Importance:The rise of patient messages sent to clinicians via a patient portal has directly led to physician burnout and dissatisfaction, prompting uptake of artificial intelligence (AI) to alleviate this burden. It is important to understand patient preferences around AI in patient-clinician communication as ethical guidelines on appropriate use and disclosure (patient notification of AI use) are developed. Objective:To analyze patient preferences regarding use of AI in electronic messages. Design, Setting, and Participants:A survey study was conducted within the Duke University Health System's patient advisory committee, consisting of individuals 18 years or older who participate in periodic surveys to inform health system patient care practices. Multiple surveys were administered to test the impact of different factors, including response author, disclosure (AI, human, or none), and seriousness of the topic. A follow-up survey assessed preferred disclosure verbiage. Surveys were administered from October 31 to December 11, 2023. Exposure:Multiple surveys. Main Outcomes and Measures:Participants rated their overall satisfaction, usefulness of the information, and perceived level of care on a 5-point Likert scale. Results:Of the 2511 members surveyed, 1455 (57.9%) responded, with respondents being older (median age, 57 [IQR, 49-70] vs 53 [IQR, 41-62] years), more educated (872 of 1083 [80.5%] vs 319 of 440 [72.5%] with a college or graduate degree), and predominantly female (921 [63.3%]). Participants preferred AI- compared with human-drafted responses, with a mean difference for satisfaction of -0.30 (95% CI, -0.37 to -0.23) points, usefulness of -0.28 (95% CI, -0.34 to -0.22) points, and perception they were cared for of -0.43 (95% CI, -0.50 to -0.37) points. Participants tended to have higher satisfaction with a human disclosure over AI disclosure, with a mean difference of 0.13 (95% CI, 0.05-0.22) points, and with no disclosure over AI authorship disclosure, with a mean difference of 0.09 (95% CI, 0.01-0.17) points. Regardless of author or disclosure type, more than 75% of respondents were satisfied (agree or strongly agree) with the response. Conclusions and Relevance:In this survey study, participants expressed a mild preference for messages written by AI but had a slightly decreased satisfaction when told AI was involved. Patient experience must be considered along with ethical implementation of AI. Although AI disclosure may slightly reduce satisfaction, disclosure should be maintained to uphold patient autonomy and empowerment.

BACKGROUND:The US healthcare system is currently facing significant challenges in quality, affordability, and labor shortages. Artificial intelligence (AI) promises to transform healthcare delivery by making it safer, more effective, less wasteful, and more patient-centered. With more than $30 billion invested in healthcare AI companies in the past three years, the proliferation of AI solutions is expected to bring much-needed relief to the strained healthcare industry. To harness the current enthusiasm for AI in healthcare, we can draw parallels to the adoption of electronic health records (EHRs) under the HITECH Act of 2009. EHR adoption has been widespread and has contributed to significant health information technology spending, but it has also brought unintended consequences, such as clinician burnout, workarounds, and mixed impacts on patient safety and quality measures. THE EHR ERA VS. THE AI ERA:DIFFERENCES: This article grounds the discussion by first reviewing the key differences between the EHR implementation era that followed the passage of HITECH and the current AI era. The authors identified three characteristics of the AI era that distinguish it from the EHR implementation era: different regulatory and legislative context, diminished capacity of the workforce to absorb new work, and an accelerated pace of change. LESSONS FROM EHR IMPLEMENTATION TO CARRY FORWARD TO AI IMPLEMENTATION: Based on the collective experience of the authorship team and published literature on EHR and AI implementation, the authors identified five critical lessons from the EHR implementation era that organizations deploying AI must consider: (1) respect the human element, (2) build strong organizational governance, (3) adapt leadership and culture, (4) ready the workforce, and (5) build for the long term. CONCLUSION:By applying these lessons, organizational leaders can realize the potential of AI to improve patient outcomes and transform healthcare delivery.

