Empirical meropenem versus piperacillin/tazobactam for adult patients with sepsis (EMPRESS) trial: Protocol.

BACKGROUND: Piperacillin/tazobactam may be associated with less favourable outcomes than carbapenems in patients with severe bacterial infections, but the certainty of evidence is low. METHODS: The Empirical Meropenem versus Piperacillin/Tazobactam for Adult Patients with Sepsis (EMPRESS) trial is an investigator-initiated, international, parallel-group, randomised, open-label, adaptive clinical trial with an integrated feasibility phase. We will randomise adult, critically ill patients with sepsis to empirical treatment with meropenem or piperacillin/tazobactam for up to 30 days. The primary outcome is 30-day all-cause mortality. The secondary outcomes are serious adverse reactions within 30 days; isolation precautions due to resistant bacteria within 30 days; days alive without life support and days alive and out of hospital within 30 and 90 days; 90- and 180-day all-cause mortality and 180-day health-related quality of life. EMPRESS will use Bayesian statistical models with weak to somewhat sceptical neutral priors. Adaptive analyses will be conducted after follow-up of the primary outcome for the first 400 participants concludes and after every 300 subsequent participants, with adaptive stopping for superiority/inferiority and practical equivalence (absolute risk difference <2.5%-points) and response-adaptive randomisation. The expected sample sizes in scenarios with no, small or large differences are 5189, 5859 and 2570 participants, with maximum 14,000 participants and ≥99% probability of conclusiveness across all scenarios. CONCLUSIONS: EMPRESS will compare the effects of empirical meropenem against piperacillin/tazobactam in adult, critically ill patients with sepsis. Due to the pragmatic, adaptive design with high probability of conclusiveness, the trial results are expected to directly inform clinical practice.

Platform trials.

Platform trials focus on the perpetual testing of many interventions in a disease or a setting. These trials have lasting organizational, administrative, data, analytic, and operational frameworks making them highly efficient. The use of adaptation often increases the probabilities of allocating participants to better interventions and obtaining conclusive results. The COVID-19 pandemic showed the potential of platform trials as a fast and valid way to improved treatments. This review gives an overview of key concepts and elements using the Intensive Care Platform Trial (INCEPT) as an example.

Development of a dynamic prediction model for unplanned ICU admission and mortality in hospitalized patients.

Frequent assessment of the severity of illness for hospitalized patients is essential in clinical settings to prevent outcomes such as in-hospital mortality and unplanned admission to the intensive care unit (ICU). Classical severity scores have been developed typically using relatively few patient features. Recently, deep learning-based models demonstrated better individualized risk assessments compared to classic risk scores, thanks to the use of aggregated and more heterogeneous data sources for dynamic risk prediction. We investigated to what extent deep learning methods can capture patterns of longitudinal change in health status using time-stamped data from electronic health records. We developed a deep learning model based on embedded text from multiple data sources and recurrent neural networks to predict the risk of the composite outcome of unplanned ICU transfer and in-hospital death. The risk was assessed at regular intervals during the admission for different prediction windows. Input data included medical history, biochemical measurements, and clinical notes from a total of 852,620 patients admitted to non-intensive care units in 12 hospitals in Denmark's Capital Region and Region Zealand during 2011-2016 (with a total of 2,241,849 admissions). We subsequently explained the model using the Shapley algorithm, which provides the contribution of each feature to the model outcome. The best model used all data modalities with an assessment rate of 6 hours, a prediction window of 14 days and an area under the receiver operating characteristic curve of 0.898. The discrimination and calibration obtained with this model make it a viable clinical support tool to detect patients at higher risk of clinical deterioration, providing clinicians insights into both actionable and non-actionable patient features.

Abnormal serum levels of magnesium, phosphate, and zinc in ICU patients-Characteristics, management, and outcomes: The WhyTrace cohort study.

