Background: Nursing home (NH) residents are at high risk of COVID-19 from exposure to infected staff and other residents. Understanding SARS-CoV-2 viral RNA kinetics in residents and staff can guide testing, isolation, and return to work recommendations. We sought to determine the duration of antigen test and polymerase chain reaction (PCR) positivity in a cohort of NH residents and staff. Methods: We prospectively collected data on SARS-CoV-2 viral kinetics from April 2023 through November 2023. Staff and residents could enroll prospectively or upon a positive test (identified through routine clinical testing, screening, or outbreak response testing). Participating facilities performed routine clinical testing; asymptomatic testing of contacts was performed within 48 hours if an outbreak or known exposure occurred and upon (re-) admission. Enrolled participants who tested positive for SARS-CoV-2 were re-tested daily for 14 days with both nasal antigen and nasal PCR tests. All PCR tests were run by a central lab with the same assay. We conducted a Kaplan-Meier survival analysis on time to first negative test restricted to participants who initially tested positive (day zero) and had at least one test ≥10 days after initially testing positive with the same test type; a participant could contribute to both antigen and PCR survival curves. We compared survival curves for staff and residents using the log-rank test. Results: Twenty-four nursing homes in eight states participated; 587 participants (275 residents, 312 staff) enrolled in the evaluation, participants were only tested through routine clinical or outbreak response testing. Seventy-two participants tested positive for antigen; of these, 63 tested PCR-positive. Residents were antigen- and PCR-positive longer than staff (Figure 1), but this finding is only statistically significant (p=0.006) for duration of PCR positivity. Five days after the first positive test, 56% of 50 residents and 59% of 22 staff remained antigen-positive; 91% of 44 residents and 79% of 19 staff were PCR-positive. Ten days after the first positive test, 22% of 50 residents and 5% of 22 staff remained antigen-positive; 61% of 44 residents and 21% of 19 staff remained PCR-positive. Conclusions: Most NH residents and staff with SARS-CoV-2 remained antigen- or PCR-positive 5 days after the initial positive test; however, differences between staff and resident test positivity were noted at 10 days. These data can inform recommendations for testing, duration of NH resident isolation, and return to work guidance for staff. Additional viral culture data may strengthen these conclusions.

Disclosure: Stefan Gravenstein: Received consulting and speaker fees from most vaccine manufacturers (Sanofi, Seqirus, Moderna, Merck, Janssen, Pfizer, Novavax, GSK, and have or expect to receive grant funding from several (Sanofi, Seqirus, Moderna, Pfizer, GSK). Lona Mody: NIH, VA, CDC, Kahn Foundation; Honoraria: UpToDate; Contracted Research: Nano-Vibronix

During the coronavirus disease 2019 pandemic, mathematical modeling has been widely used to understand epidemiological burden, trends, and transmission dynamics, to facilitate policy decisions, and, to a lesser extent, to evaluate infection prevention and control (IPC) measures. This review highlights the added value of using conventional epidemiology and modeling approaches to address the complexity of healthcare-associated infections (HAI) and antimicrobial resistance. It demonstrates how epidemiological surveillance data and modeling can be used to infer transmission dynamics in healthcare settings and to forecast healthcare impact, how modeling can be used to improve the validity of interpretation of epidemiological surveillance data, how modeling can be used to estimate the impact of IPC interventions, and how modeling can be used to guide IPC and antimicrobial treatment and stewardship decision-making. There are several priority areas for expanding the use of modeling in healthcare epidemiology and IPC. Importantly, modeling should be viewed as complementary to conventional healthcare epidemiological approaches, and this requires collaboration and active coordination between IPC, healthcare epidemiology, and mathematical modeling groups.

