Background: Vascular access devices (VAD), including central venous catheters (CVC) and peripheral intravenous catheters (PIVC), are among the most common medical intervention for hospitalized patients, yet they are also a major source of infectious and noninfectious complications. The Centers for Disease Control and Prevention (CDC) recommends prompt removal of CVC that are no longer indicated as a cornerstone of CLABSI prevention, although cessation of CVC indication is ill-defined. Our hospital implemented a Device Stewardship program to identify VAD that were likely inappropriate and to nudge providers to remove them promptly. The objective of this study is to quantify the opportunities to decrease VAD utilization and to describe early Device Stewardship efforts. Methods Study design and population. This is a retrospective cohort study at a 500-bed academic safety net hospital in Denver, CO. All inpatients hospitalized between 10/1/2025 and 12/31/2025, were eligible for inclusion. CVC placed for ECMO or rapid rewarming were excluded. Definitions. The following VAD were generally considered unnecessary: Femoral triple lumen catheters (TLC) in place for <1 day TLC present in a patient on an acute care unit Duplicate VAD (peripherally-inserted central catheter [PICC] and PIVC) in a patient on an acute care unit Intervention. Infection preventionists (IPs) identified eligible patients through an Epic worklist. Standardized text messages were used to suggest de-escalation of VAD to nursing and provider teams on weekdays (Figure 1). Results There were 576 TLC device days and 871 PICC device days in hospitalized inpatients during the study period. VAD selection could be improved in 29.6% of these patients (Table 1). Duplicate VAD was identified as the most frequent opportunity for improvement whereas femoral TLC present for <1 day was the least frequent opportunity. IPs contacted the clinical teams of 64 unique patients with duplicate VAD to suggest removal of ≥1 PIVC; 36 PIVC in 25 patients (39%) were removed as a result of intervention. Additionally, 5 femoral TLC present for <1 day and 15 TLC on acute care units were intervened upon with 1 femoral line and 3 TLC (20% each) removed during the study period. Conclusion Despite national guidelines to remove unnecessary VAD, there is ample opportunity to deescalate VAD in clinical practice. Simple criteria can be developed to standardize the selection and to prompt earlier removal of VAD, particularly duplicate access. Work can be done to improve the acceptance rate of suggestions.

Introduction: Acinetobacter infections are commonly viewed as hospital-onset events linked to critical illness and invasive devices. We evaluated whether this paradigm holds true at our institution by examining a decade of Acinetobacter infections at an urban safety-net trauma center. Methods: We conducted a retrospective analysis of patients with Acinetobacter spp. isolated from culture at a 500-bed urban academic safety-net Level I trauma center from January 2015 through December 2025. Clinical and epidemiologic variables were abstracted from the electronic health record. Hospital-onset infection was defined as culture collection on or after hospital day 4 (admission day = day 1). Infections not meeting criteria for hospital onset were classified as community-onset and subsequently assessed for healthcare-associated exposures. Healthcare-associated infection was defined by hospitalization, antibiotic exposure, dialysis, or residence in a skilled nursing facility within 60 days prior to diagnosis. Additional covariates included age, sex, birth place, race/ethnicity, intensive care unit (ICU) admission, active chemotherapy or chronic steroid use, chronic lung disease (e.g., chronic obstructive pulmonary disease or COPD), chronic liver disease (cirrhosis or hepatitis B/C), diabetes, substance use disorder (alcohol or illicit drugs), trauma within 90 days, presence of an invasive airway (endotracheal tube or tracheostomy), indwelling urinary or renal catheters, organism identification (A. baumannii vs non-baumannii species), and final antibiotic used for treatment. To identify factors associated with hospital-onset infection, we used multivariable logistic regression with generalized estimating equations (GEE) to account for within-patient clustering from repeated cultures, reporting adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Results: A total of 157 cultures with growth of Acinetobacter spp. were identified during the study period. Acinetobacter baumannii accounted for 69% of isolates. Most cases occurred among inpatients (79%). Hospital-onset infection accounted for 39%, while 61% were community-onset; among community-onset cases, 56% met criteria for healthcare-associated infection. The cohort was predominantly White (45%) or Hispanic (36%). The most common clinical syndromes were ventilator-associated pneumonia (26%), bacteremia (26%), and urinary tract infection (19%). Treatment was administered in 83% of cases, most commonly with levofloxacin (30%) or cefepime (25%). Carbapenem resistance was uncommon (3%). In multivariable GEE analysis accounting for repeated cultures, hospital-onset infection was independently associated with chronic lung disease, recent trauma, and presence of an invasive airway (Figure 2). Conclusions: In this safety-net health system, most Acinetobacter infections were community-onset rather than hospital-onset. Hospital-onset infection was strongly associated with markers of critical illness and lung injury rather than demographic factors.

