Multi-Task Learning with Recurrent Neural Networks for ARDS Prediction using only EHR Data: Model Development and Validation Study (Preprint)

BACKGROUND Acute Respiratory Distress Syndrome (ARDS) is a condition that is often considered to have broad and subjective diagnostic criteria and is associated with significant mortality and morbidity. Early and accurate prediction of ARDS and related conditions such as hypoxemia and sepsis could allow timely administration of therapies, leading to improved patient outcomes. OBJECTIVE To perform an exploration of how multi-label classification in the clinical setting can take advantage of the underlying dependencies between ARDS and related conditions to improve early prediction of ARDS. METHODS The electronic health record dataset included 40,073 patient encounters from 7 hospitals from 4/20/2018 to 3/17/2021. A recurrent neural network (RNN) was trained using data from 5 hospitals, and external validation was conducted on data from 2 hospitals. In addition to ARDS, 12 target labels for related conditions such as sepsis, hypoxemia and Covid-19 were used to train the model to classify a total of 13 outputs. As a comparator, XGBoost models were developed for each of the 13 target labels. Model performance was assessed using the area under the receiver operating characteristic (AUROC). Heatmaps to visualize attention scores were generated to provide interpretability to the NNs. Finally, cluster analysis was performed to identify potential phenotypic subgroups of ARDS patients. RESULTS The single RNN model trained to classify 13 outputs outperformed the XGBoost model for ARDS prediction, achieving an AUROC of 0.842 on the external test sets. Models trained on an increasing number of tasks resulted in increasing performance. Earlier diagnosis of ARDS nearly doubled the rate of in-hospital survival. Cluster analysis revealed distinct ARDS subgroups, some of which had similar mortality rates but different clinical presentations. CONCLUSIONS The RNN model presented in this paper can be used as an early warning system to stratify patients who are at risk of developing one of the multiple risk outcomes, hence providing practitioners with means to take early action.

Evaluation of the Effect of COVID-19 Pneumonia on Kidney Function

Background: Severe acute respiratory tract infection, pneumonia, kidney failure, and multi-organ failure may develop in cases that result in death due to COVID-19. It is emphasized that the awareness of healthcare professionals about kidney functions should be increased in cases of COVID-19 pneumonia. Quick and effective steps can be taken in the treatment of COVID-19 pneumonia with the controlling approach of nurses to changes in kidney functions. Method: This study was carried out retrospectively to evaluate the kidney functions of patients diagnosed with COVID-19 pneumonia who were hospitalized in the pandemic hospital. Hospital and nurse observation files of 120 patients who were introduced to COVID-19 pneumonia between 1 May and 30 November 2020 were examined. Categorical data were described as continuous data as median with interquartile range (IQR) and percentages (%). Results: In total, 30 patients (25.0%) required mechanical ventilation, Overall, 39.1% (47) developed acute kidney injury during hospitalization, out of which 10.8% reached stage 1, 15.0% reached stage 2, and 13.3% reached stage 3. Dialytic support was required for seven (17.1% of all patients). COVID-19 pneumonia patients had higher levels of aspartate aminotransferase (AST) (55.02±58.04), alanine aminotransferase (ALT) (74.07±140.94), lactate dehydrogenase (LDH) (483.48±477.51), C-reactive protein (CRP) (88.02±72.17), D-dimer (1023±1548.01), procalcitonin (3.70± 6.52). In addition, a proportion of COVID-19 pneumonia patients but no non-COVID-19 pneumonia patients had abnormally increased AST (10.0-274.0), ALT (7.0-854.0), LDH (164-3547), CRP (5.10- 310.90), D-dimer (151-6212), procalcitonin (195-433). SpO2 of COVID-19 pneumonia patients had 78-97%, patients who need dialysis treatment due to pneumonia, follow-up coagulation profile (Procalcitonin, LDH, D-dimer), liver-renal function (ALT, AST, Creatine, Urea, Albumin), assessing signs of DVT and psychological support. 89 patients (74.2%) received corticosteroid, 73 patients (60.8%) received expectorant, 61 patients (50.8%) received vitamin C or B complex, 110 patients (91.7%) received anticoagulant and 73 patients (60.8%) received antibiotics. All of the COVID-19 pneumonia patients received the antiviral drug. Conclusion: As the disease progresses, differences in laboratory results and radiological findings may indicate that some complications have developed. COVID-19 pneumonia draws attention with liver function tests such as AST / ALT, LDH, infection markers in the blood, and the high rate of coagulation factors such as PCT and D-dimer during the hospital stay. The fact that these elevated values ​​may cause necrosis in the kidneys also brings about the truth. Careful monitoring of laboratory findings such as elevation of AST / ALT, LDH, PCT, and D-dimer in patients who develop pneumonia due to COVID-19 may provide early action for kidney damage.

