Purpose Secondary prevention of stroke requires effective cardiovascular risk factor (RF) management, including blood pressure control, lipid lowering treatment and antithrombotic therapy to reduce the risk of recurrent stroke. Optimal management of risk factors is challenging in older, frailer people, who also have worse outcomes after a stroke. There is limited evidence evaluating secondary prevention quality and outcomes in frail populations. To understand the trends (and factors associated with inequity) in the provision of RF management we developed the `Evaluation of Population Secondary Prevention Quality and Outcomes Post-Stroke (ESP-QOPS) in Wales' e-cohort. The aims of this project include describing (i) the provision of post-stroke evidence-based risk factor management (ii) trends in frailty and outcomes, and (iii) evaluating the impact of suboptimal risk factor provision on adverse outcomes stratified by frailty status. Participants We identified 71,696 patients with a first recorded stroke diagnosis in primary or secondary care between 2010-2023 in Wales. Findings to Date The incidence of stroke gradually increased from 4,968 to 5,391 patients per year between 2010-2023, with 70% of diagnoses made in the secondary care setting. Of diagnoses made in primary care, less than 20% were subsequently admitted to secondary care.

Abstract Objective Social Determinants of Health (SDoH) are critical to patient care and population health. Despite their importance, SDoH information is frequently embedded within unstructured clinical text such as patient-reported information or social worker notes, which limits its use on clinical decision-making and resource allocation. Although transformer-based models represent the current state of the art, their scalability, computational requirements, and limited transparency pose barriers to large-scale multi-site clinical implementation. In this context, rule-based NLP systems remain valuable, particularly when explainability, reproducibility, and rapid customization are essential. Methods MedSDoH was developed within the Open Health Natural Language Processing (OHNLP) Framework using literature-derived SDoH resources, standardized domain definitions, and expert-curated rulesets. Large language models (LLMs) were used during development to assist with rule generation and lexicon expansion. Rules were iteratively refined against a gold-standard annotated corpus from two health systems and then evaluated on independent datasets. Result The final system included 942 regular expression rules spanning 22 SDoH domains. On validation on two external datasets, MedSDoH demonstrated generalizability and comparable performance across sites. The system has been made publicly available so research community can collaboratively contribute to the maintenance and extension through disease- or site-specific adaptations. Conclusion MedSDoH is a computationally efficient and open-source system for large-scale SDoH extraction from clinical text. It is well-suited for multi-site adaptation and deployment in resource-constrained settings.

Abstract Background and aims To assess the effectiveness of blood pressure (BP) and lipid management in the year following a first diagnosis of stroke. Methods A retrospective observational cohort study using routine electronic health-records of 52,400 incident stroke patients with at least one year follow-up in Wales (2010-23), including 33,080 ischaemic strokes (IS), 4,403 haemorrhagic stroke (HS) and 14,920 unspecified strokes (US) with at least one year follow-up in Wales (2010-23). Results Altogether, 17% of IS patients, 59% of HS and 23% of US were not prescribed any lipid lowering therapy (LLT) in the year post stroke. Only 22% of IS, 9% of HS and 20% of US had a documented LDL-C <1.8mmol/L. In multivariable regression analyses, a history of diabetes and prescription of LLT were independently associated with a greater likelihood of a documented LDL-C level; male sex and LLT were associated with a higher likelihood of documentation of LDL-C <1.8mmol/L. Overall, 84% ,79% and 85% of IS, HS and US patients respectively, had BP documented in their primary care record and 67%, 64% and 67% had a documented BP <140/90mmHg in the year following their stroke. Prescription for antihypertensives and history of diabetes mellitus were independently associated with a greater likelihood of a documented BP in the year following diagnosis in multivariable regression analyses. Conclusions Testing, treatment and control of lipids and blood pressure were suboptimal amongst these very high-risk patients. These data show significant opportunity for improving secondary prevention of stroke in this high risk population. Conflict of interest

