Background: Approximately half of antibiotics used in nursing home (NHs) are initiated in acute care hospitals prior to NH admission. Optimizing antibiotic prescribing on hospital discharge to these facilities presents an opportunity to improve NH antibiotic use. We aimed to identify intervention targets to optimize antibiotic use on discharge from the hospital to NHs. Methods: This was a multicenter, cross-sectional study across 9 acute care hospitals in Oregon, Wisconsin, and Washington. We selected a 20% random sample of adult (age >18) inpatients prescribed at least one antibiotic on discharge from the hospital to a NH between 2016-2018. We excluded patients discharged from the emergency department or an intensive care unit. Study data were electronically extracted from patients’ electronic health records and supplemented with manual chart review. Antibiotic optimization opportunities were determined by an infectious diseases (ID) physician or ID pharmacist and classified as definitely, possibly, or unlikely. Expert reviewers also recorded the type of optimization opportunity and the rationale for each determination. A gamma lasso algorithm was used to identify patient-level characteristics associated with definite optimization opportunity, which were then included in a logistic regression model. Results: There were 2761 antibiotic prescriptions among 2215 patients. Mean (standard deviation) age was 71.9 (14.3) years and 48.8% were male. Most discharges (83.1%) were prescribed one antibiotic, 15.2% were prescribed two antibiotics, and 1.8% were prescribed three antibiotics. The most frequently prescribed antibiotics were cephalexin (10.4%), vancomycin (9.8%), and amoxicillin clavulanate (8.4%). Among the 2761 antibiotic prescriptions, expert reviewers determined that 18.4% could definitely be optimized, 36.0% could possibly be optimized, and 45.3% unlikely could have been optimized. Among the 508 definite antibiotic optimization opportunities, 25.2% were to subtract the antibiotic, 56.3% were to change the antibiotic, 11.0% were to change the dose, 25.0% were to change the duration, 0.8% were to change the route, and 1.8% were to change the schedule. Patient-level characteristics found to be associated with definite antibiotic optimization opportunity included age over 80 years (odds ratio (OR)=1.44, 95% confidence interval (CI): [1.14, 1.82]), length of stay < 8 days (OR=1.40, 95% CI: [1.09, 1.81]), discharge with multiple antibiotic prescriptions (OR=1.92, 95% CI: [1.39, 2.63]), and discharge with prescription for oral vs intravenous (IV) antibiotics (OR=2.08, 95% CI: [1.49, 2.95]). Conclusion: We identified several patient and antibiotic characteristics which may serve as intervention targets to optimize antibiotic prescribing on discharge from the hospital to nursing homes.

Broad-spectrum antimicrobials are commonly used without indication and contribute to antimicrobial resistance (AMR). We implemented a syndrome-based stewardship intervention in a community hospital that targeted common infectious syndromes and antipseudomonal beta-lactam (APBL) use. Our intervention successfully reduced AMR, C. difficile rates, use of APBLs, and cost.

Background: Community Acquired Pneumonia (CAP) is the most common reason for antibiotic treatment in hospitalized adults. Some prior studies have found treatment differences by race/ethnicity but research on the topic is limited, results are mixed, and it is unclear if clinical outcomes are affected. We sought to examine whether guideline-concordant CAP care and patient outcomes varied by race/ethnicity. Methods: Using the Vizient clinical database, we conducted a cross sectional analysis of all hospitalized patients > = 18 years of age with a primary diagnosis of pneumonia (ICD10 codes: J12-J18) from 2018-2021. Univariate and bivariate analyses examined the distribution of demographic, clinical and hospital characteristics across race/ethnicity. The primary outcome was receipt of therapy concordant with ATS/IDSA Clinical Practice Guideline for CAP. Final models included only patients with bacterial pneumonia and examined the relationship between race/ethnicity and guideline-concordant antibiotic treatment. Secondary analysis examined the interaction between race/ethnicity and concordant antibiotic treatment with length of stay >7 days, 30-day hospital readmission, adverse events or complications in separate models. We used hierarchical multivariable regression models accounting for clustering within patients and among patients hospitalized at the same facility. Due to sample size, significance was assessed with an OR > = 1.2 and p≤ 0.05. All analyses used SAS (v.9.4, SAS Institute Inc. Cary, NC). Results: There were 1,277,770 admissions with a primary diagnosis of bacterial CAP. Sixty-nine percent of the sample was White, 18% Black, 8% Hispanic, 2% Asian and 3% identified as other. 56% of the sample received concordant care. In adjusted models Black patients had greater odds of overall concordant care (OR 1.22; p 7 days (OR 0.67 p <.0001), complication or adverse event (OR 0.75 p <.0001), but not readmission within 30 days. Conclusion: We observed differences between Black and White patients in the receipt of concordant treatment. Hospital bed size, CMI and region played an important role in both antibiotic treatment decisions and clinical outcomes, indicating that hospital and regional prescribing cultures may play in role in treatment inequities.