BACKGROUND: Abnormal serum levels of magnesium, phosphate, and zinc appear common in intensive care unit (ICU) patients, but the epidemiology, management, and associations with outcomes are less well described. We described these factors and estimated associations with outcomes in a large dataset of Danish ICU patients. METHODS: We included adults who were acutely admitted to 10 general ICUs in Denmark between October 2011 and January 2018. From the dataset, we obtained characteristics of patients who had serum levels measured of magnesium, phosphate, or zinc, including data on supplementation. We used joint models with death as a competing outcome to estimate the associations between abnormal serum levels and time to successful extubation and, for magnesium, also incident tachyarrhythmia. RESULTS: We included 16,517 of 36,514 patients in the dataset. The cumulative probability of hypomagnesemia within 28 days was 64% (95% confidence interval [CI] 62-66); of hypophosphatemia 74% (95% CI 72-75); and of hypozincemia 98% (95% CI 98-98). Supplementation of magnesium was used in 3554 out of 13,506 (26%) patients, phosphate in 2115 out of 14,148 (15%) patients, and zinc in 4465 out of 9869 (45%) patients. During ICU stay, 38% experienced hypermagnesemia, 58% hyperphosphatemia, and 1% hyperzincemia. Low serum levels of magnesium, phosphate, and zinc were associated with shorter time to successful extubation, and high serum magnesium and phosphate and low serum zinc with the competing risk of increased mortality, but too few serum measurements made the results inconclusive. CONCLUSION: In this multicenter cohort study of acutely admitted ICU patients, most experienced low serum levels of magnesium, phosphate, or zinc during ICU stay, many received supplementation, and experiencing both low and high serum levels during ICU stay was not uncommon. Associations between serum levels and clinical outcomes appeared inconclusive because the data proved unfit for these analyses.

Discrete-time survival analysis in the critically ill: a deep learning approach using heterogeneous data.

Prediction of survival for patients in intensive care units (ICUs) has been subject to intense research. However, no models exist that embrace the multiverse of data in ICUs. It is an open question whether deep learning methods using automated data integration with minimal pre-processing of mixed data domains such as free text, medical history and high-frequency data can provide discrete-time survival estimates for individual ICU patients. We trained a deep learning model on data from patients admitted to ten ICUs in the Capital Region of Denmark and the Region of Southern Denmark between 2011 and 2018. Inspired by natural language processing we mapped the electronic patient record data to an embedded representation and fed the data to a recurrent neural network with a multi-label output layer representing the chance of survival at different follow-up times. We evaluated the performance using the time-dependent concordance index. In addition, we quantified and visualized the drivers of survival predictions using the SHAP methodology. We included 37,355 admissions of 29,417 patients in our study. Our deep learning models outperformed traditional Cox proportional-hazard models with concordance index in the ranges 0.72-0.73, 0.71-0.72, 0.71, and 0.69-0.70, for models applied at baseline 0, 24, 48, and 72 h, respectively. Deep learning models based on a combination of entity embeddings and survival modelling is a feasible approach to obtain individualized survival estimates in data-rich settings such as the ICU. The interpretable nature of the models enables us to understand the impact of the different data domains.

Language-agnostic pharmacovigilant text mining to elicit side effects from clinical notes and hospital medication records.

We sought to craft a drug safety signalling pipeline associating latent information in clinical free text with exposures to single drugs and drug pairs. Data arose from 12 secondary and tertiary public hospitals in two Danish regions, comprising approximately half the Danish population. Notes were operationalised with a fastText embedding, based on which we trained 10 270 neural-network models (one for each distinct single-drug/drug-pair exposure) predicting the risk of exposure given an embedding vector. We included 2 905 251 admissions between May 2008 and June 2016, with 13 740 564 distinct drug prescriptions; the median number of prescriptions was 5 (IQR: 3-9) and in 1 184 340 (41%) admissions patients used ≥5 drugs concomitantly. A total of 10 788 259 clinical notes were included, with 179 441 739 tokens retained after pruning. Of 345 single-drug signals reviewed, 28 (8.1%) represented possibly undescribed relationships; 186 (54%) signals were clinically meaningful. Sixteen (14%) of the 115 drug-pair signals were possible interactions, and two (1.7%) were known. In conclusion, we built a language-agnostic pipeline for mining associations between free-text information and medication exposure without manual curation, predicting not the likely outcome of a range of exposures but also the likely exposures for outcomes of interest. Our approach may help overcome limitations of text mining methods relying on curated data in English and can help leverage non-English free text for pharmacovigilance.

Using Machine Learning to Identify Patients at High Risk of Inappropriate Drug Dosing in Periods with Renal Dysfunction.

