minutes long (for real time analysis), ending at the time of the alarm. For retrospective analysis, we provided a further 30 seconds of data after the alarm was triggered. A total of 750 data segments were made available for training and 500 were held back for testing. Each alarm was reviewed by expert annotators, at least two of whom agreed that the alarm was either true or false. Challenge participants were invited to submit a complete, working algorithm to distinguish true from false alarms, and received a score based on their program's performance on the hidden test set. This score was based on the percentage of alarms correct, but with a penalty that weights the suppression of true alarms five times more heavily than acceptance of false alarms. We provided three example entries based on well-known, open source signal processing algorithms, to serve as a basis for comparison and as a starting point for participants to develop their own code. A total of 38 teams submitted a total of 215 entries in this year's Challenge. This editorial reviews the background issues for this challenge, the design of the challenge itself, the key achievements, and the follow-up research generated as a result of the Challenge, published in the concurrent special issue of Physiological Measurement. Additionally we make some recommendations for future changes in the field of patient monitoring as a result of the Challenge.

MIMIC-III, a freely accessible critical care database

MIMIC-III (‘Medical Information Mart for Intensive Care’) is a large, single-center database comprising information relating to patients admitted to critical care units at a large tertiary care hospital. Data includes vital signs, medications, laboratory measurements, observations and notes charted by care providers, fluid balance, procedure codes, diagnostic codes, imaging reports, hospital length of stay, survival data, and more. The database supports applications including academic and industrial research, quality improvement initiatives, and higher education coursework.

A Visualization of Evolving Clinical Sentiment Using Vector Representations of Clinical Notes

Our objective in this paper was to visualize the evolution of clinical language and sentiment with respect to several common population-level categories including: time in the hospital, age, mortality, gender and race. Our analysis utilized seven years of unstructured free text notes from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) database. The text data was partitioned by category and used to generate several high dimensional vector space representations. We generated visualizations of the vector spaces using Distributed Stochastic Neighbor Embedding (tSNE) and Principal Component Analysis (PCA). We also investigated representative words from clusters in the vector space. Lastly, we inferred the general sentiment of the clinical notes toward each parameter by gauging the average distance between positive and negative keywords and all other terms in the space. We found intriguing differences in the sentiment of clinical notes over time, outcome, and demographic features. We noted a decrease in the homogeneity and complexity of clusters over time for patients with poor outcomes. We also found greater positive sentiment for females, unmarried patients, and patients of African ethnicity.

Hemodynamic Monitoring Using Switching Autoregressive Dynamics of Multivariate Vital Sign Time Series

In a critical care setting, shock and resuscitation endpoints are often defined based on arterial blood pressure values. Patient-specific fluctuations and interactions between heart rate (HR) and blood pressure (BP), however, may provide additional prognostic value to stratify individual patients' risks for adverse outcomes at different blood pressure targets. In this work, we use the switching autoregressive (SVAR) dynamics inferred from the multivariate vital sign time series to stratify mortality risks of intensive care units (ICUs) patients receiving vasopressor treatment. We model vital sign observations as generated from latent states from an autoregressive Hidden Markov Model (AR-HMM) process, and use the proportion of time patients stayed in different latent states to predict outcome. We evaluate the performance of our approach using minute-by-minute HR and mean arterial BP (MAP) of an ICU patient cohort while on vasopressor treatment. Our results indicate that the bivariate HR/MAP dynamics (AUC 0.74 [0.64, 0.84]) contain additional prognostic information beyond the MAP values (AUC 0.53 [0.42, 0.63]) in mortality prediction. Further, HR/MAP dynamics achieved better performance among a subgroup of patients in a low MAP range (median MAP < 65 mmHg) while on pressors. A realtime implementation of our approach may provide clinicians a tool to quantify the effectiveness of interventions and to inform treatment decisions.

Peripheral Edema, Central Venous Pressure, and Risk of AKI in Critical Illness

BACKGROUND AND OBJECTIVES

Although venous congestion has been linked to renal dysfunction in heart failure, its significance in a broader context has not been investigated.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Using an inception cohort of 12,778 critically ill adult patients admitted to an urban tertiary medical center between 2001 and 2008, we examined whether the presence of peripheral edema on admission physical examination was associated with an increased risk of AKI within the first 7 days of critical illness. In addition, in those with admission central venous pressure (CVP) measurements, we examined the association of CVPs with subsequent AKI. AKI was defined using the Kidney Disease Improving Global Outcomes criteria.