Introduction: Clinicians often order a variety of tests for hospitalized patients, including urinalysis with urine cultures. However, fever has limited diagnostic value for urinary tract infection (UTI); pooled estimates indicate that fever has likelihood ratios near 1, offering minimal discriminatory value for ruling UTI in or out. We evaluated the diagnostic yield of inpatient urine cultures by fever status in an acute care safety-net hospital in Denver, Colorado. Methods: We analyzed inpatient urine culture events from April 2016 through December 2025 using an automated electronic medical record report. Fever was captured within the National Healthcare Safety Network (NHSN) infection window period (+/- 3 calendar days). Culture positivity was defined per NHSN UTI criteria as ?10 colony forming units of one or two organisms; cultures with no growth or reported as yeast/Candida spp. or mixed flora were classified as negative. Antibiotic exposure was extracted from medications given up to 3 days after culture collection, and urinary infection-related diagnoses from problem/diagnosis entries up to 7 days after culture collection. To scale adjudication, we used deterministic text classification of medication and diagnosis fields. Medication strings were normalized and parsed with regular expressions to extract generic drug names (token preceding dose/formulation), then matched to a curated antibiotic dictionary. Urinary diagnoses were identified by regular-expression matching to UTI-related terms (e.g., UTI/urinary tract infection, cystitis, pyelonephritis, urosepsis). Results: Across 17,629 inpatient urine cultures, fever within the NHSN infection window period was documented in 37.1% of episodes and UTI diagnoses were documented in 29.6%. Overall, 7,983 cultures (42.1%) were positive for a bacterial pathogen, and antibiotics were prescribed in 77.8% of episodes. Culture positivity was similar to, or higher than, that among afebrile vs febrile episodes (44.1% vs 38.8%) and was higher with documented UTI diagnoses (58.5% vs 35.2% without). Culture positivity varied by strata: UTI diagnoses were more common among culture-positive episodes (60.9% with fever; 57.1% without fever), while febrile episodes without UTI diagnoses had the lowest positivity (29.4%) (Figure 1). Antibiotic exposure remained common across the fever–culture result combinations, including 22.5% of all episodes occurring in afebrile, culture-negative patients who nevertheless received antibiotics and 18.9% in febrile, culture-negative patients who received antibiotics (Figure 2). Conclusion: Fever was not reliably associated with urine culture positivity, whereas UTI diagnosis (clinical suspicion) was. Antibiotic use remained common after negative urine cultures, supporting opportunities to refine inpatient urine culture ordering and antimicrobial stewardship.

Background: During a measles outbreak in the United States in 2025, health care settings were on high alert for patients who may have been exposed to or infected with this disease. When an infection is found, contact tracing investigations are extensive and require significant labor. Identify, Isolate, Inform (III) has become a standard tool for special pathogen preparedness. Our facility implemented an electronic medical record (EMR) alert for patients reporting fever and rash. The objective of this work is to compare and contrast our experience in drill vs clinical practice. Methods: Setting. Denver Health is an acute-care safety net hospital with both adult and pediatric emergency departments, sharing a common security checkpoint for entrance, and a check in and arrival space before patients are triaged into individual waiting rooms. Drill. In March 2025, Denver Health conducted a drill to test the III process in the pediatric emergency department and urgent care center (PEDUC) where care was provided for a mock patient with fever, rash, and cough. Time to masking and time to isolation were collected by a drill observer. Clinical practice. In April 2025, an infant with recent international travel presented to the PEDUC with cough, fever, and rash and was subsequently diagnosed with measles by PCR. Time to isolation was collected via time stamps in the EMR. Results: During the drill, fever and rash were documented, and the alert fired appropriately. The mock patient was masked within six minutes of entry and roomed in a negative pressure room within nine minutes of entry. During clinical practice, rash was documented, but the patient was afebrile at triage, and the EMR alert did not fire. The patient spent 62 minutes in the waiting room before being roomed. Contact tracing involved individuals in both the adult and pediatric emergency departments due to the shared airspaces. In total, 149 patients were identified as contacts and were notified of the exposure by public health. No transmission was identified. Conclusions: Despite a recent successful drill in the PEDUC, when a patient presented with measles, there were challenges with identification and isolation, as well as structural limitations within the facility and gaps with technology aiding the early identification of patients at risk. Having a real-life situation resulted in an extremely large contact investigation and juxtaposed the recent successful drill, highlighting the need to continue performing “no notice drills” and work on improving each facet of preparedness.