Background: Antiviral chemoprophylaxis for influenza is recommended in nursing homes to prevent transmission and severe disease among residents with higher risk of severe influenza complications. Interim CDC guidance recommends that long-term care facilities initiate antiviral chemoprophylaxis with oral oseltamivir for all non-ill residents living in the same unit following the start of an outbreak in a facility (ie, ≥2 patients ill within 72 hours and of whom at least 1 resident has laboratory-confirmed influenza). Prophylaxis continues for a minimum of 2 weeks and for at least 7 days after the last laboratory-confirmed case. However, facilities may not strictly adhere to this guidance, with 1 study showing up to 68% of facilities were nonadherent to national guidance (Silva et al 2020). Here, we model the potential impacts of different antiviral prophylaxis strategies. Methods: We developed a susceptible–exposed–asymptomatic–infected–recovered (SEAIR) compartmental model of an average-sized nursing home comprising short-stay residents, long-stay residents, and healthcare personnel (HCP). Persons treated with antiviral chemoprophylaxis were less susceptible to infection, had a lower probability of symptoms if infected, a reduced viral load, and a shortened duration of infectiousness. We included influenza vaccination for residents and HCP through reduced probability of symptomatic infection. Coverage rates were estimated from CDC FluVaxView and CMS COVID-19 nursing home data. As a base case, we modeled a scenario with prophylaxis implemented according to guidance. We varied uptake by residents and HCP (from 10% to 90%), case thresholds for prophylaxis initiation (1–5 cases identified), and timing of prophylaxis cessation: either time dependent (ie, 10–14 days of prophylaxis) or case-dependent (ie, continuing prophylaxis for 1–7 days with no cases). Results: In the scenario based on current guidance, prophylaxis reduced resident cases by 16% and resident hospitalizations by 45%, compared to no prophylaxis (Fig. 1A). Scenarios that differed from the guidance altered case burden and timing: Time-dependent prophylaxis cessation increased resident cases and hospitalizations (Fig. 1A). Timing of prophylaxis initiation had slight effects on the timing of the epidemic and minimal effects on resident cases and hospitalizations (Fig. 1B). High resident uptake was important for reducing resident cases and hospitalizations (Fig. 1C), but increasing HCP uptake had minimal effect (Fig. 1D). Conclusions: Our findings support the current prophylaxis guidance. Promptly implementing prophylaxis reduces resident cases and hospitalizations. Continuing prophylaxis until cases are no longer identified reduces cases and hospitalizations.

Disclosure: None

Background: Clostridioides difficile and multidrug-resistant organisms (MDROs) pose challenges due to treatment complexities and substantial morbidity and mortality. Susceptibility to colonization with these organisms and potential onward transmission if colonized (ie, infectivity) is influenced by the human microbiome and its dynamics. Disruptive effects of antibiotics on the microbiome imply potential indirect effects of antibiotics on C. difficile colonization. Mathematical models can help explore the relative impact of key pathways linking antibiotic use to C. difficile colonization, including the relationship between population-level antibiotic use and colonization prevalence. Methods: We built a compartmental model of long-term C. difficile colonization prevalence of nursing home residents (though malleable for any MDRO), allowing interactions between the microbiome and the colonization process. Based on proportional abundance of microbial taxa, we classified individuals into high and low α diversity groups, each further stratified into uncolonized or colonized with C. difficile. The rate of transition from the high to low microbiome diversity group was proportional to the population-level rate of antibiotic use. Transmission dynamics followed a susceptible–infectious–susceptible framework with the possibility for increased susceptibility and infectivity for the low-diversity microbiome group. First, as a comparator, we used a “null model” in which microbiome diversity did not influence host susceptibility or infectivity. Next, we sampled from realistic (literature informed) parameter ranges to analyze how the microbiome mediates the effect of antibiotics on colonization in this population. Results: Our analysis suggests that antibiotic use can catalyze colonization with C. difficile through interactions with the host microbiome, resulting in a sharp increase in colonization with a modest increase in antibiotic use (Fig 1). Increasing the population-level antibiotic use by 5% led to a median 24% increase in long-term colonization prevalence in the model (Fig 2). In contrast, increasing susceptibility or infectivity rates by 5% resulted in slightly higher increases in total colonization (27% and 29%, respectively). However, there was considerable uncertainty around these estimates, with interquartile ranges of up to 20% for some parameters (Fig 2). Conclusions: Higher population-level antibiotic use likely increases colonization by C. difficile through indirect effects of the microbiome. The increased colonization burden attributable to increasing antibiotic use may be substantial. With high uncertainty around some estimates, conducting observational studies to better understand key colonization and microbiome parameters (eg, the relative increase in susceptibility or infectivity with lower microbiome diversity) is critical for future efforts to estimate the impact of antibiotic use on colonization with C. difficile and MDROs.

