Potential and limitations of machine meta-learning (ensemble) methods for predicting COVID-19 mortality in a large inhospital Brazilian dataset.

The majority of early prediction scores and methods to predict COVID-19 mortality are bound by methodological flaws and technological limitations (e.g., the use of a single prediction model). Our aim is to provide a thorough comparative study that tackles those methodological issues, considering multiple techniques to build mortality prediction models, including modern machine learning (neural) algorithms and traditional statistical techniques, as well as meta-learning (ensemble) approaches. This study used a dataset from a multicenter cohort of 10,897 adult Brazilian COVID-19 patients, admitted from March/2020 to November/2021, including patients [median age 60 (interquartile range 48-71), 46% women]. We also proposed new original population-based meta-features that have not been devised in the literature. Stacking has shown to achieve the best results reported in the literature for the death prediction task, improving over previous state-of-the-art by more than 46% in Recall for predicting death, with AUROC 0.826 and MacroF1 of 65.4%. The newly proposed meta-features were highly discriminative of death, but fell short in producing large improvements in final prediction performance, demonstrating that we are possibly on the limits of the prediction capabilities that can be achieved with the current set of ML techniques and (meta-)features. Finally, we investigated how the trained models perform on different hospitals, showing that there are indeed large differences in classifier performance between different hospitals, further making the case that errors are produced by factors that cannot be modeled with the current predictors.

Myocardial Injury and Prognosis in Hospitalized COVID-19 Patients in Brazil: Results From The Brazilian COVID-19 Registry. / Lesão Miocárdica e Prognóstico em Pacientes Hospitalizados com COVID-19 no Brasil: Resultados do Registro Nacional de COVID-19.

BACKGROUND: Cardiovascular complications of COVID-19 are important aspects of the disease's pathogenesis and prognosis. Evidence on the prognostic role of troponin and myocardial injury in Latin American hospitalized COVID-19 patients is still scarce. OBJECTIVES: To evaluate myocardial injury as independent predictor of in-hospital mortality and invasive mechanical ventilation support in hospitalized patients, from the Brazilian COVID-19 Registry. METHODS: This cohort study is a substudy of the Brazilian COVID-19 Registry, conducted in 31 Brazilian hospitals of 17 cities, March-September 2020. Primary outcomes included in-hospital mortality and invasive mechanical ventilation support. Models for the primary outcomes were estimated by Poisson regression with robust variance, with statistical significance of p<0.05. RESULTS: Of 2,925 patients (median age of 60 years [48-71], 57.1% men), 27.3% presented myocardial injury. The proportion of patients with comorbidities was higher among patients with cardiac injury (median 2 [1-2] vs. 1 [0-2]). Patients with myocardial injury had higher median levels of brain natriuretic peptide, lactate dehydrogenase, creatine phosphokinase, N-terminal pro-brain natriuretic peptide, and C-reactive protein than patients without myocardial injury. As independent predictors, C-reactive protein and platelet counts were related to the risk of death, and neutrophils and platelet counts were related to the risk of invasive mechanical ventilation support. Patients with high troponin levels presented a higher risk of death (RR 2.03, 95% CI 1.60-2.58) and invasive mechanical ventilation support (RR 1.87, 95% CI 1.57-2.23), when compared to those with normal troponin levels. CONCLUSION: Cardiac injury was an independent predictor of in-hospital mortality and the need for invasive mechanical ventilation support in hospitalized COVID-19 patients.

