Publications

BACKGROUND: Active surveillance pharmacovigilance is an emerging approach to identify medications with unanticipated effects. We previously developed a framework called pharmacopeia-wide association studies (PharmWAS) that limits false positive medication associations through high-dimensional confounding adjustment and set enrichment. We aimed to assess the transportability and generalizability of the PharmWAS framework by using medical claims data to reproduce known medication associations with Clostridioides difficile infection (CDI) or gastrointestinal bleeding (GIB).

METHODS: We conducted case-control studies using Optum's de-identified Clinformatics Data Mart Database of individuals enrolled in large commercial and Medicare Advantage health plans in the United States. Individuals with CDI (from 2010 to 2015) or GIB (from 2010 to 2021) were matched to controls by age and sex. We identified all medications utilized prior to diagnosis and analysed the association of each with CDI or GIB using conditional logistic regression adjusted for risk factors for the outcome and a high-dimensional propensity score.

FINDINGS: For the CDI study, we identified 55,137 cases, 220,543 controls, and 290 medications to analyse. Antibiotics with Gram-negative spectrum, including ciprofloxacin (aOR 2.83), ceftriaxone (aOR 2.65), and levofloxacin (aOR 1.60), were strongly associated. For the GIB study, we identified 450,315 cases, 1,801,260 controls, and 354 medications to analyse. Antiplatelets, anticoagulants, and non-steroidal anti-inflammatory drugs, including ticagrelor (aOR 2.81), naproxen (aOR 1.87), and rivaroxaban (aOR 1.31), were strongly associated.

INTERPRETATION: These studies demonstrate the generalizability and transportability of the PharmWAS pharmacovigilance framework. With additional validation, PharmWAS could complement traditional passive surveillance systems to identify medications that unexpectedly provoke or prevent high-impact conditions.

FUNDING: U.S. National Institute of Diabetes and Digestive and Kidney Diseases.

BACKGROUND: Prior research documented racial and ethnic disparities in health care experiences within the Veterans Health Administration (VA). Little is known about such differences in VA-funded community care programs, through which a growing number of Veterans receive health care. Community care is available to Veterans when care is not available through the VA, nearby, or in a timely manner.

OBJECTIVE: To examine differences in Veterans' experiences with VA-funded community care by race and ethnicity and assess changes in these experiences from 2016 to 2021.

DESIGN: Observational analyses of Veterans' ratings of community care experiences by self-reported race and ethnicity. We used linear and logistic regressions to estimate racial and ethnic differences in community care experiences, sequentially adjusting for demographic, health, insurance, and socioeconomic factors.

PARTICIPANTS: Respondents to the 2016-2021 VA Survey of Healthcare Experiences of Patients-Community Care Survey.

MEASURES: Care ratings in nine domains.

KEY RESULTS: The sample of 231,869 respondents included 24,306 Black Veterans (mean [SD] age 56.5 [12.9] years, 77.5% male) and 16,490 Hispanic Veterans (mean [SD] age 54.6 [15.9] years, 85.3% male). In adjusted analyses pooled across study years, Black and Hispanic Veterans reported significantly lower ratings than their White and non-Hispanic counterparts in five of nine domains (overall rating of community providers, scheduling a recent appointment, provider communication, non-appointment access, and billing), with adjusted differences ranging from - 0.04 to - 0.13 standard deviations (SDs) of domain scores. Black and Hispanic Veterans reported higher ratings with eligibility determination and scheduling initial appointments than their White and non-Hispanic counterparts, and Black Veterans reported higher ratings of care coordination, with adjusted differences of 0.05 to 0.21 SDs. Care ratings improved from 2016 to 2021, but differences between racial and ethnic groups persisted.

CONCLUSIONS: This study identified small but persistent racial and ethnic differences in Veterans' experiences with VA-funded community care, with Black and Hispanic Veterans reporting lower ratings in five domains and, respectively, higher ratings in three and two domains. Interventions to improve Black and Hispanic Veterans' patient experience could advance equity in VA community care.

BACKGROUND: Valsartan was recalled by the US Food and Drug Administration in July 2018 for carcinogenic impurities, resulting in a drug shortage and management challenges for valsartan users. The influence of the valsartan recall on clinical outcomes is unknown. We compared the risk of adverse events between hypertensive patients using valsartan and a propensity score-matched group using nonrecalled angiotensin receptor blockers and angiotensin-converting enzyme inhibitors.

