Explore the vast range of titles available.

Best Practice Alerts (BPAs) are alert messages to physicians in the electronic health record that are used to encourage appropriate use of health care resources. While these alerts are helpful in both improving care and reducing costs, BPAs are often broadly applied nonselectively across entire patient populations. The development of large language models (LLMs) provides an opportunity to selectively identify patients for BPAs. In this paper, we present an example case where an LLM screening tool is used to select patients appropriate for a BPA encouraging the prescription of deep vein thrombosis (DVT) anticoagulation prophylaxis. The artificial intelligence (AI) screening tool was developed to identify patients experiencing acute bleeding and exclude them from receiving a DVT prophylaxis BPA. Our AI screening tool used a BioMed-RoBERTa (Robustly Optimized Bidirectional Encoder Representations from Transformers Pretraining Approach; AllenAI) model to perform classification of physician notes, identifying patients without active bleeding and thus appropriate for a thromboembolism prophylaxis BPA. The BioMed-RoBERTa model was fine-tuned using 500 history and physical notes of patients from the MIMIC-III (Medical Information Mart for Intensive Care) database who were not prescribed anticoagulation. A development set of 300 MIMIC patient notes was used to determine the model's hyperparameters, and a separate test set of 300 patient notes was used to evaluate the screening tool. Our MIMIC-III test set population of 300 patients included 72 patients with bleeding (ie, were not appropriate for a DVT prophylaxis BPA) and 228 without bleeding who were appropriate for a DVT prophylaxis BPA. The AI screening tool achieved impressive accuracy with a precision-recall area under the curve of 0.82 (95% CI 0.75-0.89) and a receiver operator curve area under the curve of 0.89 (95% CI 0.84-0.94). The screening tool reduced the number of patients who would trigger an alert by 20% (240 instead of 300 alerts) and increased alert applicability by 14.8% (218 [90.8%] positive alerts from 240 total alerts instead of 228 [76%] positive alerts from 300 total alerts), compared to nonselectively sending alerts for all patients. These results show a proof of concept on how language models can be used as a screening tool for BPAs. We provide an example AI screening tool that uses a HIPAA (Health Insurance Portability and Accountability Act)-compliant BioMed-RoBERTa model deployed with minimal computing power. Larger models (eg, Generative Pre-trained Transformers-3, Generative Pre-trained Transformers-4, and Pathways Language Model) will exhibit superior performance but require data use agreements to be HIPAA compliant. We anticipate LLMs to revolutionize quality improvement in hospital medicine.

Background We identified that Stanford Health Care had a significant number of patients who after discharge are found by the utilization review committee not to meet Center for Mediare and Medicaid Services (CMS) 2-midnight benchmark for inpatient status. Some of the charges incurred during the care of these patients are written-off and known as Medicare 1-day write-offs. This study which aims to evaluate the use of a Best Practice Alert (BPA) feature on the electronic medical record, EPIC, to ensure appropriate designation of a patient’s hospitalization status as either inpatient or outpatient in accordance with Center for Medicare and Medicaid services (CMS) 2 midnight length of stay benchmark thereby reducing the number of associated write-offs. Method We incorporated a best practice alert (BPA) into the Epic Electronic Medical Record (EMR) that would prompt the discharging provider and the case manager to review the patients’ inpatient designation prior to discharge and change the patient’s designation to observation when deemed appropriate. Patients who met the inclusion criteria (Patients must have Medicare fee-for-service insurance, inpatient length of stay (LOS) less than 2 midnights, inpatient designation as hospitalization status at time of discharge, was hospitalized to an acute level of care and belonged to one of 37 listed hospital services at the time of signing of the discharge order) were randomized to have the BPA either silent or active over a three-month period from July 18, 2019, to October 18, 2019. Result A total of 88 patients were included in this study: 40 in the control arm and 48 in the intervention arm. In the intervention arm, 8 (8/48, 16.7%) had an inpatient status designation despite potentially meeting Medicare guidelines for an observation stay, comparing to 23 patients (23/40, 57.5%) patients in the control group ( p  = 0.001). The estimated number of write-offs in the control arm was 17 (73.9%, out of 23 inpatient patients) while in the intervention arm was 1 (12.5%, out of 8 inpatient patient) after accounting for patients who may have met inpatient criteria for other reasons based on case manager note review. Conclusion This is the first time to our knowledge that a BPA has been used in this manner to reduce the number of Medicare 1-day write-offs.

