Utilization Metrics Research Articles

SoK: Differentially Private Publication of Trajectory Data

Trajectory analysis holds many promises, from improvements in traffic management to routing advice or infrastructure development. However, learning users' paths is extremely privacy-invasive. Therefore, there is a necessity to protect trajectories such that we preserve the global properties, useful for analysis, while specific and private information of individuals remains inaccessible. Trajectories, however, are difficult to protect, since they are sequential, highly dimensional, correlated, bound to geophysical restrictions, and easily mapped to semantic points of interest. This paper aims to establish a systematic framework on protective masking and synthetic-generation measures for trajectory databases with syntactic and differentially private (DP) guarantees, including also utility properties, derived from ideas and limitations of existing proposals. To reach this goal, we systematize the utility metrics used throughout the literature, deeply analyze the DP granularity notions, explore and elaborate on the state of the art on privacy-enhancing mechanisms and their problems, and expose the main limitations of DP notions in the context of trajectories.

LDPTrace: Locally Differentially Private Trajectory Synthesis

Trajectory data has the potential to greatly benefit a wide-range of real-world applications, such as tracking the spread of the disease through people's movement patterns and providing personalized location-based services based on travel preference. However, privacy concerns and data protection regulations have limited the extent to which this data is shared and utilized. To overcome this challenge, local differential privacy provides a solution by allowing people to share a perturbed version of their data, ensuring privacy as only the data owners have access to the original information. Despite its potential, existing point-based perturbation mechanisms are not suitable for real-world scenarios due to poor utility, dependence on external knowledge, high computational overhead, and vulnerability to attacks. To address these limitations, we introduce LDPTrace, a novel locally differentially private trajectory synthesis framework. Our framework takes into account three crucial patterns inferred from users' trajectories in the local setting, allowing us to synthesize trajectories that closely resemble real ones with minimal computational cost. Additionally, we present a new method for selecting a proper grid granularity without compromising privacy. Our extensive experiments using real-world as well as synthetic data, various utility metrics and attacks, demonstrate the efficacy and efficiency of LDPTrace.

A broad spectrum of possibilities: Spectrum scores as a unifying metric of antibiotic utilization.

Days of therapy (DOT) currently serve as the standard antimicrobial utilization metric; however, by assigning the same weight to each agent rather than accounting for differences in spectrum of activity, DOT ignore key differences between agents that are fundamental to infectious diseases and critical to antimicrobial stewardship. Spectrum scoring assigns numeric values to individual antibiotic agents to quantify their spectrum of activity, allowing for normalization of antibiotic utilization data. When used in conjunction with traditional metrics, spectrum scores may offer further clarity to the picture of antibiotic utilization; however, issues related to development, application, and standardization of spectrum scores remain. Despite these challenges, the potential applications of spectrum scores are vast. Herein, we summarize existing data and explore the future of spectrum scoring, including application to both data analysis and routine patient care, use in inpatient and outpatient settings, integration within the electronic medical record, and opportunities for future research.

CE2RV: Commissioned energy‐efficient virtualization for large‐scale heterogeneous wireless sensor networks

SummaryThe Internet of Things (IoT) comprises sensor networks, intelligent things, devices, and humans for heterogeneous services and applications. Energy constraints in conventional wireless networks impact IoT performance resulting in service failures. For reducing the adverse impact of energy, this article introduces a commissioned energy‐efficient resource virtualization (CE2RV) scheme. This proposed scheme classifies the sensor nodes as fast and slow‐depleting for identifying service failures. The fast‐depleting nodes are discontinued from the service replications, and the remaining energy high‐node‐connected resources are identified. Such resources are virtualized for thwarting the existing energy failures over various services. The node classifications are performed using a tree‐learning algorithm. The classifications are performed for node replacement and service virtualization under different energy depletion rates. This is required for preventing sensor network disconnections between the users and service providers. The classification is required for overcoming multiple virtualizations between common nodes across different service providers. The proposed scheme's performance is analyzed using the metrics of service disconnections, energy utilization, energy efficiency, and service delay.

New Zealand (Aotearoa) clinicians' perspectives on the utility of the ICD-11 personality disorder diagnosis.

