Comparative performance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data

INTRODUCTION

Benchmarking intensive care units for audit and feedback is frequently based on comparing actual mortality versus predicted mortality. Traditionally, mortality prediction models rely on a limited number of input variables and significant manual data entry and curation. Using automatically extracted electronic health record data may be a promising alternative. However, adequate data on comparative performance between these approaches is currently lacking.

METHODS

The AmsterdamUMCdb intensive care database was used to construct a baseline APACHE IV in-hospital mortality model based on data typically available through manual data curation. Subsequently, new in-hospital mortality models were systematically developed and evaluated. New models differed with respect to the extent of automatic variable extraction, classification method, recalibration usage and the size of collection window.

RESULTS

A total of 13 models were developed based on data from 5,077 admissions divided into a train (80%) and test (20%) cohort. Adding variables or extending collection windows only marginally improved discrimination and calibration. An XGBoost model using only automatically extracted variables, and therefore no acute or chronic diagnoses, was the best performing automated model with an AUC of 0.89 and a Brier score of 0.10.

DISCUSSION

Performance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data is similar. Importantly, our results suggest that variables typically requiring manual curation, such as diagnosis at admission and comorbidities, may not be necessary for accurate mortality prediction. These proof-of-concept results require replication using multi-centre data.

Does Reinforcement Learning Improve Outcomes for Critically Ill Patients? A Systematic Review and Level-of-Readiness Assessment

OBJECTIVE

Reinforcement learning (RL) is a machine learning technique uniquely effective at sequential decision-making, which makes it potentially relevant to ICU treatment challenges. We set out to systematically review, assess level-of-readiness and meta-analyze the effect of RL on outcomes for critically ill patients.

DATA SOURCES

A systematic search was performed in PubMed, Embase.com, Clarivate Analytics/Web of Science Core Collection, Elsevier/SCOPUS and the Institute of Electrical and Electronics Engineers Xplore Digital Library from inception to March 25, 2022, with subsequent citation tracking.

DATA EXTRACTION

Journal articles that used an RL technique in an ICU population and reported on patient health-related outcomes were included for full analysis. Conference papers were included for level-of-readiness assessment only. Descriptive statistics, characteristics of the models, outcome compared with clinician's policy and level-of-readiness were collected. RL-health risk of bias and applicability assessment was performed.

DATA SYNTHESIS

A total of 1,033 articles were screened, of which 18 journal articles and 18 conference papers, were included. Thirty of those were prototyping or modeling articles and six were validation articles. All articles reported RL algorithms to outperform clinical decision-making by ICU professionals, but only in retrospective data. The modeling techniques for the state-space, action-space, reward function, RL model training, and evaluation varied widely. The risk of bias was high in all articles, mainly due to the evaluation procedure.

CONCLUSION

In this first systematic review on the application of RL in intensive care medicine we found no studies that demonstrated improved patient outcomes from RL-based technologies. All studies reported that RL-agent policies outperformed clinician policies, but such assessments were all based on retrospective off-policy evaluation.

Augmented intelligence facilitates concept mapping across different electronic health records

INTRODUCTION

With the advent of artificial intelligence, the secondary use of routinely collected medical data from electronic healthcare records (EHR) has become increasingly popular. However, different EHR systems typically use different names for the same medical concepts. This obviously hampers scalable model development and subsequent clinical implementation for decision support. Therefore, converting original parameter names to a so-called ontology, a standardized set of predefined concepts, is necessary but time-consuming and labor-intensive. We therefore propose an augmented intelligence approach to facilitate ontology alignment by predicting correct concepts based on parameter names from raw electronic health record data exports.

METHODS

We used the manually mapped parameter names from the multicenter "Dutch ICU data warehouse against COVID-19" sourced from three types of EHR systems to train machine learning models for concept mapping. Data from 29 intensive care units on 38,824 parameters mapped to 1,679 relevant and unique concepts and 38,069 parameters labeled as irrelevant were used for model development and validation. We used the Natural Language Toolkit (NLTK) to preprocess the parameter names based on WordNet cognitive synonyms transformed by term-frequency inverse document frequency (TF-IDF), yielding numeric features. We then trained linear classifiers using stochastic gradient descent for multi-class prediction. Finally, we fine-tuned these predictions using information on distributions of the data associated with each parameter name through similarity score and skewness comparisons.