Opioid mortality in the United States exhibits pronounced spatial variation within major metropolitan regions. Whether these differences shift sharply at municipal boundaries where governance structures and service systems change remains poorly understood. This study examines whether opioid involvement in drug-related deaths changes discontinuously at the boundary between the City of Chicago and suburban Cook County, Illinois and assesses whether cold-related deaths display similar spatial patterns. Mortality microdata from the Cook County Medical Examiner’s Office were linked with geocoded death locations, and a spatial regression discontinuity design was applied using the Chicago municipal boundary as a sharp institutional cutoff. The analysis reveals a consistent increase in opioid involvement on the Chicago side of the boundary, indicating that localized risk conditions may extend beyond the urban core and align with jurisdictional transitions. Cold-related deaths show a comparable boundary pattern and display additional concentrations in areas characterized by prolonged outdoor exposure. Although interaction estimates suggest that cold exposure may heighten contrasts across the boundary, this pattern is less meaningful due to the smaller number of cold-related cases. These findings illustrate that mortality risk can shift at administrative boundaries even when environmental and socioeconomic conditions remain spatially continuous. The results highlight how institutional capacity and anthropogenic environmental factors, such as inadequate shelters and warming facilities, structure localized patterns of overdose harm within a shared metropolitan region and demonstrate the value of boundary-focused approaches for understanding spatial disparities in opioid-related mortality.

Abstract Background and aims To assess the provision of antithrombotic therapy in the year following a first diagnosis of stroke, using a Wales wide databank. Methods A retrospective observational cohort study using routine electronic health records of 52,400 incident stroke patients in Wales between 2010 and 2023 with at least one year follow-up. Stroke survivors were classified as ischaemic, haemorrhagic or unspecified and were assessed being prescribed either antiplatelet (AP), antitcoagulant (AC), both or none. A further analysis in patients with atrial fibrillation (AF) was also conducted. Results In ischaemic stroke 67.5% were taking AP, 18.4% AC, 7.9% both and 6.1% none. In haemorrhagic stroke 19.2% AP, 10.4% AC, 2.8% Both and 68% none and in unspecified stroke 69.8% AP, 12.8% AT, 5.8% both and 11.6% none. In people with AF and ischaemic stroke 12.1% were taking AP, 68% AC, 16.6% both and 3.3% none. In AF and haemorrhagic stroke 13.1% AP, 44.1% AC, 9.2% both and 34.6% none and in AF and unspecified stroke 17.4% AP, 62.88% AC, 15.1% Both and 4.7% none. Conclusions Approximately two thirds of ischaemic stroke survivors with and without AF appeared to be on the appropriate antithrombotic management. Antithrombotic management was far more varied in people with haemorrhagic or unspecified stroke. These data highlight both the benefits to be gained from better targeting of treatment, identified suboptimum prescribing of treatments proven to improve outcomes in post-stroke management. Conflict of interest

Unlabelled:Public health data dashboards have substantial potential to improve transparency, understanding, and decision-making at multiple levels, from individuals to public health practitioners and policymakers. However, creating effective dashboards presents many challenges. In this case-based tutorial on public health dashboard development, we share lessons learned from our experience developing data dashboards for the HEALing Communities Study (HCS), a National Institutes of Health (NIH)-funded, community-engaged intervention to deploy evidence-based practices to reduce opioid overdose deaths in 67 communities across 4 states. We present key decision points dashboard teams must address, along with the major considerations and trade-offs that shaped our approach. First, we describe core considerations of the who, what, why, where, when, and how of data dashboard development. Second, we outline steps in data curation, including the identification of key metrics and potential data sources and developing processes to acquire the data. Third, we discuss practical aspects of developing data visualizations that can effectively communicate key messages to the end users of interest. Fourth, we describe the infrastructure considerations to host and publish data dashboards. And finally, we discuss maintenance and sustainability of the dashboard. While the material can be read sequentially as a step-by-step guide, we refer to this resource as a "playbook" because readers may engage with specific domains in a random-access fashion, that is, based on their specific needs and/or starting point rather than a fixed sequence. The information, supplemental materials, and resources will assist individuals and organizations seeking to build data dashboards by fostering context-sensitive evaluation of design and implementation choices to realize the promise of data-driven decision-making.