METHODS AND RESULTS: We used Optum's deidentified Clinformatics Datamart (July 2017-January 2019). Hypertensive patients who received valsartan or nonrecalled angiotensin receptor blockers/angiotensin-converting enzyme inhibitors for 1 year before and on the recall date were compared. Primary outcomes were measured in the 6 months following the recall and included: (1) a composite measure of all-cause hospitalization, all-cause emergency department visit, and all-cause urgent care visit, and (2) a composite cardiac event measure of hospitalizations for acute myocardial infarction and hospitalizations/emergency department visits/urgent care visits for stroke/transient ischemic attack, heart failure, or hypertension. We compared the risk of outcomes between treatment groups using Cox proportional hazard models. Of the hypertensive patients, 76 934 received valsartan, and 509 472 received a nonrecalled angiotensin receptor blocker/angiotensin-converting enzyme inhibitor. Valsartan use at the time of recall was associated with a higher risk of all-cause hospitalization, emergency department use, or urgent care use (hazard ratio [HR], 1.02 [95% CI, 1.00-1.04]) and the composite of cardiac events (HR, 1.22 [95% CI, 1.15-1.29]) within 6 months after the recall.

CONCLUSIONS: The valsartan recall and shortage affected hypertensive patients. Local- and national-level systems need to be enhanced to protect patients from drug shortages by providing safe and reliable medication alternatives.

IMPORTANCE: Accurately predicting major bleeding events in non-valvular atrial fibrillation (AF) patients on direct oral anticoagulants (DOACs) is crucial for personalized treatment and improving patient outcomes, especially with emerging alternatives like left atrial appendage closure devices. The left atrial appendage closure devices reduce stroke risk comparably but with significantly fewer non-procedural bleeding events.

OBJECTIVE: To evaluate the performance of machine learning (ML) risk models in predicting clinically significant bleeding events requiring hospitalization and hemorrhagic stroke in non-valvular AF patients on DOACs compared to conventional bleeding risk scores (HAS-BLED, ORBIT, and ATRIA) at the index visit to a cardiologist for AF management.

DESIGN: Prognostic modeling with retrospective cohort study design using electronic health record (EHR) data, with clinical follow-up at one-, two-, and five-years.

SETTING: University of Pittsburgh Medical Center (UPMC) system.

PARTICIPANTS: 24,468 non-valvular AF patients aged ≥18 years treated with DOACs, excluding those with prior history of significant bleeding, other indications for DOACs, on warfarin or contraindicated to DOACs.

EXPOSURES: DOAC therapy for non-valvular AF.

MAIN OUTCOMES AND MEASURES: The primary endpoint was clinically significant bleeding requiring hospitalization within one year of index visit. The models incorporated demographic, clinical, and laboratory variables available in the EHR at the index visit.

RESULTS: Among 24,468 patients, 553 (2.3%) had bleeding events within one year, 829 (3.5%) within two years, and 1,292 (5.8%) within five years of index visit. We evaluated multivariate logistic regression and ML models including random forest, classification trees, k-nearest neighbor, naive Bayes, and extreme gradient boosting (XGBoost) which modestly outperformed HAS-BLED, ATRIA, and ORBIT scores in predicting clinically significant bleeding at 1-year follow-up. The best performing model (random forest) showed area under the curve (AUC-ROC) 0.76 (0.70-0.81), G-Mean score of 0.67, net reclassification index 0.14 compared to 0.57 (0.50-0.63), G-Mean score of 0.57 for HASBLED score, p-value for difference <0.001. The ML models had improved performance compared to conventional risk across time-points of 2-year and 5-years and within the subgroup of hemorrhagic stroke. SHAP analysis identified novel risk factors including measures from body mass index, cholesterol profile, and insurance type beyond those used in conventional risk scores.

CONCLUSIONS AND RELEVANCE: Our findings demonstrate the superior performance of ML models compared to conventional bleeding risk scores and identify novel risk factors highlighting the potential for personalized bleeding risk assessment in AF patients on DOACs.

IMPORTANCE: Racial and ethnic disparities exist in anticoagulation therapy for atrial fibrillation (AF). Whether medical center racial and ethnic composition is associated with these disparities is unclear.

OBJECTIVE: To determine whether medical center racial and ethnic composition is associated with overall anticoagulation and disparities in anticoagulation for AF.

DESIGN, SETTING, AND PARTICIPANTS: Retrospective cohort study of Black, White, and Hispanic patients with incident AF from 2018 to 2021 at 140 Veterans Health Administration medical centers (VAMCs). Data were analyzed from March to November 2023.

EXPOSURE: VAMC racial and ethnic composition, defined as the proportion of patients from minoritized racial and ethnic groups treated at a VAMC, categorized into quartiles. VAMCs in quartile 1 (Q1) had the lowest percentage of patients from minoritized groups (ie, the reference group).