OBJECTIVES:Large language models (LLMs) are increasingly utilized in healthcare, transforming medical practice through advanced language processing capabilities. However, the evaluation of LLMs predominantly relies on human qualitative assessment, which is time-consuming, resource-intensive, and may be subject to variability and bias. There is a pressing need for quantitative metrics to enable scalable, objective, and efficient evaluation. MATERIALS AND METHODS:We propose a unified evaluation framework that bridges qualitative and quantitative methods to assess LLM performance in healthcare settings. This framework maps evaluation aspects-such as linguistic quality, efficiency, content integrity, trustworthiness, and usefulness-to both qualitative assessments and quantitative metrics. We apply our approach to empirically evaluate the Epic In-Basket feature, which uses LLM to generate patient message replies. RESULTS:The empirical evaluation demonstrates that while Artificial Intelligence (AI)-generated replies exhibit high fluency, clarity, and minimal toxicity, they face challenges with coherence and completeness. Clinicians' manual decision to use AI-generated drafts correlates strongly with quantitative metrics, suggesting that quantitative metrics have the potential to reduce human effort in the evaluation process and make it more scalable. DISCUSSION:Our study highlights the potential of a unified evaluation framework that integrates qualitative and quantitative methods, enabling scalable and systematic assessments of LLMs in healthcare. Automated metrics streamline evaluation and monitoring processes, but their effective use depends on alignment with human judgment, particularly for aspects requiring contextual interpretation. As LLM applications expand, refining evaluation strategies and fostering interdisciplinary collaboration will be critical to maintaining high standards of accuracy, ethics, and regulatory compliance. CONCLUSION:Our unified evaluation framework bridges the gap between qualitative human assessments and automated quantitative metrics, enhancing the reliability and scalability of LLM evaluations in healthcare. While automated quantitative evaluations are not ready to fully replace qualitative human evaluations, they can be used to enhance the process and, with relevant benchmarks derived from the unified framework proposed here, they can be applied to LLM monitoring and evaluation of updated versions of the original technology evaluated using qualitative human standards.

IMPORTANCE:The US healthcare system faces significant challenges, including clinician burnout, operational inefficiencies, and concerns about patient safety. Artificial intelligence (AI), particularly generative AI, has the potential to address these challenges, but its adoption, effectiveness, and barriers to implementation are not well understood. OBJECTIVE:To evaluate the current state of AI adoption in US healthcare systems, assess successes and barriers to implementation during the early generative AI era. DESIGN, SETTING, AND PARTICIPANTS:This cross-sectional survey was conducted in Fall 2024, and included 67 health systems members of the Scottsdale Institute, a collaborative of US non-profit healthcare organizations. Forty-three health systems completed the survey (64% response rate). Respondents provided data on the deployment status and perceived success of 37 AI use cases across 10 categories. MAIN OUTCOMES AND MEASURES:The primary outcomes were the extent of AI use case development, piloting, or deployment, the degree of reported success for AI use cases, and the most significant barriers to adoption. RESULTS:Across the 43 responding health systems, AI adoption and perceptions of success varied significantly. Ambient Notes, a generative AI tool for clinical documentation, was the only use case with 100% of respondents reporting adoption activities, and 53% reported a high degree of success with using AI for Clinical Documentation. Imaging and radiology emerged as the most widely deployed clinical AI use case, with 90% of organizations reporting at least partial deployment, although successes with diagnostic use cases were limited. Similarly, many organizations have deployed AI for clinical risk stratification such as early sepsis detection, but only 38% report high success in this area. Immature AI tools were identified a significant barrier to adoption, cited by 77% of respondents, followed by financial concerns (47%) and regulatory uncertainty (40%). CONCLUSIONS AND RELEVANCE:Ambient Notes is rapidly advancing in US healthcare systems and demonstrating early success. Other AI use cases show varying degrees of adoption and success, constrained by barriers such as immature AI tools, financial concerns, and regulatory uncertainty. Addressing these challenges through robust evaluations, shared strategies, and governance models will be essential to ensure effective integration and adoption of AI into healthcare practice.

Ambient digital scribing (ADS) tools alleviate clinician documentation burden, reducing burnout and enhancing efficiency. As AI-driven ADS tools integrate into clinical workflows, robust governance is essential for ethical and secure deployment. This study proposes a comprehensive ADS evaluation framework incorporating human evaluation, automated metrics, simulation testing, and large language models (LLMs) as evaluators. Our framework assesses transcription, diarization, and medical note generation across criteria such as fluency, completeness, and factuality. To demonstrate its effectiveness, we developed an ADS tool and applied our framework to evaluate the tool's performance on 40 real clinical visit recordings. Our evaluation revealed strengths, such as fluency and clarity, but also highlighted weaknesses in factual accuracy and the ability to capture new medications. These findings underscore the value of structured ADS evaluation in improving healthcare delivery while emphasizing the need for strong governance to ensure safe, ethical integration.