BACKGROUND:Despite naloxone's proven effectiveness in reversing opioid poisoning, receipt rates among the high-risk Medicaid population remain insufficient and the patient- and system-level characteristics associated with this gap are poorly understood. OBJECTIVES:To describe the patient-and system-level characteristics associated with a naloxone prescription fill within 7 days of a nonfatal opioid poisoning among Medicaid beneficiaries. METHODS:Using the Merative™ MarketScan Multi-State Medicaid Database (2018-2022), we identified adult (≥18 years) beneficiaries with a nonfatal opioid poisoning. The primary outcome was a naloxone prescription claim within 7 days of the index poisoning. Multivariable logistic regression characterized associations between beneficiary characteristics and naloxone prescription fill. Robustness was assessed using last-event indexing RESULTS: Among 64,332 beneficiaries (mean [SD] age 38.6 [12.1] years), 2590 (4.0%) received naloxone within 7 days. After adjustment, the odds of naloxone receipt were higher among Black (adjusted odds ratio [aOR] 1.26, 95% CI 1.13-1.41) and Hispanic (aOR 1.22, 95% CI 1.01-1.47) beneficiaries relative to White, and among those with a prior mental health disorder (aOR 1.23, 95% CI 1.12-1.34). Greater comorbidity was associated with higher odds relative to a Charlson score of 0 (1-2: aOR 1.14, 95% CI 1.03-1.25; 3-4: aOR 1.23, 95% CI 1.05-1.45 and 1.20, 95% CI 1.01-1.42). Odds were lower among older beneficiaries, particularly those aged ≥65 years (aOR 0.33, 95% CI 0.22-0.48), and those in capitated (managed care) plans (aOR 0.57, 95% CI 0.53-0.62). Findings were largely consistent across the sensitivity analysis. Prior opioid poisoning, more common under last-event indexing (≥1 episode in 65.5%, vs. 0.8%), was not associated with naloxone receipt (aOR 1.04, 95% CI 0.96-1.11). CONCLUSIONS:Naloxone remains underprescribed for Medicaid beneficiaries after nonfatal opioid poisoning, with receipt varying by demographic, clinical, and insurance characteristics. Timely naloxone should be offered to all in need, regardless of patient or system characteristics.

Public health data and tools have proliferated, yet practical guidance for community-engaged data-driven decision making is limited. The HEALing Communities Study (HCS) was a randomized, wait-list controlled trial to assess the impact of an intervention to reduce fatal opioid overdoses in 67 highly affected communities across 4 sites (Kentucky, Massachusetts, New York, and Ohio). HCS researchers implemented the Communities That HEAL intervention, a phased approach which included a coalition-engaged, data-driven approach to selection of evidence-based practice strategies to reduce fatal opioid overdoses. Core steps to the data-driven approach included data selection, access, display, and engagement. Staff selected metrics that aligned with study goals, accessed data from numerous sources, created visualizations, and engaged coalition members to assess resource gaps and intervention opportunities. At the intervention conclusion, all 4 sites' staff collectively workshopped best practices and barriers encountered to data-driven decision making. This article explains the data-driven decision-making approach implemented, assessment results, alterations for subsequent implementation, and guidance for future implementations.