Disclosures: None

Background: Multimodal approaches are often used to prevent transmission of antimicrobial-resistant pathogens among patients in healthcare settings; understanding the effect of individual interventions is challenging. We designed a model to compare the effectiveness of hand hygiene (HH) with or without decolonization in reducing patient colonization with carbapenem-resistant Enterobacterales (CRE). Methods: We developed an agent-based model to represent transmission of CRE in an acute-care hospital comprising 3 general wards and 2 ICUs, each with 20 single-occupancy rooms, located in a community of 85,000 people. The model accounted for the movement of healthcare personnel (HCP), including their visits to patients. CRE dynamics were modeled using a susceptible–infectious–susceptible framework with transmission occurring via HCP–patient contacts. The mean time to clearance of CRE colonization without intervention was 387 days (Zimmerman et al, 2013). Our baseline included a facility-level HH compliance of 30%, with an assumed efficacy of 50%. Contact precautions were employed for patients with CRE-positive cultures with assumed adherence and efficacy of 80% and 50%, respectively. Intervention scenarios included decolonization of culture-positive CRE patients, with a mean time to decolonization of 3 days. We considered 2 hypothetical intervention scenarios: (A) decolonization of patients with the baseline HH compliance and (B) decolonization with a slightly improved HH compliance of 35%. The hospital-level CRE incidence rate was used to compare the results from these intervention scenarios. Results: CRE incidence rates were lower in intervention scenarios than the baseline scenario (Fig. 1). The baseline mean incidence rate was 29.1 per 10,000 patient days. For decolonization with the baseline HH, the mean incidence rate decreased to 14.5 per 10,000 patient days, which is a 50.2% decrease relative to the baseline incidence (Table 1). The decolonization scenario with a slightly improved HH compliance of 35% produced a relative reduction of 71.9% relative to the baseline incidence. Conclusions: Our analysis shows that decolonization, combined with modest improvement in HH compliance, could lead to large decreases in pathogen transmission. In turn, this model implies that efforts to identify and improve decolonization strategies for better patient safety in health care may be needed and are worth exploring.

Disclosures: None

Background: Pathogen transmission among staff and residents in nursing homes can vary depending on their interactions and by the amount of time a resident receives care in the facility. Understanding the relative differences in transmission rates between and among staff and residents can identify the pathways that contributed most to the spread of SARS-CoV-2 in US nursing homes. Further exploring relative differences by categorizing facilities by residents’ lengths of stay can identify priority categories for intervention. Methods: Using US National Healthcare Safety Network (NHSN) surveillance data on resident and staff cases, vaccination, and resident deaths during June 2020–June 2022, we estimated SARS-CoV-2 transmission among and between residents and staff. We used a Bayesian inversion of a susceptible–exposed–infected–removed–virus–death (SEIRVD) compartmental model to produce the estimates. The facilities were divided into those with median length of stay (LOS) among the residents of 10 weeks. Additional inputs included the incidence and vaccination levels of the county where each facility was located. For the compartmental model, all data were averaged to form a representative facility for each category. Transmission was estimated separately for 3 periods: (1) June 2020–March 2021 as before the SARS-CoV-2 delta variant, (2) April 2021—October 2021 during SARS-CoV-2 delta variant dominance, and (3) November 2021—June 2022 during the prevalence of the SARS-CoV-2 omicron variant. Results: Regardless of facility category, transmission was highest from staff to residents or resident to resident (Fig.). These estimates of transmission were highest during the pre–SARS-CoV-2 delta variant phase. Transmission in that phase was highest in the facilities with LOS >10 weeks from staff to residents at 0.88 per week (95% credible interval [CrI], 0.06–1.85), in the facilities with LOS 6–10 weeks from staff to residents at 0.68 per week (95% CrI, 0.03–1.78), and in the facilities with LOS <6 weeks between residents at 0.47 per week (95% CrI, 0.02–0.95). Conclusions: Staff-to-resident or resident-to-resident transmission were the dominant pathways of spread of SARS-CoV-2 across the periods or the facility categories. Facilities with LOS 6 weeks or longer had higher median transmission estimates across the periods and transmission routes compared to facilities with LOS less than 6 weeks, implying that when prioritization of intervention resources is needed, facilities caring for populations with longer stays could be prioritized.

