ABSTRACT
Passenger comfort systems such as Heating, Ventilation, and Air-Conditioning units (HVACs) usually lack the data monitoring quality enjoyed by mission-critical systems in trains. But climate change, in addition to the high ventilation standards enforced by authorities due to the COVID pandemic, have increased the importance of HVACs worldwide. We propose a machine learning (ML) approach to the challenge of failure detection from incomplete data, consisting of two steps: 1. human-annotation bootstrapping, on a fraction of temperature data, to detect ongoing functional loss and build an artificial ground truth (AGT);2. failure prediction from digital-data, using the AGT to train an ML model based on failure diagnose codes to foretell functional loss. We exercise our approach in trains of Dutch Railways, showing its implementation, ML-predictive capabilities (the ML model for the AGT can detect HVAC malfunctions online), limitations (we could not foretell failures from our digital data), and discussing its application to other assets. © 2022, Springer Nature Switzerland AG.
ABSTRACT
TOPIC: Chest Infections TYPE: Original Investigations PURPOSE: Certain risk factors and prognostic indicators have been identified in patients with coronavirus disease 2019 (COVID-19). However, no study has addressed the risk factor distribution pattern among patients who die from COVID-19. In this study, we used latent class analysis (LCA) to identify phenotypes and risk factor distribution patterns in hospitalized patients who died from COVID-19. METHODS: We reviewed the charts of patients who died from COVID-19 at Greenwich Hospital from February 1 to May 30, 2020. We performed LCA based on well-documented prognostic factors of COVID-19. We also compared the in-hospital laboratory results, and treatment information among the different clusters identified by LCA. To validate of the clustering results, we conducted a robust LCA of the entire COVID-19 cohort. RESULTS: 483 patients who were admitted for COVID-19 infection from February 1 to May 30, 2020, 81 patients died. Using latent cluster analysis, we identified two risk factor clusters among COVID-19 death: C1 (n=58) and C2 (n=23). In C1, patients were older (p<0.001) with a higher proportion of comorbidities such as hypertension (82.8% vs. 39.1%, p<0.001), CAD (43.1% vs. 0%, p<0.001), CHF (22.4% vs. 0%, p=0.015), and pre-existing respiratory disease (32.8% vs. 0%, p=0.004) than in C2. In C2, patients were significantly younger and were more likely to be obese (BMI ≥30 kg/m2;56.5% vs. 24.1%, p=0.012), male (87.0% vs. 58.6%, p=0.018), and non-white (60.9% vs. 15.8%, p<0.001) than in C1. Compared with patients in C1, patients in C2 showed a pattern of increased expression of inflammatory and hypercoagulable markers, including C-reactive protein (84.4±117 vs. 12.7±10.7 mg/L, p=0.008) and D-dimer (17.8±13.6 vs. 8.0±9.9 mg/L, p=0.004). The robust-test by using the entire cohort of patients hospitalized with COVID-19 had identified two clusters of risk factor patterns similar to those identified in the cohort of patients who died from COVID-19. CONCLUSIONS: Our study suggests that there are two patterns of risk factors that contributed to death in patients with COVID-19. These results indicate that different pathophysiologic processes lead to COVID-19 death and may be useful in identifying treatment targets and selecting patients with severe COVID-19 disease for future clinical trials. CLINICAL IMPLICATIONS: The results indicate that each phenotype of patients who died from COVID-19 has its distinct feature and underlying pathophysiology. Research focused on targeted therapy for each phenotype may help decrease the mortality rate among patients with severe COVID-19. DISCLOSURES: No relevant relationships by Deepa Jansen, source=Web Response No relevant relationships by Pengyang Li, source=Web Response No relevant relationships by Catherine Teng, source=Web Response