Background: A major problem in treating acute kidney injury (AKI) is that clinical criteria for recognition are markers of established kidney damage or impaired function; treatment before such damage manifests is desirable. Clinicians could intervene during what may be a crucial stage for preventing permanent kidney injury if patients with incipient AKI and those at high risk of developing AKI could be identified. Objective: In this study, we evaluate a machine learning algorithm for early detection and prediction of AKI. Design: We used a machine learning technique, boosted ensembles of decision trees, to train an AKI prediction tool on retrospective data taken from more than 300 000 inpatient encounters. Setting: Data were collected from inpatient wards at Stanford Medical Center and intensive care unit patients at Beth Israel Deaconess Medical Center. Patients: Patients older than the age of 18 whose hospital stays lasted between 5 and 1000 hours and who had at least one documented measurement of heart rate, respiratory rate, temperature, serum creatinine (SCr), and Glasgow Coma Scale (GCS). Measurements: We tested the algorithm’s ability to detect AKI at onset and to predict AKI 12, 24, 48, and 72 hours before onset. Methods: We tested AKI detection and prediction using the National Health Service (NHS) England AKI Algorithm as a gold standard. We additionally tested the algorithm’s ability to detect AKI as defined by the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines. We compared the algorithm’s 3-fold cross-validation performance to the Sequential Organ Failure Assessment (SOFA) score for AKI identification in terms of area under the receiver operating characteristic (AUROC). Results: The algorithm demonstrated high AUROC for detecting and predicting NHS-defined AKI at all tested time points. The algorithm achieves AUROC of 0.872 (95% confidence interval [CI], 0.867-0.878) for AKI detection at time of onset. For prediction 12 hours before onset, the algorithm achieves an AUROC of 0.800 (95% CI, 0.792-0.809). For 24-hour predictions, the algorithm achieves AUROC of 0.795 (95% CI, 0.785-0.804). For 48-hour and 72-hour predictions, the algorithm achieves AUROC values of 0.761 (95% CI, 0.753-0.768) and 0.728 (95% CI, 0.719-0.737), respectively. Limitations: Because of the retrospective nature of this study, we cannot draw any conclusions about the impact the algorithm’s predictions will have on patient outcomes in a clinical setting. Conclusions: The results of these experiments suggest that a machine learning–based AKI prediction tool may offer important prognostic capabilities for determining which patients are likely to suffer AKI, potentially allowing clinicians to intervene before kidney damage manifests.

High rates of mental health symptoms such as depression, anxiety, and posttraumatic stress disorder (PTSD) have been found in patients hospitalized with traumatic injuries, but little is known about these problems in patients hospitalized with acute illnesses. A similarly high prevalence of mental health problems in patients hospitalized with acute illness would have significant public health implications because acute illness and injury are both common, and mental health problems of depression, anxiety, and PTSD are highly debilitating. In patients admitted after emergency care for Acute Illness (N = 656) or Injury (N = 661) to three hospitals across the United States, symptoms of depression, anxiety, and posttraumatic stress were compared acutely (Acute Stress Disorder) and two months post-admission (PTSD). Patients were ethnically/racially diverse and 54% female. No differences were found between the Acute Illness and Injury groups in levels of any symptoms acutely or two months post-admission. At two months post-admission, at least one symptom type was elevated for 37% of the Acute Illness group and 39% of the Injury group. Within racial/ethnic groups, PTSD symptoms were higher in Black patients with injuries than for Black patients with acute illness. A disproportionate number of Black patients had been assaulted. This study found comparable levels of mental health sequelae in patients hospitalized after emergency care for acute illness as in patients hospitalized after emergency care for injury. Findings of significantly higher symptoms and interpersonal violence injuries in Black patients with injury suggest that there may be important and actionable differences in mental health sequelae across ethnic/racial identities and/or mechanisms of injury or illness. Routine screening for mental health risk for all patients admitted after emergency care could foster preventive care and reduce ethnic/racial disparities in mental health responses to acute illness or injury.