Disclosures: None

Background: The CDC’s new Public Health Strategies to Prevent the Spread of Novel and Targeted Multidrug-Resistant Organisms (MDROs) were informed by mathematical models that assessed the impact of implementing preventive strategies directed at a subset of healthcare facilities characterized as influential or highly connected based on their predicted role in the regional spread of MDROs. We developed an interactive tool to communicate mathematical modeling results and visualize the regional patient transfer network for public health departments and healthcare facilities to assist in planning and implementing prevention strategies. Methods: An interactive RShiny application is currently hosted in the CDC network and is accessible to external partners through the Secure Access Management Services (SAMS). Patient transfer volumes (direct and indirect, that is, with up to 30 days in the community between admissions) were estimated from the CMS fee-for-service claims data from 2019. The spread of a carbapenem-resistant Enterobacterales (CRE)–like MDROs within a US state was simulated using a deterministic model with susceptible and infectious compartments in the community and healthcare facilities interconnected through patient transfers. Individuals determined to be infectious through admission screening, point-prevalence surveys (PPSs), or notified from interfacility communication were assigned lower transmissibility if enhanced infection prevention and control practices were in place at a facility. Results: The application consists of 4 interactive tabs. Users can visualize the statewide patient-sharing network for any US state and select territories in the first tab (Fig. 1). A feature allows users to highlight a facility of interest and display downstream or upstream facilities that received or sent transfers from the facility of interest, respectively. A second tab lists influential facilities to aid in prioritizing screening and prevention activities. A third tab lists all facilities in the state in descending order of their dispersal rate (ie, the rate at which patients are shared downstream to other facilities), which can help identify highly connected facilities. In the fourth tab, an interactive graph displays the predicted reduction of MDRO prevalence given a range of intervention scenarios (Fig. 2). Conclusions: Our RShiny application, which can be accessed by public health partners, can assist healthcare facilities and public health departments in planning and tailoring MDRO prevention activity bundles.

