COVID-PA Bulletin: reports on artificial intelligence-based forecasting in coping with COVID-19 pandemic in the state of Pará, Brazil. / Boletim COVID-PA: relatos sobre projeções baseadas em inteligência artificial no enfrentamento da pandemia de COVID-19 no estado do Pará.

OBJECTIVE: To report the university extension research result entitled 'The COVID-PA Bulletin', which presented forecasts on the behavior of the pandemic in the state of Pará, Brazil. METHODS: The artificial intelligence technique also known as 'artificial neural networks' was used to generate 13 bulletins with short-term forecasts based on historical data from the State Department of Public Health information system. RESULTS: After eight months of predictions, the technique generated reliable results, with an average accuracy of 97% (observed for147 days) for confirmed cases, 96% (observed for 161 days) for deaths and 86% (observed for 72 days) for Intensive Care Unit bed occupancy. CONCLUSION: These bulletins have become a useful decision-making tool for public managers, assisting in the reallocation of hospital resources and optimization of COVID-19 control strategies in various regions of the state of Pará.

Bacterial infections and death among patients with Covid-19 versus non Covid-19 patients with pneumonia.

BACKGROUND: Many patients with Coronavirus disease-2019 (Covid-19) present with radiological evidence of pneumonia. Because it is difficult to determine co-existence of bacterial pneumonia, many of these patients are initially treated with antibiotics. We compared the rates of bacterial infections and mortality in Covid-19 patients with pulmonary infiltrates versus patients diagnosed with 'pneumonia' the year previously. METHODS: We conducted a medical record review of patients admitted with Covid-19 and a pulmonary infiltrate and compared them with patients diagnosed with pneumonia admitted in the prior year before the pandemic. Data abstracted included baseline demographics, comorbidities, signs and symptoms, laboratory and microbiological results, and imaging findings. Outcomes were bacterial infections and mortality. Patients presenting with and without Covid-19 were compared using univariable and multivariable analyses. RESULTS: There were 1398 and 1001 patients admitted through the emergency department (ED) with and without Covid-19 respectively. Compared with non-Covid-19 patients, those with Covid-19 were younger (61±18 vs. 65±25 years, P < 0.001) and had a lower Charlson Comorbidity Index (0.7 vs. 1.2, P < 0.001). Bacterial infections were present in fewer Covid-19 than non-Covid-19 patients (8% vs. 13%, P < 0.001), and most infections in Covid-19 were nosocomial as opposed to community acquired in non-Covid-19 patients. CXR was more often read as abnormal and with bilateral infiltrates in patients with Covid-19 (82% vs. 70%, P < 0.001 and 81% vs. 48%, P < 0.001, respectively). Mortality was higher in patients with Covid-19 vs. those without (15% vs. 9%, P < 0.001). Multivariable predictors (OR [95%CI]) of mortality were age (1.04 [1.03-1.05]/year), tachypnea (1.55 [1.12-2.14]), hypoxemia (2.98 [2.04-4.34]), and bacterial infection (2.80 [1.95-4.02]). Compared with non-Covid-19 patients with pneumonia, patients with Covid-19 were more likely to die (2.68 [1.97-3.63]). CONCLUSIONS: The rate of bacterial infections is lower in Covid-19 patients with pulmonary infiltrates compared with patients diagnosed with pneumonia prior to the pandemic and most are nosocomial. Mortality was higher in Covid-19 than non-Covid-19 patients even after adjusting for age, tachypnea, hypoxemia, and bacterial infection.

Artificial neural networks for short-term forecasting of cases, deaths, and hospital beds occupancy in the COVID-19 pandemic at the Brazilian Amazon.

The first case of the novel coronavirus in Brazil was notified on February 26, 2020. After 21 days, the first case was reported in the second largest State of the Brazilian Amazon. The State of Pará presented difficulties in combating the pandemic, ranging from underreporting and a low number of tests to a large territorial distance between cities with installed hospital capacity. Due to these factors, mathematical data-driven short-term forecasting models can be a promising initiative to assist government officials in more agile and reliable actions. This study presents an approach based on artificial neural networks for the daily and cumulative forecasts of cases and deaths caused by COVID-19, and the forecast of demand for hospital beds. Six scenarios with different periods were used to identify the quality of the generated forecasting and the period in which they start to deteriorate. Results indicated that the computational model adapted capably to the training period and was able to make consistent short-term forecasts, especially for the cumulative variables and for demand hospital beds.

Outcomes in Hispanics With COVID-19 Are Similar to Those of Caucasian Patients in Suburban New York.

BACKGROUND: Despite reported higher rates and worse outcomes due to COVID-19 in certain racial and ethnic groups, much remains unknown. We explored the association between Hispanic ethnicity and outcomes in COVID-19 patients in Long Island, New York. METHODS: We conducted a retrospective cohort study of 2,039 Hispanic and non-Hispanic Caucasian patients testing positive for SARS-CoV-2 between March 7 and May 23, 2020, at a large suburban academic tertiary care hospital near New York City. We explored the association of ethnicity with need for intensive care unit (ICU), invasive mechanical ventilation (IMV), and mortality. RESULTS: Of all patients, 1,079 (53%) were non-Hispanic Caucasians and 960 (47%) were Hispanic. Hispanic patients presented in higher numbers than expected for our catchment area. Compared with Caucasians, Hispanics were younger (45 years vs. 59 years), had fewer comorbidities (66% with no comorbidities vs. 40%), were less likely to have commercial insurance (35% vs. 59%), or were less likely to come from a nursing home (2% vs. 10%). In univariate comparisons, Hispanics were less likely to be admitted (37% vs. 59%) or to die (3% vs. 10%). Age, shortness of breath, congestive heart failure (CHF), coronary artery disease (CAD), hypoxemia, and presentation from nursing homes were associated with admission. Male sex and hypoxemia were associated with ICU admission. Male sex, chronic obstructive pulmonary disease, and hypoxemia were associated with IMV. Male sex, CHF, CAD, and hypoxemia were associated with mortality. After other factors were adjusted for, Hispanics were less likely to be admitted (odds ratio = 0.62, 95% confidence interval = 0.52 to 0.92) but Hispanic ethnicity was not associated with ICU admission, IMV, or mortality. CONCLUSIONS: Hispanics presented at higher rates than average for our population but outcomes among Hispanic patients with COVID-19 were similar to those of Caucasian patients.

