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.

Prognostic models of in-hospital mortality of intensive care patients using neural representation of unstructured text: A systematic review and critical appraisal.

OBJECTIVE: To review and critically appraise published and preprint reports of prognostic models of in-hospital mortality of patients in the intensive-care unit (ICU) based on neural representations (embeddings) of clinical notes. METHODS: PubMed and arXiv were searched up to August 1, 2022. At least two reviewers independently selected the studies that developed a prognostic model of in-hospital mortality of intensive-care patients using free-text represented as embeddings and extracted data using the CHARMS checklist. Risk of bias was assessed using PROBAST. Reporting on the model was assessed with the TRIPOD guideline. To assess the machine learning components that were used in the models, we present a new descriptive framework based on different techniques to represent text and provide predictions from text. The study protocol was registered in the PROSPERO database (CRD42022354602). RESULTS: Eighteen studies out of 2,825 were included. All studies used the publicly-available MIMIC dataset. Context-independent word embeddings are widely used. Model discrimination was provided by all studies (AUROC 0.75-0.96), but measures of calibration were scarce. Seven studies used both structural clinical variables and notes. Model discrimination improved when adding clinical notes to variables. None of the models was externally validated and often a simple train/test split was used for internal validation. Our critical appraisal demonstrated a high risk of bias in all studies and concerns regarding their applicability in clinical practice. CONCLUSION: All studies used a neural architecture for prediction and were based on one publicly available dataset. Clinical notes were reported to improve predictive performance when used in addition to only clinical variables. Most studies had methodological, reporting, and applicability issues. We recommend reporting both model discrimination and calibration, using additional data sources, and using more robust evaluation strategies, including prospective and external validation. Finally, sharing data and code is encouraged to improve study reproducibility.

Dynamic prediction of mortality in COVID-19 patients in the intensive care unit: A retrospective multi-center cohort study.

Background: The COVID-19 pandemic continues to overwhelm intensive care units (ICUs) worldwide, and improved prediction of mortality among COVID-19 patients could assist decision making in the ICU setting. In this work, we report on the development and validation of a dynamic mortality model specifically for critically ill COVID-19 patients and discuss its potential utility in the ICU. Methods: We collected electronic medical record (EMR) data from 3222 ICU admissions with a COVID-19 infection from 25 different ICUs in the Netherlands. We extracted daily observations of each patient and fitted both a linear (logistic regression) and non-linear (random forest) model to predict mortality within 24 h from the moment of prediction. Isotonic regression was used to re-calibrate the predictions of the fitted models. We evaluated the models in a leave-one-ICU-out (LOIO) cross-validation procedure. Results: The logistic regression and random forest model yielded an area under the receiver operating characteristic curve of 0.87 [0.85; 0.88] and 0.86 [0.84; 0.88], respectively. The recalibrated model predictions showed a calibration intercept of -0.04 [-0.12; 0.04] and slope of 0.90 [0.85; 0.95] for logistic regression model and a calibration intercept of -0.19 [-0.27; -0.10] and slope of 0.89 [0.84; 0.94] for the random forest model. Discussion: We presented a model for dynamic mortality prediction, specifically for critically ill COVID-19 patients, which predicts near-term mortality rather than in-ICU mortality. The potential clinical utility of dynamic mortality models such as benchmarking, improving resource allocation and informing family members, as well as the development of models with more causal structure, should be topics for future research.

Can we reliably automate clinical prognostic modelling? A retrospective cohort study for ICU triage prediction of in-hospital mortality of COVID-19 patients in the Netherlands.

BACKGROUND: Building Machine Learning (ML) models in healthcare may suffer from time-consuming and potentially biased pre-selection of predictors by hand that can result in limited or trivial selection of suitable models. We aimed to assess the predictive performance of automating the process of building ML models (AutoML) in-hospital mortality prediction modelling of triage COVID-19 patients at ICU admission versus expert-based predictor pre-selection followed by logistic regression. METHODS: We conducted an observational study of all COVID-19 patients admitted to Dutch ICUs between February and July 2020. We included 2,690 COVID-19 patients from 70 ICUs participating in the Dutch National Intensive Care Evaluation (NICE) registry. The main outcome measure was in-hospital mortality. We asessed model performance (at admission and after 24h, respectively) of AutoML compared to the more traditional approach of predictor pre-selection and logistic regression. FINDINGS: Predictive performance of the autoML models with variables available at admission shows fair discrimination (average AUROC = 0·75-0·76 (sdev = 0·03), PPV = 0·70-0·76 (sdev = 0·1) at cut-off = 0·3 (the observed mortality rate), and good calibration. This performance is on par with a logistic regression model with selection of patient variables by three experts (average AUROC = 0·78 (sdev = 0·03) and PPV = 0·79 (sdev = 0·2)). Extending the models with variables that are available at 24h after admission resulted in models with higher predictive performance (average AUROC = 0·77-0·79 (sdev = 0·03) and PPV = 0·79-0·80 (sdev = 0·10-0·17)). CONCLUSIONS: AutoML delivers prediction models with fair discriminatory performance, and good calibration and accuracy, which is as good as regression models with expert-based predictor pre-selection. In the context of the restricted availability of data in an ICU quality registry, extending the models with variables that are available at 24h after admission showed small (but significantly) performance increase.