Disclosures: None

Background: Multidrug-resistant organisms (MDROs), such as carbapenem-resistant Enterobacterales (CRE), can spread rapidly in a region. Facilities that care for high-acuity patients with long average lengths of stay (eg, long-term acute-care hospitals or LTACHs and ventilator-capable skilled nursing facilities or vSNFs) may amplify this spread. We assessed the impact of interventions on CRE spread within a region individually, bundled, and implemented at different facility types. Methods: We developed a deterministic compartmental model, parametrized using CRE data reported to the NHSN and patient transfer data from the CMS specific to a US state. The model includes the community and the healthcare facilities within the state. Individuals may be either susceptible or infected and infectious. Infected patients determined to have CRE through admission screening or point-prevalence surveys at a facility are placed in a state of lower transmissibility if enhanced infection prevention and control (IPC) practices are in place. Results: Intervention bundles that included periodic point-prevalence surveys and enhanced IPC at high-acuity postacute-care facilities had the greatest impact on regional prevalence 10 years into an outbreak; the benefits of including admission screening and improved interfacility communication were more modest (Fig. 1A). Delaying interventions by 3 years is predicted to result in smaller reductions in prevalence (Fig. 1B). Increasing the frequency of point-prevalence surveys from biannually to quarterly resulted in a substantial relative reduction in prevalence (from 25% to 44%) if conducted from the start of an outbreak. IPC improvements in vSNFs resulted in greater relative reductions than in LTACHs. Admission screening at LTACHs and vSNFs was predicted to have a greater impact on prevalence if in place prior to CRE introduction (~20% reduction), and the impact decreased by approximately half if implementation was delayed until 3 years after CRE introduction. In contrast, the effect of admission screening in ACH was less (~10% reduction in prevalence) and did not change with implementation delays. Conclusions: Our model suggests that interventions that limit unrecognized MDRO introduction to, or dispersal from, LTACHs and vSNFs through screening are predicted to slow distribution regionally. Interventions to detect colonization and improve IPC practices within LTACHs and vSNFs may substantially reduce the regional burden. Prevention strategies are predicted to have the greatest impact when interventions are bundled and implemented before an MDRO is identified in a region, but reduction in overall prevalence is still possible if implemented after initial MDRO spread.

Funding: None

Disclosures: None

Background: Carbapenem-resistant Enterobacteriaceae (CRE) are endemic in the Chicago region. We assessed the regional impact of a CRE control intervention targeting high-prevalence facilities; that is, long-term acute-care hospitals (LTACHs) and ventilator-capable skilled nursing facilities (vSNFs). Methods: In July 2017, an academic–public health partnership launched a regional CRE prevention bundle: (1) identifying patient CRE status by querying Illinois’ XDRO registry and periodic point-prevalence surveys reported to public health, (2) cohorting or private rooms with contact precautions for CRE patients, (3) combining hand hygiene adherence, monitoring with general infection control education, and guidance by project coordinators and public health, and (4) daily chlorhexidine gluconate (CHG) bathing. Informed by epidemiology and modeling, we targeted LTACHs and vSNFs in a 13-mile radius from the coordinating center. Illinois mandates CRE reporting to the XDRO registry, which can also be manually queried or generate automated alerts to facilitate interfacility communication. The regional intervention promoted increased automation of alerts to hospitals. The prespecified primary outcome was incident clinical CRE culture reported to the XDRO registry in Cook County by month, analyzed by segmented regression modeling. A secondary outcome was colonization prevalence measured by serial point-prevalence surveys for carbapenemase-producing organism colonization in LTACHs and vSNFs. Results: All eligible LTACHs (n = 6) and vSNFs (n = 9) participated in the intervention. One vSNF declined CHG bathing. vSNFs that implemented CHG bathing typically bathed residents 2–3 times per week instead of daily. Overall, there were significant gaps in infection control practices, especially in vSNFs. Also, 75 Illinois hospitals adopted automated alerts (56 during the intervention period). Mean CRE incidence in Cook County decreased from 59.0 cases per month during baseline to 40.6 cases per month during intervention (P < .001). In a segmented regression model, there was an average reduction of 10.56 cases per month during the 24-month intervention period (P = .02) (Fig. 1), and an estimated 253 incident CRE cases were averted. Mean CRE incidence also decreased among the stratum of vSNF/LTACH intervention facilities (P = .03). However, evidence of ongoing CRE transmission, particularly in vSNFs, persisted, and CRE colonization prevalence remained high at intervention facilities (Table 1). Conclusions: A resource-intensive public health regional CRE intervention was implemented that included enhanced interfacility communication and targeted infection prevention. There was a significant decline in incident CRE clinical cases in Cook County, despite high persistent CRE colonization prevalence in intervention facilities. vSNFs, where understaffing or underresourcing were common and lengths of stay range from months to years, had a major prevalence challenge, underscoring the need for aggressive infection control improvements in these facilities.