The ICD-11 has now taken effect and includes a new dimensional personality disorder (PD) diagnosis. The current study aimed to examine Aotearoa/New Zealand practitioners' perceptions of the clinical utility of the new PD system. A sample of 124 psychologists and psychiatrists completed a survey, applying the DSM-5 and ICD-11 PD diagnostic systems to a current patient, and completed clinical utility metrics on the DSM-5 and ICD-11 models. Additional open-ended questions further elicited clinicians' perceptions of the strengths, weaknesses and potential application issues of the ICD-11 PD diagnosis, and these responses were analysed through thematic analysis. Overall, the ICD-11 system was rated higher than the DSM-5 on all six clinical metrics, with no significant difference between psychologists' and psychiatrists' ratings. Five themes emerged: appreciation for an alternative to DSM-5, structural barriers preventing ICD-11 PD implementation, personal barriers to ICD-11 implementation, diagnoses viewed as low utility, clinician preference for formulation and cultural safety considerations for implementation of ICD-11 PD in Aotearoa/NZ. Overall, clinicians had positive opinions of the clinical utility of the ICD-11 PD diagnosis, although expressed some concerns about its implementation. The study expands upon initial evidence that mental health practitioners have generally positive perceptions of the ICD-11 PDs' clinical utility.

Postoperative Outcomes and Resource Utilization Following Open vs Endoscopic Far Lateral Lumbar Discectomy.

Operative approaches for far lateral disc herniation (FLDH) repair may be classified as open or minimally invasive. The present study aims to compare postoperative outcomes and resource utilization between patients undergoing open and endoscopic (one such minimally invasive approach) FLDH surgeries. A total of 144 consecutive adult patients undergoing FLDH repair at a single, university health system over an 8-year period (2013-2020) were retrospectively reviewed. Patients were divided into 2 cohorts: "open" (n = 92) and "endoscopic" (n = 52). Logistic regression was performed to evaluate the impact of procedural type on postoperative outcomes, and resource utilization metrics were compared between cohorts using χ 2 test (for categorical variables) or t test (for continuous variables). Primary postsurgical outcomes included readmissions, reoperations, emergency department visits, and neurosurgery outpatient office visits within 90 days of the index operation. Primary resource utilization outcomes included total direct cost of the procedure and length of stay. Secondary measures included discharge disposition, operative length, and duration of follow-up. No differences were observed in adverse postoperative events. Patients undergoing open FLDH surgery were more likely to attend outpatient visits within 30 days (P = 0.016). Although direct operating room cost was lower (P < 0.001) for open procedures, length of hospital stay was longer (P < 0.001). Patients undergoing open surgery also demonstrated less favorable discharge dispositions, longer operative length, and greater duration of follow-up. While both procedure types represent viable options for FLDH, endoscopic surgeries appear to achieve comparable clinical outcomes with decreased perioperative resource utilization. The present study suggests that endoscopic FLDH repairs do not lead to inferior outcomes but may decrease utilization of perioperative resources.

Impact of Performance Status on Survival Outcomes and Health Care Utilization in Patients With Advanced NSCLC Treated With Immune Checkpoint Inhibitors

IntroductionLandmark trials testing immune checkpoint inhibitors (ICIs) in advanced NSCLC are difficult to extrapolate to real-world practice given the exclusion of patients with poor (i.e., ≥2) Eastern Cooperative Oncology Group performance status (ECOG PS). We sought to evaluate the impact of ECOG PS on clinical outcomes and health care utilization in patients with NSCLC treated with ICIs in real-world practice. MethodsPatients with advanced NSCLC who received at least one dose of pembrolizumab or nivolumab were retrospectively identified from the Alberta Immunotherapy Database. The primary outcome was median overall survival, as stratified by ECOG PS. Secondary outcomes included median time-to-treatment failure and metrics of health care utilization, including emergency department visits, hospitalizations, and death in hospital. ResultsA total of 790 patients were included, with 29.2% having poor ECOG PS at initiation of ICI. These patients had significantly lower median overall survival (3.3 versus 13.4 mo) and median time-to-treatment failure (1.4 versus 4.9 mo) compared with those with favorable ECOG PS (p < 0.0001 for both outcomes). Patients with poor ECOG PS were also more likely to present to the emergency department, be admitted to the hospital, and die in the hospital during their first admission (risk ratio = 1.6, 2.3–2.7, p < 0.001). ConclusionsPatients with NSCLC with poor ECOG PS treated with ICI had significantly worse survival outcomes and were significantly more likely to use health care services than those with favorable ECOG PS. The large proportion of patients with poor ECOG PS further justifies the urgent need for randomized trials evaluating the efficacy of ICI in this high-risk population.