OBJECTIVE:The complexity and rapid pace of development of algorithmic technologies pose challenges for their regulation and oversight in healthcare settings. We sought to improve our institution's approach to evaluation and governance of algorithmic technologies used in clinical care and operations by creating an Implementation Guide that standardizes evaluation criteria so that local oversight is performed in an objective fashion. MATERIALS AND METHODS:Building on a framework that applies key ethical and quality principles (clinical value and safety, fairness and equity, usability and adoption, transparency and accountability, and regulatory compliance), we created concrete guidelines for evaluating algorithmic technologies at our institution. RESULTS:An Implementation Guide articulates evaluation criteria used during review of algorithmic technologies and details what evidence supports the implementation of ethical and quality principles for trustworthy health AI. Application of the processes described in the Implementation Guide can lead to algorithms that are safer as well as more effective, fair, and equitable upon implementation, as illustrated through 4 examples of technologies at different phases of the algorithmic lifecycle that underwent evaluation at our academic medical center. DISCUSSION:By providing clear descriptions/definitions of evaluation criteria and embedding them within standardized processes, we streamlined oversight processes and educated communities using and developing algorithmic technologies within our institution. CONCLUSIONS:We developed a scalable, adaptable framework for translating principles into evaluation criteria and specific requirements that support trustworthy implementation of algorithmic technologies in patient care and healthcare operations.

Introduction Despite a higher risk of severe COVID-19 disease in individuals with HIV, the interactions between SARS-CoV-2 and HIV infections remain unclear. To delineate these interactions, multicentre Electronic Health Records (EHR) hold existing promise to provide full-spectrum and longitudinal clinical data, demographics and sociobehavioural data at individual level. Presently, a comprehensive EHR-based cohort for the HIV/SARS-CoV-2 coinfection has not been established; EHR integration and data mining methods tailored for studying the coinfection are urgently needed yet remain underdeveloped. Methods and analysis The overarching goal of this exploratory/developmental study is to establish an EHR-based cohort for individuals with HIV/SARS-CoV-2 coinfection and perform large-scale EHR-based data mining to examine the interactions between HIV and SARS-CoV-2 infections and systematically identify and validate factors contributing to the severe clinical course of the coinfection. We will use a nationwide EHR database in the USA, namely, National COVID Cohort Collaborative (N3C). Ultimately, collected clinical evidence will be implemented and used to pilot test a clinical decision support prototype to assist providers in screening and referral of at-risk patients in real-world clinics. Ethics and dissemination The study was approved by the institutional review boards at the University of South Carolina (Pro00121828) as non-human subject study. Study findings will be presented at academic conferences and published in peer-reviewed journals. This study will disseminate urgently needed clinical evidence for guiding clinical practice for individuals with the coinfection at Prisma Health, a healthcare system in collaboration.

Artificial intelligence/machine learning models are being rapidly developed and used in clinical practice. However, many models are deployed without a clear understanding of clinical or operational impact and frequently lack monitoring plans that can detect potential safety signals. There is a lack of consensus in establishing governance to deploy, pilot, and monitor algorithms within operational healthcare delivery workflows. Here, we describe a governance framework that combines current regulatory best practices and lifecycle management of predictive models being used for clinical care. Since January 2021, we have successfully added models to our governance portfolio and are currently managing 52 models.

Duke Health Technology Solutions, Duke University Health System, Durham, North Carolina, USA, Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA, Department of Biomedical Informatics, Vanderbilt University School of Medicine, Nashville, Tennessee, USA, Department of Pediatrics & Department of Internal Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA, and Department of Informatics, University of California, Irvine, Irvine, California, USA

The novel coronavirus disease-19 (COVID-19) pandemic has altered our economy, society, and healthcare system. While this crisis has presented the U.S. healthcare delivery system with unprecedented challenges, the pandemic has catalyzed rapid adoption of telehealth, or the entire spectrum of activities used to deliver care at a distance. Using examples reported by U.S. healthcare organizations, including ours, we describe the role that telehealth has played in transforming healthcare delivery during the 3 phases of the U.S. COVID-19 pandemic: (1) stay-at-home outpatient care, (2) initial COVID-19 hospital surge, and (3) postpandemic recovery. Within each of these 3 phases, we examine how people, process, and technology work together to support a successful telehealth transformation. Whether healthcare enterprises are ready or not, the new reality is that virtual care has arrived.

