Publications

IMPORTANCE: Efforts to improve heart failure (HF) outcomes have focused on prescribing guideline-directed medical therapy at hospital discharge. Whether new prescriptions translate to sustained medication use after HF hospitalization is largely unknown.

OBJECTIVE: To characterize medication-use patterns following HF hospitalizations.

DESIGN, SETTING, AND PARTICIPANTS: A cohort study of patients in a regional US health system discharged home after an HF hospitalization between July 1, 2017, and March 30, 2023. Data were analyzed from August 2024 to February 2026.

MAIN OUTCOMES AND MEASURES: Medications of interest included β-blockers, renin-angiotensin system inhibitors (RASis), mineralocorticoid receptor antagonists (MRAs), and sodium-glucose co-transporter 2 inhibitors (SGLT2is). New prescriptions were identified by in-hospital administration or discharge medication orders. Using electronic health records linked to pharmacy dispensation records, medication initiation (prescription dispensation) and primary nonadherence (discharge prescriptions that were not filled) were assessed at 7 and 90 days postdischarge. Adherence (proportion of days covered ≥80%) and persistence (continuous days' supply) were assessed over 6 months postdischarge.

RESULTS: This cohort study included 6111 patients with HF hospitalizations (mean [SD] age, 69.9 [13.6] years; 37.3% female [2277]) of whom 51% of were prescribed at least 1 new HF medication at discharge. At admission, 58.1% of patients were prescribed β-blockers (3549), 41.4% RASis (2530), 16.3% MRAs (997), and 4.9% SGLT2is (299); and at discharge, use increased to 73.3% (4481), 52.9% (3232), 28.5% (1740), and 9.0% (548), respectively. Primary nonadherence was observed in 51% (972 of 1904) of new prescriptions for β-blockers, 48% (659 of 1361) for RASis, 39% (482 of 1225) for MRAs, and 35% (134 of 383) for SGLT2is. Thus, of 4873 new discharge prescriptions, only 54% (2626) were initiated within 7 days of discharge and an additional 20% (954) had delayed initiation, between 7 and 90 days postdischarge. At 6 months, persistence to β-blockers was 70% (3127 of 4481), for RASis was 60% (1952 of 3232), for MRAs was 55% (961 of 1740), and for SGLT2is was 56% (306 of 548). As a result, at 6 months, only 42% (2188 of 5183) of patients were adherent and 51% (2620 of 5183) were persistent to all HF medications used at discharge.

CONCLUSIONS AND RELEVANCE: This cohort study found that following HF hospitalization, most new guideline-directed medical therapy prescriptions went unfilled. Moreover, most newly initiated prescriptions did not persist at 6 months, suggesting a need to develop interventions to support patients' medication use beyond the initial prescription at discharge.

IMPORTANCE: The 2026 American Heart Association/American College of Cardiology/multisociety guideline on the management of dyslipidemia issued new recommendations on estimating atherosclerotic cardiovascular disease (ASCVD) risk and on populations eligible for statins for primary prevention.

OBJECTIVE: To assess the population health impact of the 2026 guideline on primary prevention statin therapy.

DESIGN, SETTING, AND PARTICIPANTS: Nationally representative, cross-sectional sample of nonpregnant adults aged 30 to 79 years without known ASCVD, who participated in the National Health and Nutrition Examination Survey from 2017 to 2023. Data were analyzed from March to May 2026.

MAIN OUTCOMES AND MEASURES: Changes in eligibility for primary prevention statin therapy, comparing the 2026 and 2018 lipid guidelines.