The Association Between Family Medicine Appointment Cancellations and Hospital Utilization in 2019 and 2020.

The COVID-19 pandemic caused a disruption in the usual primary care services offered and received by patients. The objective of this study was to compare the impact of family medicine appointment cancellations on hospital utilization metrics both before and during the COVID-19 pandemic within a family medicine residency clinic. This study is a retrospective chart review of cohorts of patients with a family medicine clinic cancellation who presented to the emergency department during a similar time period before and during the pandemic (March-May of 2019 vs March-May 2020). The patient population studied has multiple chronic diagnoses and prescriptions. Hospital admission, hospital readmission, and length of stay for hospitalizations during these periods were compared. The impacts of appointment cancellations on the emergency department presentation with subsequent inpatient admission, readmission, and length of stay were examined using generalized estimating equation (GEE) logistic or Poisson regression models to account for the lack of independence between patient outcomes. A total of 1878 patients were included in the final cohorts. Of these patients, 101 (5.7%) presented to the emergency department and/or hospital in both 2019 and 2020. An increased odds of readmission was associated with family medicine appointment cancellation regardless of year. The effects of appointment cancellations were not associated with admissions or length of stay between 2019 and 2020. Between the 2019 and 2020 cohorts, appointment cancellations were not associated with significant differences in likelihood of admission, readmission, or length of stay. A higher risk of readmission was associated with patients with a recent family medicine appointment cancellation.

Emergency Vehicle Driving Assistance System Using Recurrent Neural Network with Navigational Data Processing Method

Emergency vehicle transportation is important for responding to and transporting individuals during emergencies. This type of transportation faces several issues, such as road safety, navigation and communication, time-critical operations, resource utilisation, traffic congestion, data processing and analysis, and individual safety. Vehicle navigation and coordination is a critical aspect of emergency response that involves guiding emergency vehicles, such as ambulances, to the location of an emergency or medical centre as quickly and safely as possible. Therefore, it requires additional effort to reduce driving risks. The roadside units support emergency vehicles and infrastructure to decrease collisions and enhance optimal navigation routes. However, during the emergency vehicle’s data communication and navigation process, communication is interrupted due to vehicle outages. Therefore, this study proposes the Navigation Data Processing for Assisted Driving (NDP-AD) method to address the problem. The proposed approach assimilates infrastructure and neighbouring location information during driving. The integrated information is processed for distance and traffic during the previous displacement interval. The NDP-AD method employs a recurrent neural network learning approach to analyse opposing vehicle distance and traffic to provide accurate, independent guidance. This effective learning-based guidance process minimises false navigations and deviation in displacement. System efficiency is evaluated based on processing latency, displacement error, data utilisation, false rate, and accuracy metrics.

Outcomes for Hospitalized Aggressive and Violent Patients When Physical Restraints Are Introduced.