RESULTS

Of the 18% (n=2338) of patients with peripheral edema on admission, 27% (n=631) developed AKI, compared with 16% (n=1713) of those without peripheral edema. In a model that included adjustment for comorbidities, severity of illness, and the presence of pulmonary edema, peripheral edema was associated with a 30% higher risk of AKI (95% confidence interval [95% CI], 1.15 to 1.46; P<0.001), whereas pulmonary edema was not significantly related to risk. Peripheral edema was also associated with a 13% higher adjusted risk of a higher AKI stage (95% CI, 1.07 to 1.20; P<0.001). Furthermore, levels of trace, 1+, 2+, and 3+ edema were associated with 34% (95% CI, 1.10 to 1.65), 17% (95% CI, 0.96 to 1.14), 47% (95% CI, 1.18 to 1.83), and 57% (95% CI, 1.07 to 2.31) higher adjusted risk of AKI, respectively, compared with edema-free patients. In the 4761 patients with admission CVP measurements, each 1 cm H2O higher CVP was associated with a 2% higher adjusted risk of AKI (95% CI, 1.00 to 1.03; P=0.02).

CONCLUSIONS

Venous congestion, as manifested as either peripheral edema or increased CVP, is directly associated with AKI in critically ill patients. Whether treatment of venous congestion with diuretics can modify this risk will require further study.

Tracking progression of patient state of health in critical care using inferred shared dynamics in physiological time series

Physiologic systems generate complex dynamics in their output signals that reflect the changing state of the underlying control systems. In this work, we used a switching vector autoregressive (switching VAR) framework to systematically learn and identify a collection of vital sign dynamics, which can possibly be recurrent within the same patient and shared across the entire cohort. We show that these dynamical behaviors can be used to characterize and elucidate the progression of patients' states of health over time. Using the mean arterial blood pressure time series of 337 ICU patients during the first 24 hours of their ICU stays, we demonstrated that the learned dynamics from as early as the first 8 hours of patients' ICU stays can achieve similar hospital mortality prediction performance as the well-known SAPS-I acuity scores, suggesting that the discovered latent dynamics structure may yield more timely insights into the progression of a patient's state of health than the traditional snapshot-based acuity scores.

The PhysioNet/Computing in Cardiology Challenge 2015: Reducing False Arrhythmia Alarms in the ICU

High false alarm rates in the ICU decrease quality of care by slowing staff response times while increasing patient delirium through noise pollution. The 2015 Physio-Net/Computing in Cardiology Challenge provides a set of 1,250 multi-parameter ICU data segments associated with critical arrhythmia alarms, and challenges the general research community to address the issue of false alarm suppression using all available signals. Each data segment was 5 minutes long (for real time analysis), ending at the time of the alarm. For retrospective analysis, we provided a further 30 seconds of data after the alarm was triggered. A collection of 750 data segments was made available for training and a set of 500 was held back for testing. Each alarm was reviewed by expert annotators, at least two of whom agreed that the alarm was either true or false. Challenge participants were invited to submit a complete, working algorithm to distinguish true from false alarms, and received a score based on their program's performance on the hidden test set. This score was based on the percentage of alarms correct, but with a penalty that weights the suppression of true alarms five times more heavily than acceptance of false alarms. We provided three example entries based on well-known, open source signal processing algorithms, to serve as a basis for comparison and as a starting point for participants to develop their own code. A total of 38 teams submitted a total of 215 entries in this year's Challenge.

Increased incidence of diuretic use in critically ill obese patients

OBJECTIVE

Sodium retention occurs commonly in cardiac and liver disease, requiring the administration of diuretics to restore fluid balance. Whether obesity is associated with sodium retention has not been fully evaluated.

METHODS

In a large single-center cohort of critically ill patients, we evaluated whether admission body mass index was associated with the administration of either oral or intravenous diuretics during the intensive care unit (ICU) stay.

MAIN RESULTS

Of 7724 critically ill patients, 3946 (51.1%) were prescribed diuretics during the ICU stay. Overweight, class I obesity, and class II/III obesity were associated with a 1.35 (95% confidence interval [CI], 1.20-1.53; P < .001), 1.56 (95% CI, 1.35-1.80; P < .001), and 1.91 (95% CI, 1.61-2.26; P < .001) adjusted risk of receiving diuretics within the ICU, respectively. In adjusted analysis, a 5-kg/m(2) increment of body mass index was associated with a 1.19 (95% CI, 1.14-1.23; P < .001) increased adjusted risk of within-ICU diuretics. Among those patients receiving loop diuretics, obese patients received significantly larger daily diuretic doses.