Herpes zoster (HZ) is a painful condition that is managed with a combination of antivirals, steroids, and pain medications. Nursing home residents with Alzheimer’s disease and related dementias (ADRD) may have difficulty expressing HZ symptoms, thus affecting medication management. This study compared prescription and non-prescription medication use in nursing home residents with HZ by ADRD diagnosis. We conducted a retrospective cohort study using nursing home electronic health record data from 2017-2022. Residents with HZ and ADRD were identified using ICD-10 codes as of the HZ diagnosis date. Administration for pain medications, antivirals, and steroids were measured before and after the HZ diagnosis date in periods of time (e.g., 1-7, 8-14, and 15-30 days before/after diagnosis). To assess relative changes in medication prevalence over time, administrations 15-30 days prior to HZ represented the baseline period. We identified 4,719 residents with HZ (31.5% had ADRD). Among residents without ADRD, oxycodone increased from 1.6% to 3.4% (vs. ADRD=0.5% to 0.8%) and gabapentin from 13.4% to 33.9% (vs. ADRD=12.6% to 25.6%) between baseline and 0-7 days after HZ diagnosis. In the 0-7 days after HZ diagnosis, prednisone use was 1.74 times greater in residents without ADRD (10.3%) vs. with ADRD (5.9%), and acyclovir use was similar between groups (without ADRD=28.5%; with ADRD=27.1%). Differences in medications, especially prednisone, between residents with and without ADRD may reflect undertreatment of HZ symptoms in residents with ADRD.

Objectives: Access to substance use treatment is critical to reduce the risk of opioid overdose. However, analyzing geographic accessibility using singular measures may provide imprecise estimates. Our objective was to examine changes in geographic accessibility to buprenorphine providers across three HEALing Communities Study (HCS) states, Kentucky, Massachusetts, and Ohio in the US, using multiple spatial analyses and determining whether disparities in access existed across geographic regions and sociodemographic indicators. Methods: We used buprenorphine provider data from the Drug Enforcement Administration database in January 2020 (reference point) and June 2022. We used four methods to examine changes in geographic accessibility to buprenorphine: 1) kernel density analysis; 2) change in rates analysis (at the county or municipality level); 3) minkowski distance (drive-time to the nearest provider); 4) enhanced 2-step floating catchment area (E2SFCA) and spatial regression analyses, controlling for HCS communities (Wave 1 or Wave 2) and other sociodemographic factors. Results: We identified an increase in geographic accessibility to buprenorphine across all three states. Improvements in accessibility were concentrated in areas that had higher reference point access. While our regression model of change in the E2SFCA accessibility index shows that Wave 1 communities in Kentucky had a higher increase in access compared to the state average increase, we were not able to establish consistent associations between HCS communities and changes in the E2SFCA index. Conclusion: Since geographic accessibility to buprenorphine treatment is critical, additional efforts are needed to improve access to buprenorphine providers in rural areas and areas with limited baseline access.

Background: Atherosclerotic cardiovascular disease (ASCVD) most often manifests initially with a myocardial infarction or stroke. While the aetiology of atherosclerosis and benefits of risk factor (lipids and blood pressure) management are well established, less is known about the effectiveness of ASCVD risk assessment and management in the period leading up to such events. Aims: To identify patients with a major coronary or cerebrovascular event at first presentation of ASCVD between 2010-23 and assess identification and management of ASCVD risk in the year preceding the event. Methods: We conducted a retrospective, population-level,observational study using linked anonymised health record data amongst 102,148 patients with a fatal or non-fatal major coronary or cerebrovascular event at first ASCVD diagnosisin Wales (UK). Trends in the documentation of QRISK 10-year risk score, management of LDL-C and blood pressure(BP), and prescription of lipid lowering therapy (LLT) were identified in the year preceding presentation. Results: Documentation of LDL-C in the year prior to presentation increased from 27.8% to 38.5% between 2010-23, of which the proportion with a level of <1.8 mmol/L increased from 4.5% to 9.5% (Figure 1A). Prescription of LLT increased from 26.0% to 33.8% between 2010-23, with an increase in the prescribing of high-intensity statin from 2.0% to 6.7% (Figure 1B). Documentation of BP decreased from 55.2% to 49.8% between 2010-23, and achievement of systolic and diastolic BP <140&90 mmHg increased, thoughonly from 26.5% to 27.4% (Figure 1C). Documentation of QRISK score within 5-years prior to presentation increased from 16.1% to 27.7% between 2014-23, of which 28.1% scored >20% in 2023 (Figure 1D). Among patients with a recorded QRISK of >20%, only 19.1% were prescribed LLTin 2014, increasing to 27.5% in 2023 (Figure 1E). Conclusions: Although management of lipids has improved over time in the primary prevention setting, the effectiveness of control of patient's risk factors remains suboptimal at a population level. The majority of patients suffering MI or stroke as their initial presentation of CVD do not have their BP or lipids tested (or controlled when tested) in the year preceding initial presentation with these acute vascular events.Improvements are required in identification of those at risk and management of modifiable risk factors to meet guideline-recommended standards of care at the population level.