Hospital outcomes among patients acting aggressively or violently have not been extensively studied in the United States. The aims of the study are to determine rates of physical restraint use among hospitalized patients who are aggressive/violent and to characterize associations with mortality and utilization metrics. National Inpatient Sample databases from 2016 to 2019 were analyzed with multivariable regression to compare aggressive/violent patients in whom physical restraints were or were not used. Prevalence of physical restraint use, in-hospital mortality, length of stay, and total hospital charges were measures. A total of 40,309 hospitalized patients were coded as having aggressive/violent behavior, of whom 4475 (11.1%) were physically restrained. Physically restrained patients were younger (mean age ± standard error, 42.6 ± 0.64 versus 45.7± 0.41; P < 0.01), more frequently male (71.0% versus 65.4%; P < 0.01), and had less comorbidity (Charlson Index score >3: 7.9% versus 12.5%; P < 0.01) than unrestrained patients. Patients with physical restraints had higher odds of in-hospital mortality (adjusted odds ratio, 2.4, confidence interval [CI], 1.0-5.7; P = 0.04) and lower odds of being discharged to home (adjusted odds ratio, 0.46; CI, 0.38-0.56; P < 0.01) compared with unrestrained patients. Longer hospital stays (adjusted mean difference, 4.1 days CI, 2.1-6.0; P < 0.01) and higher hospitalization charges (adjusted mean difference, $16,996; CI, 6883-27,110; P < 0.01) were observed for those who were physically restrained. Physically restrained aggressive/violent patients had worse in-hospital outcomes compared with their unrestrained counterparts. Avoiding physical restraints whenever possible should be considered when managing this confrontational yet vulnerable patient population. When restraints are needed, providers must thoughtfully bear in mind heightened risks for worse outcomes.

A comparison of operative and anesthetic techniques for inguinal hernia repair in infants

BackgroundPediatric inguinal hernia repair (IHR) is increasingly performed using minimally invasive surgery (MIS) but has only recently been described using caudal block without endotracheal intubation. We evaluated the surgical outcomes and resource utilization of infants undergoing hernia repair, comparing both the operative approach (open/MIS) and anesthetic technique (general anesthesia [GA]/caudal). MethodsAll infants <1 year-of-age undergoing elective IHR without concomitant procedures from July 2016 to July 2021 at a single tertiary care teaching center were retrospectively reviewed. Eight surgeons and 25 anesthesiologists contributed patients, with approach dictated by practitioner preference.Data collected included patient demographics, surgical and anesthetic details, and operating room (OR) utilization metrics. Post-operative complications were evaluated and aggregated, including recurrent hernia, metachronous hernia, hematoma, hydrocele, testicular atrophy, and acquired cryptorchidism. Descriptive statistics were performed with R Studios (p < 0.05). ResultsOf the 338 patients included for analysis, most underwent an open procedure (n = 275) while anesthetic technique was evenly split between GA (n = 185) and caudal (n = 153). Most patients were male (87.6%) and born premature: mean gestational age of 31.4 ± 4.1 weeks. MIS-to-Open conversion was noted once (3.3%) in the GA MIS group, but none in caudal. Median follow up was 2.5 (1.4–3.8) years. No differences were noted in aggregate surgical complication rates (p = 0.4). The Caudal Open group had the shortest total OR time (p < 0.01); caudal anesthesia shortened post-procedure times (p < 0.01). ConclusionMIS IHR performed under caudal block and sedation yields comparable complication rates compared to the open approach or GA. Open IHR with caudal blockade was the most efficient operative room utilization. Type of StudyOriginal Article, Clinical Research. Levels of EvidenceLevel III.

Data-Driven Approaches to Hospital Capacity Planning and Management

Data-driven approaches to hospital capacity planning and management involve using historical and real-time data to identify patterns and trends in patient demand, resource utilization, and other key metrics. This information provides to develop predictive models, forecast patient demand, optimize staffing levels, and improve patient outcomes. Electronic health record systems and Internet of Things devices can also be used to monitor hospital operations in real-time and identify areas of inefficiency. Hospital capacity planning and management are critical to ensuring that healthcare facilities have enough resources to meet the needs of their patients. Data-driven approaches can be helpful in addressing these challenges by providing insights into patient demand, resource utilization, and other key metrics. This article discusses the various data-driven approaches to hospital capacity planning and management and their potential benefits. It also highlights the importance of having the right infrastructure and expertise effectively collect, analyse, and act on this data.