CONCLUSION

Critically ill obese patients are more likely to receive diuretics during their stay in the ICU and to receive higher dosages of diuretics. Our data suggest that obesity is an independent risk factor for sodium retention.

Association between fluid balance and survival in critically ill patients

OBJECTIVE

Although the consequences of chronic fluid retention are well known, those of iatrogenic fluid retention that occurs during critical illness have not been fully determined. Therefore, we investigated the association between fluid balance and survival in a cohort of almost 16,000 individuals who survived an intensive care unit (ICU) stay in a large, urban, tertiary medical centre.

DESIGN

Longitudinal analysis of fluid balance at ICU discharge and 90-day post-ICU survival.

MEASUREMENTS

Associations between fluid balance during the ICU stay, determined from the electronic bedside record, and survival were tested using Cox proportional hazard models adjusted for severity of critical illness.

RESULTS

There were 1827 deaths in the first 90 days after ICU discharge. Compared with the lowest quartile of discharge fluid balance [median (interquartile range) -1.5 (-3.1, -0.7) L], the highest quartile [7.6 (5.7, 10.8) L] was associated with a 35% [95% confidence interval (CI) 1.13-1.61)] higher adjusted risk of death. Fluid balance was not associated with outcome amongst individuals without congestive heart failure or renal dysfunction. Amongst patients with either comorbidity, however, fluid balance was strongly associated with outcome, with the highest quartile having a 55% (95% CI 1.24-1.95) higher adjusted risk of death than the lowest quartile. Isotonic fluid balance, defined as the difference between intravenous isotonic fluid administration and urine output, was similarly associated with 90-day outcomes.

CONCLUSION

Positive fluid balance at the time of ICU discharge is associated with increased risk of death, after adjusting for markers of illness severity and chronic medical conditions, particularly in patients with underlying heart or kidney disease. Restoration of euvolaemia prior to discharge may improve survival after acute illness.

Proton pump inhibitor use is not associated with cardiac arrhythmia in critically ill patients

Hypomagnesemia can lead to cardiac arrythmias. Recently, observational data have linked chronic proton pump inhibitor (PPI) exposure to hypomagnesemia. Whether PPI exposure increases the risk for arrhythmias has not been well studied. Using a large, single-center inception cohort of critically ill patients, we examined whether PPI exposure was associated with admission electrocardiogram readings of a cardiac arrhythmia in more than 8000 patients. There were 25.4% PPI users, whereas 6% were taking a histamine 2 antagonist. In all, 14.0% had a cardiac arrhythmia. PPI use was associated with an unadjusted risk of arrhythmia of 1.15 (95% CI,1.00-1.32; P =.04) and an adjusted risk of arrhythmia of 0.91 (95% CI, 0.77-1.06; P =.22). Among diuretic users (n = 2476), PPI use was similarly not associated with an increased risk of cardiac arrhythmia. In summary, in a large cohort of critically ill patients, PPI exposure is not associated with an increased risk of cardiac arrhythmia.

Making big data useful for health care: a summary of the inaugural mit critical data conference

With growing concerns that big data will only augment the problem of unreliable research, the Laboratory of Computational Physiology at the Massachusetts Institute of Technology organized the Critical Data Conference in January 2014. Thought leaders from academia, government, and industry across disciplines—including clinical medicine, computer science, public health, informatics, biomedical research, health technology, statistics, and epidemiology—gathered and discussed the pitfalls and challenges of big data in health care. The key message from the conference is that the value of large amounts of data hinges on the ability of researchers to share data, methodologies, and findings in an open setting. If empirical value is to be from the analysis of retrospective data, groups must continuously work together on similar problems to create more effective peer review. This will lead to improvement in methodology and quality, with each iteration of analysis resulting in more reliability.