RESULTS: The weighted sample included 4366 NHANES participants representative of 154.5 million US adults (weighted mean age, 51 years; 52.0% female). Of these, 5.5% (95% CI, 4.7%-6.6%) had untreated low-density lipoprotein cholesterol below 70 mg/dL, 17.8% (95% CI, 16.3%-19.5%) reported currently taking statins, and 8.6% (95% CI, 7.6%-9.8%) met criteria for statin eligibility independent of ASCVD risk estimation based on a low-density lipoprotein cholesterol of 190 mg/dL or greater, diabetes, or chronic kidney disease. The remaining 68.0% of patients (95% CI, 65.9%-70.0%) met guideline criteria for using ASCVD risk estimation to guide statin decisions. In total, an estimated 87.5 million (56.6% [95% CI, 54.2%-58.9%]) nonpregnant US adults aged 30 to 79 years were statin eligible based on the 2026 guideline, including 21.5 million (13.9% [95% CI, 12.5%-15.5%]) who were newly statin eligible. More than 93% of adults aged 70 to 79 years and 85% of adults aged 60 to 69 years are eligible for primary prevention statin therapy compared with 11% of adults aged 30 to 39 years. Newly statin-eligible populations were largely younger and lower risk than populations previously recommended statin therapy (mean estimated 10-year ASCVD risk, 3.1% [95% CI, 2.7%-3.5%] for newly statin-eligible individuals vs 6.1% [95% CI, 5.8%-6.4%] for individuals previously eligible for statin therapy).

CONCLUSIONS AND RELEVANCE: The 2026 dyslipidemia guideline substantially expands the US population recommended for primary prevention statin therapy, predominantly in lower-risk individuals.

Substance use disorder (SUD) is a complex chronic condition requiring a multi-disciplinary approach to both research and treatment. Randomized controlled trials (RCTs) are gold standard methodologies for inferring causal relationships between an intervention and treatment outcomes but often face challenges in generalizability, scalability and real-world implementation. Target trial emulation (TTE) is a powerful methodological framework that uses observational or real-world data sources to emulate the methodology of these gold standard target trials to complement the learning from RCTs and enhance translation to real world evidence. An additional methodological innovation is the translational testing of clinical- and community-based digital health systems to provide new insights into SUD in the real world and provide scalable access to therapeutic resources. To explore these methodological innovations in SUD research, the National Institute on Drug Abuse Center for the Clinical Trials Network convened a variety of experts for a virtual workshop titled "Target Trial Emulation in Observational Research and Translational Testing of Advanced Digital Health Tools for Substance Use Disorder Prevention and Treatment." This article summarizes the discourse of the workshop, focused on three thematic areas: TTE using real-world healthcare data, SUD evidence from nationwide data sources that may be useful in TTE analyses, and translational testing of clinical- and community-based digital health systems. The workshop also highlighted various exemplars of digital health systems that demonstrate success in translational research addressing SUDs, key methodological and translational challenges, importance of rigorous study design, robust data linkages and expanding use of common data elements, and the integration of digital health tools to enhance causal inference and clinical impact. Future research directions are outlined to refine these approaches, address barriers, and maximize the utility of real-world data in shaping effective SUD prevention and treatment strategies.

IMPORTANCE: There are large racial and ethnic differences in the use of medications for opioid use disorder (MOUD). Whether differences in geographic availability of MOUD providers (defined in this study as buprenorphine prescribers, methadone dispensing opioid treatment programs, and naltrexone prescribers) contribute to these differences in Medicaid is unknown.

OBJECTIVE: To examine differential geographic availability of MOUD in Medicaid and whether it is associated with MOUD use.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study analyzed the geographic availability of Medicaid prescribers of MOUD in 2021, spanning 10 states (Delaware, Kentucky, Maryland, Maine, Michigan, North Carolina, Pennsylvania, Tennessee, Virginia, and West Virginia) in the Medicaid Outcomes Distributed Research Network. The study population included Medicaid enrollees aged 18 to 64 not enrolled in Medicare and their MOUD providers. The data analysis was conducted from December 2022 to April 2026.

EXPOSURES: Geographic availability was measured at the zip code level (number of MOUD providers available in Medicaid within a 15-minute drive time per 100 Medicaid enrollees).