MEFood: A Large-Scale Representative Benchmark of Quotidian Foods for the Middle East

Automatic food recognition systems have been receiving increasing attention in the research community with the advancements in inductive learning (e.g., classification in computer vision) due to their applicability in the healthcare and hospitality industry. However, food recognition is challenging due to its fine-grained nature and its high correlation with culture, geo-location, and language. To make food recognition systems feasible for the Middle Eastern region, we present a large-scale dataset (MEFood) of commonly consumed food items in the Middle East, thereby providing a dataset for current development and establishing a benchmark for future research. We have also thoroughly examined the MEFood dataset highlighting its challenging aspects and its real-world nature. Additionally, we have conducted a thorough experimental study benchmarking the mainstream computer vision and mobile networks on classification, runtime, and resource utilization metrics. Our results highlight that EfficientNet-V2 achieves performance closer to the best-performing individual model on the MEFood dataset while having the least resource utilization and minimal inference times. Finally, we have performed a thorough error analysis study to glean additional insights about the networks and MEFood dataset.

Novel integration of governmental data sources using machine learning to identify super-utilization among U.S. counties

BackgroundSuper-utilizers consume the greatest share of resource intensive healthcare (RIHC) and reducing their utilization remains a crucial challenge to healthcare systems in the United States (U.S.). The objective of this study was to predict RIHC among U.S. counties, using routinely collected data from the U.S. government, including information on consumer spending, offering an alternative method for identifying super-utilization among population units rather than individuals. MethodsCross-sectional data from 5 governmental sources in 2017 were used in a machine learning pipeline, where target-prediction features were selected and used in 4 distinct algorithms. Outcome metrics of RIHC utilization came from the American Hospital Association and included yearly: (1) emergency rooms visit, (2) inpatient days, and (3) hospital expenditures. Target-prediction features included: 149 demographic characteristics from the U.S. Census Bureau, 151 adult and child health characteristics from the Centers for Disease Control and Prevention, 151 community characteristics from the American Community Survey, and 571 consumer expenditures from the Bureau of Labor Statistics. SHAP analysis identified important target-prediction features for 3 RIHC outcome metrics. Results2475 counties with emergency rooms and 2491 counties with hospitals were included. The median yearly emergency room visits per capita was 0.450 [IQR:0.318, 0.618], the median inpatient days per capita was 0.368 [IQR: 0.176, 0.826], and the median hospital expenditures per capita was $2104 [IQR: $1299.93, 3362.97]. The coefficient of determination (R2), calculated on the test set, ranged between 0.267 and 0.447. Demographic and community characteristics were among the important predictors for all 3 RIHC outcome metrics. ConclusionsIntegrating diverse population characteristics from numerous governmental sources, we predicted 3-outcome metrics of RIHC among U.S. counties with good performance, offering a novel and actionable tool for identifying super-utilizer segments in the population. Wider integration of routinely collected data can be used to develop alternative methods for predicting RIHC among population units.

AI in Healthcare Management and Accounting: Novelties and Trends from a Literature Review and Illustrative Cases

Through a mixed-methods approach based on a critical analysis of both literature and industry trends (referred to real-world practices / illustrative cases), as rapidly emerging in recent years, we investigate the transformative impact of Artificial Intelligence on healthcare sector, notably the powerful intersection – still too little explored by extant scientific doctrine – of health management and accounting system AI related. As the technological advancements continue to reshape the healthcare landscape, we found an AI expanding as a pivotal instrument in managing, innovating, and optimising strategic and operational processes (by enhancing executives’ decision-making). We drew as main effect of such a process – when enabled by a fruitful integration of managers and experts (data scientists / IT professionals) – an improvement of both patient care and financial efficiency as well (i.e., overall healthcare delivery in order to generate business value) thanks to a more profound real-time KPIs monitoring (patient outcomes, resource utilisation, and financial metrics). We also critically review new challenges such as data privacy, security, and ethical considerations emphasising the importance of responsible AI deployment for building patient trust and ensuring regulatory compliance. In this regard, a public awareness of AI and the establishment of standardised guidelines will be additionally required. By delving into the novel applications of AI in healthcare administration, financial management, and synergies between these domains, this work can benefit a diverse audience across various sectors of the healthcare industry, technology, public administration, and academia. Limitations and future directions of research are provided.

Racial disparities in arthroscopic rotator cuff repair: an analysis of utilization and perioperative outcomes.