Background: Midlines and peripherally-inserted central catheters (PICC) have increased in use due to their ease of insertion, low insertion risk to patients, and relatively low cost. Three common complications are central line-associated bloodstream infection (CLABSI), upper extremity deep venous thrombosis (DVT), and lumen occlusion. The literature reports PICC-associated CLABSI rates of 0.5-2.1 per 1000 catheter days, DVT rates of 1.4 to 9.5%, and lumen occlusion 5 to 16.1%. Both CLABSI and lumen occlusion are more common in multi-lumen than in single lumen (SL) PICC. Our objective is to evaluate the appropriateness of midline and PICC lumen selection, to assess for rates of complication, and to identify targets for clinical decision support intervention. Methods: Setting. This is a retrospective cohort study of adult patients ≥18 y.o. with a midline or PICC placed by the Vascular Access Team (VAT) between 1/1/24 and 6/30/25 at a 500-bed academic safety net and teaching hospital. Data abstraction. Indication for line placement was determined by VAT consultation order or manual chart review. Upper extremity DVT was screened by ICD10 codes and confirmed through manual chart review. CLABSI was determined through cross-reference to National Healthcare Safety Network data. Alteplase administration was used as a surrogate for lumen occlusion. Appropriateness criteria. Appropriateness of midline and PICC lumen selection was determined by adapting the Michigan Appropriateness Guide for Intravenous Catheters and Michigan Multi-Lumen Appropriateness Criteria (Figure 1). Results: Six hundred twelve midline (n=97, all SL) or PICCs (n=515) were placed into 524 patients in the study period (SL PICC, n=213; double lumen [DL] PICC, n=243; triple lumen [TL] PICC, n=59). The indication for line differed based on the line type (Table 1). The majority of midlines (95.9%) and SL PICCs (99.5%) were deemed appropriate; in contrast, 74.1% of DL PICC and 3.4% of TL PICC were considered appropriate. Complications of midline and PICC were rare (CLABSI: 0.19 per 1000 line days; DVT: 1.1%; lumen occlusion: 11.8%). Any complication occurred in 12.9% of cases and was more common in multi-lumen as compared to SL devices (i.e. midline or SL PICC) (19.9% vs 6.1%, P). Conclusion: Opportunities exist to reduce multi-lumen PICC insertion. This may decrease device-related complications. Optimizing lumen selection through clinical decision support may improve patient safety and vascular access outcomes.

We evaluated large language model (LLM)-based agents integrated with the electronic medical record to assess blood culture appropriateness. While sensitivity was high, specificity remained low. Performance was shaped by prompt phrasing, sycophantic behavior, and semantic triggers, reflecting both the potential and limitations of LLMs in real-world clinical decision support.

