Study design and approval

This was a retrospective cohort study of all unique, consecutive adult patients admitted to University of Utah Health with uncomplicated Streptococcal BSI. Patients were identified for chart review from the University of Utah Health Enterprise Data Warehouse with admissions from January 1, 2018 to July 15, 2023 with at least one positive blood culture with a Streptococcal species. Inclusion was limited to the first eligible encounter for patients with more than one encounter meeting criteria for inclusion during the study period. The first positive blood culture for the encounter was defined as the index blood culture. Patients were included if they were at least 18 years of age, diagnosed with uncomplicated Streptococcal BSI, and received in vitro active antibiotics for at least 72 hours from the calendar date of the index blood culture. Patients were excluded if they received antibiotics for more than 21 consecutive days to assist in excluding those with complicated BSI, were diagnosed with co-infections with non-Streptococcal organisms requiring antibiotic treatment, had polymicrobial BSI, or died or were transitioned to hospice during antibiotic therapy. Uncomplicated Streptococcal BSI was defined as negative follow-up blood cultures (if collection deemed necessary by the treating clinician), clinical resolution by day 3 (defined as no fever [<38°C],] hemodynamic stability, and a trend towards a normal WBC within 72 hours of starting active antibiotic therapy), no evidence of metastatic foci, source control (if indicated), and absence of complicated infection, including endocarditis, bone and joint infections, and/or central nervous system infections. Source control was defined as the removal of any infected hardware, catheters, or devices and drainage procedure was performed of infected fluid collections, as well as imaging assurance, as needed, of no residual or metastatic sites of infection. The inclusion and exclusion criteria were confirmed by manual chart review. Patients were assigned to either the short treatment group or the long treatment group based on the number of consecutive days they received in vitro active antibiotics (including inpatient and discharge antibiotics) for 10 days or fewer (short duration) or more than 10 days (long duration).

Patient consent was not required for this retrospective chart review study. The University of Utah Institutional Review reviewed this study and deemed exempt from oversight.

Outcomes

The primary outcome was a composite of incidence of recurrent BSI, all-cause mortality, and unplanned readmissions at 30 days from the end of antibiotic treatment (EOT). Recurrent BSI was defined as at least one positive blood culture with the same Streptococcal species collected between 72 hours after the index blood culture and the end of the 30 or 90 day follow up period. Secondary outcomes included hospital length of stay, 90-day all-cause mortality from EOT, 90-day recurrent BSI from EOT, 90-day unplanned readmissions from EOT.

Data collection

Data were collected from the Enterprise Data Warehouse and by chart review by study personnel (JG) using REDCap®. Data gathered from the Enterprise Data Warehouse included patient demographics (age at admission, gender, and self-reported ethnic group), hospital length of stay, intensive care unit (ICU) admission status, receipt of Infectious Diseases (ID) consultation, receipt of echocardiogram, positive blood culture with Streptococcal species, index vitals and labs, date of death, and date of readmission. Data gathered by manual chart review included chronic conditions present at the time of hospital admission, presence of indwelling lines, cardiac prosthetics or devices or other prosthetics, immunosuppression status (Table S1), history of intravenous drug use (IVDU), vitals and labs on day 3 from index blood culture, index absolute neutrophil count (ANC), repeat blood cultures, infection source, source control procedures, inpatient and outpatient antibiotics, duration of antibiotic therapy, and date of repeat BSI. The source of infection was determined from ID consultation documentation, if available, or primary team providers’ discharge documentation.

Statistical analysis

To account for potential confounders in the decision to treat Streptococcal BSI with a short versus long antibiotic duration, we used inverse probability of treatment weighting (IPTW) to estimate the average treatment effects (ATE) of in vitro active antibiotics administered for 10 days or fewer versus more than 10 days, adjusting for pretreatment variables. Potential confounders were identified a priori and include source of infection, index ANC, ID consultation, chronic conditions, cardiac and other prosthetic devices, history of IVDU, index blood pressure and respiratory rate, receipt of immunosuppressive medication, age and admission to ICU (Figure S2).

We first assessed balance in these variables between the treatment groups using absolute standardized mean differences (ASMDs). ^{Reference Austin16,Reference Chesnaye, Stel, Tripepi, Dekker, Fu, Zoccali and Jager17} Variables with the highest ASMDs (>0.2 or close to 0.2) were included in a multivariable logistic regression model to predict treatment assignment. The predicted probabilities (ie propensity scores) were then used to construct the IPT weights which were truncated at 99^th percentile to minimize the impact of individuals with disproportionately large weights. We then reassessed ASMDs of all variables between the weighted groups. If there were still variables with large ASMDs, we revised the propensity score model by including additional variables and removing correlated variables then recalculated propensity scores and IPT weights. This iterative process continued until all variables were adequately balanced between treatment groups, indicated by ideally, ASMDs smaller than or close to 0.1. ^{Reference Garrido, Kelley, Paris, Roza, Meier, Morrison and Aldridge18} We then examined the distributions of calculated propensity scores to ensure that the ranges of scores mostly overlapped between treatment groups (Figure 1). ^{Reference Garrido, Kelley, Paris, Roza, Meier, Morrison and Aldridge18,Reference Austin and Stuart19} Additionally, we performed a separate sensitivity analysis comparing outcomes from index blood culture rather than from EOT.

Figure 1.

Absolute standardized mean difference.

With subjects weighted by IPTW, we estimated the ATE of short antibiotic duration using logistic regression for binary outcomes and gamma regression (a generalized linear model assuming a gamma distribution of the response variable given the predictors) for length of stay (LOS) because of its positive nature and right-skewed distribution. Odds ratios (or ratio for the LOS outcome) were reported with 95% confidence intervals (CIs) and P-values. Predicted outcomes by treatment groups were also reported. All analyses were conducted in R v4.3 and all tests were two sided. Total counts and percentages, median and interquartile ranges, and means and standard deviations were used to summarize data as appropriate.