There remains a paucity of literature addressing racial disparities in utilization and perioperative metrics in arthroscopic rotator cuff repair procedures. The American College of Surgeons National Surgical Quality Improvement Program database was used to evaluate patients undergoing arthroscopic rotator cuff repair from 2010 to 2019. Baseline demographics, utilization trends, and perioperative measures, including adverse events, operative time, length of hospital stay, days from operation to discharge, and readmission, were analyzed. Of 42,443 included patients, 38,090 (89.7%) were White, and 4353 (10.3%) were Black or African American. Black or African American patients had a significantly higher percentage of diabetes mellitus (23.6% vs. 15.6%), smoking (16.9% vs. 14.8%), congestive heart failure (0.3% vs. 0.1%), and hypertension (59.2% vs. 45.9%). In addition, logistic regression showed that Black or African American patients had increased odds of longer operative time (adjusted rate ratio 1.07, 95% confidence interval 1.05-1.08) and time from operation to discharge (adjusted rate ratio 1.19, 95% confidence interval 1.04-1.37). Disparities in relative utilization decreased as the proportion of Black or African American patients undergoing arthroscopic rotator cuff repair increased (7.4% in 2010 vs. 10.4% in 2019) compared with White patients (P trend < .0001). Racial disparities exist regarding baseline comorbidities and perioperative metrics in arthroscopic rotator cuff repair. Further investigation is needed to fully understand and address the causes of these inequalities to provide equitable care.

SPACE: Segment Prefetching and Caching at the Edge for Adaptive Video Streaming

Multi-access Edge Computing (MEC) is a new paradigm that brings storage and computing close to the clients. MEC enables the deployment of complex network-assisted mechanisms for video streaming that improve clients’ Quality of Experience (QoE). One of these mechanisms is segment prefetching, which transmits the future video segments in advance closer to the client to serve content with lower latency. In this work, for HAS-based (HTTP Adaptive Streaming) video streaming and specifically considering a cellular ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g</i> ., 5G) network edge, we present our approach <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Segment Prefetching and Caching at the Edge for Adaptive Video Streaming (SPACE)</i> . We propose and analyze different segment prefetching policies that differ in resource utilization, player and radio metrics needed, and deployment complexity. This variety of policies can dynamically adapt to the network’s current conditions and the service provider’s needs. We present segment prefetching policies based on diverse approaches and techniques: past segment requests, segment transrating ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e</i> ., reducing segment bitrate/quality), Markov prediction model, machine learning to predict future segment requests, and super-resolution.We study their performance and feasibility using metrics such as QoE characteristics, computing times, prefetching hits, and link bitrate consumption. We analyze and discuss which segment prefetching policy is better under which circumstances, as well as the influence of the client-side Adaptive Bit Rate (ABR) algorithm and the set of available representations (“bitrate ladder”) in segment prefetching. Moreover, we examine the impact on segment prefetching of different caching policies for (pre-)fetched segments, including Least Recently Used (LRU), Least Frequently Used (LFU), and our proposed popularity-based caching policy Least Popular Used (LPU).

A trusted IoT data sharing and secure oracle based access for agricultural production risk management

Agricultural risks associated with weather events, soil conditions, diseases, and pests have risen in recent years due to climate change. These risks have burdened farmers with the shock of financial stress and endangered the region’s food security. Despite increasing risks, notable risk management tools like agricultural insurance have not effectively reached marginal farmers in developing economies. This can be attributed to the fact that existing centralized agricultural information systems lack trust factors in sharing and accessing agricultural risk data, which is accompanied by processing delays in insurance payouts. To this end, the emergence of promising technologies like blockchain helps establish trust and automates insurance-based payments with smart contracts. This work proposes a framework called AgriSSIOracle with two key contributions. The first is to have a trusted agricultural Internet of Things (IoT) data sharing that employs blockchain-based Self-Sovereign Identity (SSI) and provable credentials. Second is the decentralized oracle-based access control mechanism for smart contracts in agricultural insurance. A method for privacy-preserved authentication and data provenance for agricultural IoT devices is developed with SSI-based Decentralized Identifiers (DIDs) and Verifiable Credentials (VCs). Next, the IoT data sharing scheme is illustrated using publish–subscribe and request–response communication patterns based on the device data credentials. Finally, the decentralized oracle mechanism for smart contracts is built based on multiparty computation for risk-related data access. The implementation of the framework is realized using a permissioned identity blockchain called Hyperledger Indy and Ethereum, which is a public blockchain for smart contracts. The security and privacy analysis confirms that the AgriSSIOracle framework ensures trust factors, particularly authenticity, privacy, data provenance, data integrity, and access control. The framework is evaluated with respect to the transaction throughput, latency, credential issue-verify-revoke time, and resource utilization metrics. The evaluation results prove that the AgriSSIOracle solution is scalable and efficient and meets the requirements of real-time deployment.