Comparison of COVID-19 infections among healthcare workers and non-healthcare workers.

OBJECTIVES: Healthcare workers face distinct occupational challenges that affect their personal health, especially during a pandemic. In this study we compare the characteristics and outcomes of Covid-19 patients who are and who are not healthcare workers (HCW). METHODS: We retrospectively analyzed a cohort of 2,842 adult patients with known HCW status and a positive SARS-CoV-2 RT-PCR test presenting to a large academic medical center emergency department (ED) in New York State from March 21 2020 through June 2020. Early in the pandemic we instituted a policy to collect data on patient occupation and exposures to suspected Covid-19. The primary outcome was hospital admission. Secondary outcomes were ICU admission, need for invasive mechanical ventilation (IMV), and mortality. We compared baseline characteristics and outcomes of Covid-19 adult patients based on whether they were or were not HCW using univariable and multivariable analyses. RESULTS: Of 2,842 adult patients (mean age 53+/-19 years, 53% male) 193 (6.8%) were HCWs and 2,649 (93.2%) were not HCWs. Compared with non-HCW, HCWs were younger (43 vs 53 years, P<0.001), more likely female (118/193 [61%] vs 1211/2649 [46%], P<0.001), and more likely to have a known Covid-19 exposure (161/193 [83%] vs 946/2649 [36%], P<0.001), but had fewer comorbidities. On presentation to the ED, HCW also had lower frequencies of tachypnea (12/193 [6%] vs 426/2649 [16%], P<0.01), hypoxemia (15/193 [8%] vs 564/2649 [21%], P<0.01), bilateral opacities on imaging (38/193 [20%] vs 1189/2649 [45%], P<0.001), and lymphocytopenia (6/193 [3%] vs 532/2649 [20%], P<0.01) compared to non-HCWs. Direct discharges home from the ED were more frequent in HCW 154/193 (80%) vs 1275/2649 (48%) p<0.001). Hospital admissions (38/193 [20%] vs 1264/2694 [47%], P<0.001), ICU admissions (7/193 [3%] vs 321/2694 [12%], P<0.001), need for IMV (6/193 [3%] vs 321/2694 [12%], P<0.001) and mortality (2/193 [1%] vs 219/2694 [8%], P<0.01) were lower than among non-HCW. After controlling for age, sex, comorbidities, presenting vital signs and radiographic imaging, HCW were less likely to be admitted (OR 0.6, 95%CI 0.3-0.9) than non HCW. CONCLUSIONS: Compared with non HCW, HCW with Covid-19 were younger, had less severe illness, and were less likely to be admitted.

Cohort of Four Thousand Four Hundred Four Persons Under Investigation for COVID-19 in a New York Hospital and Predictors of ICU Care and Ventilation.

Study objective: Most coronavirus disease 2019 (COVID-19) reports have focused on severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) positive patients. However, at initial presentation, most patients' viral status is unknown. Determination of factors that predict initial and subsequent need for ICU and invasive mechanical ventilation is critical for resource planning and allocation. We describe our experience with 4,404 persons under investigation and explore predictors of ICU care and invasive mechanical ventilation at a New York COVID-19 epicenter. Methods: We conducted a retrospective cohort study of all persons under investigation and presenting to a large academic medical center emergency department (ED) in New York State with symptoms suggestive of COVID-19. The association between patient predictor variables and SARS-CoV-2 status, ICU admission, invasive mechanical ventilation, and mortality was explored with univariate and multivariate analyses. Results: Between March 12 and April 14, 2020, we treated 4,404 persons under investigation for COVID-19 infection, of whom 68% were discharged home, 29% were admitted to a regular floor, and 3% to an ICU. One thousand six hundred fifty-one of 3,369 patients tested have had SARS-CoV-2-positive results to date. Of patients with regular floor admissions, 13% were subsequently upgraded to the ICU after a median of 62 hours (interquartile range 28 to 106 hours). Fifty patients required invasive mechanical ventilation in the ED, 4 required out-of-hospital invasive mechanical ventilation, and another 167 subsequently required invasive mechanical ventilation in a median of 60 hours (interquartile range 26 to 99) hours after admission. Testing positive for SARS-CoV-2 and lower oxygen saturations were associated with need for ICU and invasive mechanical ventilation, and with death. High respiratory rates were associated with the need for ICU care. Conclusion: Persons under investigation for COVID-19 infection contribute significantly to the health care burden beyond those ruling in for SARS-CoV-2. For every 100 admitted persons under investigation, 9 will require ICU stay, invasive mechanical ventilation, or both on arrival and another 12 within 2 to 3 days of hospital admission, especially persons under investigation with lower oxygen saturations and positive SARS-CoV-2 swab results. This information should help hospitals manage the pandemic efficiently.