Use of a Large Language Model to Assess Clinical Acuity of Adults in the Emergency Department.

Importance: The introduction of large language models (LLMs), such as Generative Pre-trained Transformer 4 (GPT-4; OpenAI), has generated significant interest in health care, yet studies evaluating their performance in a clinical setting are lacking. Determination of clinical acuity, a measure of a patient's illness severity and level of required medical attention, is one of the foundational elements of medical reasoning in emergency medicine. Objective: To determine whether an LLM can accurately assess clinical acuity in the emergency department (ED). Design, Setting, and Participants: This cross-sectional study identified all adult ED visits from January 1, 2012, to January 17, 2023, at the University of California, San Francisco, with a documented Emergency Severity Index (ESI) acuity level (immediate, emergent, urgent, less urgent, or nonurgent) and with a corresponding ED physician note. A sample of 10 000 pairs of ED visits with nonequivalent ESI scores, balanced for each of the 10 possible pairs of 5 ESI scores, was selected at random. Exposure: The potential of the LLM to classify acuity levels of patients in the ED based on the ESI across 10 000 patient pairs. Using deidentified clinical text, the LLM was queried to identify the patient with a higher-acuity presentation within each pair based on the patients' clinical history. An earlier LLM was queried to allow comparison with this model. Main Outcomes and Measures: Accuracy score was calculated to evaluate the performance of both LLMs across the 10 000-pair sample. A 500-pair subsample was manually classified by a physician reviewer to compare performance between the LLMs and human classification. Results: From a total of 251 401 adult ED visits, a balanced sample of 10 000 patient pairs was created wherein each pair comprised patients with disparate ESI acuity scores. Across this sample, the LLM correctly inferred the patient with higher acuity for 8940 of 10 000 pairs (accuracy, 0.89 [95% CI, 0.89-0.90]). Performance of the comparator LLM (accuracy, 0.84 [95% CI, 0.83-0.84]) was below that of its successor. Among the 500-pair subsample that was also manually classified, LLM performance (accuracy, 0.88 [95% CI, 0.86-0.91]) was comparable with that of the physician reviewer (accuracy, 0.86 [95% CI, 0.83-0.89]). Conclusions and Relevance: In this cross-sectional study of 10 000 pairs of ED visits, the LLM accurately identified the patient with higher acuity when given pairs of presenting histories extracted from patients' first ED documentation. These findings suggest that the integration of an LLM into ED workflows could enhance triage processes while maintaining triage quality and warrants further investigation.

Innovation and adaptation: The rise of a fentanyl smoking culture in San Francisco.

BACKGROUND: Illicitly manufactured fentanyls and stimulants are implicated in the escalating US mortality from drug overdose. San Francisco, California (SF) has seen declining fentanyl injection while smoking has increased. Beliefs and behaviors surrounding this development are not well understood. METHODS: The study used rapid ethnography to explore fentanyl and methamphetamine use in SF. The team conducted semi-structured interviews (n = 34) with participants recruited from syringe service programs. Video-recorded smoking sequences (n = 12), photography and daily field notes supplemented interview data. RESULTS: Difficulty injecting and fear of overdose motivated transitions from injecting to smoking. Fentanyl was extremely cheap-$10/gram-with variability in quality. Foil was the most commonly used smoking material but glass bubbles, bongs and dabbing devices were also popular. No reliable visible methods for determining fentanyl quality existed, however, participants could gauge potency upon inhalation, and developed techniques to regulate dosage. Several participants reported at least hourly use, some reporting one or more grams of daily fentanyl consumption. Smoking was also very social, with people sharing equipment, drugs and information. Participants raised concerns about hygiene and overdose risk to others arising from shared equipment. Reportedly potent fentanyl 'residue' accumulated on smoking materials and was commonly shared/traded/stolen or consumed accidentally with diverse preferences for its use. CONCLUSION: Our data highlight fentanyl residue as a new overdose risk with potential mismatch between the potency of the residual drug and the recipient's tolerance. Further, large doses of fentanyl are being consumed (estimated at approximately 50 mg of pure fentanyl/day). Smoking fentanyl has potential health benefits over injecting and may be protective against overdose, but substantial uncertainty exists. However, SF overdose mortality hit a record high in 2023. Recommendations to reduce fentanyl smoking overdose risks through pacing, greater awareness of dosages consumed and checking tolerance of residue recipients are potentially viable interventions deserving further exploration.

