Brendan Jacka

Introduction: People who inject drugs (PWID) face a high risk for serious infections, yet International Classification of Diseases (ICD) codes fail to identify this population. Large language models (LLM) offer a promising alternative by extracting information from unstructured clinical text. This study evaluated the diagnostic performance of off‐the‐shelf LLMs in identifying PWID and related attributes from hospital discharge summaries. Methods: In this cross‐sectional study, discharge summaries from the Infectious Diseases service at St Vincent's Public Hospital, Sydney, between 2018 and 2022 were reviewed. A single reviewer manually annotated each de‐identified summary for PWID status, drugs reported, injection recency and opioid agonist therapy. Eight LLMs (Gemma3, Llama 3.3, Mistral, Phi4, hippomistral, llama3‐med [8B and 70B] and OpenBioLLM) were compared using prevalence‐weighted average‐F1 scores. Diagnostic metrics with bootstrapped 95% confidence intervals were calculated for each annotated category. Results: Of 859 first admissions, manual review identified 149 (17.1%) PWID. ICD codes showed low sensitivity (≤ 0.32) but high specificity (≥ 0.97) for identifying PWID. The best‐performing model (Llama 3.3) achieved a prevalence‐weighted average‐F1 of 0.845 (0.733, 0.927). For injecting drug use, sensitivity was 0.819 (95% CI 0.753, 0.879) and specificity 0.999 (0.996, 1.00). Identification of heroin, methamphetamine, cannabis and methadone was near perfect (F1 > 0.973), while illicit prescription opioid and benzodiazepine use were identified less accurately (F1 = 0.400 and 0.606). Discussion and Conclusions: LLMs accurately identify PWID from discharge summaries, outperforming ICD codes. Challenges remain for certain substances, underscoring the need for task‐specific tuning, external validation and integration with structured data to enhance surveillance and interventions.

Background: Despite increasing hospitalization for injecting-related infection, there has been limited large-scale evaluation of in-hospital and post-discharge outcomes. This study aimed to characterize population-level injecting-related infection hospitalization and correlates associated with patient-directed discharge, readmission, and all-cause mortality among persons who inject drugs with hepatitis C in New South Wales, Australia, between 2001 and 2022. Methods: Hepatitis C notifications in New South Wales were linked to data involving hospitalizations, opioid agonist treatment, incarceration, and death registration. Hospitalizations among people who inject drugs with injecting-related infections were identified by ICD-10 code algorithms. Incidence of patient-directed discharge, readmission, and mortality was calculated, and correlates associated with each outcome were assessed by extension of a Cox proportional hazards model for recurrent events. Results: In total, 18074 injecting-related infection hospitalizations were included among 9045 individuals, predominantly males (64%) with an average age of 41 years. The incidence was 47.2 per 100 person-years and increased over time. The proportion of hospitalizations ending in patient-directed discharge was 18% and was associated with stimulant use and incarceration, and lower in those with severe disease and opiate agonist therapy. The proportions of hospitalizations that were followed by 30-day and 1-year readmission were 25% and 61%, respectively, and had a strong association with patient-directed discharge. Mortality was 2% at 30 days and 15% at 1 year post-discharge. Conclusions: Patient-directed discharge was common among people admitted with injecting-related infections and was associated with readmission but not mortality. Person-centered models of care are necessary to address the health inequity experienced by people who inject drugs.

Background: Individuals with postacute sequelae of coronavirus disease 2019 (COVID-19), or long COVID, experience substantial burden of illness many months after initial infection. Few studies have comprehensively and longitudinally assessed health outcomes for people with long COVID following mild/moderate infection. We applied the Wilson-Cleary model of health-related quality of life (HRQOL) to describe the impact of long COVID on multiple health dimensions up to 24 months following mild/moderate COVID-19. Methods: Participants within the ADAPT post-COVID study (N = 172, 86% mild/moderate infection) completed structured patient-reported outcome measures at 4, 8, 12, 18, and 24 months postinfection. Following the Wilson-Cleary model, questionnaires assessed symptoms (anxiety/depression, chronic fatigue, breathlessness), return to pre-COVID-19 functioning, perceived health status (EuroQol Visual Analogue Scale), and wellbeing (EuroQol EQ-5D-5L, Personal Wellbeing Index). Temporal trends were assessed using general estimating equations and ordinal logistic regression, including time × long COVID interactions. Results: Thirty-seven percent of participants were diagnosed with long COVID (≥1 new/persisting symptoms of chest pain, breathlessness, or fatigue/malaise at least 12 weeks after infection). Long COVID was associated with poorer health outcomes across all domains at first assessment. Over 2 years, participants with long COVID reported improvement in return to pre-COVID-19 work and Somatic and Psychological Health Report chronic fatigue but sustained impairment was observed in all other health domains, compared to participants recovered from COVID-19. Conclusions: Substantial long-term impairment in various health domains were observed for individuals with long COVID following mild/moderate initial infection, with little improvement over time in most. Multimodal interventions must address impairment in multiple domains of HRQOL in individuals with long COVID.