Introduction: The healthcare sector contributes significantly to global greenhouse gas (GHG) emissions, accounting for 8.5% of the total emissions in the United States alone. Infection control practices designed to prevent disease transmission contribute to this substantial carbon footprint. These practices, which include enhanced ventilation requirements, extensive sterilization processes, and laundry services, inherently increase energy consumption and GHG emissions. The approach to making healthcare more sustainable has been multipronged, including initiatives to reduce waste and energy use. We explored environmental cleaning practices and the energy use associated with ultraviolent (UV) light disinfection. Methods: A retrospective analysis was conducted on the energy consumption of three different UV light disinfection machines used at Stanford Health Care from September 2023 to August 2024. Annual run time data was obtained from vendor-provided logs, and energy use was calculated using the equipment wattage specifications provided with each machine. Results: We found that UV light disinfection utilized approximately 7,300 kWh constituting less than 1% of Stanford Health Care total energy use for the period. This energy consumption equates to the charge required for 3.5 round trips from San Francisco to New York City in an electric vehicle. Discussion: UV light has been widely used in healthcare over the last decade. Recent data suggests that there may be no additional benefit to UV light disinfection when other enhanced cleaning methods, such as sporicidal cleaners, are utilized. Therefore, using UV light in addition to sporicidal cleaners may be redundant. Infection prevention practices often incorporate redundancies given the dependence on human behavior and the high consequences of practice failures. As the healthcare industry continues to work towards reducing greenhouse gas emissions it will be important to consider all energy reductions and any redundancies in practices. In the future, a life cycle analysis could be conducted to compare UV light disinfection and sporicidal cleaning methods to evaluate each practice’s impact on sustainability efforts. Our evaluation showed that UV light disinfection results in modest energy usage reductions. However, as the healthcare industry continues to work towards reducing greenhouse gas emissions it will be important to consider all energy reductions and any redundancies in practices.

Background: Duplicative laboratory testing is prevalent in health care. Prior research surrounding repeat urine cultures showed that when a negative index culture is repeated within 48 hours, less than 5% of repeat urine cultures show a new bacteriuria. We evaluated the diagnostic yield of repeating urine cultures at longer time intervals, and of repeating a positive urine culture. Methods: We conducted a retrospective study of adult inpatients at Stanford Healthcare who had more than one urine culture collected during hospitalization between January 2023 and February 2024. We included urine cultures that were collected with or without urinary catheters; nephrostomy tubes were excluded. Urine cultures were classified as index or repeat. We analyzed the diagnostic yield of the repeat urine culture, defined as the percent of repeat urine cultures that detected a new bacteriuria not detected in the index culture. Bacteriuria was defined as growth of a bacterial species in quantities >100,000 CFU/mL. A negative urine culture was defined as one that did not have bacteriuria meeting this threshold. Sensitivity analyses used a threshold of 10,000 CFU/mL as the threshold for significant bacteriuria. Results: Overall, 6,955 urine cultures were performed from 6,058 patients. Of these, 864 (12%) urine cultures were repeats. Of the 864 index cultures, 75% were negative. The median time to repeat urine culture was 4 days. When negative index cultures were repeated at 0-3 days, the diagnostic yield for detecting a new bacteriuria was only 9%. Diagnostic yield at 3-6 days was 10%, not significantly higher compared to 0-3 days (p=0.620). Diagnostic yield at 6-9 days was 19%; this increase was significant compared to the 0-3 days group (p=0.014). When positive index cultures were repeated at 0-3 days, the diagnostic yield for detecting a new bacteriuria was only 8%. Diagnostic yield at 3-6 days was also 8%. Yield increased significantly to 15% at 6-9 days from index culture (p=0.013). When the threshold for significant bacteriuria was adjusted to 10,000 CFU/mL, more bacteriuria was detected overall, but primarily of gram-positive organisms. Whether the threshold for significant bacteriuria was 100,000 CFU/mL or 10,000 CFU/mL, the rate of detection of new gram-negative bacteriuria was similar, and remained less than 10% until 6-9 days from index culture (Figure 1). Conclusions: Among inpatients, most urine cultures repeated at less than 6 days provide redundant information. This unnecessary retesting offers an opportunity for diagnostic stewardship.