The COVID-19 pandemic has accelerated changes in health care across the nation. Particularly, infection prevention programs have been subjected to pressures and increased responsibilities with no expansion in support. In addition, there is a rapid trend for health systems to merge to ensure long term sustainability. Based on our experience leading infection prevention at one of the largest health systems in the United States, we outline how systems can provide and increase capacity to optimize and enhance the hospital level infection prevention programs and outcomes. In this commentary, “Ten Pillars for the Expansion of Health System Infection Prevention Capacity” we offer 10 categories of what we have found to establish a successful and functioning infection prevention program. The pillars to support the infection prevention programs focus on structure, processes, empowerment, and partnerships, and the elements and strategies that comprise them.

Background: The National Healthcare Safety Network (NHSN) provides detailed surveillance case definitions for healthcare-associated infections (HAI), including central line-associated bloodstream infections (CLABSI). CLABSI data are used for several purposes, including improving patient safety, value-based purchasing, and comparing hospitals’ performance. Our Infection Prevention (IP) team conducts house-wide HAI surveillance. To ensure that our hospital CLABSI reporting is accurate and that staff are implementing case definitions consistently and systematically, we conducted an internal validation of CLABSI. This undertaking allowed us to identify educational opportunities for IPs and improve surveillance data consistency. Methods: At UNC Hospitals, data on all positive blood cultures collected in the inpatient setting from July 2022 – June 2023 were obtained from electronic medical records. A random number generator was used to select 16 records per quarter. Each record was then randomly assigned to two different IPs (out of 8 total inpatient IPs) for review. Concordance of CLABSI classification was summarized across the two reviews and compared to the initial review. Discordant cases were then reviewed by the Associate IP Director (a certified IP with 15 years of experience) for final adjudication. A summary of findings and discordant cases details were discussed at regular IP educational meetings. Results: From July 2022-June 2023, there were 1658 positive blood cultures collected in the inpatient setting. Of the 64 randomly selected blood cultures, total concordance amongst all reviewers occurred 65.6% of the time. Concordance improved in the 2nd half of FY23 compared to the 1st half (72% vs, 59%, p>0.05). Amongst the 33% of blood culture results with reviewer discrepancy, the most common reasons were related to distinction of a bloodstream infection secondary to another infection site (32%) and application of the repeat infection timeframe (18%). Importantly, there was only one instance where a blood culture result was categorized by all 3 reviewers as present on admission, but upon Associate Director review, actually represented a CLABSI (i.e., false negative). Conclusions: Standardized case definitions remain open to interpretation. At our hospital, we experienced discordance in approximately one-third of instances during review of blood culture data amongst trained infection preventionists. Reviewing all blood culture data is key for validation so that both false positives and false negative CLABSIs can be identified. Identifying the most common reasons for discordance and using specific examples when case disagreement occurred for educational purposes may lead to improved reliability and accuracy of application of the NHSN surveillance defintions.

Antimicrobial-product package inserts and insufficient staffing impede routine carbapenem monitoring in the inpatient setting in Japan. The collaboration between antimicrobial stewardship teams and clinical pharmacists was associated with a sustained improvement in carbapenem dosing optimization. Our findings could be of use to countries with inadequate monitoring of carbapenem antimicrobial use.