RESULTS

The initial model, trained using data from one hospital organization for each of three EHR systems, scored an overall top 1 precision of 0.744, recall of 0.771, and F1-score of 0.737 on a total of 58,804 parameters. Leave-one-hospital-out analysis returned an average top 1 recall of 0.680 for relevant parameters, which increased to 0.905 for the top 5 predictions. When reducing the training dataset to only include relevant parameters, top 1 recall was 0.811 and top 5 recall was 0.914 for relevant parameters. Performance improvement based on similarity score or skewness comparisons affected at most 5.23% of numeric parameters.

CONCLUSION

Augmented intelligence is a promising method to improve concept mapping of parameter names from raw electronic health record data exports. We propose a robust method for mapping data across various domains, facilitating the integration of diverse data sources. However, recall is not perfect, and therefore manual validation of mapping remains essential.

Evolution of Clinical Phenotypes of COVID-19 Patients During Intensive Care Treatment: An Unsupervised Machine Learning Analysis

BACKGROUND

Identification of clinical phenotypes in critically ill COVID-19 patients could improve understanding of the disease heterogeneity and enable prognostic and predictive enrichment. However, previous attempts did not take into account temporal dynamics with high granularity. By including the dimension of time, we aim to gain further insights into the heterogeneity of COVID-19.

METHODS

We used granular data from 3202 adult COVID patients in the Dutch Data Warehouse that were admitted to one of 25 Dutch ICUs between February 2020 and March 2021. Parameters including demographics, clinical observations, medications, laboratory values, vital signs, and data from life support devices were selected. Twenty-one datasets were created that each covered 24 h of ICU data for each day of ICU treatment. Clinical phenotypes in each dataset were identified by performing cluster analyses. Both evolution of the clinical phenotypes over time and patient allocation to these clusters over time were tracked.

RESULTS

The final patient cohort consisted of 2438 COVID-19 patients with a ICU mortality outcome. Forty-one parameters were chosen for cluster analysis. On admission, both a mild and a severe clinical phenotype were found. After day 4, the severe phenotype split into an intermediate and a severe phenotype for 11 consecutive days. Heterogeneity between phenotypes appears to be driven by inflammation and dead space ventilation. During the 21-day period, only 8.2% and 4.6% of patients in the initial mild and severe clusters remained assigned to the same phenotype respectively. The clinical phenotype half-life was between 5 and 6 days for the mild and severe phenotypes, and about 3 days for the medium severe phenotype.

CONCLUSIONS

Patients typically do not remain in the same cluster throughout intensive care treatment. This may have important implications for prognostic or predictive enrichment. Prominent dissimilarities between clinical phenotypes are predominantly driven by inflammation and dead space ventilation.

INCIDENCE, RISK FACTORS, AND OUTCOME OF SUSPECTED CENTRAL VENOUS CATHETER-RELATED INFECTIONS IN CRITICALLY ILL COVID-19 PATIENTS: A MULTICENTER RETROSPECTIVE COHORT STUDY

ABSTRACT

Background: Aims of this study were to investigate the prevalence and incidence of catheter-related infection, identify risk factors, and determine the relation of catheter-related infection with mortality in critically ill COVID-19 patients. Methods: This was a retrospective cohort study of central venous catheters (CVCs) in critically ill COVID-19 patients. Eligible CVC insertions required an indwelling time of at least 48 hours and were identified using a full-admission electronic health record database. Risk factors were identified using logistic regression. Differences in survival rates at day 28 of follow-up were assessed using a log-rank test and proportional hazard model. Results: In 538 patients, a total of 914 CVCs were included. Prevalence and incidence of suspected catheter-related infection were 7.9% and 9.4 infections per 1,000 catheter indwelling days, respectively. Prone ventilation for more than 5 days was associated with increased risk of suspected catheter-related infection; odds ratio, 5.05 (95% confidence interval 2.12-11.0). Risk of death was significantly higher in patients with suspected catheter-related infection (hazard ratio, 1.78; 95% confidence interval, 1.25-2.53). Conclusions: This study shows that in critically ill patients with COVID-19, prevalence and incidence of suspected catheter-related infection are high, prone ventilation is a risk factor, and mortality is higher in case of catheter-related infection.