How do we understand the value of drug checking as a component of harm reduction services? A qualitative exploration of client and provider perspectives.

BACKGROUND: Mortality related to opioid overdose in the U.S. has risen sharply in the past decade. In California, opioid overdose death rates more than tripled from 2018 to 2021, and deaths from synthetic opioids such as fentanyl increased more than seven times in those three years alone. Heightened attention to this crisis has attracted funding and programming opportunities for prevention and harm reduction interventions. Drug checking services offer people who use drugs the opportunity to test the chemical content of their own supply, but are not widely used in North America. We report on qualitative data from providers and clients of harm reduction and drug checking services, to explore how these services are used, experienced, and considered. METHODS: We conducted in-depth semi-structured key informant interviews across two samples of drug checking stakeholders: "clients" (individuals who use drugs and receive harm reduction services) and "providers" (subject matter experts and those providing clinical and harm reduction services to people who use drugs). Provider interviews were conducted via Zoom from June-November, 2022. Client interviews were conducted in person in San Francisco over a one-week period in November 2022. Data were analyzed following the tenets of thematic analysis. RESULTS: We found that the value of drug checking includes but extends well beyond overdose prevention. Participants discussed ways that drug checking can fill a regulatory vacuum, serve as a tool of informal market regulation at the community level, and empower public health surveillance systems and clinical response. We present our findings within three key themes: (1) the role of drug checking in overdose prevention; (2) benefits to the overall agency, health, and wellbeing of people who use drugs; and (3) impacts of drug checking services at the community and systems levels. CONCLUSION: This study contributes to growing evidence of the effectiveness of drug checking services in mitigating risks associated with substance use, including overdose, through enabling people who use and sell drugs to test their own supply. It further contributes to discussions around the utility of drug checking and harm reduction, in order to inform legislation and funding allocation.

Conceptual framework for preterm birth review in San Francisco.

Preterm birth persists as a leading cause of infant mortality and morbidity despite decades of intervention effort. Intervention null effects may reflect failure to account for social determinants of health (SDH) or jointly acting risk factors. In some communities, persistent preterm birth trends and disparities have been consistently associated with SDH such as race/ethnicity, zip code, and housing conditions. Health authorities recommend conceptual frameworks for targeted action on SDH and precision public health approaches for preterm birth prevention. We document San Francisco, California's experience identifying the need, rationale, methods, and pilot work for developing a conceptual framework for preterm birth review (PTBR) in San Francisco. The PTBR conceptual framework is intended to enable essential public health services in San Francisco that prevent a range of preterm birth phenotypes by guiding plans for data collection, hypothesis testing, analytical methods, reports, and intervention strategy. Key elements of the PTBR conceptual framework are described including, 10 domains of SDH, 9 domains at the whole person level, such as lived experience and health behaviors, 8 domains at the within-person level, such as biomarkers and clinical measures, 18 preterm birth phenotypes, and the interconnections between domains. Assumptions for the PTBR conceptual framework were supported by a scoping review of literature on SDH effects on preterm birth, health authority consensus reports, and PTBR pilot data. Researcher and health authority interest in each of the domains warrants the framework to prompt systematic consideration of variables in each proposed domain. PTBR pilot data, illustrated in heatmaps, confirm the feasibility of data collection based on the framework, prevalence of co-occurring risk factors, potential for joint effects on specific preterm birth phenotypes, and opportunity for intervention to block SDH effects on preterm birth. The proposed PTBR conceptual framework has practical implications for specifying (1) population groups at risk, (2) grids or heatmap visualization of risk factors, (3) multi-level analyses, and (4) multi-component intervention design in terms of patterns of co-occurring risk factors. Lessons learned about PTBR data collection logistics, variable choice, and data management will be incorporated into future work to build PTBR infrastructure based on the PTBR conceptual framework.