Monitoring, management, and outcome of hypotension in Intensive Care Unit patients, an international survey of the European Society of Intensive Care Medicine.

INTRODUCTION: Hypotension in the ICU is common, yet management is challenging and variable. Insight in management by ICU physicians and nurses may improve patient care and guide future hypotension treatment trials and guidelines. METHODS: We conducted an international survey among ICU personnel to provide insight in monitoring, management, and perceived consequences of hypotension. RESULTS: Out of 1464 respondents, 1197 (81.7%) were included (928 physicians (77.5%) and 269 nurses (22.5%)). The majority indicated that hypotension is underdiagnosed (55.4%) and largely preventable (58.8%). Nurses are primarily in charge of monitoring changes in blood pressure, physicians are in charge of hypotension treatment. Balanced crystalloids, dobutamine, norepinephrine, and Trendelenburg position were the most frequently reported fluid, inotrope, vasopressor, and positional maneuver used to treat hypotension. Reported complications believed to be related to hypotension were AKI and myocardial injury. Most ICUs do not have a specific hypotension treatment guideline or protocol (70.6%), but the majority would like to have one in the future (58.1%). CONCLUSIONS: Both physicians and nurses report that hypotension in ICU patients is underdiagnosed, preventable, and believe that hypotension influences morbidity. Hypotension management is generally not protocolized, but the majority of respondents would like to have a specific hypotension management protocol.

Definition and incidence of hypotension in intensive care unit patients, an international survey of the European Society of Intensive Care Medicine.

INTRODUCTION: Although hypotension in ICU patients is associated with adverse outcome, currently used definitions are unknown and no universally accepted definition exists. METHODS: We conducted an international, peer-reviewed survey among ICU physicians and nurses to provide insight in currently used definitions, estimations of incidence, and duration of hypotension. RESULTS: Out of 1394 respondents (1055 physicians (76%) and 339 nurses (24%)), 1207 (82%) completed the questionnaire. In all patient categories, hypotension definitions were predominantly based on an absolute MAP of 65 mmHg, except for the neuro(trauma) category (75 mmHg, p < 0.001), without differences between answers from physicians and nurses. Hypotension incidence was estimated at 55%, and time per day spent in hypotension at 15%, both with nurses reporting higher percentages than physicians (estimated mean difference 5%, p = 0.01; and 4%, p < 0.001). CONCLUSIONS: An absolute MAP threshold of 65 mmHg is most frequently used to define hypotension in ICU patients. In neuro(trauma) patients a higher threshold was reported. The majority of ICU patients are estimated to endure hypotension during their ICU admission for a considerable amount of time, with nurses reporting a higher estimated incidence and time spent in hypotension than physicians.

Postoperative microcirculatory perfusion and endothelial glycocalyx shedding following cardiac surgery with cardiopulmonary bypass.

We investigated microcirculatory perfusion disturbances following cardiopulmonary bypass in the early postoperative period and whether the course of these disturbances mirrored restoration of endothelial glycocalyx integrity. We performed sublingual sidestream dark field imaging of the microcirculation during the first three postoperative days in patients who had undergone on-pump coronary artery bypass graft surgery. We calculated the perfused vessel density, proportion of perfused vessels and perfused boundary region. Plasma was obtained to measure heparan sulphate and syndecan-1 levels as glycocalyx shedding markers. We recruited 17 patients; the mean (SD) duration of non-pulsatile cardiopulmonary bypass was 103 (18) min, following which 491 (29) ml autologous blood was transfused through cell salvage. Cardiopulmonary bypass immediately decreased both microcirculatory perfused vessel density; 11 (3) vs. 16 (4) mm.mm-2 , p = 0.052 and the proportion of perfused vessels; 92 (5) vs. 69 (9) %, p < 0.0001. The proportion of perfused vessels did not increase after transfusion of autologous salvaged blood following cardiopulmonary bypass; 72 (7) %, p = 0.19 or during the first three postoperative days; 71 (5) %, p < 0.0001. The perfused boundary region increased after cardiopulmonary bypass; 2.2 (0.3) vs. 1.9 (0.3) µm, p = 0.037 and during the first three postoperative days; 2.4 (0.3) vs. 1.9 (0.3) µm, p = 0.003. Increased plasma heparan sulphate levels were inversely associated with the proportion of perfused vessels during cardiopulmonary bypass; R = -0.49, p = 0.02. Plasma syndecan-1 levels were inversely associated with the proportion of perfused vessels during the entire study period; R = -0.51, p < 0.0001. Our study shows that cardiopulmonary bypass-induced acute microcirculatory perfusion disturbances persist in the first three postoperative days, and are associated with prolonged endothelial glycocalyx shedding. This suggests prolonged impairment and delayed recovery of both microcirculatory perfusion and function after on-pump cardiac surgery.

