Multi-objective Symbolic Regression to Generate Data-driven, Non-fixed Structure and Intelligible Mortality Predictors using EHR: Binary Classification Methodology and Comparison with State-of-the-art

Symbolic Regression (SR) is a data-driven methodology based on Genetic Programming, and it is widely used to produce arithmetic expressions for modelling learning tasks. Compared to other popular statistical techniques, SR outcomes are given by an arbitrary set of mathematical operations, representing arbitrarily complex linear and non-linear functions without a predefined fixed structure. Another advantage is that, unlike other machine learning algorithms, SR produces interpretable results. In this paper, we explore the qualities and limitations of this technique in a novel implementation as a binary classifier for in-hospital or short-term mortality prediction in patients with Covid-19. Our results highlight that SR provides a competitive alternative to popular statistical and machine learning methodologies to model relevant clinical phenomena thanks to good classification performance, stability in unbalanced dataset management, and intrinsic interpretability.

Changing epidemiology of acute kidney injury in critically ill patients with COVID-19: a prospective cohort.

BACKGROUND: Acute kidney injury (AKI) is common in critically ill patients with coronavirus disease-19 (COVID-19). We aimed to explore the changes in AKI epidemiology between the first and the second COVID wave in the United Kingdom (UK). METHODS: This was an observational study of critically ill adult patients with COVID-19 in an expanded tertiary care intensive care unit (ICU) in London, UK. Baseline characteristics, organ support, COVID-19 treatments, and patient and kidney outcomes up to 90 days after discharge from hospital were compared. RESULTS: A total of 772 patients were included in the final analysis (68% male, mean age 56 ± 13.6). Compared with wave 1, patients in wave 2 were older, had higher body mass index and clinical frailty score, but lower baseline serum creatinine and C-reactive protein (CRP). The proportion of patients receiving invasive mechanical ventilation (MV) on ICU admission was lower in wave 2 (61% vs 80%; p < 0.001). AKI incidence within 14 days of ICU admission was 76% in wave 1 and 51% in wave 2 (p < 0.001); in wave 1, 32% received KRT compared with 13% in wave 2 (p < 0.001). Patients in wave 2 had significantly lower daily cumulative fluid balance (FB) than in wave 1. Fewer patients were dialysis dependent at 90 days in wave 2 (1% vs. 4%; p < 0.001). CONCLUSIONS: In critically ill adult patients admitted to ICU with COVID-19, the risk of AKI and receipt of KRT significantly declined in the second wave. The trend was associated with less MV, lower PEEP and lower cumulative FB. TRIAL REGISTRATION: NCT04445259.

Ethnic inequalities in the impact of COVID-19 on primary care consultations: a time series analysis of 460,084 individuals with multimorbidity in South London.

BACKGROUND: The COVID-19 pandemic caused rapid changes in primary care delivery in the UK, with concerns that certain groups of the population may have faced increased barriers to access. This study assesses the impact of the response to the COVID-19 pandemic on primary care consultations for individuals with multimorbidity and identifies ethnic inequalities. METHODS: A longitudinal study based on monthly data from primary care health records of 460,084 patients aged ≥18 years from 41 GP practices in South London, from February 2018 to March 2021. Descriptive analysis and interrupted time series (ITS) models were used to analyse the effect of the pandemic on primary care consultations for people with multimorbidity and to identify if the effect varied by ethnic groups and consultation type. RESULTS: Individuals with multimorbidity experienced a smaller initial fall in trend at the start of the pandemic. Their primary care consultation rates remained stable (879 (95% CI 869-890) per 1000 patients in February to 882 (870-894) March 2020), compared with a 7% decline among people without multimorbidity (223 consultations (95% CI 221-226) to 208 (205-210)). The gap in consultations between the two groups reduced after July 2020. The effect among individuals with multimorbidity varied by ethnic group. Ethnic minority groups experienced a slightly larger fall at the start of the pandemic. Individuals of Black, Asian, and Other ethnic backgrounds also switched from face-to-face to telephone at a higher rate than other ethnic groups. The largest fall in face-to-face consultations was observed among people from Asian backgrounds (their consultation rates declined from 676 (659-693) in February to 348 (338-359) in April 2020), which may have disproportionately affected their quality of care. CONCLUSIONS: The COVID-19 pandemic significantly affected primary care utilisation in patients with multimorbidity. While there is evidence of a successful needs-based prioritisation of multimorbidity patients within primary care at the start of the pandemic, inequalities among ethnic minority groups were found. Strengthening disease management for these groups may be necessary to control widening inequalities in future health outcomes.

Real-world effectiveness of steroids in severe COVID-19: a retrospective cohort study.