Environmental Justice and Systems Analysis for Air Quality Planning in the Port of Oakland in California.

Many frontline communities experience adverse health impacts from living in proximity to high-polluting industrial sources. Securing environmental justice requires, in part, a comprehensive set of quantitative indicators. We incorporate environmental justice and life-cycle thinking into air quality planning to assess fine particulate matter (PM2.5) exposure and monetized damages from operating and maintaining the Port of Oakland, a major multimodal marine port located in the historically marginalized West Oakland community in the San Francisco Bay Area. The exposure domain for the assessment is the entire San Francisco Bay Area, a home to more than 7.5 million people. Of the more than 14 sources included in the emissions inventory, emissions from large container ships, or ocean-going vessels (OGVs), dominate the PM2.5 intake, and supply chain sources (material production and delivery, fuel production) represent between 3.5% and 7.5% of annual intake. Exposure damages, which model the costs from excess mortalities resulting from exposure from the study's emission sources, range from USD 100 to 270 million per annum. Variations in damages are due to the use of different concentration-response relationships, hazard ratios, and Port resurfacing area assumptions. Racial and income-based exposure disparities are stark. The Black population and people within the lowest income quintile are 2.2 and 1.9 times more disproportionately exposed, respectively, to the Port's pollution sources relative to the general population. Mitigation efforts focused on electrifying in-port trucking operations yield modest reductions (3.5%) compared to strategies that prioritize emission reductions from OGVs and commercial harbor craft operations (8.7-55%). Our recommendations emphasize that a systems-based approach is critical for identifying all relevant emission sources and mitigation strategies for improving equity in civil infrastructure systems.

Soft Tissue Reconstruction After Sacral Neoplasm Resection: The University of California San Francisco Experience.

PURPOSE: Resection of sacral neoplasms such as chordoma and chondrosarcoma with subsequent reconstruction of large soft tissue defects is a complex multidisciplinary process. Radiotherapy and prior abdominal surgery play a role in reconstructive planning; however, there is no consensus on how to maximize outcomes. In this study, we present our institution's experience with the reconstructive surgical management of this unique patient population. METHODS: We conducted a retrospective review of patients who underwent reconstruction after resection of primary or recurrent pelvic chordoma or chondrosarcoma between 2002 and 2019. Surgical details, hospital stay, and postoperative outcomes were assessed. Patients were divided into 3 groups for comparison based on reconstruction technique: gluteal-based flaps, vertical rectus abdominus myocutaneous (VRAM) flaps, and locoregional fasciocutaneous flaps. RESULTS: Twenty-eight patients (17 males, 11 females), with mean age of 62 years (range, 34-86 years), were reviewed. Twenty-two patients (78.6%) received gluteal-based flaps, 3 patients (10.7%) received VRAM flaps, and 3 patients (10.7%) were reconstructed with locoregional fasciocutaneous flaps. Patients in the VRAM group were significantly more likely to have undergone total sacrectomy (P < 0.01) in a 2-stage operation (P < 0.01) compared with patients in the other 2 groups. Patients in the VRAM group also had a significantly greater average number of reoperations (2 ± 3.5, P = 0.04) and length of stay (29.7 ± 20.4 days, P = 0.01) compared with the 2 other groups. The overall minor and major wound complication rates were 17.9% and 42.9%, respectively, with 17.9% of patients experiencing at least 1 infection or seroma. There was no association between prior abdominal surgery, surgical stages, or radiation therapy and an increased risk of wound complications. CONCLUSIONS: Vertical rectus abdominus myocutaneous flaps are a more suitable option for patients with larger defects after total sacrectomy via 2-staged anteroposterior resections, whereas gluteal myocutaneous flaps are effective options for posterior-only resections. For patients with small- to moderate-sized defects, local fasciocutaneous flaps are a less invasive and effective option. Paraspinous flaps may be used in combination with other techniques to provide additional bulk and coverage for especially long postresection wounds. Furthermore, mesh is a useful adjunct for any reconstruction aimed at protecting against intra-abdominal complications.

Overdose from Unintentional Fentanyl Use when Intending to Use a Non-opioid Substance: An Analysis of Medically Attended Opioid Overdose Events.