Background: Indiscriminate urine culturing of patients with indwelling urinary catheters may lead to overdiagnosis of urinary tract infections, resulting in unnecessary antibiotic treatment and inaccurate reporting of catheter-associated urinary tract infections (CAUTIs) as a hospital quality metric. We evaluated the impact of a computerized diagnostic stewardship intervention to improve urine culture testing among patients with indwelling urinary catheters. Methods: We performed a single-center retrospective observational study at Rush University Medical Center from April 2018 – July 2023. In February 2021, we implemented a computerized clinical decision support tool to promote adherence to our internal urine culture guidelines for patients with indwelling urinary catheters. Providers were required to select one guideline criteria: 1) neutropenia, 2) kidney transplant, 3) recent urologic procedure, 4) urinary tract obstruction; or if none of the criteria were met, then an infectious diseases consultation was required for approval. We compared facility-wide CAUTI rate per 10,000 catheter days and standardized infection ratio (SIR) during baseline and intervention periods using ecologic models, controlling for time and for monthly Covid-19 hospitalizations. In the intervention period, we evaluated how providers responded to the intervention. Potential harm was defined as collection of a urine culture within 7 days of the intervention that resulted in a change in clinical management. Results: In unadjusted models, CAUTI rate decreased from 12.5 to 7.6 per 10,000 catheter days (p=0.04) and SIR decreased from 0.77 to 0.49 (p=0.09) during baseline vs intervention periods. In adjusted models, the CAUTI rate decreased from 6.9 to 5.5 per 10,000 catheter days (p=0.60) (Figure 1) and SIR decreased from 0.41 to 0.35 (p=0.65) during baseline vs intervention periods. Urine catheter standard utilization ratio (SUR) did not change (p=0.36). There were 598 patient encounters with ≥1 intervention. Selecting the first intervention for each encounter, 284 (47.5%) urine cultures met our guidelines for testing and 314 (52.5%) were averted (Figure 2). Of these, only 3 ( < 1 %) had a urine culture collected in the subsequent 7 days that resulted in change in clinical management. Conclusion: We observed a trend of decreased CAUTIs over time, but effect of our diagnostic stewardship intervention was difficult to assess due to healthcare disruption caused by Covid-19. Adverse outcomes were rare among patients who had a urine culture averted. A computerized clinical decision support tool may be safe and effective as part of a multimodal program to reduce unnecessary urine cultures in patients with indwelling urinary catheters.

Background: Healthcare contributes significantly to waste production and greenhouse gas emissions. This became especially apparent during the COVID-19 pandemic. Yet there is modest recognition of this issue, particularly within decision-making in Infection Prevention and Control (IPC). The aim of our study was to gauge general knowledge and attitudes of hospital epidemiologists (HEs) and infection preventionists (IPs) around the intersection of environmental sustainability and IPC, and to identify related institutional practices. Methods: An online survey, composed of ten questions related to environmental sustainability in IPC, was created and emailed to members of the SHEA Research Network (SRN), a national consortium of healthcare facilities collaborating on IPC research, from August - October 2023. Survey answers were collated via Redcap© and descriptive results were obtained. Results: Forty-two individuals (33 HEs, 7 directors of IPC, and 2 IPs) from unique institutions completed the survey. Thirty (71.4%) were from academic medical centers, 5 (11.9%) were from VA medical centers and 7 (16.7%) were from community hospitals. Over half of participants correctly estimated the amount of waste and carbon emissions produced annually by the US healthcare system (6 million tons and 8.5% of national emissions, respectively). However, only 42.9% considered environmental sustainability concerns important or very important when making IPC decisions. Fifteen (34.9%) had an environmental sustainability committee at their institution and of these, 8 had an established relationship with the IPC department. The most common techniques to promote sustainability amongst institutions were water/energy conservation (59.5%), reusable personal protective equipment (52.4%) and Leadership in Energy and Environmental Design (LEED) certification (47.6%). When asked which efforts they would support at their institution, 28.6% would eliminate the use of single-use endoscopes and one third would avoid use of ethylene oxide for sterilization. In deciding whether to support environmental sustainability measures, key considerations participants articulated were patient safety concerns, knowledge about effectiveness and costs, and administrative support. Conclusion: Although there is growing awareness around the contribution of the healthcare industry to carbon emissions and waste production, IPC professionals have not yet universally adopted measures to promote environmental sustainability. In our survey, many participants acknowledged the importance of balancing patient safety and sustainability concerns. Our study demonstrates the need for more research and education to inform decisions around environmentally sustainable efforts in IPC that also preserve patient safety. Additionally, professional and regulatory bodies must acknowledge and promote the importance of environmental sustainability in IPC decision-making.