Background: Racial and ethnic disparities have been demonstrated across a range of healthcare outcomes and services, including infectious disease burden. Individuals who reside in communities with increased social vulnerability are more likely to experience worse infection-related outcomes. These disparities are likely exacerbated by structural and systemic inequities experienced during inpatient care, even for common diagnoses, such as urinary tract infections. We explored the relationship between race, population-level social vulnerability, and urine culture results in the inpatient setting. Methods: We conducted a retrospective cohort study at Stanford Health Care from January–December, 2023. We included all adult inpatients who had a urine culture collected. The Center for Disease Control’s Social Vulnerability Index (SVI) was used as a composite measure of social vulnerability. Patient addresses from electronic health records (EHR) were geocoded to determine census tract-level SVI designation using the California-specific SVI database, and out-of-state addresses were excluded. Unhoused patients were identified by a discrete field in the EHR. We included demographics, urine culture results, and time of urine culture collection. Community-associated positive urine cultures were defined by a collection time ≤48 hours from admission, while healthcare-associated positive cultures were collected >48 hours from admission. Results: There were 5,374 admissions with urine cultures collected in 2023. The overall median SVI was 0.34. Compared to the statewide median overall vulnerability of 0.50[IQR: 0.25–0.75], our inpatient population resided in less vulnerable areas. When comparing patients with positive and negative urine cultures, the overall SVI, the four-specific SVI themes and SVI quartiles were similar. Unhoused patients were more likely to have a negative culture than a positive culture. Patients who identified as Asian were more likely to have a healthcare-associated positive urine culture than a community-associated positive culture. Patients who identified as Hispanic were more likely to have a community-associated positive culture than a healthcare-associated positive urine culture. Patients who identify as white or black had similar likelihood of developing a community-associated or healthcare-associated positive culture. Discussion: We did not find any differences in SVI among patients based on urine culture positivity. However, when stratified by community- vs healthcare-associated we found that patients who identify as Asian or Hispanic may be more likely to have a positive urine culture. These differences in outcomes are likely complex and multifaceted, potentially related to various social drivers of health present both before and during admission. Further exploration is needed to understand what is contributing to these findings.

Background: Wastewater-based epidemiology has demonstrated effectiveness in monitoring trends of viral infections at the city, state, and national levels. It captures data independent of testing intensity, providing a comprehensive biological sample of pathogens excreted in all secretions, that is unaffected by individual testing behaviors. Traditional healthcare-associated infection surveillance relies on case-based approaches, which can be resource-intensive, prone to misclassification, and may miss patients who are colonized. We aimed to evaluate the feasibility of implementing wastewater-based epidemiology in an acute care hospital for monitoring pathogens relevant to infection prevention and control. Methods: In this pilot study, we deployed a Teledyne ISCO 5800™ wastewater autosampler to collect weekly composite 1000 mL samples (15 mL every 151 minutes) from the final Stanford Hospital outflow point before wastewater merged with the community system. Wastewater samples were processed within 48 hours of collection. The solid phase was separated via centrifugation, followed by nucleic acid extraction employing silica-based purification techniques optimized for efficient inhibitor removal. Droplet digital PCR was conducted targeting pathogens previously validated by the WastewaterSCAN program (https://www.wastewaterscan.org/en/pathogens). We compared hospital wastewater nucleic acid concentrations with the number of positive tests/cultures at Stanford Hospital during the same period and with Wastewaterscan community wastewater data. Results: We collected three weekly composite samples: Jun 20–26, Jul 10–17, and Jul 18–25. Challenges included the location of the final outflow, and the autosampler’s size (132 x 74 x 84 cm and 88.5 kg). The outflow point was situated in a high-traffic area for patients and staff, requiring barricades to ensure safety and prevent interference with sampling equipment. In terms of interpreting results, viral nucleic acid concentrations (e.g., influenza, SARS-CoV-2) appeared to parallel the number of clinical cases and were similar to community wastewater trends (Figure 1). Most antimicrobial resistance genes, including vanA (Figure 2) and carbapenemase genes (KPC, NDM, OXA-48, VIM) (Figure 3), showed limited alignment with clinical cases; however, mecA exhibited some alignment (Figure 2). Hospital wastewater had higher resistance gene concentrations than community wastewater from San Mateo County (Figure 4). Conclusion: Continuous collection of hospital wastewater proved challenging, mainly from logistical issues such as equipment size and access limitations. Clinical respiratory virus trends appeared to be reflected in wastewater data. However, trends for antimicrobial resistance genes may be influenced by additional factors, such as the number of colonized patients, bacterial load in the hospital sewage system, hospital antimicrobial use, and antibiotic residues in wastewater.