Fentanyl-mixed and substituted heroin is well-documented, but less is known about unintentional fentanyl use among people using stimulants. To determine the prevalence of and racial and ethnic disparities in unintentional fentanyl use among people experiencing a medically attended opioid overdose, we reviewed 448 suspected non-fatal overdose cases attended by a community paramedic overdose response team in San Francisco from June to September 2022. We applied a case definition for opioid overdose to paramedic records and abstracted data on intended substance use prior to overdose. Among events meeting case criteria with data on intended substance use, intentional opioid use was reported by 57.3%, 98.0% of whom intended to use fentanyl. No intentional opioid use was reported by 42.7%, with most intending to use stimulants (72.6%), including methamphetamine and cocaine. No intentional opioid use was reported by 58.5% of Black, 52.4% of Latinx, and 28.8% of White individuals (p = 0.021), and by 57.6% of women and 39.5% of men (p = 0.061). These findings suggest that unintentional fentanyl use among people without opioid tolerance may cause a significant proportion of opioid overdoses in San Francisco. While intentional fentanyl use might be underreported, the magnitude of self-reported unintentional use merits further investigation to confirm this phenomenon, explore mechanisms of use and disparities by race and ethnicity, and deploy targeted overdose prevention interventions.

HIV Viral Suppression Among Psychiatric Inpatients with Schizophrenia in San Francisco: A Retrospective Cohort Study.

People with schizophrenia are at increased risk for contracting HIV and face higher mortality rates compared with the general population. Viral suppression is key to HIV care, yet little is known about this metric among people with HIV and schizophrenia. A chart review was conducted among people with HIV/AIDS and schizophrenia living in San Francisco who had received inpatient mental health services between 2010 and 2016. Demographic, laboratory, medication, encounter, and discharge data were collected, and were compared with all people living with HIV in San Francisco (PLWH-SF). Among 153 people living with HIV and comorbid schizophrenia, 77% were virally suppressed, compared to 67% for all PLWH-SF. Viral suppression for people with comorbid HIV and schizophrenia living in San Francisco appears higher than PLWH-SF. Further research is needed to confirm the association and mechanisms behind better treatment outcomes for people living with HIV and comorbid schizophrenia.

First record of a structural infestation by the bird flea Dasypsyllus gallinulae perpinnatus (Siphonaptera: Ceratophyllidae) in the San Francisco Bay Area, California, USA.

Herein we present the first-known case report of a structural infestation by the bird flea (Dasypsyllus gallinulae perpinnatus) (Dale 1878) in the United States. In March of 2023, the San Mateo County Mosquito and Vector Control District was contacted by a resident in South San Francisco, CA who reported the presence of fleas inside their condominium. The resident had 2 dogs who were on oral flea medication and only 1 inhabitant out of 4 reported receiving flea bites. The front walkway, backyard, and garage were flagged and a small passerine nest was removed from a fire alarm bell in the front walkway. A total of 31 fleas (13 males, 18 females) were collected by flagging from the front entryway underneath the nest. One female flea was collected from the backyard, and 20 (9 males, 11 females) were collected by the resident from inside the house. A total of 387 fleas (163 males, 224 females) were collected from the nest of a small passerine. All life stages (egg, larvae, pupae, and adult) were observed within the nest. Additional parasites in the nest included bird lice and larval western black-legged ticks (Ixodes pacificus, Cooley and Kohls 1943). Research should be conducted into whether bird fleas are capable of transmitting avian pathogens to humans and this information should be communicated to healthcare providers as part of a One Health approach. Additionally, resources to aid in species-level flea identification should be made accessible to pest control operators as this will aid the development of targeted treatments as part of an integrated pest management plan.

An Historical Case Study of Trans Exclusion and Empowerment: Implications for Transgender Older Adults and Aging.