A physiological time series dynamics-based approach to patient monitoring and outcome prediction

Cardiovascular variables such as heart rate (HR) and blood pressure (BP) are regulated by an underlying control system, and therefore, the time series of these vital signs exhibit rich dynamical patterns of interaction in response to external perturbations (e.g., drug administration), as well as pathological states (e.g., onset of sepsis and hypotension). A question of interest is whether "similar" dynamical patterns can be identified across a heterogeneous patient cohort, and be used for prognosis of patients' health and progress. In this paper, we used a switching vector autoregressive framework to systematically learn and identify a collection of vital sign time series dynamics, which are possibly recurrent within the same patient and may be shared across the entire cohort. We show that these dynamical behaviors can be used to characterize the physiological "state" of a patient. We validate our technique using simulated time series of the cardiovascular system, and human recordings of HR and BP time series from an orthostatic stress study with known postural states. Using the HR and BP dynamics of an intensive care unit (ICU) cohort of over 450 patients from the MIMIC II database, we demonstrate that the discovered cardiovascular dynamics are significantly associated with hospital mortality (dynamic modes 3 and 9, p=0.001, p=0.006 from logistic regression after adjusting for the APACHE scores). Combining the dynamics of BP time series and SAPS-I or APACHE-III provided a more accurate assessment of patient survival/mortality in the hospital than using SAPS-I and APACHE-III alone (p=0.005 and p=0.045). Our results suggest that the discovered dynamics of vital sign time series may contain additional prognostic value beyond that of the baseline acuity measures, and can potentially be used as an independent predictor of outcomes in the ICU.

Discovering shared dynamics in physiological signals: application to patient monitoring in ICU

Modern clinical databases include time series of vital signs, which are often recorded continuously during a hospital stay. Over several days, these recordings may yield many thousands of samples. In this work, we explore the feasibility of characterizing the "state of health" of a patient using the physiological dynamics inferred from these time series. The ultimate objective is to assist clinicians in allocating resources to high-risk patients. We hypothesize that "similar" patients exhibit similar dynamics and the properties and duration of these states are indicative of health and disease. We use Bayesian nonparametric machine learning methods to discover shared dynamics in patients' blood pressure (BP) time series. Each such "dynamic" captures a distinct pattern of evolution of BP and is possibly recurrent within the same time series and shared across multiple patients. Next, we examine the utility of this low-dimensional representation of BP time series for predicting mortality in patients. Our results are based on an intensive care unit (ICU) cohort of 480 patients (with 16% mortality) and indicate that the dynamics of time series of vital signs can be an independent useful predictor of outcome in ICU.

Proton-pump inhibitor use is associated with low serum magnesium concentrations

Although case reports link proton-pump inhibitor (PPI) use and hypomagnesemia, no large-scale studies have been conducted. Here we examined the serum magnesium concentration and the likelihood of hypomagnesemia (<1.6 mg/dl) with a history of PPI or histamine-2 receptor antagonist used to reduce gastric acid, or use of neither among 11,490 consecutive adult admissions to an intensive care unit of a tertiary medical center. Of these, 2632 patients reported PPI use prior to admission, while 657 patients were using a histamine-2 receptor antagonist. PPI use was associated with 0.012 mg/dl lower adjusted serum magnesium concentration compared to users of no acid-suppressive medications, but this effect was restricted to those patients taking diuretics. Among the 3286 patients concurrently on diuretics, PPI use was associated with a significant increase of hypomagnesemia (odds ratio 1.54) and 0.028 mg/dl lower serum magnesium concentration. Among those not using diuretics, PPI use was not associated with serum magnesium levels. Histamine-2 receptor antagonist use was not significantly associated with magnesium concentration without or with diuretic use. The use of PPI was not associated with serum phosphate concentration regardless of diuretic use. Thus, we verify case reports of the association between PPI use and hypomagnesemia in those concurrently taking diuretics. Hence, serum magnesium concentrations should be followed in susceptible individuals on chronic PPI therapy.

Accessing the public MIMIC-II intensive care relational database for clinical research

Background

The Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II) database is a free, public resource for intensive care research. The database was officially released in 2006, and has attracted a growing number of researchers in academia and industry. We present the two major software tools that facilitate accessing the relational database: the web-based QueryBuilder and a downloadable virtual machine (VM) image.

Results

QueryBuilder and the MIMIC-II VM have been developed successfully and are freely available to MIMIC-II users. Simple example SQL queries and the resulting data are presented. Clinical studies pertaining to acute kidney injury and prediction of fluid requirements in the intensive care unit are shown as typical examples of research performed with MIMIC-II. In addition, MIMIC-II has also provided data for annual PhysioNet/Computing in Cardiology Challenges, including the 2012 Challenge “Predicting mortality of ICU Patients”.

Conclusions

QueryBuilder is a web-based tool that provides easy access to MIMIC-II. For more computationally intensive queries, one can locally install a complete copy of MIMIC-II in a VM. Both publicly available tools provide the MIMIC-II research community with convenient querying interfaces and complement the value of the MIMIC-II relational database.

