Bloodstream Infection Risk, Incidence, and Deaths for Hospitalized Patients During Coronavirus Disease Pandemic

Bhavarth S. Shukla; Prem R. Warde; Eric Knott; Sebastian Arenas; Darryl Pronty; Reinaldo Ramirez; Arely Rego; Miriam Levy; Martin Zak; Dipen J. Parekh; Tanira Ferreira; Hayley B. Gershengorn


Emerging Infectious Diseases. 2021;27(10):2588-2594. 

In This Article


Study Design and Cohort

We conducted a retrospective cohort study of adult hospitalizations during March 25–October 27, 2020, at an academic, tertiary, acute-care facility in Miami, Florida, USA, which lacks capacity to give care with extracorporeal membrane oxygenation. Patients were included in the cohort if they had ≥1 reverse transcription PCR completed; patients could be included more than once if they were admitted to the hospital more than once over the study period. During the study period, all patients were screened by reverse transcription PCR before hospital admission. Although there were no specific exclusion criteria, the facility does not offer pediatric or obstetric services, so pregnant woman and patients <18 years of age were not included. A restricted cohort of patients that had central venous catheters at any point during their hospital stay was also considered.

Exposure and Outcomes

Our exposure of interest was COVID-19 positivity (determined by SARS-CoV-2 PCR testing) during the hospital stay. Patients who had ≥1 positive test result (from 7 days before admission up until discharge with or without preceding negative test results) were considered positive for COVID-19. Our primary outcome was LCBI. Secondary outcomes were death, time to LCBI (time from hospital admission to first positive blood culture per patient admission), and development of central line–associated bloodstream infection (CLABSI) evaluated by using the restricted cohort of patients who had a central venous catheter. We defined LCBIs and CLABSIs according to National Healthcare Safety Network (NHSN) 2020 criteria.[9] In brief, LCBI is defined in these criteria as a single positive blood culture or molecular test result for a pathogen or 2 positive blood cultures for a commensal organism. CLABSI is defined as an LCBI associated with a central venous catheter in place for ≥2 calendar days.[9]

Data Sources and Variables

We obtained information for each patient from the hospital system's electronic medical record by using EPIC software ( In addition to COVID-19 infection status and outcomes (including organism identification), we abstracted information on demographics (age, sex, race, primary insurance provider), organisms isolated from blood cultures, chronic health conditions, Elixhauser comorbidity conditions,[10] body mass index, severity of acute illness (sequential organ failure assessment score [SOFA] during hospital day 1),[11] renal replacement therapy (either intermittent or continuous), invasive mechanical ventilation, care in the ICU, prone positioning (including persons using mechanical ventilation), central venous catheters, urinary catheters, systemic corticosteroids, tocilizumab, and remdesivir. For each of the resources other than prone positioning, we identified whether they were used and the total duration of use. For prone positioning, we were able to identify only use (not duration or timing of use). We also obtained data on hospital length of stay.

Statistical Analysis and Ethics

We described the cohort by using standard summary statistics. We compared characteristics by outcome by using χ 2 and Kruskal-Wallis tests as appropriate. Our primary analysis was a risk for LCBI assessment by using multivariable logistic regression modeling with an exposure of COVID-19 status. We included all data elements except prone positioning, remdesivir, and tocilizumab as covariables, and resource elements were modeled as receipt/nonreceipt before development of LCBI (or hospital discharge if no LCBI). To ensure our results would not be confounded by deaths in hospitals, we recreated the same models for hospital survivors and decedents separately.

To consider secondary outcomes, we first used multivariable Cox proportional hazards modeling with censoring at hospital discharge and a competing risk for death to assess the association of COVID-19 positivity and time to LCBI. We then constructed a multivariable logistic regression model to assess the association of COVID-19 positivity with risk for death by hospital discharge. For this model, we included days of resource use as covariates.

We also constructed 3 models to evaluate for LCBI that developed ≥2 calendar days from admission, indicated as LCBI HAI. Next, we reconstructed the models (for our primary and 2 secondary outcomes) by using the restricted cohort of patients who had used central venous catheters to assess risk for and time to CLABSI and death. Finally, to identify risk factors for infection among COVID-19 patients, we constructed 3 multivariable logistic regression models: for LCBI among all COVID-19 patients, for LCBI HAI among all COVID-19 patients, and for CLABSI among COVID-19 patients who had central venous catheters.

Because of a large number of patients who had missing data regarding calculation of SOFA, we imputed this score for each model (using multivariable regression modeling, including all covariates and outcome). We conducted 2 sensitivity analyses to assess the robustness of our primary analysis: only patients with an available SOFA score, and all patients but not including SOFA as a model covariate.

Study approval was obtained from the University of Miami Institutional Review Board (#20200739). We performed all analyses by using the programming language R (R Foundation for Statistical Computing, Results were considered significant if p<0.05. Because we did not adjust for multiple comparisons, we considered all nonprimary analyses to be hypothesis generating.