Unplanned hospital readmissions represent a significant health care value problem with high costs and poor quality of care. A significant percentage of readmissions could be prevented if clinical inpatient teams were better able to predict which patients were at higher risk for readmission. Many of the current clinical decision support models that predict readmissions are not configured to integrate closely with the electronic health record or alert providers in real-time prior to discharge about a patient’s risk for readmission. We report on the implementation and monitoring of the Epic electronic health record—“Unplanned readmission model version 1”—over 2 years from 1/1/2018–12/31/2019. For patients discharged during this time, the predictive capability to discern high risk discharges was reflected in an AUC/C-statistic at our three hospitals of 0.716–0.760 for all patients and 0.676–0.695 for general medicine patients. The model had a positive predictive value ranging from 0.217–0.248 for all patients. We also present our methods in monitoring the model over time for trend changes, as well as common readmissions reduction strategies triggered by the score.

BACKGROUND Successful integrations of machine learning into routine clinical care are exceedingly rare, and barriers to its adoption are poorly characterized in the literature. OBJECTIVE This study aims to report a quality improvement effort to integrate a deep learning sepsis detection and management platform, Sepsis Watch, into routine clinical care. METHODS In 2016, a multidisciplinary team consisting of statisticians, data scientists, data engineers, and clinicians was assembled by the leadership of an academic health system to radically improve the detection and treatment of sepsis. This report of the quality improvement effort follows the learning health system framework to describe the problem assessment, design, development, implementation, and evaluation plan of Sepsis Watch. RESULTS Sepsis Watch was successfully integrated into routine clinical care and reshaped how local machine learning projects are executed. Frontline clinical staff were highly engaged in the design and development of the workflow, machine learning model, and application. Novel machine learning methods were developed to detect sepsis early, and implementation of the model required robust infrastructure. Significant investment was required to align stakeholders, develop trusting relationships, define roles and responsibilities, and to train frontline staff, leading to the establishment of 3 partnerships with internal and external research groups to evaluate Sepsis Watch. CONCLUSIONS Machine learning models are commonly developed to enhance clinical decision making, but successful integrations of machine learning into routine clinical care are rare. Although there is no playbook for integrating deep learning into clinical care, learnings from the Sepsis Watch integration can inform efforts to develop machine learning technologies at other health care delivery systems.

There is little known about how academic medical centers (AMCs) in the US develop, implement, and maintain predictive modeling and machine learning (PM and ML) models. We conducted semi-structured interviews with leaders from AMCs to assess their use of PM and ML in clinical care, understand associated challenges, and determine recommended best practices. Each transcribed interview was iteratively coded and reconciled by a minimum of 2 investigators to identify key barriers to and facilitators of PM and ML adoption and implementation in clinical care. Interviews were conducted with 33 individuals from 19 AMCs nationally. AMCs varied greatly in the use of PM and ML within clinical care, from some just beginning to explore their utility to others with multiple models integrated into clinical care. Informants identified 5 key barriers to the adoption and implementation of PM and ML in clinical care: (1) culture and personnel, (2) clinical utility of the PM and ML tool, (3) financing, (4) technology, and (5) data. Recommendation to the informatics community to overcome these barriers included: (1) development of robust evaluation methodologies, (2) partnership with vendors, and (3) development and dissemination of best practices. For institutions developing clinical PM and ML applications, they are advised to: (1) develop appropriate governance, (2) strengthen data access, integrity, and provenance, and (3) adhere to the 5 rights of clinical decision support. This article highlights key challenges of implementing PM and ML in clinical care at AMCs and suggests best practices for development, implementation, and maintenance at these institutions.