[Chronic thromboembolic pulmonary hypertension: recognition, diagnosis and treatment]. / Chronische trombo-embolische pulmonale hypertensie.

- Chronic thromboembolic pulmonary hypertension (CTEPH), characterised by pulmonary hypertension and persistent perfusion defects despite adequate anticoagulation, causes significant morbidity and mortality.- Persistent dyspnoea after acute pulmonary embolism is frequent and an indication for additional diagnostics. Only a minority of these patients develop CTEPH.- Echocardiography and perfusion scintigraphy are the cornerstone of diagnostics when suspecting CTEPH. Right-heart catheterisation and pulmonary angiography should confirm the diagnosis.- The diagnostic phase is preferably carried out in an expert centre in order to optimise the diagnosis and choose the optimal treatment for each individual patient.- Treatment of patients with CTEPH is a multidisciplinary team effort.- Pulmonary endarterectomy is the only potentially curative treatment; perioperative mortality is less than 5% in experienced centres. Inoperable patients can be treated with medication that specifically targets pulmonary arterial hypertension, but a survival benefit has not yet been shown for this medication.- Balloon pulmonary angioplasty has recently become available in the Netherlands as a treatment option, but the exact role of this new technique in the treatment of patients with CTEPH still needs to be investigated.

[Healthcare professionals back in the driving seat]. / Zorgprofessionals moeten stuur weer overnemen.

This commentary discusses the increasingly observed managerilisation of healthcare. Managerilisation frequently results in a framework of rules, regulations and accompanying time-consuming forms and procedures to guide decision-making. Although likely developed with the best of intentions in mind, this framework may be of limited value and tends to leave healthcare professionals feeling frustrated and distrusted. In addition, overzealous bureaucracy and rigid adherence to protocols may be disadvantageous to patient care and outcomes. Instead, we advocate a renewed focus on common sense and in particular on a renewed trust in healthcare professionals. Their professional judgement is based on many years of education and bedside experience. Hospital management should once again seek to embrace their expertise, while healthcare professionals should actively seek to regain the reins when it comes to delivering healthcare.

[Should every doctor be allowed to use ultrasound? Dilemmas surrounding broad application of ultrasound in clinical practice]. / Mag iedere arts zonder meer echo's maken?

Ultrasound is rapidly gaining ground in clinical medicine. This offers distinct advantages for diagnosis and treatment. This is notably the case when moving images are created by the treating physician, who can integrate them immediately with all other clinical information. The downside of a broad application of ultrasound is an increase in the number of incidental findings and missed diagnoses. This is amplified by the frequent lack of formal requirements for training and skills. Storage of ultrasound images may furthermore lead to verifiable misinterpretations. We are of the opinion that responsible integration of ultrasound in clinical practice requires clear peer agreements without sectarian thinking. To illustrate this, we discuss the dilemmas surrounding increased use of ultrasound from a medicolegal perspective.

Lung ultrasound: routine practice for the next generation of internists.

BACKGROUND: The lung is at the crossroads of ventilation and circulation and can provide a wealth of diagnostic information. In the past, lung ultrasound (LUS) was considered impossible. However, the interplay between air, fluid and pleurae creates distinctive artefacts. Combinations of these artefacts can help differentiate between various pathological processes, including pulmonary oedema, pneumonia, pulmonary embolism, obstructive airway disease and pneumothorax. LUS, when used by experienced physicians, is superior to chest X-ray and comparable to computed tomography for establishing a diagnosis in acutely dyspnoeic patients. LUS allows for rapid, non-invasive and bedside patient assessment. It is therefore unfortunate that unlike many other medical specialists in the Netherlands, internists have not yet incorporated LUS into their daily practice. OBJECTIVES: This review aims to be the starting point for internists wanting to acquire competence in LUS. REVIEW CONTENT: This narrative review describes the principles of ultrasound equipment, LUS artefacts, gives practical guidance to perform LUS and provides a road map towards LUS competence. Furthermore, it presents a decision tree to differentiate between causes of acute dyspnoea. AUTHORS CONCLUSIONS: LUS is a promising diagnostic technique that can be of great help for the internist. It can be applied directly at the bedside and can also be used to follow up on disease progression and therapy. It is our belief that it will replace the stethoscope and that it will be the most used imaging technique in the near future, especially in dyspnoeic patients.