Funding: The Centers for Disease Control and Prevention (SHEPheRD Contract No. 200-2011-42037)

Disclosures: M.Y.L. has received research support in the form of contributed product from OpGen and Sage Products (now part of Stryker Corporation), and has received an investigator-initiated grant from CareFusion Foundation (now part of BD).

Background: Successful containment of regional outbreaks of emerging multidrug-resistant organisms (MDROs) relies on early outbreak detection. However, deploying regional containment is resource intensive; understanding the distribution of different types of outbreaks might aid in further classifying types of responses. Objective: We used a stochastic model of disease transmission in a region where healthcare facilities are linked by patient sharing to explore optimal strategies for early outbreak detection. Methods: We simulated the introduction and spread of Candida auris in a region using a lumped-parameter stochastic adaptation of a previously described deterministic model (Clin Infect Dis 2019 Mar 28. doi:10.1093/cid/ciz248). Stochasticity was incorporated to capture early-stage behavior of outbreaks with greater accuracy than was possible with a deterministic model. The model includes the real patient sharing network among healthcare facilities in an exemplary US state, using hospital claims data and the minimum data set from the CMS for 2015. Disease progression rates for C. auris were estimated from surveillance data and the literature. Each simulated outbreak was initiated with an importation to a Dartmouth Atlas of Health Care hospital referral region. To estimate the potential burden, we quantified the “facility-time” period during which infectious patients presented a risk of subsequent transmission within each healthcare facility. Results: Of the 28,000 simulated outbreaks initiated with an importation to the community, 2,534 resulted in patients entering the healthcare facility network. Among those, 2,480 (98%) initiated a short outbreak that died out or quickly attenuated within 2 years without additional intervention. In the simulations, if containment responses were initiated for each of those short outbreaks, facility time at risk decreased by only 3%. If containment responses were initiated for the 54 (2%) outbreaks lasting 2 years or longer, facility time at risk decreased by 79%. Sentinel surveillance through point-prevalence surveys (PPSs) at the 23 skilled-nursing facilities caring for ventilated patients (vSNF) in the network detected 50 (93%) of the 54 longer outbreaks (median, 235 days to detection). Quarterly PPSs at the 23 largest acute-care hospitals (ie, most discharges) detected 48 longer outbreaks (89%), but the time to detection was longer (median, 716 days to detection). Quarterly PPSs also identified 76 short-term outbreaks (in comparison to only 14 via vSNF PPS) that self-terminated without intervention. Conclusions: A vSNF-based sentinel surveillance system likely provides better information for guiding regional intervention for the containment of emerging MDROs than a similarly sized acute-care hospital–based system.

Funding: None

Disclosures: None

Distinguished Oral

Background: Shared Healthcare Intervention to Eliminate Life-threatening Dissemination of MDROs in Orange County, California (SHIELD OC) was a CDC-funded regional decolonization intervention from April 2017 through July 2019 involving 38 hospitals, nursing homes (NHs), and long-term acute-care hospitals (LTACHs) to reduce MDROs. Decolonization in NH and LTACHs consisted of universal antiseptic bathing with chlorhexidine (CHG) for routine bathing and showering plus nasal iodophor decolonization (Monday through Friday, twice daily every other week). Hospitals used universal CHG in ICUs and provided daily CHG and nasal iodophor to patients in contact precautions. We sought to evaluate whether decolonization reduced hospitalization and associated healthcare costs due to infections among residents of NHs participating in SHIELD compared to nonparticipating NHs. Methods: Medicaid insurer data covering NH residents in Orange County were used to calculate hospitalization rates due to a primary diagnosis of infection (counts per member quarter), hospital bed days/member-quarter, and expenditures/member quarter from the fourth quarter of 2015 to the second quarter of 2019. We used a time-series design and a segmented regression analysis to evaluate changes attributable to the SHIELD OC intervention among participating and nonparticipating NHs. Results: Across the SHIELD OC intervention period, intervention NHs experienced a 44% decrease in hospitalization rates, a 43% decrease in hospital bed days, and a 53% decrease in Medicaid expenditures when comparing the last quarter of the intervention to the baseline period (Fig. 1). These data translated to a significant downward slope, with a reduction of 4% per quarter in hospital admissions due to infection (P < .001), a reduction of 7% per quarter in hospitalization days due to infection (P < .001), and a reduction of 9% per quarter in Medicaid expenditures (P = .019) per NH resident. Conclusions: The universal CHG bathing and nasal decolonization intervention adopted by NHs in the SHIELD OC collaborative resulted in large, meaningful reductions in hospitalization events, hospitalization days, and healthcare expenditures among Medicaid-insured NH residents. The findings led CalOptima, the Medicaid provider in Orange County, California, to launch an NH incentive program that provides dedicated training and covers the cost of CHG and nasal iodophor for OC NHs that enroll.