OBJECTIVE While electronic health record (EHR) systems store copious amounts of patient data, aggregating those data across patients can be challenging. Visual analytic tools that integrate with EHR systems allow clinicians to gain better insight and understanding into clinical care and management. We report on our experience building Tableau-based visualizations and integrating them into our EHR system. MATERIALS AND METHODS Visual analytic tools were created as part of 12 clinician-initiated quality improvement projects. We built the visual analytic tools in Tableau and linked it within our EPIC environment. We identified 5 visual themes that spanned the various projects. To illustrate these themes, we choose 1 exemplary project which aimed to improve obstetric operating room efficiency. RESULTS Across our 12 projects, we identified 5 visual themes that are integral to project success: scheduling & optimization (in 11/12 projects); provider assessment (10/12); executive assessment (8/12); patient outcomes (7/12); and control and goal charts (2/12). DISCUSSION Many visualizations share common themes. Identification of these themes has allowed our internal team to be more efficient and directed in developing visualizations for future projects. CONCLUSION Organizing visual analytics into themes can allow informatics teams to more efficiently provide visual products to clinical collaborators.

Background Health systems often employ interruptive alerts through the electronic health record to improve patient care. However, concerns of "alert fatigue" have been raised, highlighting the importance of understanding the time burden and impact of these alerts on providers. Objectives Our main objective was to determine the total time providers spent on interruptive alerts in both inpatient and outpatient settings. Our secondary objectives were to analyze dwell time for individual alerts and examine both provider and alert-related factors associated with dwell time variance. Methods We retrospectively evaluated use and response to the 75 most common interruptive ("popup") alerts between June 1st, 2015 and November 1st, 2016 in a large academic health care system. Alert "dwell time" was calculated as the time between the alert appearing on a provider's screen until it was closed. The total number of alerts and dwell times per provider per month was calculated for inpatient and outpatient alerts and compared across alert type. Results The median number of alerts seen by a provider was 12 per month (IQR 4-34). Overall, 67% of inpatient and 39% of outpatient alerts were closed in under 3 seconds. Alerts related to patient safety and those requiring more than a single click to proceed had significantly longer median dwell times of 5.2 and 6.7 seconds, respectively. The median total monthly time spent by providers viewing alerts was 49 seconds on inpatient alerts and 28 seconds on outpatient alerts. Conclusion Most alerts were closed in under 3 seconds and a provider's total time spent on alerts was less than 1 min/mo. Alert fatigue may lie in their interruptive and noncritical nature rather than time burden. Monitoring alert interaction time can function as a valuable metric to assess the impact of alerts on workflow and potentially identify routinely ignored alerts.

Objective As available data increases, so does the opportunity to develop risk scores on more refined patient populations. In this paper we assessed the ability to derive a risk score for a patient no-showing to a clinic visit. Methods Using data from 2 264 235 outpatient appointments we assessed the performance of models built across 14 different specialties and 55 clinics. We used regularized logistic regression models to fit and assess models built on the health system, specialty, and clinic levels. We evaluated fits based on their discrimination and calibration. Results Overall, the results suggest that a relatively robust risk score for patient no-shows could be derived with an average C-statistic of 0.83 across clinic level models and strong calibration. Moreover, the clinic specific models, even with lower training set sizes, often performed better than the more general models. Examination of the individual models showed that risk factors had different degrees of predictability across the different specialties. Implementation of optimal modeling strategies would lead to capturing an additional 4819 no-shows per-year. Conclusion Overall, this work highlights both the opportunity for and the importance of leveraging the available electronic health record data to develop more refined risk models.

SummaryObjective: To understand emergency department (ED) physicians’ use of electronic documentation in order to identify usability and workflow considerations for the design of future ED information system (EDIS) physician documentation modules.Methods: We invited emergency medicine resident physicians to participate in a mixed methods study using task analysis and qualitative interviews. Participants completed a simulated, standardized patient encounter in a medical simulation center while documenting in the test environment of a currently used EDIS. We recorded the time on task, type and sequence of tasks performed by the participants (including tasks performed in parallel). We then conducted semi-structured interviews with each participant. We analyzed these qualitative data using the constant comparative method to generate themes.Results: Eight resident physicians participated. The simulation session averaged 17 minutes and participants spent 11 minutes on average on tasks that included electronic documentation. Participants performed tasks in parallel, such as history taking and electronic documentation. Five of the 8 participants performed a similar workflow sequence during the first part of the session while the remaining three used different workflows. Three themes characterize electronic documentation: (1) physicians report that location and timing of documentation varies based on patient acuity and workload, (2) physicians report a need for features that support improved efficiency; and (3) physicians like viewing available patient data but struggle with integration of the EDIS with other information sources.Conclusion: We confirmed that physicians spend much of their time on documentation (65%) during an ED patient visit. Further, we found that resident physicians did not all use the same work-flow and approach even when presented with an identical standardized patient scenario. Future EHR design should consider these varied workflows while trying to optimize efficiency, such as improving integration of clinical data. These findings should be tested quantitatively in a larger, representative study.Citation: Neri PM, Redden L, Poole S, Pozner CN, Horsky J, Raja AS, Poon E, Schiff G, Landman A. Emergency medicine resident physicians’ perceptions of electronic documentation and workflow – a mixed methods study. Appl Clin Inf 2015; 6: 27–41http://dx.doi.org/10.4338/ACI-2014-08-RA-0065