Background: In patients with atherosclerotic cardiovascular disease (ASCVD), guidelines recommend a reduction in low-density lipoprotein cholesterol (LDL-C) to target levels with statin therapy as required. Where targets are not achieved, escalation to higher intensity statin +/- additional lipid-lowering therapy (LLT) is recommended. It is unknown to what extent adjustments in LLT are made after LLT initiation in routine practice, in relation to measured LDL-C levels. Aims: To document (i) testing and (ii) achievement of ESC guideline-recommended LDL-C levels (<1.8 mmol/L), in relation to (iii) prescribed LLT regimen during the first two years post-ASCVD diagnosis. Methods: A retrospective, population-level, observational study using linked anonymised population electronic health record (EHR) data amongst 178696 patients diagnosed with ASCVD between 2010-22, with at least one year of follow-up data. Patients were followed up for two years post-diagnosis, and censored at the point of death, migration, or loss of linked primary care data. Results: Within 90-days post-diagnosis only 36318 (20.3%) patients had LDL-C levels documented in their EHR, of which 7.3% were <1.8 mmol/L. Of those 23446 (13.1%) patients with LDL-C ≥1.8 mmol/L, 12.7% were receiving high-intensity statin therapy or combination LLT (HI), 40% had lower-intensity statin therapy or other LLT monotherapy (LI) and 47.3% were not treated (Figure 1). During the first-year post-diagnosis, only 4222 (2.4%) had LDL-C tested and 929 (22%) of these achieved target. Of those not achieving target, 15.9% were prescribed HI-LLT, 43.4% LI-LLT, and 40.7% were not treated (Figure 1). Among patients with no test at one-year, 48.1% were prescribed LLT and 51.9% were not treated. Of those not treated, 78.7% received no test or treatment by two-years post-diagnosis. By two-years post-ASCVD diagnosis, 12.5%, 27.2%, and 54.5% were prescribed HI, LI, and no LLT respectively. Figure 1 illustrates that most patients were not tested following 90-days post-diagnosis and remained on the initial LLT regime over the subsequent two-years. Conclusions: Management of lipids was below guideline-recommended standards in the early years following ASCVD diagnosis in the population, with low levels of testing and escalation in lipid treatment regimen uncommon, even when indicated by documented LDL-C levels. A more rigorous system-wide approach to LLT is required to maximise the potential benefits among these very high-risk patients.

The ability to predict drug overdose risk from a patient's medical records is crucial for timely intervention and prevention. Traditional machine learning models have shown promise in analyzing longitudinal medical records for this task. However, recent advancements in large language models (LLMs) offer an opportunity to enhance prediction performance by leveraging their ability to process long textual data and their inherent prior knowledge across diverse tasks. In this study, we assess the effectiveness of Open AI's GPT-4o LLM in predicting drug overdose events using patients' longitudinal insurance claims records. We evaluate its performance in both fine-tuned and zero-shot settings, comparing them to strong traditional machine learning methods as baselines. Our results show that LLMs not only outperform traditional models in certain settings but can also predict overdose risk in a zero-shot setting without task-specific training. These findings highlight the potential of LLMs in clinical decision support, particularly for drug overdose risk prediction.