Fatal and nonfatal overdoses involving opioids have increased to crisis levels in recent years. Laypersons have been increasingly tasked with responding to these events by administering naloxone, performing rescue breaths/cardiopulmonary resuscitation, and calling for medical assistance. However, little is known about the development of posttraumatic stress disorder (PTSD) related to opioid overdose responding among laypersons. To this end, we sought to determine the factors associated with PTSD stemming from responding to an opioid overdose event. From April 2021 to October 2021, structured interviews were conducted with layperson responders who had responded to an opioid overdose. Participants were administered structured diagnostic interviews, Clinician-Administered PTSD Scale for Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5), and completed validated self-report measures. In total, 101 layperson adults who responded to an opioid overdose were recruited for the present study (M_age: 34, 65% identified as women, 79% as White, and 11% as Hispanic/Latino). Of the 80 participants who completed the Clinician-Administered PTSD Scale for DSM-5, 100% met Criteria A for PTSD related to overdose responding, and over one quarter (27.5%) met current PTSD diagnosis criteria related to overdose responding. Current PTSD related to overdose responding was associated with depression symptoms, generalized anxiety symptoms, and presence of law enforcement or professional first responders during the most distressing overdose responding event. Responding to opioid overdoses is traumatizing for many and results in a considerable burden of PTSD among layperson responders. As such, we call for trauma-informed interventions that cater to the unique experiences of layperson opioid overdose responders. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

This study investigated the humoral and cellular immune responses in individuals with long COVID (LC) compared to age and gender matched recovered COVID-19 controls (MC) over 24-months. LC participants showed elevated spike and nucleocapsid IgG levels, higher neutralizing capacity, and increased spike- and nucleocapsid-specific CD4+ T cells, PD-1, and TIM-3 expression on CD4+ and CD8+ T cells at 3- and 8-months, but these differences did not persist at 24-months. Some LC participants had detectable IFN-γ and IFN-β, that was attributed to reinfection and antigen re-exposure. Single-cell RNA sequencing at 24-month timepoint revealed similar immune cell proportions and reconstitution of naïve T and B cell subsets in LC. No significant differences in exhaustion scores or antigen-specific T cell clones were observed. These findings suggest resolution of immune activation in LC and return to comparable immune responses between LC and MC over time. Improvement in self-reported health-related quality of life at 24-months was also evident in the majority of LC (62%). PTX3, CRP levels and platelet count were associated with improvements in health-related quality of life.

Background: Increasing numbers of opioid overdoses have been observed during the COVID-19 pandemic, likely reflecting the pandemic’s multiple effects on this already vulnerable population. People in recovery from opioid use disorder (OUD) have reported disproportionate psychosocial distress and isolation, as well as significant disruptions in access to treatment, including peer support, during the COVID-19 pandemic. Peer support is a key component of many evidence-based OUD recovery programs; it improves recovery capital, treatment engagement, and perceived social support and reduces psychosocial distress, particularly when used in conjunction with other evidence-based treatments, such as medication for OUD. Objective: This study aims to evaluate a novel mobile peer support app platform among a national sample of individuals in recovery from OUD as an adjunct to usual care during the COVID-19 pandemic. Methods: Individuals residing in the United States who are aged ≥18 years; own a smartphone; and self-report being in recovery for an OUD, being in treatment for an OUD (ie, in the past 30 days received prescribed methadone, naltrexone, or buprenorphine), or currently receiving some form of assisted recovery support (n=1300) will be recruited through online, targeted social media advertisements. Eligible participants will be randomly assigned (1:1) to a mobile peer recovery support intervention utilizing a novel smartphone-based app or to a control. Participants will complete 1 baseline survey and then a follow-up survey 1, 3, and 6 months after randomization. The primary aim of recovery capital will be determined by the change in recovery capital between study groups over the 6-month study period. We will also examine treatment engagement by using administrative data from a subset of individuals (n=650) residing in Rhode Island and Indiana. Results: As of June 2022, we enrolled 43 participants. Conclusions: If this mobile app demonstrates efficacy among a large national sample of patients, it has the potential to augment existing treatment programs, improve recovery capital, and reduce the disproportionate impacts of COVID-19 on this vulnerable population. Trial Registration: ClinicalTrials.gov NCT05405712; https://clinicaltrials.gov/ct2/show/NCT05405712

Study Objective: Emergency department (ED)‐based naloxone distribution and peer‐based behavioral counseling have been shown to be feasible, but little is known about utilization maintenance over time and clinician, patient, and visit level factors influencing implementation.Methods: We conducted a retrospective cohort study of an ED overdose prevention program providing take‐home naloxone, behavioral counseling, and treatment linkage for patients treated for an opioid overdose at two Rhode Island EDs from 2017 to 2020: one tertiary referral center and a community hospital. Utilizing a Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE‐AIM) framework, we evaluated program reach, adoption, implementation modifiers, and maintenance using logistic and Poisson regression.Results: Seven hundred forty two patients were discharged after an opioid overdose, comprising 966 visits (median: 32 visits per month; interquartile range: 29, 41). At least one intervention was provided at most (86%, 826/966) visits. Take‐home naloxone was provided at 69% of visits (637/919). Over half (51%, 495/966) received behavioral counseling and treatment referral (65%, 609/932). Almost all attending physicians provided take‐home naloxone (97%, 105/108), behavioral counseling (95%, 103/108), or treatment referral (95%, 103/108) at least once. Most residents and advanced practice practitioners (APPs) provided take home naloxone (78% residents; 72% APPs), behavioral counseling (76% residents; 67% APPs), and treatment referral (80% residents; 81% APPs) at least once. Most clinicians provided these services for over half of the opioid overdose patients they cared for. Patients were twice as likely to receive behavioral counseling when treated by an attending in combination with a resident and/or APP (adjusted odds ratio: 2.29; 95% confidence interval, 1.68, 3.12) compared to an attending alone. There was no depreciation in use over time.Conclusions: ED naloxone distribution, behavioral counseling, and referral to treatment can be successfully integrated into usual emergency care and maintained over time with high reach and adoption. Further work is needed to identify low‐cost implementation strategies to improve services use and dissemination across clinical settings.