The adoption of electronic health records (EHRs) in US hospitals continues to rise steeply, with nearly 60% of all hospitals having at least a basic EHR as of 2014. EHRs bring with them the ability to inform and guide clinicians as they make decisions. In theory, this form of clinical decision support (CDS) ensures quality of care, reduces adverse events, and improves efficiency; in practice, experience in the field paints a mixed picture. This issue of the Journal of Hospital Medicine presents 3 examples of CDS that illustrate the distance between what we see as CDS’ full potential and current limitations. In the study by Herzig et al. investigators took on the challenge of implementing stress ulcer prophylaxis guidelines developed by the Society of Hospital Medicine. The investigators first demonstrated that targeted electronic prompts captured patients’ indications for acid suppressive therapy, and could be used to prohibit prescribers from ordering acid suppressive therapy among patients outside the intensive care unit (ICU) setting. Through an elegant interrupted time series study design deployed across 2 hospital campuses, the investigators were able to demonstrate immediate and clinically significant reduction in acid suppressive therapy outside the ICU. They further found that the impact of this reduction was augmented over time, suggesting that the electronic prompts had a sustained impact on provider ordering behavior. However, below the headline—and relevant to the limitations of CDS— the investigators noted that much of the reduction in the use of acid suppressive therapy for stress ulcer prophylaxis could be accounted for by providers’ choice of another acceptable indication (eg, continuing preadmission medication). The authors speculated that the CDS intervention prompted providers to more accurately record the indication for acid suppressive therapy. It is also possible that providers simply chose an alternate indication to circumvent the decision-support step. Perhaps as a result of these 2 offsetting factors, the actual use of acid suppressive therapy, regardless of indication, only decreased in a modest and statistically nonsignificant way, casting the true effectiveness of this CDS intervention into question. Two other studies in this issue of the Journal of Hospital Medicine provide valuable insights into interactions between social and technical factors that determine the success or failure in the use of technology such as CDS to drive organizational performance. At the technical end of this sociotechnical spectrum, the study by Knight et al. illustrated that a minimally configured and visually unintuitive medication decisionsupport system resulted in a high number of alerts (approximately 17% of studied orders), leading to the well-reported phenomena of alert fatigue and substantially lower response rate compared to those reported in the literature. Moreover, the analysis suggested that response to these alerts were particularly muted among situations that were particularly high risk, including the patient being older, patient having a greater length of stay, care being delivered in the internal medicine service, resident physician being the prescriber, and the medication being on the Institute for Safe Medication Practices list of high-alert medications. The investigators concluded that a redesign of the medication decision-support system was needed. The study by Chen et al. illuminated how social factors pose challenges in implementing CDS. Investigators in this study were previously successful in using a combination of an education campaign and interruptive decision-support prompts to reduce the inappropriate ordering of blood transfusions. However, even with a successful intervention, up to 30% of transfusions occurred outside of recommended guidelines. This finding prompted the investigators to analyze the free-text reasons offered by providers for overriding the recommended guidelines. Two key patterns emerged from their structured analysis. First, many of the apparently inappropriate transfusions occurred under officially sanctioned protocols (such as stem cell transplant) that the computer system was not able to take into account in generating alerts. Second, many orders that reflected questionable practices were being entered by resident physicians, physician assistants, nurse practitioners, and nurses who were least empowered to challenge requests from senior staff. Several practical and actionable lessons can be drawn from the 3 sets of investigators featured in this issue of the Journal of Hospital Medicine. First, all *Address for correspondence and reprint requests: Eric G. Poon, MD, Infectious Boston Medical Center, BCD 5002, 800 Harrison Ave., Boston, MA 02118; E-mail: eric.poon@bmc.org