The overuse of antipsychotics in persons with dementia in long-term care (LTC) has been a source of clinical concern, public attention, and policy intervention for over 30 years. Targeted quality improvement, broader awareness of risks, and other initiatives have resulted in substantial reductions in antipsychotic use in LTC settings in North America and elsewhere. Limited evidence suggests that reductions in antipsychotic use may be resulting in unintended consequences, such as substitution with alternate, but similarly harmful, psychotropic medications. We used a retrospective, matched cohort study design using linked population-based health care databases held at ICES. LTC residents 66 years or older with dementia who were prescribed an inappropriate (i.e., without an indication aligning with the definition of appropriate antipsychotic use in LTC, including schizophrenia, Huntington’s disorder, hallucinations, delusions, or end-of-life care) antipsychotic medication with at least 6-months of continuous use were identified over a 10-year period (2008-18). Antipsychotic users who subsequently discontinued an antipsychotic medication were matched 1:1 to persistent users on key variables and followed for up to 1-year following antipsychotic discontinuation for new prescriptions of one or more psychotropic medications and clinical outcomes. Among 26,092 LTC residents with dementia (mean age of 84 years) who were prescribed an inappropriate antipsychotic medication, 5,854 (22%) discontinued during the follow-up period. After adjusting for key variables, new psychotropic medication prescription was not more common among antipsychotic discontinuers in the 6-months following antipsychotic discontinuation compared to those who continued an antipsychotic medication (hazard ratio (HR) 0.90; 95% CI, 0.70-1.15). Mortality was similar between the two groups (HR 1.01; 95% CI 0.90-1.15) at up to 1-year following antipsychotic discontinuation. Antipsychotic discontinuation in this study was not associated with medication substitution. These results supports other studies indicating that antipsychotic discontinuation in dementia can be safe but has questionable effect on mortality. While medication use trends in LTC have shown increases in other psychotropic medication use alongside antipsychotic reductions, this study suggests that this may be driven by factors other than substitution, such as increasing complexity of LTC residents, including higher prevalence of mental health disorders other than dementia.

Importance: Ecological and observational studies have shown a protective association between herpes zoster (HZ) vaccination and dementia risk, yet many had methodological limitations or examined the live HZ vaccine that is no longer available in the US. Improved access to linked electronic health records for patients receiving post-acute care and long-term care permit robust comparisons of dementia risk in adults eligible to receive the recombinant HZ vaccine. Objective: Emulate a randomized trial in observational data to estimate the association of the recombinant HZ vaccine (RZV) with incident dementia risk among older adults newly admitted for post-acute or long-term care in nursing homes (NHs). Design: Retrospective cohort study with target trial emulation and the clone censor approach. Setting: U.S. NHs that use PointClickCare as their electronic health record. Participants: Individuals who were admitted to a NH between 01/01/2017-12/31/2022; Medicare fee-for-service beneficiaries; did not have prevalent dementia; and eligible to receive RZV as of admission. Exposures: Receive one or more RZV doses within one year of admission vs. do not receive any RZV over four years of follow-up. Results: We identified 509,926 eligible NH residents (mean age 79 years; 36% men). Among those alive, uncensored, and without dementia at 12 months of follow-up, 8,843 received one or more doses of RZV. Receipt of RZV within one year of NH admission was associated with a 5.8% lower absolute risk (95%CI: -3.9% to -7.5%) of newly diagnosed dementia over four years (risk ratio [RR] = 0.76 [95%CI: 0.69-0.84]; cumulative incidence in 1+ RZV vs. no RZV: 18.8% vs. 24.6%). Associations were smaller in men (RR=0.82 [95%CI: 0.68-1.01]) and those with prior live HZ vaccination (RR=0.86 [95%CI: 0.65-1.09]). Bias analyses based on two negative control outcomes (NCOs) attenuated, but did not fully explain, the main effect of RZV on dementia risk (bias-adjusted RR = 0.82 [wellness visit NCO] and RR = 0.88 [hip fracture NCO]). Conclusions and Relevance: Administering RZV within 1 year of NH admission may reduce dementia risk. As RZV uptake was low overall, new NH residents would benefit from increased RZV vaccination uptake.