911. Impact of Intensified Clindamycin Stewardship Initiatives in Three Phases: A Quasi-Experimental Study

Abstract Background Clindamycin utilization has provoked gram-positive and anaerobic bacterial resistance and is associated with Clostridioides difficile infections (CDI). The purpose of this study was to evaluate if implementing local clindamycin-focused stewardship initiatives impacts clindamycin utilization and outcome metrics. Methods This multicenter, retrospective quasi-experimental study was conducted in adult hospitalized patients who received clindamycin from 2018 to 2020. Outcomes were compared in three phases between an intervention and control group (Figure 1). The primary outcome was inappropriate utilization: a resistant pathogen, concurrent antibiotic(s) with pathogen activity, a non-necrotizing infection, an alternative agent was preferred, or alternative reasons per expert discretion. Secondary outcomes included clindamycin duration of therapy per 1000 patient hospital days (DOT/1000 PD), 30-day CDI, 30-day readmission, and in-hospital mortality. Outcomes were compared between groups using the chi-square, Fisher’s exact, or Kruskal-Wallis tests as appropriate. Figure 1.Study DesignID: infectious diseases, I: intervention group, C: control group Results The study included 481 patients (Table 1). Inappropriate clindamycin use in the intervention group was highest in Phase 1 (94%), then decreased in Phase 2 (72%) and Phase 3 (74%) (p&amp;lt; 0.01). Comparing Phase 1, Phase 2, and Phase 3 between the intervention and control groups, respectively, the primary outcome occurred in 94% vs 93% (p=0.80), 72% vs 88% (p=0.02), and 74% vs 89% (p=0.03). The DOT/1000 PD was 10 in Phase 1, 9.2 in Phase 2, and 6.2 in Phase 3. Initiatives were not associated with significant reductions in 30-day CDI (1% in Phase 1, 0% in Phase 2, 0% in Phase 3, p=0.64), 30-day readmission (18% in Phase 1, 31% in Phase 2, 22% in Phase 3, p=0.13), or in-hospital mortality (0% in Phase 1, 2% in Phase 2, 0% in Phase 3, p=0.33). Table 1.OutcomesI: intervention group, C: control group, DOT/1000 PD: days of therapy per 1000 patient days, CDI: Clostridioides difficile infection Conclusion Clindamycin-focused stewardship initiatives reduced inappropriate prescribing patterns by approximately 20% but still remained high in Phase 3. Clindamycin was primarily ordered in the emergency department (ED). Additional initiatives may include formal criteria for use and removing clindamycin from automated dispensing cabinets in the ED. Our initiatives may serve as a model for other institutions to reduce clindamycin utilization and its associated complications. Disclosures All Authors: No reported disclosures.

A Multifaceted benchmarking of synthetic electronic health record generation models.

Synthetic health data have the potential to mitigate privacy concerns in supporting biomedical research and healthcare applications. Modern approaches for data generation continue to evolve and demonstrate remarkable potential. Yet there is a lack of a systematic assessment framework to benchmark methods as they emerge and determine which methods are most appropriate for which use cases. In this work, we introduce a systematic benchmarking framework to appraise key characteristics with respect to utility and privacy metrics. We apply the framework to evaluate synthetic data generation methods for electronic health records data from two large academic medical centers with respect to several use cases. The results illustrate that there is a utility-privacy tradeoff for sharing synthetic health data and further indicate that no method is unequivocally the best on all criteria in each use case, which makes it evident why synthetic data generation methods need to be assessed in context.