SAGE advice and political decision-making: ‘Following the science’ in times of epistemic uncertainty

This article presents a preliminary analysis of the advice provided by the UK government’s Scientific Advisory Group for Emergencies (SAGE) held between 22 January and 23 March 2020 in response to the emerging coronavirus pandemic. Drawing on the published minutes of the group’s meetings, the article examines what was known and not known, the assumptions and working practices that shaped their work, and how this knowledge was reflected in the decisions made by the government. In doing so, the article critically examines what it means for policy making to be ‘led by the science’ when the best available science is provisional and uncertain. Using ideas of ‘externality’ and ‘evidential significance’, the article argues that the apparent desire for high levels of certainty by both scientists and political decision-makers made early action impossible as the data needed were not, and could not be, available in time. This leads to an argument for changes to the institutions that provide scientific advice based on sociologically informed expectations of science in which expert judgement plays a more significant role.

Early Warning Early Action for the Banking Solvency Risk in the COVID-19 Pandemic Era: A Case Study of Indonesia

The COVID-19 pandemic has affected people’s lives and increased the banking solvency risk. This research aimed to build an early warning and early action simulation model to mitigate the solvency risk using the system dynamics methodology and the Powersim Studio 10© software. The addition of an early action simulation updates the existing early warning model. Through this model, the effect of policy design and options on potential solvency risks is known before implementation. The trials conducted at Bank BRI (BBRI) and Bank Mandiri (BMRI) showed that the model had the ability to provide an early warning of the potential increase in bank solvency risk when the loan restructuring policy is revoked. It also simulates the effectiveness of management’s policy options to mitigate these risks. This research used publicly accessible banking data and analysis. Bank management could also take advantage of this model through a self-stimulation facility developed in this study to accommodate their needs using the internal data.

Trajectory and Correlation of Intrinsic Capacity and Frailty in a Beijing Elderly Community

Objectives: The World Health Organization proposed intrinsic capacity (IC) model to guide the implementation of person-centered care plan aimed at preserving or reserving functional ability, especially in frail older adults. We aimed to show the trajectory of IC and the overlap between IC impairment and frailty and investigate the correlation between IC domains and frailty status transitions.Method: Longitudinal observational study covering 230 community-dwelling older adults (mean age 84.0 ± 4.5 years) at baseline, and transition information at 2-year follow-up (n = 196). IC was measured by five domains: locomotion, cognition, vitality, psychological, and sensory. Frailty was defined by FRAIL Scale. IC and frailty status transitions were assessed. Logistic regression, odds ratios (OR) and 95% confidence interval (CI) were used for the analysis.Results: The prevalence of frailty was 23.0% and increased up to 41.8% over two years. Regarding frailty transitions, 38.3% of older adults progressed to more frailty status, and 8.6% regressed to lesser frailty status. The prevalence of IC impairment was 67.9% and increased to 81.6% over two years. Regarding IC transitions, 49.2% of adults with no IC impairment at baseline kept stable, and 50.8% developed new IC impairment. Among individuals with IC impairment at baseline, 57.9% worsened, and 13.5% improved. Importantly, IC impairment at baseline existed in 42.4% robust adults, 83.3% pre-frail adults, and 93.3% frail adults. 47.1% individuals who kept non-frail status within two years experienced IC worsened transition. Univariable analysis illustrated that new impaired locomotion, vitality, cognition, and sensory domains increased the risk of non-frail progressed to frail status. After adjusting for covariables, new impaired locomotion (OR = 3.625, 95% CI: 1.348–9.747) and vitality domains (OR = 3.034, 95% CI: 1.229–7.487) were associated with a higher possibility of non-frail progressed to frail status.Conclusion: IC impairment and frailty overlap and co-exist in older adults. IC impairment, especially new impairment in locomotion and vitality are associated with the transitions from non-frail to frail status. It is important that geriatricians tightly monitor IC trajectory and find the new impaired domains to take early action to minimize the public health burden of frailty.