Transgender older adults have a long history of exclusion that shapes current experiences with social services. However, scant gerontological research uses archival data, which can provide critical context for service providers. Moreover, sparse research examines how exclusion can be a catalyst for change that social workers could leverage. Empowerment theory provides a theoretical tool to explain how this is possible. This multidisciplinary case study blends community member interviews and archival data to answer this question: How did exclusion shape empowerment and social change for transgender Americans? This study focuses on the events before and after the Compton's Cafeteria Riot, one of the first acts of LGBTQIA+ collective resistance that led to new services for and by transgender Americans. Data reveal how exclusion facilitated the emergence of collective empowerment among transgender women and queer youth in San Francisco. Archival data shows how exclusion preceded self-efficacy, critical consciousness, involvement with similar others, acquisition of new skills, and ultimately action to eliminate social, economic, and political barriers and power imbalances. This study provides both empirical and theoretical tools to contribute new data and perspectives on trans exclusion and empowerment and its implications for social workers serving transgender older adults.

Antigenic Characterization of Novel Human Norovirus GII.4 Variants San Francisco 2017 and Hong Kong 2019.

Norovirus is a major cause of acute gastroenteritis; GII.4 is the predominant strain in humans. Recently, 2 new GII.4 variants, Hong Kong 2019 and San Francisco 2017, were reported. Characterization using GII.4 monoclonal antibodies and serum demonstrated different antigenic profiles for the new variants compared with historical variants.

Trends in presumed drug overdose out-of-hospital cardiac arrests in San Francisco, 2015-2023.

INTRODUCTION: Estimates of the prevalence of drug-related out of hospital cardiac arrest (OHCA) vary, ranging from 1.8% to 10.0% of medical OHCA. However, studies conducted prior to the recent wave of fentanyl deaths likely underestimate the current prevalence of drug-related OHCA. We evaluated recent trends in drug-related OHCA, hypothesizing that the proportion of presumed drug-related OHCA treated by emergency medical services (EMS) has increased since 2015. METHODS: We conducted a retrospective analysis of OHCA patients treated by EMS providers in San Francisco, California between 2015 and 2023. Participants included OHCA cases in which resuscitation was attempted by EMS. The study exposure was the year of arrest. Our primary outcome was the occurrence of drug-related OHCA, defined as the EMS impression of OHCA caused by a presumed or known overdose of medication(s) or drug(s). RESULTS: From 2015 to 2023, 5044 OHCA resuscitations attended by EMS (average 561 per year) met inclusion criteria. The median age was 65 (IQR 50-79); 3508 (69.6%) were male. The EMS impression of arrest etiology was drug-related in 446/5044 (8.8%) of OHCA. The prevalence of presumed drug-related OHCA increased significantly each year from 1% in 2015 to 17.6% in 2023 (p-value for trend = 0.0001). After adjustment, presumed drug-related OHCA increased by 30% each year from 2015-2023. CONCLUSION: Drug-related OHCA is an increasingly common etiology of OHCA. In 2023, one in six OHCA was presumed to be drug related. Among participants less than 60 years old, one in three OHCA was presumed to be drug related.

Radical hospitality: Innovative programming to build community and meet the needs of people who use drugs at a government-sanctioned overdose prevention site in San Francisco, California.

BACKGROUND: The Tenderloin Center (TLC), a multi-service center where people could receive or be connected to basic needs, behavioral health care, housing, and medical services, was open in San Francisco for 46 weeks in 2022. Within a week of operation, services expanded to include an overdose prevention site (OPS), also known as safe consumption site. OPSs have operated internationally for over three decades, but government-sanctioned OPSs have only recently been implemented in the United States. We used ethnographic methods to understand the ways in which a sanctioned OPS, situated in a multi-service center, impacts the lives of people who use drugs (PWUD). METHODS: We conducted participant observation and in-depth interviews June-December 2022. Extensive field notes and 39 in-depth interviews with 24 TLC guests and 15 TLC staff were analyzed using an inductive analysis approach. Interviewees were asked detailed questions about their experiences using and working at the TLC. RESULTS: TLC guests and staff described an atmosphere where radical hospitality-welcoming guests with extraordinary warmth, generosity, and unconditional acceptance-was central to the culture. We found that the co-location of an OPS within a multi-service agency (1) allowed for the culture of radical hospitality to flourish, (2) yielded a convenient one-stop shop model, (3) created a space for community building, and (4) offered safety and respite to guests. CONCLUSIONS: The co-location of an OPS within a multi-service drop-in center is an important example of how such an organization can build positive sociality among PWUD while protecting autonomy and reducing overdose mortality. Overdose response and reversal is an act of relational accountability in which friends, peers, and even strangers intervene to protect and revive one another. This powerful intervention was operationalized as an anti-oppressive, horizontal activity through radical hospitality with a built environment that allowed PWUD to be both social and safe.