Private hospital rooms offer potential advantages over semiprivate rooms, but the impact of room type on patient experience across multiple dimensions of care remains understudied. A retrospective study was conducted to investigate how room type influenced patients’ perception of their experience at Stanford Health Care, a large university medical centre in California, USA. Hospital Consumer Assessment of Healthcare Providers and Systems survey data from medicine patients discharged from January 2018 to January 2020 (n=891) was analysed. The percentage of top responses was calculated for 18 survey sections including overall assessment. Patients in private rooms were more likely to give a top response (aOR, 1.30; 95% CI, 1.24 to 1.36), rating overall assessment and 10 other sections significantly higher than patients in semiprivate rooms. The greatest differences were in survey sections related to the room (room, hospital environment and visitors/family). However, private rooms also performed better on sections not directly related to room type (tests/treatments, care transitions and discharge). These widespread improvements suggest a ‘halo effect’, in which a patient’s positive impression of their room may enhance their perception of overall care. These findings underscore the substantial influence of the care environment on patients’ perceptions of their overall hospital experience.

BackgroundResident work hour restrictions have led to the creation of the ‘night float’ to care for the patients of multiple primary teams after hours. These residents are often inundated with acute issues in the numerous patients they cover and are less able to address non-urgent issues that arise at night. Further, non-urgent pages may contribute to physician alarm fatigue and negatively impact patient outcomes.ObjectiveTo delineate the burden of non-urgent paging at night and propose solutions.MethodsWe performed a resident review and categorisation of 1820 pages to night floats between September 2014 and December 2014. Both attending and nursing review of 10% of pages was done and compared.ResultsOf reviewed pages, 62.1% were urgent and 27.7% were non-urgent. Attending review of random page samples correlated well with resident review. Common reasons for non-urgent pages were non-urgent patient status updates, low-priority order requests and non-critical lab values.ConclusionsA significant number of non-urgent pages are sent at night. These pages likely distract from acute issues that arise at night and place an unnecessary burden on night floats. Both behavioural and systemic adjustments are needed to address this issue. Possible interventions include integrating low-priority messaging into the electronic health record system and use of charge nurses to help determine urgency of issues and batch non-urgent pages.

Background Quality improvement (QI) is a systematic approach to improving healthcare delivery with applications across all fields of medicine. However, exposure to QI is minimal in early medical education. We evaluated the effectiveness of an elective QI curriculum in teaching preclinical health professional students foundational QI concepts. Methods This prospective controlled cohort study was conducted at a single academic institution. The elective QI curriculum consisted of web-based video didactics and exercises, supplemented with in-person classroom discussions. An optional hospital-based QI project was offered. Assessments included pre- and post-intervention surveys evaluating QI skills and beliefs and attitudes, quizzes, and Quality Improvement Knowledge Application Tool-Revised (QIKAT-R) cases. Within-group pre-post and between-group comparisons were performed using descriptive statistics. Results Overall, 57 preclinical medical or physician assistant students participated under the QI curriculum group ( N  = 27) or control group ( N  = 30). Twenty-three (85%) curriculum students completed a QI project. Mean quiz scores were significantly improved in the curriculum group from pre- to post-assessment (Quiz 1: 2.0, P  < 0.001; Quiz 2: 1.7, P  = 0.002), and the mean differences significantly differed from those in the control group (Quiz 1: P  < 0.001; Quiz 2: P  = 0.010). QIKAT-R scores also significantly differed among the curriculum group versus controls ( P  = 0.012). In the curriculum group, students had improvements in their confidence with all 10 QI skills assessed, including 8 that were significantly improved from pre- to post-assessment, and 4 with significant between-group differences compared with controls. Students in both groups agreed that their medical education would be incomplete without a QI component and that they are likely to be involved in QI projects throughout their medical training and practice. Conclusions The elective QI curriculum was effective in guiding preclinical students to develop their QI knowledge base and skillset. Preclinical students value QI as an integral component of their medical training. Future directions involve evaluating the impact of this curriculum on clinical clerkship performance and across other academic institutions.