INTRODUCTION: Randomised controlled trials have shown that steroids reduce the risk of dying in patients with severe Coronavirus disease 2019 (COVID-19), whilst many real-world studies have failed to replicate this result. We aim to investigate real-world effectiveness of steroids in severe COVID-19. METHODS: Clinical, demographic, and viral genome data extracted from electronic patient record (EPR) was analysed from all SARS-CoV-2 RNA positive patients admitted with severe COVID-19, defined by hypoxia at presentation, between March 13th 2020 and May 27th 2021. Steroid treatment was measured by the number of prescription-days with dexamethasone, hydrocortisone, prednisolone or methylprednisolone. The association between steroid > 3 days treatment and disease outcome was explored using multivariable cox proportional hazards models with adjustment for confounders (including age, gender, ethnicity, co-morbidities and SARS-CoV-2 variant). The outcome was in-hospital mortality. RESULTS: 1100 severe COVID-19 cases were identified having crude hospital mortality of 15.3%. 793/1100 (72.1%) individuals were treated with steroids and 513/1100 (46.6%) received steroid ≤ 3 days. From the multivariate model, steroid > 3 days was associated with decreased hazard of in-hospital mortality (HR: 0.47 (95% CI: 0.31-0.72)). CONCLUSION: The protective effect of steroid treatment for severe COVID-19 reported in randomised clinical trials was replicated in this retrospective study of a large real-world cohort.

Descriptive comparison of admission characteristics between pandemic waves and multivariable analysis of the association of the Alpha variant (B.1.1.7 lineage) of SARS-CoV-2 with disease severity in inner London.

BACKGROUND: The Alpha variant (B.1.1.7 lineage) of SARS-CoV-2 emerged and became the dominant circulating variant in the UK in late 2020. Current literature is unclear on whether the Alpha variant is associated with increased severity. We linked clinical data with viral genome sequence data to compare admitted cases between SARS-CoV-2 waves in London and to investigate the association between the Alpha variant and the severity of disease. METHODS: Clinical, demographic, laboratory and viral sequence data from electronic health record systems were collected for all cases with a positive SARS-CoV-2 RNA test between 13 March 2020 and 17 February 2021 in a multisite London healthcare institution. Multivariate analysis using logistic regression assessed risk factors for severity as defined by hypoxia at admission. RESULTS: There were 5810 SARS-CoV-2 RNA-positive cases of which 2341 were admitted (838 in wave 1 and 1503 in wave 2). Both waves had a temporally aligned rise in nosocomial cases (96 in wave 1 and 137 in wave 2). The Alpha variant was first identified on 15 November 2020 and increased rapidly to comprise 400/472 (85%) of sequenced isolates from admitted cases in wave 2. A multivariate analysis identified risk factors for severity on admission, such as age (OR 1.02, 95% CI 1.01 to 1.03, for every year older; p<0.001), obesity (OR 1.70, 95% CI 1.28 to 2.26; p<0.001) and infection with the Alpha variant (OR 1.68, 95% CI 1.26 to 2.24; p<0.001). CONCLUSIONS: Our analysis is the first in hospitalised cohorts to show increased severity of disease associated with the Alpha variant. The number of nosocomial cases was similar in both waves despite the introduction of many infection control interventions before wave 2.

Procalcitonin Has Good Accuracy for Prognosis of Critical Condition and Mortality in COVID-19: A Diagnostic Test Accuracy Systematic Review and Meta-analysis.

Several reports have determined that changes in white blood cell counts and inflammatory biomarkers are related to disease outcome of coronavirus disease 2019 (COVID-19) and they can be utilized as prognostic biomarkers. For introducing a factor as a diagnostic/prognostic biomarker, diagnostic test accuracy (DTA) systematic review and meta-analysis are recommended. For the first time, we aimed to determine the accuracies of white blood cell counts and inflammatory biomarkers for prognosis of COVID-19 patient's outcome by a DTA meta-analysis. Until August24, 2020, we searched Web of Sciences, Scopus, and MEDLINE/PubMed databases to achieve related papers. Summary points and lines of included studies were calculated from 2×2 tables by bivariate/hierarchical models. Critical condition and mortality were considered as outcomes. A total of 13387 patients from 28 studies were included in this study. Six biomarkers containing leukocytosis, neutrophilia, lymphopenia, increased level of C-reactive protein, procalcitonin (PCT), and ferritin met the inclusion criteria. Analysis of the area under the curve (AUCHSROC) indicated that the PCT was the only applicable prognostic biomarker for critical condition and mortality (AUCHSROC=0.80 for both conditions). Pooled-diagnostic odds ratios were 6.78 (95% CI, 3.65-12.61) for prognosis of critical condition and 13.21 (95% CI, 3.95-44.19) for mortality. Other biomarkers had insufficient accuracies for both conditions (AUCHSROC< 0.80). Among evaluated biomarkers, only PCT has good accuracy for the prognosis of both critical condition and mortality in COVID-19 and it can be considered as a single prognostic biomarker for poor outcomes. Also, PCT has more accuracy for the prognosis of mortality in comparison to critical condition.