PURPOSE: Dosing of renally cleared drugs in patients with kidney failure often deviates from clinical guidelines, so we sought to elicit predictors of receiving inappropriate doses of renal risk drugs. PATIENTS AND METHODS: We combined data from the Danish National Patient Register and in-hospital data on drug administrations and estimated glomerular filtration rates for admissions between 1 October 2009 and 1 June 2016, from a pool of about 2.6 million persons. We trained artificial neural network and linear logistic ridge regression models to predict the risk of five outcomes (>0, ≥1, ≥2, ≥3 and ≥5 inappropriate doses daily) with index set 24 hours after admission. We used time-series validation for evaluating discrimination, calibration, clinical utility and explanations. RESULTS: Of 52,451 admissions included, 42,250 (81%) were used for model development. The median age was 77 years; 50% of admissions were of women. ≥5 drugs were used between admission start and index in 23,124 admissions (44%); the most common drug classes were analgesics, systemic antibacterials, diuretics, antithrombotics, and antacids. The neural network models had better discriminative power (all AUROCs between 0.77 and 0.81) and were better calibrated than their linear counterparts. The main prediction drivers were use of anti-inflammatory, antidiabetic and anti-Parkinson's drugs as well as having a diagnosis of chronic kidney failure. Sex and age affected predictions but slightly. CONCLUSION: Our models can flag patients at high risk of receiving at least one inappropriate dose daily in a controlled in-silico setting. A prospective clinical study may confirm that this holds in real-life settings and translates into benefits in hard endpoints.

Using machine learning for predicting intensive care unit resource use during the COVID-19 pandemic in Denmark.

The COVID-19 pandemic has put massive strains on hospitals, and tools to guide hospital planners in resource allocation during the ebbs and flows of the pandemic are urgently needed. We investigate whether machine learning (ML) can be used for predictions of intensive care requirements a fixed number of days into the future. Retrospective design where health Records from 42,526 SARS-CoV-2 positive patients in Denmark was extracted. Random Forest (RF) models were trained to predict risk of ICU admission and use of mechanical ventilation after n days (n = 1, 2, …, 15). An extended analysis was provided for n = 5 and n = 10. Models predicted n-day risk of ICU admission with an area under the receiver operator characteristic curve (ROC-AUC) between 0.981 and 0.995, and n-day risk of use of ventilation with an ROC-AUC between 0.982 and 0.997. The corresponding n-day forecasting models predicted the needed ICU capacity with a coefficient of determination (R2) between 0.334 and 0.989 and use of ventilation with an R2 between 0.446 and 0.973. The forecasting models performed worst, when forecasting many days into the future (for large n). For n = 5, ICU capacity was predicted with ROC-AUC 0.990 and R2 0.928, and use of ventilator was predicted with ROC-AUC 0.994 and R2 0.854. Random Forest-based modelling can be used for accurate n-day forecasting predictions of ICU resource requirements, when n is not too large.

Developing and validating COVID-19 adverse outcome risk prediction models from a bi-national European cohort of 5594 patients.

Patients with severe COVID-19 have overwhelmed healthcare systems worldwide. We hypothesized that machine learning (ML) models could be used to predict risks at different stages of management and thereby provide insights into drivers and prognostic markers of disease progression and death. From a cohort of approx. 2.6 million citizens in Denmark, SARS-CoV-2 PCR tests were performed on subjects suspected for COVID-19 disease; 3944 cases had at least one positive test and were subjected to further analysis. SARS-CoV-2 positive cases from the United Kingdom Biobank was used for external validation. The ML models predicted the risk of death (Receiver Operation Characteristics-Area Under the Curve, ROC-AUC) of 0.906 at diagnosis, 0.818, at hospital admission and 0.721 at Intensive Care Unit (ICU) admission. Similar metrics were achieved for predicted risks of hospital and ICU admission and use of mechanical ventilation. Common risk factors, included age, body mass index and hypertension, although the top risk features shifted towards markers of shock and organ dysfunction in ICU patients. The external validation indicated fair predictive performance for mortality prediction, but suboptimal performance for predicting ICU admission. ML may be used to identify drivers of progression to more severe disease and for prognostication patients in patients with COVID-19. We provide access to an online risk calculator based on these findings.

Dynamic and explainable machine learning prediction of mortality in patients in the intensive care unit: a retrospective study of high-frequency data in electronic patient records.