Telemedicine has been adopted in the inpatient setting to facilitate clinical interactions between on-site clinicians and isolated hospitalized patients. Such remote interactions have the potential to reduce pathogen exposure and use of personal protective equipment but may also pose new safety concerns given prior evidence that isolated patients can receive suboptimal care. Formal evaluations of the use and practical acceptance of inpatient telemedicine among hospitalized patients are lacking. We aimed to evaluate the experience of patients hospitalized for COVID-19 with inpatient telemedicine introduced as an infection control measure during the pandemic. We conducted a qualitative evaluation in a COVID-19 designated non-intensive care hospital unit at a large academic health center (Stanford Health Care) from October 2020 through January 2021. Semistructured qualitative interviews focused on patient experience, impact on quality of care, communication, and mental health. Purposive sampling was used to recruit participants representing diversity across varying demographics until thematic saturation was reached. Interview transcripts were qualitatively analyzed using an inductive-deductive approach. Interviews with 20 hospitalized patients suggested that nonemergency clinical care and bridging to in-person care comprised the majority of inpatient telemedicine use. Nurses were reported to enter the room and call on the tablet far more frequently than physicians, who typically entered the room at least daily. Patients reported broad acceptance of the technology, citing improved convenience and reduced anxiety, but preferred in-person care where possible. Quality of care was believed to be similar to in-person care with the exception of a few patients who wanted more frequent in-person examinations. Ongoing challenges included low audio volume, shifting tablet location, and inconsistent verbal introductions from the clinical team. Patient experiences with inpatient telemedicine were largely favorable. Although most patients expressed a preference for in-person care, telemedicine was acceptable given the circumstances associated with the COVID-19 pandemic. Improvements in technical and care team use may enhance acceptability. Further evaluation is needed to understand the impact of inpatient telemedicine and the optimal balance between in-person and virtual care in the hospital setting.

Introduction: Many central line-associated bloodstream infections are identified in patients nearing the end of life. Stanford Health Care recently introduced the General Inpatient Hospice program. This program offers inpatient hospice care for patients who, due to uncontrolled symptoms, cannot be discharged to a hospice facility or receive home hospice care. We investigated whether this program would impact blood cultures practices near the time of death. Methods: We performed a retrospective cohort study at Stanford Health Care using records of blood culture events from May 2019 to October 2023. We defined a blood culture near-death as those collected within 2 days before the date of death. We performed an interrupted time series linear regression before and after the implementation of the General Inpatient Hospice program on July 1, 2022 to assess blood culture intensity near-death. Blood culture intensity was defined as the proportion of cultures collected near-death in relation to the total number of blood cultures. Additionally, we calculated blood culture positivity rate, which was defined as the proportion of positive blood cultures among all those collected during our study period. Results: Out of 220,269 blood cultures from 24,955 unique patients, a total of 6,147 cultures (9%) were obtained near the time of death. Among these subjects, the median age was 65 years (range 20–102), with 43% identifying as being of White race-ethnicity and 57% as male. Of these cultures, 3044 were positive (49.5%), with Escherichia coli (618, 24%), Klebsiella pneumoniae (341, 13%), and Staphylococcus aureus (166, 10%) being the most common organisms. After the implementation of the General Inpatient Hospice program, the median enrollment was 12 patients (range 3–18) and the median mortality rate was 2.3% (range 2–3%). The blood culture intensity near death decreased by 0.81%, a change that was not statistically significant (95% CI -2.4% to 0.8%, p=.32; Figure 1). Subsequently, the blood culture intensity showed a non-significant increasing trend of 0.05% (95% CI -0.1% to 0.2%, p=0.53). The blood culture positivity rate near the time of death increased by 16% following the intervention, but this increase was not statistically significant (95% CI – 11.8% to 43.3%, p=.26; Figure 2), and it was followed by a non-significant downtrend of 1.9% (95% CI -3.9% to 1.4%, p=.36). Conclusion: We found no significant association between the implementation of an inpatient hospice program and blood culture practices near the time of death, likely due to low patient enrollment.

A patient's likelihood to recommend a hospital is used to assess the quality of their experience. This study investigated whether room type influences patients’ likelihood to recommend Stanford Health Care using Hospital Consumer Assessment of Healthcare Providers and Systems survey data from November 2018 to February 2021 (n = 10,703). The percentage of patients who gave the top response was calculated as a top box score, and the effects of room type, service line, and the COVID-19 pandemic were represented as odds ratios (ORs). Patients in private rooms were more likely to recommend than patients in semi-private rooms (aOR: 1.32; 95% CI: 1.16–1.51; 86% vs 79%, p < .001), and service lines with only private rooms had the greatest increases in odds of a top response. The new hospital had significantly higher top box scores than the original hospital (87% vs 84%, p < .001), indicating that room type and hospital environment impact patients’ likelihood to recommend.