Inappropriate antibiotic use may lead to increased adverse drug events (ADEs). This study assessed whether an antimicrobial stewardship recommendation to discontinue antibiotics in patients with low likelihood for bacterial infection reduced antibiotic duration and antibiotic-associated ADEs. At a 4-hospital system, hospitalized adult patients receiving empiric antibiotics for suspected infection were identified between May 2, 2016 and June 30, 2018. For those patients who were deemed unlikely to have a bacterial infection, a note was left by an infectious diseases physician recommending antibiotic discontinuation. Patient cases were considered “adherent” to recommendations if antibiotics were discontinued within 48 hours of the note and “non-adherent” to recommendations if antibiotics were continued beyond this. Duration of antibiotics and potential antibiotic-associated ADEs were collected retrospectively. Attribution of the adverse event to the antibiotic was decided upon by the investigators. The incidence of ADEs and duration of antibiotics between the adherent and non-adherent groups were compared. Of 253 patients deemed unlikely to have a bacterial infection, 114 (45%) treatment teams stopped antibiotics within 48 hours of the recommendation, and 139 (55%) continued antibiotics. The total number of ADEs was significantly higher in the non-adherent group compared to the adherent group (34 ADEs vs 9 ADEs, P = .001). The median number of total prescribed antibiotic days was higher in the non-adherent group than in the adherent group (5 days vs 1 day, P < .001). This study demonstrates that stewardship programs may prevent ADEs by recommending antibiotic discontinuation in patients with low suspicion for bacterial infection.

Background: Measuring and reporting antibiotic use are essential to antimicrobial stewardship activities. The most common metric to assess facility-level use is days of antibiotic therapy per 1000 days of care (DOT/1000 DOC). This metric may be difficult to calculate, not be readily comparable, or not provide actionable data to individual prescribers, particularly those that work in post-acute and long-term care (PALTC) settings. Here we use data from a centralized dispensing pharmacy to develop antibiotic use metrics suitable for offering individualized feedback to prescribers working in PALTC settings. Methods: We obtained medication dispensing data and resident census data for 13 PALTC settings within the same network. After omitting non-pharmacologic items and limiting the data to medications dispensed from 1/2020 – 12/2022, we determined the following metrics by month: days of antibiotic therapy (DOT), number of medications prescribed, number of antibiotic courses prescribed (antibiotic starts), and the number of individual residents issued a prescription for any medication (unique residents). These metrics were assessed for each facility (2020 – 2022) and for prescribers responsible for > 1% of prescriptions within that facility (2022 only). Prescriber-level unique residents was the number of residents issued a prescription by the given provider. We obtained facility-level census data to calculate antibiotic DOT/1000 resident days of care (DOC) as a standard to which we compared novel metrics. Results: During the 3-year study period, 1718 prescribers at 13 PALTC settings wrote for 672256 medications, including 31087 antibiotic courses. At the facility level, the correlation between monthly antibiotic starts (courses)/unique residents and antibiotic DOT/1000 DOC was 0.83 (p < 0 .0001). The correlation between monthly antibiotic DOT/unique residents and antibiotic DOT/1000 DOC was 0.98 (p < 0 .0001). Trends in monthly values of both novel metrics and DOT/1000 DOC were consistent across the examined period (Figure 1). For individual prescribers, both novel metrics permit assessment and comparison of antibiotic prescription rates over time (Figure 2). Conclusions: Pharmacy dispensing data can be used to determine antibiotic DOT/unique residents and antibiotic starts/unique residents at the facility level and for individual providers. The novel metric antibiotic DOT/unique residents demonstrated strong correlation with antibiotic DOT/1000 DOC at the facility level. In addition to supporting tracking and reporting of antibiotic use among PALTC settings, these new metrics permit visualization of the antibiotic prescribing rates of individuals prescribers, as well as peer comparison, which in turn can lead to actionable feedback that helps improve antibiotic use in the care of PALTC.