BACKGROUND: Many mortality prediction models have been developed for patients in intensive care units (ICUs); most are based on data available at ICU admission. We investigated whether machine learning methods using analyses of time-series data improved mortality prognostication for patients in the ICU by providing real-time predictions of 90-day mortality. In addition, we examined to what extent such a dynamic model could be made interpretable by quantifying and visualising the features that drive the predictions at different timepoints. METHODS: Based on the Simplified Acute Physiology Score (SAPS) III variables, we trained a machine learning model on longitudinal data from patients admitted to four ICUs in the Capital Region, Denmark, between 2011 and 2016. We included all patients older than 16 years of age, with an ICU stay lasting more than 1 h, and who had a Danish civil registration number to enable 90-day follow-up. We leveraged static data and physiological time-series data from electronic health records and the Danish National Patient Registry. A recurrent neural network was trained with a temporal resolution of 1 h. The model was internally validated using the holdout method with 20% of the training dataset and externally validated using previously unseen data from a fifth hospital in Denmark. Its performance was assessed with the Matthews correlation coefficient (MCC) and area under the receiver operating characteristic curve (AUROC) as metrics, using bootstrapping with 1000 samples with replacement to construct 95% CIs. A Shapley additive explanations algorithm was applied to the prediction model to obtain explanations of the features that drive patient-specific predictions, and the contributions of each of the 44 features in the model were analysed and compared with the variables in the original SAPS III model. FINDINGS: From a dataset containing 15 615 ICU admissions of 12 616 patients, we included 14 190 admissions of 11 492 patients in our analysis. Overall, 90-day mortality was 33·1% (3802 patients). The deep learning model showed a predictive performance on the holdout testing dataset that improved over the timecourse of an ICU stay: MCC 0·29 (95% CI 0·25-0·33) and AUROC 0·73 (0·71-0·74) at admission, 0·43 (0·40-0·47) and 0·82 (0·80-0·84) after 24 h, 0·50 (0·46-0·53) and 0·85 (0·84-0·87) after 72 h, and 0·57 (0·54-0·60) and 0·88 (0·87-0·89) at the time of discharge. The model exhibited good calibration properties. These results were validated in an external validation cohort of 5827 patients with 6748 admissions: MCC 0·29 (95% CI 0·27-0·32) and AUROC 0·75 (0·73-0·76) at admission, 0·41 (0·39-0·44) and 0·80 (0·79-0·81) after 24 h, 0·46 (0·43-0·48) and 0·82 (0·81-0·83) after 72 h, and 0·47 (0·44-0·49) and 0·83 (0·82-0·84) at the time of discharge. INTERPRETATION: The prediction of 90-day mortality improved with 1-h sampling intervals during the ICU stay. The dynamic risk prediction can also be explained for an individual patient, visualising the features contributing to the prediction at any point in time. This explanation allows the clinician to determine whether there are elements in the current patient state and care that are potentially actionable, thus making the model suitable for further validation as a clinical tool. FUNDING: Novo Nordisk Foundation and the Innovation Fund Denmark.

Heterogeneity of treatment effect of prophylactic pantoprazole in adult ICU patients: a post hoc analysis of the SUP-ICU trial.

PURPOSE: The Stress Ulcer Prophylaxis in the Intensive Care Unit (SUP-ICU) trial compared prophylactic pantoprazole with placebo in 3291 adult ICU patients at risk of clinically important gastrointestinal bleeding (CIB). As a predefined subgroup analysis suggested increased 90-day mortality with pantoprazole in the most severely ill patients, we aimed to further explore whether heterogenous treatment effects (HTE) were present. METHODS: We assessed HTE in subgroups defined according to illness severity by SAPS II quintiles and the total number of risk factors for CIB using Bayesian hierarchical models, and on the continuous scale using Bayesian logistic regression models with interactions. Estimates were presented as posterior probability distributions of odds ratios (ORs), probabilities of different effect sizes, and marginal effects plots. RESULTS: We observed potential HTE for 90-day mortality according to illness severity (median subgroup OR range 0.90-1.09) with higher risk in the most severely ill, but not with different numbers of risk factors (1.00-1.02). We observed potential HTE of pantoprazole for clinically important events (0.86-1.18) and infectious adverse events (0.88-1.27) with higher risk in patients with greater illness severity and in those with more risk factors for CIB. Pantoprazole substantially and consistently reduced the risk of CIB with no indications of HTE (0.53-0.63). CONCLUSIONS: In this post hoc analysis of the SUP-ICU trial, we found indications of HTE with increased risks of serious adverse events in patients with greater illness severity or more risk factors for CIB allocated to pantoprazole. These findings are hypothesis-generating and warrant further prospective investigation. CLINICALTRIALS. GOV IDENTIFIER: NCT02467621.