Signal quality estimation with multichannel adaptive filtering in intensive care settings

A signal quality estimate of a physiological waveform can be an important initial step for automated processing of real-world data. This paper presents a new generic point-by-point signal quality index (SQI) based on adaptive multichannel prediction that does not rely on ad hoc morphological feature extraction from the target waveform. An application of this new SQI to photoplethysmograms (PPG), arterial blood pressure (ABP) measurements, and ECG showed that the SQI is monotonically related to signal-to-noise ratio (simulated by adding white Gaussian noise) and to subjective human quality assessment of 1361 multichannel waveform epochs. A receiver-operating-characteristic (ROC) curve analysis, with the human "bad" quality label as positive and the "good" quality label as negative, yielded areas under the ROC curve of 0.86 (PPG), 0.82 (ABP), and 0.68 (ECG).

Open-access MIMIC-II database for intensive care research

The critical state of intensive care unit (ICU) patients demands close monitoring, and as a result a large volume of multi-parameter data is collected continuously. This represents a unique opportunity for researchers interested in clinical data mining. We sought to foster a more transparent and efficient intensive care research community by building a publicly available ICU database, namely Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II). The data harnessed in MIMIC-II were collected from the ICUs of Beth Israel Deaconess Medical Center from 2001 to 2008 and represent 26,870 adult hospital admissions (version 2.6). MIMIC-II consists of two major components: clinical data and physiological waveforms. The clinical data, which include patient demographics, intravenous medication drip rates, and laboratory test results, were organized into a relational database. The physiological waveforms, including 125 Hz signals recorded at bedside and corresponding vital signs, were stored in an open-source format. MIMIC-II data were also deidentified in order to remove protected health information. Any interested researcher can gain access to MIMIC-II free of charge after signing a data use agreement and completing human subjects training. MIMIC-II can support a wide variety of research studies, ranging from the development of clinical decision support algorithms to retrospective clinical studies. We anticipate that MIMIC-II will be an invaluable resource for intensive care research by stimulating fair comparisons among different studies.

Collective Experience: A Database-Fuelled, Inter-Disciplinary Team-Led Learning System

We describe the framework of a data-fuelled, interdisciplinary team-led learning system. The idea is to build models using patients from one's own institution whose features are similar to an index patient as regards an outcome of interest, in order to predict the utility of diagnostic tests and interventions, as well as inform prognosis. The Laboratory of Computational Physiology at the Massachusetts Institute of Technology developed and maintains MIMIC-II, a public deidentified high- resolution database of patients admitted to Beth Israel Deaconess Medical Center. It hosts of teams of clinicians (nurses, doctors, pharmacists) and scientists (database engineers, modelers, epidemiologists) who translate the day-to-day questions during rounds that have no clear answers in the current medical literature into study designs, perform the modeling and the analysis and publish their findings. The studies fall into the following broad categories: identification and interrogation of practice variation, predictive modeling of clinical outcomes within patient subsets and comparative effectiveness research on diagnostic tests and therapeutic interventions. Clinical databases such as MIMIC-II, where recorded health care transactions - clinical decisions linked with patient outcomes - are constantly uploaded, become the centerpiece of a learning system.

Predicting In-Hospital Mortality of ICU Patients: The PhysioNet/Computing in Cardiology Challenge 2012

Acuity scores, such as APACHE, SAPS, MPM, and SOFA, are widely used to account for population differences in studies aiming to compare how medications, care guidelines, surgery, and other interventions impact mortality in Intensive Care Unit (ICU) patients. By contrast, the focus of the PhysioNet/CinC Challenge 2012 is to develop methods for patient-specific prediction of in-hospital mortality. The data used for the challenge consisted of 5 general descriptors and 36 time series (measurements of vital signs and laboratory results) from the first 48 hours of the first available ICU stay of 12,000 adult patients from the MIMIC II database. The challenge was organized as two events: event 1 measured performance of a binary classifier, and event 2 measured performance of a risk estimator. The score of event 1 was the lower of sensitivity and positive predictive value. The score for event 2 was a range-normalized Hosmer-Lemeshow statistic. A baseline algorithm (using SAPS-1) obtained event 1 and 2 scores of 0.3125 and 68.58 respectively. Most participants submitted entries that outperformed the baseline algorithm. The top final scores for events 1 and 2 were 0.5353 and 17.88 respectively.