Public Institutional Structures for Disaster Preparedness in the Cereal Value Chain: A Zambian Case Study

Increasing extreme climate events and cyclonic activities provide clear evidence that the Southern African Development Community (SADC) region is a hotspot for climate change-driven natural disasters which critically disrupt agricultural production cycles. This is especially true with regard to the production of cereal, produce widely used to represent food security. Although studies have attempted to disentangle the effect of demand vis à vis projected population growth on cereal production across the region, the contradiction between cereal production and climate disaster preparedness remains poorly resolved. Therefore, literature on the subject matter is scanty. The present study is motivated by the need to overcome this paucity of literature and thus, deepen our understanding of cereal production and climate disaster preparedness in the region. Therefore, the main aim of this study is to assess public institutional support structures that are currently being employed for climate disaster preparedness in the cereal value chain across Zambia as perceived by small scale farmers. After a comprehensive assessment of focus group discussions (FGDs), several points emerge specifically highlighting four salient findings: first, results show that a government-led Farmer Input Support Programme (FISP) is the only strategy particularly targeted at disaster preparedness. All other initiatives are targeted at improving or safeguarding livelihoods with some components having a ripple effect on the cereal value chain. Second, results show that climate forecasts that are supposed to trigger early action are generally characterized by low prediction skill with more false alarms and misses than hits. Third, forecasts were found to lack geographical specificity with generalities over large areas being common thus, diminishing their usefulness at the local scale. Fourth, end-users found forecasts to usually contain technical jargon that is difficult to decipher especially that most small-scale farmers are illiterate. This study concludes that to fully support the cereal value chain and realize food security in Zambia, policy formulation that champion the establishment of an effective early warning and early action system (EWEAS) involving multiple interest groups and actors should be considered a matter of urgency.

Assessment of Impact of Patient Recruitment Volume on Risk Profile, Outcomes, and Treatment Effect in a Randomized Trial of Ticagrelor Versus Prasugrel in Acute Coronary Syndromes

BACKGROUND Whether there are differences in the risk profile and treatment effect in patients recruited in a low recruitment center (LRC) versus patients recruited in a high recruitment center (HRC) in a randomized multicenter trial remains unknown. METHODS AND RESULTS This study included 4018 patients with acute coronary syndrome recruited in the ISAR‐REACT 5 (Intracoronary Stenting and Antithrombotic Regimen: Rapid Early Action for Coronary Treatment 5) trial. The primary end point was a composite of all‐cause death, myocardial infarction, or stroke. Overall, 3011 patients (75%) were recruited in the HRCs (7 centers recruiting 258 to 628 patients; median, 413 patients) and 1007 patients (25%) were recruited in the LRCs (16 centers recruiting 5 to 201 patients; median, 52 patients). Patients recruited in the LRCs had more favorable cardiovascular risk profiles than patients recruited in the HRCs. The primary end point occurred in 72 patients in the LRCs and 249 patients in the HRCs (cumulative incidence, 7.3% and 8.4%; P =0.267). All‐cause mortality was lower among patients recruited in the LRCs (n=29) than among patients recruited in the HRCs (n=134; cumulative incidence 2.9% versus 4.5%; P =0.031). There was no significant interaction between the treatment effect of ticagrelor versus prasugrel and patient recruitment category (LRC versus HRC) regarding the primary efficacy end point (LRC: hazard ratio [HR], 1.42 [95% CI, 0.89–2.28]; HRC: HR, 1.33 [95% CI, 1.04−1.72]; P for interaction=0.800). CONCLUSIONS Patients with acute coronary syndrome recruited in a LRC appear to have more favorable cardiovascular risk profiles and lower 1‐year mortality rates compared with patients recruited in a HRC. The recruitment volume did not interact with the treatment effect of ticagrelor versus prasugrel. REGISTRATION URL: https://www.clinicaltrials.gov ; Unique identifier: NCT01944800.

Path to net zero is critical to climate outcome

Net zero greenhouse gas targets have become a central element for climate action. However, most company and government pledges focus on the year that net zero is reached, with limited awareness of how critical the emissions pathway is in determining the climate outcome in both the near- and long-term. Here we show that different pathways of carbon dioxide and methane—the most prominent long-lived and short-lived greenhouse gases, respectively—can lead to nearly 0.4 °C of warming difference in midcentury and potential overshoot of the 2 °C target, even if they technically reach global net zero greenhouse gas emissions in 2050. While all paths achieve the Paris Agreement temperature goals in the long-term, there is still a 0.2 °C difference by end-of-century. We find that early action to reduce both emissions of carbon dioxide and methane simultaneously leads to the best climate outcomes over all timescales. We therefore recommend that companies and countries supplement net zero targets with a two-basket set of interim milestones to ensure that early action is taken for both carbon dioxide and methane. A one-basket approach, such as the standard format for Nationally Determined Contributions, is not sufficient because it can lead to a delay in methane mitigation.