MAIN OUTCOMES AND MEASURES: Main outcomes included the probability of buprenorphine, methadone, and naltrexone use as a function of enrollee race and ethnicity, whether they had above-median geographic availability, and interactions between above-median geographic availability and race and ethnicity.

RESULTS: The sample included 8 081 899 Medicaid enrollees; 472 409 (5.8%) had an OUD diagnosis. The population was 58.7% female and 41.3% male. Among the study population, 11.7% were aged 18 to 20 years, 39.5% were aged 21 to 34 years, 21.4% were aged 35 to 44 years, 14.5% were aged 45 to 54 years, and 12.9% were aged 55 to 64 years. Overall, 7.3% of the sample were Hispanic enrollees, 27.4% were non-Hispanic Black enrollees, 56.2% were non-Hispanic White enrollees, and 7.2% were enrollees from other racial and ethnic groups. Overall, 13 575 buprenorphine prescribers, 516 methadone dispensers, and 4801 naltrexone prescribers billed Medicaid. Median Medicaid MOUD providers available within a 15-minute drive were 0.89 per 100 enrollees for buprenorphine, 0.03 for methadone, and 0.32 for naltrexone. Above-median geographic availability of methadone was associated with a 0.99 (95% CI, 0.54-1.42)-percentage point increase in methadone use for non-Hispanic White enrollees; there was no such increase for non-Hispanic Black or Hispanic enrollees. Evidence of similar differences was limited for naltrexone. Above-median availability of buprenorphine was not associated with increased use of MOUD for any racial or ethnic group.

CONCLUSIONS AND RELEVANCE: In this cross-sectional study of 10 state Medicaid programs, greater geographic availability of MOUD was associated with increased use only for methadone and, to a lesser extent, naltrexone. No racial and ethnic groups experienced gains in use associated with improved access. Additional strategies beyond addressing geographic access may be needed to close racial and ethnic gaps in MOUD.

OBJECTIVES: The 2023 Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) equations for 10-year cardiovascular disease (CVD) were developed by the American Heart Association to improve upon the 2013 Pooled Cohort Equations (PCEs). This study sought to compare risk reclassification using the PREVENT equations with and without the use of optional predictors of hemoglobin A1c (HbA1c) and urine albumin-creatinine ratio (UACR).

RESEARCH DESIGN AND METHODS: This was a cross-sectional study of adults aged 30 to 79 years participating in the National Health and Nutrition Examination Survey 2015-2020. Ten-year CVD estimates stratified by diabetes status using the base PREVENT equations and the enhanced PREVENT equations including HbA1c and UACR were examined.

RESULTS: The study included 8,293 participants (weighted mean age 51 years; 52% female; 13% with diabetes) representing 147.9 million US adults. Estimated 10-year CVD risks using the base and enhanced PREVENT equations were within 5 percentage points for 99% of adults without diabetes and 82% of adults with diabetes. Individuals with HbA1c ≥9.0% had a higher enhanced equations' mean risk of 17.8% (95% CI, 15.9%-19.7%) compared to the base equations' mean risk of 11.8% (95% CI, 10.7%-12.9%). This was mirrored with UACR, as individuals UACR ≥300 mg/g had a higher enhanced equations' mean risk at 36.1% (95% CI, 33.2%-39.1%) compared to the based equations' mean risk of 23.4% (95% CI, 20.0%-26.9%).

CONCLUSIONS: Inclusion of HbA1c and UACR in the PREVENT equations provides opportunity for greater individualization of CVD risk estimation in adults with diabetes but little benefit in populations without diabetes.

BACKGROUND: Physical restraints and antipsychotics should only be used as a last resort in hospitalized individuals at risk of harm to themselves or others. Since verbal reorientation and de-escalation are first-line management in these scenarios, patients with limited English proficiency (LEP) may be at higher risk of physical restraint and antipsychotic use. We aimed to examine the relationship between English proficiency, physical restraints, and antipsychotic use in hospitalized adults.