Effect of Selepressin vs Placebo on Ventilator- and Vasopressor-Free Days in Patients With Septic Shock: The SEPSIS-ACT Randomized Clinical Trial.

IMPORTANCE: Norepinephrine, the first-line vasopressor for septic shock, is not always effective and has important catecholaminergic adverse effects. Selepressin, a selective vasopressin V1a receptor agonist, is a noncatecholaminergic vasopressor that may mitigate sepsis-induced vasodilatation, vascular leakage, and edema, with fewer adverse effects. OBJECTIVE: To test whether selepressin improves outcome in septic shock. DESIGN, SETTING, AND PARTICIPANTS: An adaptive phase 2b/3 randomized clinical trial comprising 2 parts that included adult patients (n = 868) with septic shock requiring more than 5 µg/min of norepinephrine. Part 1 used a Bayesian algorithm to adjust randomization probabilities to alternative selepressin dosing regimens and to trigger transition to part 2, which would compare the best-performing regimen with placebo. The trial was conducted between July 2015 and August 2017 in 63 hospitals in Belgium, Denmark, France, the Netherlands, and the United States, and follow-up was completed by May 2018. INTERVENTIONS: Random assignment to 1 of 3 dosing regimens of selepressin (starting infusion rates of 1.7, 2.5, and 3.5 ng/kg/min; n = 585) or to placebo (n = 283), all administered as continuous infusions titrated according to hemodynamic parameters. MAIN OUTCOMES AND MEASURES: Primary end point was ventilator- and vasopressor-free days within 30 days (deaths assigned zero days) of commencing study drug. Key secondary end points were 90-day mortality, kidney replacement therapy-free days, and ICU-free days. RESULTS: Among 868 randomized patients, 828 received study drug (mean age, 66.3 years; 341 [41.2%] women) and comprised the primary analysis cohort, of whom 562 received 1 of 3 selepressin regimens, 266 received placebo, and 817 (98.7%) completed the trial. The trial was stopped for futility at the end of part 1. Median study drug duration was 37.8 hours (IQR, 17.8-72.4). There were no significant differences in the primary end point (ventilator- and vasopressor-free days: 15.0 vs 14.5 in the selepressin and placebo groups; difference, 0.6 [95% CI, -1.3 to 2.4]; P = .30) or key secondary end points (90-day mortality, 40.6% vs 39.4%; difference, 1.1% [95% CI, -6.5% to 8.8%]; P = .77; kidney replacement therapy-free days: 18.5 vs 18.2; difference, 0.3 [95% CI, -2.1 to 2.6]; P = .85; ICU-free days: 12.6 vs 12.2; difference, 0.5 [95% CI, -1.2 to 2.2]; P = .41). Adverse event rates included cardiac arrhythmias (27.9% vs 25.2% of patients), cardiac ischemia (6.6% vs 5.6%), mesenteric ischemia (3.2% vs 2.6%), and peripheral ischemia (2.3% vs 2.3%). CONCLUSIONS AND RELEVANCE: Among patients with septic shock receiving norepinephrine, administration of selepressin, compared with placebo, did not result in improvement in vasopressor- and ventilator-free days within 30 days. Further research would be needed to evaluate the potential role of selepressin for other patient-centered outcomes in septic shock. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT02508649.

Heterogeneity of treatment effect of stress ulcer prophylaxis in ICU patients: A secondary analysis protocol.

BACKGROUND: In the Stress Ulcer Prophylaxis in the Intensive Care Unit (SUP-ICU) trial, 3291 adult ICU patients at risk for gastrointestinal (GI) bleeding were randomly allocated to intravenous pantoprazole 40 mg or placebo once daily in the ICU. No difference was observed between the groups in the primary outcome 90-day mortality or the secondary outcomes, except for clinically important gastrointestinal bleeding. However, heterogeneity of treatment effect (HTE) not detected by conventional subgroup analyses could be present. METHODS: This is a protocol and statistical analysis plan for a secondary, post hoc, exploratory analysis of the SUP-ICU trial. We will explore HTE in one set of subgroups based on severity of illness (using the Simplified Acute Physiology Score [SAPS] II) and another set of subgroups based on the total number of risk factors for GI bleeding in each patient using Bayesian hierarchical models. We will summarise posterior probability distributions using medians and 95% credible intervals and present probabilities for different levels of benefit and harm of the intervention in each subgroup. Finally, we will assess if the treatment effect interacts with SAPS II and the number of risk factors separately on the continuous scale using marginal effects plots. CONCLUSIONS: The outlined post hoc analysis will explore whether HTE was present in the SUP-ICU trial and may help answer some of the remaining questions regarding the balance between benefits and harms of pantoprazole in ICU patients at risk of GI bleeding. CLINICALTRIALS. GOV REGISTRATION: NCT02467621.