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning

Background

For mechanically ventilated critically ill COVID-19 patients, prone positioning has quickly become an important treatment strategy, however, prone positioning is labor intensive and comes with potential adverse effects. Therefore, identifying which critically ill intubated COVID-19 patients will benefit may help allocate labor resources.

Methods

From the multi-center Dutch Data Warehouse of COVID-19 ICU patients from 25 hospitals, we selected all 3619 episodes of prone positioning in 1142 invasively mechanically ventilated patients. We excluded episodes longer than 24 h. Berlin ARDS criteria were not formally documented. We used supervised machine learning algorithms Logistic Regression, Random Forest, Naive Bayes, K-Nearest Neighbors, Support Vector Machine and Extreme Gradient Boosting on readily available and clinically relevant features to predict success of prone positioning after 4 h (window of 1 to 7 h) based on various possible outcomes. These outcomes were defined as improvements of at least 10% in PaO₂/FiO₂ ratio, ventilatory ratio, respiratory system compliance, or mechanical power. Separate models were created for each of these outcomes. Re-supination within 4 h after pronation was labeled as failure. We also developed models using a 20 mmHg improvement cut-off for PaO₂/FiO₂ ratio and using a combined outcome parameter. For all models, we evaluated feature importance expressed as contribution to predictive performance based on their relative ranking.

Results

The median duration of prone episodes was 17 h (11–20, median and IQR, N = 2632). Despite extensive modeling using a plethora of machine learning techniques and a large number of potentially clinically relevant features, discrimination between responders and non-responders remained poor with an area under the receiver operator characteristic curve of 0.62 for PaO₂/FiO₂ ratio using Logistic Regression, Random Forest and XGBoost. Feature importance was inconsistent between models for different outcomes. Notably, not even being a previous responder to prone positioning, or PEEP-levels before prone positioning, provided any meaningful contribution to predicting a successful next proning episode.

Conclusions

In mechanically ventilated COVID-19 patients, predicting the success of prone positioning using clinically relevant and readily available parameters from electronic health records is currently not feasible. Given the current evidence base, a liberal approach to proning in all patients with severe COVID-19 ARDS is therefore justified and in particular regardless of previous results of proning.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13613-022-01070-0.

Assess and validate predictive performance of models for in-hospital mortality in COVID-19 patients: A retrospective cohort study in the Netherlands comparing the value of registry data with high-granular electronic health records

PURPOSE

To assess, validate and compare the predictive performance of models for in-hospital mortality of COVID-19 patients admitted to the intensive care unit (ICU) over two different waves of infections. Our models were built with high-granular Electronic Health Records (EHR) data versus less-granular registry data.

METHODS

Observational study of all COVID-19 patients admitted to 19 Dutch ICUs participating in both the national quality registry National Intensive Care Evaluation (NICE) and the EHR-based Dutch Data Warehouse (hereafter EHR). Multiple models were developed on data from the first 24 h of ICU admissions from February to June 2020 (first COVID-19 wave) and validated on prospective patients admitted to the same ICUs between July and December 2020 (second COVID-19 wave). We assessed model discrimination, calibration, and the degree of relatedness between development and validation population. Coefficients were used to identify relevant risk factors.

RESULTS

A total of 1533 patients from the EHR and 1563 from the registry were included. With high granular EHR data, the average AUROC was 0.69 (standard deviation of 0.05) for the internal validation, and the AUROC was 0.75 for the temporal validation. The registry model achieved an average AUROC of 0.76 (standard deviation of 0.05) in the internal validation and 0.77 in the temporal validation. In the EHR data, age, and respiratory-system related variables were the most important risk factors identified. In the NICE registry data, age and chronic respiratory insufficiency were the most important risk factors.