Background: Machine-learning (ML) models, such as neural networks (NNs), have been proposed to predict antimicrobial susceptibility at the patient level while incorporating patient-level information from electronic medical record (EMR) systems. However, NNs often do not perform well in predicting rare outcomes, such as carbapenem resistance. We aimed to apply a novel multitask NN to create personalized antibiograms for individual patients with Escherichia coli clinical isolates to predict antimicrobial resistance (AMR) for four major antimicrobial classes simultaneously with improved accuracy for carbapenem resistance by using shared hidden layers (Figure 1). Methods: We analyzed all E. coli clinical isolates from the US Veterans Health Administration’s network from January 1, 2017, to December 31, 2019, focusing on AMR profiles of aminopenicillins, narrow-spectrum (NS) cephalosporins, extended-spectrum (ES) cephalosporins, and carbapenems. Patient-level clinical data (demographics, antimicrobial exposure history, previous isolates (if any), comorbidities, and recent procedures) were extracted from EMR. Antibiograms for all hospitals were generated using standard methods for the preceding calendar years. We employed logistic regression to evaluate the efficacy of conventional antibiograms in predicting AMR profiles. We adopted the ML approach using conventional NNs and novel multitask NNs on all extracted clinical data and hospital antibiograms. The models were trained with data from 2017 and 2018 and then tested on 2019 data, assessing their performance using the area under the receiver-operating curve (AUC). Results: The study included 257,968 E. coli isolates, split into 171,391 for training and 86,577 for validation. The prevalence of AMR in the test data from 2019 was 49.8% for aminopenicillins, 28.4% for NS cephalosporins, 10.7% for ES cephalosporins, and 0.2% for carbapenems, respectively. Conventional hospital antibiograms showed low prediction accuracy with AUC scores of 0.56 for aminopenicillins, 0.67 for NS cephalosporins, 0.61 for ES cephalosporin, and 0.67 for carbapenem. AUC scores from preliminary models for conventional and multitask NNs were 0.78/0.78 for aminopenicillins, 0.83/0.82 for NS cephalosporins, 0.84/0.85 for ES cephalosporins, 0.68/0.75 for carbapenems. While producing improved accuracy for carbapenem and comparable accuracies for three other classes, multitask NNs took approximately 66% less time for model training than conventional NNs. Conclusions: Integrating EMR data with NNs improved their predictive accuracy, potentially leading to a decision-support tool for better empirical antimicrobial therapy guidance in the window between species identification and confirmed susceptibilities. Multitask NNs can potentially improve the prediction accuracy of uncommon AMRs while maintaining comparable prediction accuracies for common AMRs and optimizing the efficiency of model training.

Disclosure: Michi Goto: Contracted Research - Merck

Background: Neonatal intensive care units (NICU) outbreaks caused by methicillin-susceptible Staphylococcus aureus (MSSA) are less commonly reported than outbreaks caused by methicillin-resistant S. aureus. We report an unusual outbreak of MSSA skin and soft tissue infections (SSTIs) in a level IV NICU investigated by whole genome sequencing (WGS) and molecular typing. Methods: An investigation was initiated in a 56-single-bed NICU after four patients developed MSSA SSTIs in Week 1. Case-patients had positive MSSA cultures identified by clinical cultures or surveillance sampling (bilateral nares, axillae, umbilicus, and groin), and antibiotic susceptibility testing was performed. WGS and assessment of isolate relatedness through mutation event analysis and multi-locus sequence typing (MLST) was performed by the NYS DOH. Demographic and isolate characteristics were compared using Wilcoxon rank sum test, Fisher’s exact test and Pearson’s Chi-squared test, as appropriate. Results: From Week 2 to Week 32, 9 rounds of surveillance for MSSA colonization were conducted. In all, 30 case-patients had MSSA colonization and 16 infants developed infections including impetigo (n=7), pustules (n=5), staphylococcal scalded skin syndrome (SSSS, n=2), abscess (n=1), and bacteremia (n=1). All SSTI cases presented on infants’ faces, all of whom were on non-invasive respiratory support. MLST identified 4 distinct types including MLST 121 (n=12), MLST 398 (n=10), MLST 30 (n=6), and MLST 15 (n=6). Eight isolates were unrelated to other isolates. MLST 398 and MLST 30 included isolates not closely related (>9 mutation events). The 12 MLST 121 isolates were closely related (≤9 mutational events between all isolates), harbored the mupA gene, and were mupirocin-resistant (MIC>1024 ug/ml). Clinical infection and mupirocin resistance were associated with MLST 121 (Table 1). Multiple infection control measures were implemented, including increased availability of alcohol-based hand sanitizers, introducing bare-below-the-elbows practice for staff, contact precautions for case-patients, decolonization with mupirocin and chlorhexidine baths, environmental cleaning/disinfection, and removing excess equipment and supplies. No new cases of mupirocin-resistant or MLST 121 SSTIs occurred after Week 25. Conclusion: We report a MSSA outbreak associated with multiple MLST types and a predominant mupirocin-resistant strain. This report highlights the ability of molecular typing to characterize strains causing infections versus colonization and the potential loss of mupirocin as a control measure when outbreaks are caused by mupirocin-resistant strains. WGS analysis allows for increased discrimination of mutation events allowing for improved resolution of case relatedness compared to other typing methods. Successful control of this outbreak was achieved with a multitude of infection prevention and control.

We assessed the impact of metagenomic next-generation sequencing (mNGS) on patient care using previously established criteria. Among 37 patients receiving mNGS testing, 16% showed results that had a positive clinical impact. While mNGS results may offer valuable supplementary information, results should be interpreted within the broader clinical context and evaluation.