Funding: None

Disclosures: Gabrielle M. Gussin, University of California, Irvine, Stryker (Sage Products): Conducting studies in which contributed antiseptic product is provided to participating hospitals and nursing homes. Clorox: Conducting studies in which contributed antiseptic product is provided to participating hospitals and nursing homes. Medline: Conducting studies in which contributed antiseptic product is provided to participating hospitals and nursing homes. Xttrium: Conducting studies in which contributed antiseptic product is provided to participating hospitals and nursing homes.

Background: Epidemiological studies have utilized administrative discharge diagnosis codes to identify methicillin-resistant and methicillin-sensitive Staphylococcus aureus (MRSA and MSSA) infections and trends, despite debate regarding the accuracy of utilizing codes for this purpose. We assessed the sensitivity and positive predictive value (PPV) of MRSA- and MSSA-specific diagnosis codes, trends, characteristics, and outcomes of S. aureus hospitalizations by method of identification. Methods: Clinical micro biology results and discharge data from geographically diverse US hospitals participating in the Premier Healthcare Database from 2012–2017 were used to identify monthly rates of MRSA and MSSA. Positive MRSA or MSSA clinical cultures and/or a MRSA- or MSSA-specific International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9/10 CM) diagnosis codes from adult inpatients (aged ≥18 years) were included as S. aureus hospitalizations. Septicemia was defined as a positive blood culture or a MRSA or MSSA septicemia code. Sensitivity and PPV for codes were calculated for hospitalizations where admission status was not listed as transfer; true infection was considered a positive clinical culture. Negative binominal regression models measured trends in rates of MRSA and MSSA per 1,000 hospital discharges. Results: We identified 168,634 MRSA and 148,776 MSSA hospitalizations in 256 hospitals; 17% of MRSA and 21% of MSSA were septicemia. Less than half of all S. aureus hospitalizations (49% MRSA, 46% MSSA) and S. aureus septicemia hospitalizations (37% MRSA, 38% MSSA) had both a positive culture and diagnosis code (Fig. 1). Sensitivity of MRSA codes in identifying positive cultures was 61% overall and 56% for septicemia, PPV was 62% overall and 53% for septicemia. MSSA codes had a sensitivity of 49% in identifying MSSA cultures and 52% for MSSA septicemia; PPV was 69% overall and 62% for septicemia. Despite low sensitivity, MRSA trends are similar for cultures and codes, and MSSA trends are divergent (Fig. 2). For hospitalizations with septicemia, mortality was highest among those with a blood culture only (31.3%) compared to hospitalizations with both a septicemia code and blood culture (16.6%), and septicemia code only (14.7%). Conclusions: ICD diagnosis code sensitivity and PPV for identifying infections were consistently poor in recent years. Less than half of hospitalizations have concordant microbiology laboratory results and diagnosis codes. Rates and trend estimates for MSSA differ by method of identification. Using diagnosis codes to identify S. aureus infections may not be appropriate for descriptive epidemiology or assessing trends due to significant misclassification.

Funding: None

Disclosures: Scott Fridkin reports that his spouse receives consulting fees from the vaccine industry.