Introduction: Enterococci are the third most common healthcare-associated pathogen, with 30% of isolates resistant to vancomycin (VRE). Resistance is often conferred by the vanA gene cluster on transposon Tn1546 and is frequently plasmid-borne. The suspected role of person-to-person transmission prompted the recommendation for VRE isolation precautions in 1995. However, quasi-experimental studies in hospitals discontinuing these precautions found no significant increases in VRE infections or bacterial clone spread using short-read whole genome sequencing (WGS). We used long-read WGS to analyze vanA plasmid transmission dynamics after discontinuing isolation precautions for VRE at Stanford University Hospital. Methods: This study was conducted at Stanford University Hospital, an 800-bed quaternary referral and transplant center. Routine contact precautions for VRE were discontinued on October 1, 2021. Blood culture Enterococcus faecalis and E. faecium isolates collected during 2021 (prior to and following discontinuation) were included, along with additional isolates retrieved from January–October 2023. Bacterial whole genome sequencing with long-read nanopore technology was performed. Custom analyses were performed on the assembled genomes. Patient data were collected retrospectively. Results: We retrieved 105 blood culture isolates (36.2%) from the isolation period (January–October 2021) and 185 isolates (63.8%) from the no isolation period (October–December 2021, January–October 2023), representing 202 unique patients. Patient characteristics and microbiological findings are shown in Table 1. Only 4.3% (7/171) of E. faecalis and 70.5% (84/119) of E. faecium isolates were vancomycin-resistant. Long-read WGS revealed no clustering between the isolation and no isolation periods. (Figure 1A and B); however, a dominant E. faecium ST117 cluster was seen, while E. faecalis showed greater diversity (Figure 1C). There were only four pairs of putative transmissions Conclusion: The discontinuation of contact isolation precautions at Stanford Hospital did not result in an increase in genetically related Enterococci or genetically related vanA plasmids among patients with Enterococcal bacteremia.

Background: Candida auris is an emerging fungal pathogen with potential to cause outbreaks. To mitigate transmission, the California Department of Public Health (CDPH) recommends considering implementation of an active case detection process in acute care hospitals to identify high-risk patients who may be colonized on admission. Methods: From 1/5/2024 to 9/20/2024, Stanford Health Care piloted an active surveillance program to identify high-risk patients for C. auris – defined as patients coming from (1) long term acute care hospitals (LTACHs), (2) ventilator skilled nursing facilities (vSNFs), (3) outside institutions with known C. auris outbreaks, (4) hospitals in Nevada, (5) recent international hospitalizations, and (6) patients with carbapenemase-producing organism (CPO) colonization.

Patients were identified for screening via daily review of custom-designed lists from the electronic medical record (EMR). A list of patients admitted from skilled nursing facilities (SNFs) and a list of patients transferred to Stanford from an outside facility were cross-referenced with a published list of high-risk facilities provided by CDPH. A list of inpatients flagged for CPOs was reviewed daily. Infection prevention was also notified by the transfer center or the care team if a patient had a recent international hospitalization. Screening was via superficial skin specimens from the axilla and groin. Culture-based testing was performed with identification of any fungal growth via MALDI-TOF. Results: During the pilot period, 1159 patients were evaluated for high-risk criteria; 58 (5%) met criteria for C. auris testing. One of 58 patients (colonization (Figure 1). There were 5 clinical cases during the pilot period, including the patient identified via screening. Active surveillance required 5-7 hours per week of infection preventionist effort, plus an hour to educate nursing staff when screening tests were performed. Conclusions: Our experience with performing active surveillance for C. auris resulted in one positive case, suggesting that this approach may have a lower yield in regions with low prevalence. Since surveillance can often be a time-intensive task for infection preventionists and nursing staff, it is important we continue to improve our knowledge about when and what surveillance is the most effective.

We analyzed the diagnostic yield of repeat urine cultures in a retrospective study of adult inpatients. Most urine cultures repeated at less than 6 days provided redundant information. This was true whether the index culture was positive or negative, and whether the threshold for positivity was 10,000 or 100,000 CFU/mL.

We investigated the accuracy and completeness of four large language model (LLM) artificial intelligence tools. Most LLMs provided acceptable answers to commonly asked infection prevention questions (accuracy 98.9%, completeness 94.6%). The use of LLMs to supplement infection prevention consults should be further explored.

We evaluated one of the first secure large language models approved for protected health information, for identifying central line-associated bloodstream infections (CLABSIs) using real clinical notes. Despite no pretraining, the model demonstrated rapid assessment and high sensitivity for CLABSI identification. Performance would improve with access to more patient data.