FUNDAMENTO: As complicações cardiovasculares da COVID-19 são aspectos importantes da patogênese e do prognóstico da doença. Evidências do papel prognóstico da troponina e da lesão miocárdica em pacientes hospitalizados com COVID-19 na América Latina são ainda escassos. OBJETIVOS: Avaliar a lesão miocárdica como preditor independente de mortalidade hospitalar e suporte ventilatório mecânico em pacientes hospitalizados, do registro brasileiro de COVID-19. MÉTODOS: Este estudo coorte é um subestudo do registro brasileiro de COVID-19, conduzido em 31 hospitais brasileiros de 17 cidades, de março a setembro de 2020. Os desfechos primários incluíram mortalidade hospitalar e suporte ventilatório mecânico invasivo. Os modelos para os desfechos primários foram estimados por regressão de Poisson com variância robusta, com significância estatística de p<0,05. RESULTADOS: Dos 2925 pacientes [idade mediana de 60 anos (48-71), 57,1%], 27,3% apresentaram lesão miocárdica. A proporção de pacientes com comorbidades foi maior nos pacientes com lesão miocárdica [mediana 2 (1-2) vs. 1 (0-20)]. Os pacientes com lesão miocárdica apresentaram maiores valores medianos de peptídeo natriurético cerebral, lactato desidrogenase, creatina fosfoquinase, N-terminal do pró-peptídeo natriurético tipo B e proteína C reativa em comparação a pacientes sem lesão miocárdica. Como fatores independentes, proteína C reativa e contagem de plaquetas foram relacionados com o risco de morte, e neutrófilos e contagem de plaquetas foram relacionados ao risco de suporte ventilatório mecânico invasivo. Os pacientes com níveis elevados de troponina apresentaram um maior risco de morte (RR 2,03, IC95% 1,60-2,58) e suporte ventilatório mecânico (RR 1,87;IC95% 1,57-2,23), em comparação àqueles com níveis de troponina normais. CONCLUSÃO: Lesão cardíaca foi um preditor independente de mortalidade hospitalar e necessidade de suporte ventilatório mecânico em pacientes hospitalizados com COVID-19.

Mechanical ventilation and death in pregnant patients admitted for COVID-19: a prognostic analysis from the Brazilian COVID-19 registry score.

BACKGROUND: The assessment of clinical prognosis of pregnant COVID-19 patients at hospital presentation is challenging, due to physiological adaptations during pregnancy. Our aim was to assess the performance of the ABC2-SPH score to predict in-hospital mortality and mechanical ventilation support in pregnant patients with COVID-19, to assess the frequency of adverse pregnancy outcomes, and characteristics of pregnant women who died. METHODS: This multicenter cohort included consecutive pregnant patients with COVID-19 admitted to the participating hospitals, from April/2020 to March/2022. Primary outcomes were in-hospital mortality and the composite outcome of mechanical ventilation support and in-hospital mortality. Secondary endpoints were pregnancy outcomes. The overall discrimination of the model was presented as the area under the receiver operating characteristic curve (AUROC). Overall performance was assessed using the Brier score. RESULTS: From 350 pregnant patients (median age 30 [interquartile range (25.2, 35.0)] years-old]), 11.1% had hypertensive disorders, 19.7% required mechanical ventilation support and 6.0% died. The AUROC for in-hospital mortality and for the composite outcome were 0.809 (95% IC: 0.641-0.944) and 0.704 (95% IC: 0.617-0.792), respectively, with good overall performance (Brier = 0.0384 and 0.1610, respectively). Calibration was good for the prediction of in-hospital mortality, but poor for the composite outcome. Women who died had a median age 4 years-old higher, higher frequency of hypertensive disorders (38.1% vs. 9.4%, p < 0.001) and obesity (28.6% vs. 10.6%, p = 0.025) than those who were discharged alive, and their newborns had lower birth weight (2000 vs. 2813, p = 0.001) and five-minute Apgar score (3.0 vs. 8.0, p < 0.001). CONCLUSIONS: The ABC2-SPH score had good overall performance for in-hospital mortality and the composite outcome mechanical ventilation and in-hospital mortality. Calibration was good for the prediction of in-hospital mortality, but it was poor for the composite outcome. Therefore, the score may be useful to predict in-hospital mortality in pregnant patients with COVID-19, in addition to clinical judgment. Newborns from women who died had lower birth weight and Apgar score than those who were discharged alive.