A Co-Created Tool to Help Counter Health Misinformation for Spanish-Speaking Communities in the San Francisco Bay Area.

BACKGROUND: Health misinformation, which was particularly prevalent during the COVID-19 pandemic, hampers public health initiatives. Spanish-speaking communities in the San Francisco Bay Area may be especially affected due to low digital health literacy and skepticism towards science and healthcare experts. Our study aims to develop a checklist to counter misinformation, grounded in community insights. METHODS: We adopted a multistage approach to understanding barriers to COVID-19 vaccine uptake in Spanish-speaking populations in Alameda and San Francisco counties. Initial work included key informant and community interviews. Partnering with a community-based organization (CBO), we organized co-design workshops in July 2022 to develop a practical tool for identifying misinformation. Template analysis identified key themes for actionable steps, such as source evaluation and content assessment. From this, we developed a Spanish-language checklist. FINDINGS: During formative interviews, misinformation was identified as a major obstacle to vaccine uptake. Three co-design workshops with 15 Spanish-speaking women resulted in a 10-step checklist for tackling health misinformation. Participants highlighted the need for scrutinizing sources and assessing messenger credibility, and cues in visual content that could instill fear. The checklist offers a pragmatic approach to source verification and information assessment, supplemented by resources from local CBOs. CONCLUSION: We have co-created a targeted checklist for Spanish-speaking communities to identify and counter health misinformation. Such specialized tools are essential for populations that are more susceptible to misinformation, enabling them to differentiate between credible and non-credible information.

Pragmatic phase II clinical trial to improve depression care in a real-world diverse MS cohort from an academic MS centre in Northern California: MS CATCH study protocol.

INTRODUCTION: Depression occurs in over 50% of individuals living with multiple sclerosis (MS) and can be treated using many modalities. Yet, it remains: under-reported by patients, under-ascertained by clinicians and under-treated. To enhance these three behaviours likely to promote evidence-based depression care, we engaged multiple stakeholders to iteratively design a first-in-kind digital health tool. The tool, MS CATCH (Care technology to Ascertain, Treat, and engage the Community to Heal depression in patients with MS), closes the communication loop between patients and clinicians. Between clinical visits, the tool queries patients monthly about mood symptoms, supports patient self-management and alerts clinicians to worsening mood via their electronic health record in-basket. Clinicians can also access an MS CATCH dashboard displaying patients' mood scores over the course of their disease, and providing comprehensive management tools (contributing factors, antidepressant pathway, resources in patient's neighbourhood). The goal of the current trial is to evaluate the clinical effect and usability of MS CATCH in a real-world clinical setting. METHODS AND ANALYSIS: MS CATCH is a single-site, phase II randomised, delayed start, trial enrolling 125 adults with MS and mild to moderately severe depression. Arm 1 will receive MS CATCH for 12 months, and arm 2 will receive usual care for 6 months, then MS CATCH for 6 months. Clinicians will be randomised to avoid practice effects. The effectiveness analysis is superiority intent-to-treat comparing MS CATCH to usual care over 6 months (primary outcome: evidence of screening and treatment; secondary outcome: Hospital Anxiety Depression Scale-Depression scores). The usability of the intervention will also be evaluated (primary outcome: adoption; secondary outcomes: adherence, engagement, satisfaction). ETHICS AND DISSEMINATION: University of California, San Francisco Institutional Review Board (22-36620). The findings of the study are planned to be shared through conferences and publishments in a peer-reviewed journal. The deidentified dataset will be shared with qualified collaborators on request, provision of CITI and other certifications, and data sharing agreement. We will share the results, once the data are complete and analysed, with the scientific community and patient/clinician participants through abstracts, presentations and manuscripts. TRIAL REGISTRATION NUMBER: NCT05865405.