CONCLUSION

In our study, prognostic models built on less-granular but readily-available registry data had similar performance to models built on high-granular EHR data and showed similar transportability to a prospective COVID-19 population. Future research is needed to verify whether this finding can be confirmed for upcoming waves.

Rapid Evaluation of Coronavirus Illness Severity (RECOILS) in intensive care: Development and validation of a prognostic tool for in-hospital mortality

Background

The prediction of in‐hospital mortality for ICU patients with COVID‐19 is fundamental to treatment and resource allocation. The main purpose was to develop an easily implemented score for such prediction.

Methods

This was an observational, multicenter, development, and validation study on a national critical care dataset of COVID‐19 patients. A systematic literature review was performed to determine variables possibly important for COVID‐19 mortality prediction. Using a logistic multivariable model with a LASSO penalty, we developed the Rapid Evaluation of Coronavirus Illness Severity (RECOILS) score and compared its performance against published scores.

Results

Our development (validation) cohort consisted of 1480 (937) adult patients from 14 (11) Dutch ICUs admitted between March 2020 and April 2021. Median age was 65 (65) years, 31% (26%) died in hospital, 74% (72%) were males, average length of ICU stay was 7.83 (10.25) days and average length of hospital stay was 15.90 (19.92) days. Age, platelets, PaO2/FiO2 ratio, pH, blood urea nitrogen, temperature, PaCO2, Glasgow Coma Scale (GCS) score measured within +/−24 h of ICU admission were used to develop the score. The AUROC of RECOILS score was 0.75 (CI 0.71–0.78) which was higher than that of any previously reported predictive scores (0.68 [CI 0.64–0.71], 0.61 [CI 0.58–0.66], 0.67 [CI 0.63–0.70], 0.70 [CI 0.67–0.74] for ISARIC 4C Mortality Score, SOFA, SAPS‐III, and age, respectively).

Conclusions

Using a large dataset from multiple Dutch ICUs, we developed a predictive score for mortality of COVID‐19 patients admitted to ICU, which outperformed other predictive scores reported so far.

Predictors for extubation failure in COVID-19 patients using a machine learning approach

INTRODUCTION

Determining the optimal timing for extubation can be challenging in the intensive care. In this study, we aim to identify predictors for extubation failure in critically ill patients with COVID-19.

METHODS

We used highly granular data from 3464 adult critically ill COVID patients in the multicenter Dutch Data Warehouse, including demographics, clinical observations, medications, fluid balance, laboratory values, vital signs, and data from life support devices. All intubated patients with at least one extubation attempt were eligible for analysis. Transferred patients, patients admitted for less than 24 h, and patients still admitted at the time of data extraction were excluded. Potential predictors were selected by a team of intensive care physicians. The primary and secondary outcomes were extubation without reintubation or death within the next 7 days and within 48 h, respectively. We trained and validated multiple machine learning algorithms using fivefold nested cross-validation. Predictor importance was estimated using Shapley additive explanations, while cutoff values for the relative probability of failed extubation were estimated through partial dependence plots.

RESULTS

A total of 883 patients were included in the model derivation. The reintubation rate was 13.4% within 48 h and 18.9% at day 7, with a mortality rate of 0.6% and 1.0% respectively. The grandient-boost model performed best (area under the curve of 0.70) and was used to calculate predictor importance. Ventilatory characteristics and settings were the most important predictors. More specifically, a controlled mode duration longer than 4 days, a last fraction of inspired oxygen higher than 35%, a mean tidal volume per kg ideal body weight above 8 ml/kg in the day before extubation, and a shorter duration in assisted mode (< 2 days) compared to their median values. Additionally, a higher C-reactive protein and leukocyte count, a lower thrombocyte count, a lower Glasgow coma scale and a lower body mass index compared to their medians were associated with extubation failure.