METHODS: Retrospective cohort study of adults discharged from an academic medical center between 1/2019 and 6/2023. LEP was defined by non-English primary language listed in the electronic medical record. The primary outcome was use of either physical restraints or antipsychotics during the hospitalization, defined by orders and pharmacy charges, respectively. Multivariable generalized estimating equations were used to examine the relationship between primary language and outcomes, controlling for demographic, hospitalization, and clinical characteristics.

RESULTS: The cohort consisted of 132,767 hospitalizations (mean age 62.9 years, 48.3% female, 10.8% with LEP). Either physical restraints or antipsychotics were used in 14,802 (11.2%) hospitalizations. Unadjusted outcome rates in LEP and non-LEP groups were similar (10.1% vs. 11.3% for either physical restraints or antipsychotics, 7.9% vs. 8.9% for physical restraints, and 4.9% vs. 4.9% for antipsychotics). However, LEP patients had higher adjusted odds of either physical restraint or antipsychotic use (aOR 1.18; 95% CI, 1.08-1.28), physical restraint use (aOR 1.30; 95% CI 1.17-1.43), and antipsychotic use (aOR 1.12; 95% CI 1.01-1.25). Subgroup analysis among patients with diagnosed delirium revealed an even stronger association between LEP and the primary outcome (aOR 1.42; 95% CI 1.22-1.65).

CONCLUSION: This single center study found language-based disparities in orders for physical restraints and antipsychotics among hospitalized patients. Given the potential risks associated with both physical restraints and antipsychotics, causes and mitigators of this disparity should be investigated.

IMPORTANCE: Drug shortages are an ongoing challenge in health care delivery. Research has focused on factors associated with manufacturer-level shortage reports; however, these reports do not reflect population-level drug availability. It is essential to determine the impact of supply chain events on trends in drug utilization and identify factors associated with decreases in use to ascertain shortage risk and inform future policy prioritization.

OBJECTIVE: To quantify incident shortages following Canadian supply chain events and determine drug characteristics associated with increased shortage risk.

DESIGN, SETTING, AND PARTICIPANTS: This matched cohort study analyzed monthly trends in drug purchasing of Canadian pharmaceutical products from the IQVIA MIDAS database from January 2017 to December 2021 with a 6-month follow-up. Data were analyzed from September 2023 to September 2024.

EXPOSURE: Reporting of a supply chain issue in Drug Shortages Canada.

MAIN OUTCOMES AND MEASURES: The main outcome was an incident shortage, defined as a decrease of 33% or more in purchase volume within 6 months following the supply chain issue. A logistic regression with random effects compared the odds of experiencing an incident shortage for cases and controls. A secondary outcome, shortage intensity, defined as the decrease in purchases multiplied by the duration, was analyzed using a zero-inflated beta regression with random effects.

RESULTS: Of 1591 drugs, 972 unique drugs (61%) were exposed to 1919 supply chain events. Incident shortages were observed in 11% of cases (216 drugs), compared with 7% of controls (1271 comparators). Drugs with baseline sales less than $100 000 (odds ratio [OR], 4.01; 95% CI, 3.03-5.30), anti-infective drugs (OR, 3.07; 95% CI, 2.33-4.04), and over-the-counter drugs (OR, 3.05; 95% CI, 2.38-3.90) had greater odds of experiencing an incident shortage. Schedule G drugs, which include controlled drugs (OR, 3.61; 95% CI, 1.66-7.77), drugs with more than 25 unique products available (OR, 2.86; 95% CI, 1.89-4.44), and those with equal dominance in both retail and hospital sectors (OR, 2.43; 95% CI, 1.93-3.01) had the largest magnitudes of association with higher shortage intensity.

CONCLUSIONS AND RELEVANCE: In this study of shortage incidence following Canadian supply chain events, 9 in 10 reported supply chain events did not experience true shortages. These findings highlight a need for a unified shortage definition focusing on real disruptions in supply and can be used to guide policies to manage drug shortages, improve patient outcomes, and enhance health care delivery.