Predictive Modeling of Drug-Related Adverse Events with Real-World Data: A Case Study of Linezolid Hematologic Outcomes.

Electronic health records (EHRs) provide meaningful knowledge of drug-related adverse events (AEs) that are not captured in standard drug development and postmarketing surveillance. Using variables obtained from EHR data in the University of California San Francisco de-identified Clinical Data Warehouse, we aimed to evaluate the potential of machine learning to predict two hematological AEs, thrombocytopenia and anemia, in a cohort of patients treated with linezolid for 3 or more days. Features for model input were extracted at linezolid initiation (index), and outcomes were characterized from index to 14 days post-treatment. Random forest classification (RFC) was used for AE prediction, and reduced feature models were evaluated using cumulative importance (cImp) for feature selection. Grade 3+ thrombocytopenia and anemia occurred in 31% of 2,171 and 56% of 2,170 evaluable patients, respectively. Of the total 53 features, as few as 7 contributed at least 50% cImp, resulting in prediction accuracies of 70% or higher and area under the receiver operating characteristic curves of 0.886 for grade 3+ thrombocytopenia and 0.759 for grade 3+ anemia. Sensitivity analyses in strictly defined patient subgroups revealed similarly high predictive performance in full and reduced feature models. A logistic regression model with the same 50% cImp features showed similar predictive performance as RFC and good concordance with RFC probability predictions after isotonic calibration, adding interpretability. Collectively, this work demonstrates potential for machine learning prediction of AE risk in real-world patients using few variables regularly available in EHRs, which may aid in clinical decision making and/or monitoring.

Leveraging electronic health records and knowledge networks for Alzheimer's disease prediction and sex-specific biological insights.

Identification of Alzheimer's disease (AD) onset risk can facilitate interventions before irreversible disease progression. We demonstrate that electronic health records from the University of California, San Francisco, followed by knowledge networks (for example, SPOKE) allow for (1) prediction of AD onset and (2) prioritization of biological hypotheses, and (3) contextualization of sex dimorphism. We trained random forest models and predicted AD onset on a cohort of 749 individuals with AD and 250,545 controls with a mean area under the receiver operating characteristic of 0.72 (7 years prior) to 0.81 (1 day prior). We further harnessed matched cohort models to identify conditions with predictive power before AD onset. Knowledge networks highlight shared genes between multiple top predictors and AD (for example, APOE, ACTB, IL6 and INS). Genetic colocalization analysis supports AD association with hyperlipidemia at the APOE locus, as well as a stronger female AD association with osteoporosis at a locus near MS4A6A. We therefore show how clinical data can be utilized for early AD prediction and identification of personalized biological hypotheses.

Participant experiences in a combination HIV cure-related trial with extended analytical treatment interruption in San Francisco, United States.

BACKGROUND: There is limited systematic information available about the perspectives of participants enrolled in intensive combination HIV cure-related trials inclusive of an extended analytical treatment interruption (ATI). OBJECTIVE: To assess and understand experiences of people with HIV involved in a combination HIV cure-related trial with an extended ATI. METHODS: The trial included five interventions and was followed by an ATI lasting up to 52 wk. From 2022 - 2023, we conducted in-depth interviews with study participants following their extended ATIs. Interviews were audio-recorded, transcribed, and analyzed via conventional thematic analysis. RESULTS: We interviewed seven participants. The majority were male, White, and non-Hispanic, with a median age of 37 years. Trust in the research team, scientific altruism and hope of becoming a post-intervention controller were key motivators for joining the trial. Interviewees reported being satisfied with their decision to participate in the trial and the extended ATI. Most recounted feelings of worry related to viral rebound during the ATI. Participants reported both defeat and relief with ART restart. Four faced challenges with protecting partners from HIV during their ATI, such as trying to find out if their partner(s) were using pre-exposure prophylaxis. CONCLUSIONS: Our findings demonstrate potential improvements for future ATI trial participant experiences, such as more robust resources for psychosocial support and partner protections. Dedicating greater effort to understanding participant ATI experiences can inform the design of future participant-centered HIV cure trial protocols.