Background: Central line-associated bloodstream infections (CLABSIs) and catheter-associated urinary tract infections (CAUTIs) cause significant morbidity and mortality among hospitalized patients. Over the last 10 to 20 years, hospital accountability for the prevention of device-associated infections increased at the state and national levels. For example, the Centers for Medicare & Medicaid Services implemented the Hospital Inpatient Quality Reporting Program in 2015. The objective of this study was to assess the impact of increased federal attention on infection prevention using longitudinal data from the National Healthcare Safety Network (NHSN). We hypothesize that there was a significant decrease in statewide standardized infection ratios (SIRs) for CLABSI and CAUTI over the last 8 years. Methods: We collected SIRs for CLABSI and CAUTI in acute care hospitals for all 50 states and Washington D.C. from the NHSN database from 2015 to 2022. For CLABSI, we performed unique analyses for critical care units, wards, and neonatal intensive care (NICU) locations. For CAUTI, we stratified by critical care units and wards. We included all states with more than 5 hospitals reporting data. Those with fewer than 5 were excluded by listwise deletion in the corresponding analysis. We tested trends over time using linear mixed effect models with year as fixed effect and state as random effect. We also included an indicator variable representing the influence of SARS-CoV-2 (Covid-19) on healthcare-associated infections (HAIs). We elected an alpha of 0.05 as the threshold for statistical significance. Results: Overall, CLABSI and CAUTI SIRs exhibited significant negative slopes (Figures 1-5) after controlling for the influence of Covid-19. Each analysis revealed progressively lower SIRs compared to the previous year except for the 2019-2020 interval. Interestingly, the linear trend resumed after 2020 with subsequently lower SIRs in 2021 and 2022. Covid-19 had a greater influence on CLABSI SIRs in critical care settings compared to ward or NICU locations. The slope of CAUTI SIRs were impacted less by Covid-19 in wards compared to critical care settings. Conclusion: The results of the analysis demonstrate that CLABSI and CAUTI are trending in the desired direction despite the HAI spike during Covid-19. Government and hospital stakeholders in the United States should be encouraged by the reported trends and continue to prioritize the funding and use of resources for evidence-based device-associated infection prevention.

Background: Nonsterile ice is frequently used in healthcare settings for a wide array of patient care activities and clinical procedures. However, this ice can harbor pathogenic organisms which can threaten patient safety and cause outbreaks. We sought to characterize recent Centers for Disease Control and Prevention (CDC) consultations involving ice leading to healthcare-associated infections (HAIs). Methods: We reviewed internal CDC records from the Division of Healthcare Quality Promotion (DHQP) to identify investigations of outbreaks and potential outbreaks involving the use of ice in healthcare facilities. We searched records from January 1, 2016, through November 30, 2023, for keywords related to ice. We excluded consultations in which ice was not thought to be a potential transmission pathway as well as those in which only sterile ice products (e.g., saline slush) were investigated. Results: We identified 45 consultations for ice-related investigations, involving a total of 533 patients. Nontuberculous mycobacteria were the most frequently implicated organisms, appearing in 40% (n=18) of investigations. Eighty-four percent (n=38) of investigations occurred in acute care hospitals. The most frequently implicated hospital settings were intensive care units (13%, n=6), operating rooms (13%, n=6), and bronchoscopy suites (13%, n=6). We identified a variety of plausible exposure pathways, including direct ingestion of ice by patients, use of ice during the bronchoscopy procedure, use of nonsterile ice in heater-cooler devices during cardiothoracic surgery, and the use of ice to chill saline for respiratory care. Environmental sampling directly of ice machines was performed in 62% of investigations (n=28) and nonsterile ice from these machines was sampled in 9% of investigations (n=4). Among those investigations in which ice machines were sampled, the organism implicated in the outbreak was isolated in 54% of investigations (n=15). Among those investigations in which ice itself was sampled, the organism implicated in the outbreak was isolated in 75% of investigations (n=3). These organisms included Mycobacterium mucogenicum, Burkholderia multivorans, and Acanthamoeba spp. Conclusions: The use of nonsterile ice during clinical care is a potential source of pathogens that cause patient infections and HAI outbreaks. Healthcare personnel should be aware of the risk posed by nonsterile ice and consider avoiding its use, especially when caring for patients who are critically ill or immunocompromised. Healthcare facilities should ensure regular cleaning and disinfection of ice machines to decrease their microbial burden. When HAI outbreaks involving water-associated organisms are identified, nonsterile ice should be considered as a potential mode of transmission.