Frequency and burden of neurological manifestations upon hospital presentation in COVID-19 patients: Findings from a large Brazilian cohort.

BACKGROUND: Scientific data regarding the prevalence of COVID-19 neurological manifestations and prognosis in Latin America countries is still lacking. Therefore, the study aims to understand neurological manifestations of SARS-CoV 2 infection and outcomes in the Brazilian population. METHODS: This study is part of the Brazilian COVID-19 Registry, a multicentric cohort, including data from 37 hospitals. For the present analysis, patients were grouped according to the presence of reported symptoms (i.e., headache; anosmia and ageusia; syncope and dizziness) vs. clinically-diagnosed neurological manifestations (clinically-defined neurological syndrome: neurological signs or diagnoses captured by clinical evaluation) and matched with patients without neurological manifestations by age, sex, number of comorbidities, hospital of admission, and whether or not patients had underlying neurological disease. RESULTS: From 6,635 hospitalized patients with COVID-19, 30.8% presented reported neurological manifestations, 10.3% were diagnosed with a neurological syndrome and 60.1% did not show any neurological manifestations. In patients with reported symptoms, the most common ones were headache (20.7%), ageusia (11.1%) and anosmia (8.0%). In patients with neurological syndromes, acute encephalopathy was the most common diagnosis (9.7%). In the matched analysis, patients with neurological syndromes presented more cases of septic shock (17.0 vs. 13.0%, p = 0.045), intensive care unit admission (45.3 vs. 38.9%, p = 0.023), and mortality (38.7 vs. 32.6%, p = 0.026; and 39.2 vs. 30.3%, p < 0.001) when compared to controls. CONCLUSION: COVID-19 in-hospital patients with clinically defined neurological syndromes presented a higher incidence of septic shock, ICU admission and death when compared to controls.

Development and validation of the MMCD score to predict kidney replacement therapy in COVID-19 patients.

BACKGROUND: Acute kidney injury (AKI) is frequently associated with COVID-19, and the need for kidney replacement therapy (KRT) is considered an indicator of disease severity. This study aimed to develop a prognostic score for predicting the need for KRT in hospitalised COVID-19 patients, and to assess the incidence of AKI and KRT requirement. METHODS: This study is part of a multicentre cohort, the Brazilian COVID-19 Registry. A total of 5212 adult COVID-19 patients were included between March/2020 and September/2020. Variable selection was performed using generalised additive models (GAM), and least absolute shrinkage and selection operator (LASSO) regression was used for score derivation. Accuracy was assessed using the area under the receiver operating characteristic curve (AUC-ROC). RESULTS: The median age of the model-derivation cohort was 59 (IQR 47-70) years, 54.5% were men, 34.3% required ICU admission, 20.9% evolved with AKI, 9.3% required KRT, and 15.1% died during hospitalisation. The temporal validation cohort had similar age, sex, ICU admission, AKI, required KRT distribution and in-hospital mortality. The geographic validation cohort had similar age and sex; however, this cohort had higher rates of ICU admission, AKI, need for KRT and in-hospital mortality. Four predictors of the need for KRT were identified using GAM: need for mechanical ventilation, male sex, higher creatinine at hospital presentation and diabetes. The MMCD score had excellent discrimination in derivation (AUROC 0.929, 95% CI 0.918-0.939) and validation (temporal AUROC 0.927, 95% CI 0.911-0.941; geographic AUROC 0.819, 95% CI 0.792-0.845) cohorts and good overall performance (Brier score: 0.057, 0.056 and 0.122, respectively). The score is implemented in a freely available online risk calculator ( https://www.mmcdscore.com/ ). CONCLUSIONS: The use of the MMCD score to predict the need for KRT may assist healthcare workers in identifying hospitalised COVID-19 patients who may require more intensive monitoring, and can be useful for resource allocation.