CONCLUSION

The most important predictors for extubation failure in critically ill COVID-19 patients include ventilatory settings, inflammatory parameters, neurological status, and body mass index. These predictors should therefore be routinely captured in electronic health records.

The Dutch Data Warehouse, a multicenter and full-admission electronic health records database for critically ill COVID-19 patients

BACKGROUND

The Coronavirus disease 2019 (COVID-19) pandemic has underlined the urgent need for reliable, multicenter, and full-admission intensive care data to advance our understanding of the course of the disease and investigate potential treatment strategies. In this study, we present the Dutch Data Warehouse (DDW), the first multicenter electronic health record (EHR) database with full-admission data from critically ill COVID-19 patients.

METHODS

A nation-wide data sharing collaboration was launched at the beginning of the pandemic in March 2020. All hospitals in the Netherlands were asked to participate and share pseudonymized EHR data from adult critically ill COVID-19 patients. Data included patient demographics, clinical observations, administered medication, laboratory determinations, and data from vital sign monitors and life support devices. Data sharing agreements were signed with participating hospitals before any data transfers took place. Data were extracted from the local EHRs with prespecified queries and combined into a staging dataset through an extract-transform-load (ETL) pipeline. In the consecutive processing pipeline, data were mapped to a common concept vocabulary and enriched with derived concepts. Data validation was a continuous process throughout the project. All participating hospitals have access to the DDW. Within legal and ethical boundaries, data are available to clinicians and researchers.

RESULTS

Out of the 81 intensive care units in the Netherlands, 66 participated in the collaboration, 47 have signed the data sharing agreement, and 35 have shared their data. Data from 25 hospitals have passed through the ETL and processing pipeline. Currently, 3464 patients are included in the DDW, both from wave 1 and wave 2 in the Netherlands. More than 200 million clinical data points are available. Overall ICU mortality was 24.4%. Respiratory and hemodynamic parameters were most frequently measured throughout a patient's stay. For each patient, all administered medication and their daily fluid balance were available. Missing data are reported for each descriptive.

CONCLUSIONS

In this study, we show that EHR data from critically ill COVID-19 patients may be lawfully collected and can be combined into a data warehouse. These initiatives are indispensable to advance medical data science in the field of intensive care medicine.

Electrocardiographic changes are strongly correlated with the extent of cardiac inflammation in mice with Coxsackievirus B3-induced viral myocarditis

OBJECTIVE

Viral myocarditis (VM) can induce changes in myocardial electrical conduction and arrhythmia. However, their relationship with myocarditis-associated arrhythmic substrates in the heart such as inflammation and fibrosis is relatively unknown. This we have analyzed in the present study.

METHODS

C3H mice were infected with 1×10⁵ plaque-forming units Coxsackievirus B3 (CVB3, n=68) and were compared with uninfected control mice (n=10). Electrocardiograms (ECGs) were recorded in all conscious mice shortly before sacrifice and included heart rate; P-R interval; QRS duration; QTc interval and R-peak amplitude of lead II and aVF. Mice were sacrificed at 4, 7, 10, 21, 35 or 49 days post-infection. Cardiac lesion size, calcification, fibrosis and cellular infiltration of CD45+ lymphocytes, MAC3+ macrophages, Ly6G+ neutrophils and mast cells were quantitatively determined in cross-sections of the ventricles. Putative relations between ECG changes and lesion size and/or cardiac inflammation were then analyzed.

RESULTS

Significant transient reductions in QRS duration and R-peak amplitude occurred between 4 and 14 days post-infection and returned to baseline values thereafter. The magnitude of these ECG changes strongly correlated to the extent of lymphocyte (days 7 and 14), macrophage (days 7 and 10) and neutrophil (days 7) infiltration. The ECG changes did not significantly correlate with lesion size and fibrosis.

CONCLUSION

VM induces transient changes in myocardial electrical conduction that are strongly related to cellular inflammation of the heart. These data show that even in mild VM, with relatively little cardiac damage, the inflammatory infiltrate can form an important arrhythmogenic substrate.