No reduction in mortality after centralisation in treatment of patients with ruptured abdominal aneurism.

INTRODUCTION: Ruptured abdominal aortic aneurism (rAAA) is a severe condition with all-cause mortality rates reaching 80%. We speculated whether the 2008 centralisation of the treatment of patients with rAAA in Denmark had improved outcome as suggested in other surgical specialties. Accordingly, our aim was to describe temporal changes in mortality for patients undergoing surgery for rAAA in the Capital Region of Denmark between 2009 and 2015. METHODS: This was a retrospective population-based cohort study of patients in the intensive care unit diagnosed and treated for rAAA at Rigshospitalet from 1 January 2009 to 31 December 2015. Patient characteristics and procedure-related variables were obtained from the medical records. The primary outcome measure was death within 90 days of the primary surgical procedure. RESULTS: A total of 339 patients were diagnosed with rAAA, and 275 patients were included in the final study population; 26.9% of the patients died within 90 days of the primary surgical procedure, whereas the 30-day and one-year mortality was 18.5% and 31.6%, respectively. No consistent reduction in mortality was observed throughout the observation period. CONCLUSIONS: In this population-based cohort study of patients surgically treated for rAAA, we found no consistent reduction in mortality over time following centralisation of treatment. FUNDING: none. TRIAL REGISTRATION: not relevant.

Survival prediction in intensive-care units based on aggregation of long-term disease history and acute physiology: a retrospective study of the Danish National Patient Registry and electronic patient records.

BACKGROUND: Intensive-care units (ICUs) treat the most critically ill patients, which is complicated by the heterogeneity of the diseases that they encounter. Severity scores based mainly on acute physiology measures collected at ICU admission are used to predict mortality, but are non-specific, and predictions for individual patients can be inaccurate. We investigated whether inclusion of long-term disease history before ICU admission improves mortality predictions. METHODS: Registry data for long-term disease histories for more than 230 000 Danish ICU patients were used in a neural network to develop an ICU mortality prediction model. Long-term disease histories and acute physiology measures were aggregated to predict mortality risk for patients for whom both registry and ICU electronic patient record data were available. We compared mortality predictions with admission scores on the Simplified Acute Physiology Score (SAPS) II, the Acute Physiologic Assessment and Chronic Health Evaluation (APACHE) II, and the best available multimorbidity score, the Multimorbidity Index. An external validation set from an additional hospital was acquired after model construction to confirm the validity of our model. During initial model development data were split into a training set (85%) and an independent test set (15%), and a five-fold cross-validation was done during training to avoid overfitting. Neural networks were trained for datasets with disease history of 1 month, 3 months, 6 months, 1 year, 2·5 years, 5 years, 7·5 years, 10 years, and 23 years before ICU admission. FINDINGS: Mortality predictions with a model based solely on disease history outperformed the Multimorbidity Index (Matthews correlation coefficient 0·265 vs 0·065), and performed similarly to SAPS II and APACHE II (Matthews correlation coefficient with disease history, age, and sex 0·326 vs 0·347 and 0·300 for SAPS II and APACHE II, respectively). Diagnoses up to 10 years before ICU admission affected current mortality prediction. Aggregation of previous disease history and acute physiology measures in a neural network yielded the most precise predictions of in-hospital mortality (Matthews correlation coefficient 0·391 for in-hospital mortality compared with 0·347 with SAPS II and 0·300 with APACHE II). These results for the aggregated model were validated in an external independent dataset of 1528 patients (Matthews correlation coefficient for prediction of in-hospital mortality 0·341). INTERPRETATION: Longitudinal disease-spectrum-wide data available before ICU admission are useful for mortality prediction. Disease history can be used to differentiate mortality risk between patients with similar vital signs with more precision than SAPS II and APACHE II scores. Machine learning models can be deconvoluted to generate novel understandings of how ICU patient features from long-term and short-term events interact with each other. Explainable machine learning models are key in clinical settings, and our results emphasise how to progress towards the transformation of advanced models into actionable, transparent, and trustworthy clinical tools. FUNDING: Novo Nordisk Foundation and Innovation Fund Denmark.