MAIN OUTCOMES AND MEASURES: The odds of initiating any anticoagulant, direct-acting oral anticoagulant (DOAC), or warfarin therapy within 90 days of an index AF diagnosis, adjusting for sociodemographics, medical comorbidities, and facility factors.

RESULTS: The cohort comprised 89 791 patients with a mean (SD) age of 73.0 (10.1) years; 87 647 (97.6%) were male, 9063 (10.1%) were Black, 3355 (3.7%) were Hispanic, and 77 373 (86.2%) were White. Overall, 64 770 individuals (72.1%) initiated any anticoagulant, 60 362 (67.2%) initiated DOAC therapy, and 4408 (4.9%) initiated warfarin. Compared with White patients, Black and Hispanic patients had lower rates of any anticoagulant and DOAC therapy initiation but higher rates of warfarin initiation across all quartiles of VAMC racial and ethnic composition. Any anticoagulant therapy initiation was lower in Q4 than Q1 (69.8% vs 74.9%; adjusted odds ratio [aOR], 0.80; 95% CI, 0.69-0.92; P < .001). DOAC and warfarin initiation were also lower in Q4 than in Q1 (DOAC, 69.4% vs 65.3%; aOR, 0.85; 95% CI, 0.74-0.97; P < .001; warfarin, 5.4% vs 4.5%; aOR, 0.82; 95% CI, 0.67-1.00; P < .001). In adjusted models, patients in Q4 were significantly less likely to initiate any anticoagulant therapy than those in Q1 (aOR, 0.88; 95% CI, 0.78-0.99). Patients in Q3 (aOR, 0.75; 95% CI, 0.60-0.93) and Q4 (aOR, 0.69; 95% CI, 0.55-0.87) were significantly less likely to initiate warfarin therapy than those in Q1. There was no significant difference in the adjusted odds of initiating DOAC therapy across racial and ethnic composition quartiles. Although significant Black-White and Hispanic-White differences in initiation of any anticoagulant, DOAC, and warfarin therapy were observed, interactions between patient race and ethnicity and VAMC racial composition were not significant.

CONCLUSIONS AND RELEVANCE: In a national cohort of VA patients with AF, initiation of any anticoagulant and warfarin, but not DOAC therapy, was lower in VAMCs serving more minoritized patients.

IMPORTANCE: Drug shortages are a chronic and worsening issue that compromises patient safety. Despite the destabilizing impact of the COVID-19 pandemic on pharmaceutical production, it remains unclear whether issues affecting the drug supply chain were more likely to result in meaningful shortages during the pandemic.

OBJECTIVE: To estimate the proportion of supply chain issue reports associated with drug shortages overall and with the COVID-19 pandemic.

DESIGN, SETTING, AND PARTICIPANTS: This longitudinal cross-sectional study used data from the IQVIA Multinational Integrated Data Analysis database, comprising more than 85% of drug purchases by US pharmacies from wholesalers and manufacturers, from 2017 to 2021. Data were analyzed from January to May 2023.

EXPOSURE: Presence of a supply chain issue report to the US Food and Drug Administration or the American Society of Health-Systems Pharmacists (ASHP).

MAIN OUTCOMES AND MEASURES: The main outcome was drug shortage, defined as at least 33% decrease in units purchased within 6 months of a supply chain issue report. Random-effects logistic regression models compared the marginal odds of shortages for drugs with vs without reports. Interaction terms assessed heterogeneity prior to vs during the COVID-19 pandemic and by drug characteristics (formulation, age, essential medicine status, clinician- vs self-administered, sales volume, and number of manufacturers).

RESULTS: A total of 571 drugs exposed to 731 supply chain issue reports were matched to 7296 comparison medications with no reports. After adjusting for drug characteristics, 13.7% (95% CI, 10.4%-17.8%) of supply chain issue reports were associated with subsequent drug shortages vs 4.1% (95% CI, 3.6%-4.8%) of comparators (marginal odds ratio [mOR], 3.7 [95% CI, 2.6-5.1]). Shortages increased among both drugs with and without reports in February to April 2020 (34.2% of drugs with supply chain issue reports and 9.5% of comparison drugs; mOR, 4.9 [95% CI, 2.1-11.6]), and then decreased after May 2020 (9.8% of drugs with reports and 3.6% of comparison drugs; mOR, 2.9 [95% CI, 1.6-5.3]). Significant associations were identified by formulation (parenteral mOR, 1.9 [95% CI, 1.1-3.2] vs oral mOR, 5.4 [95% CI, 3.3-8.8]; P for interaction = .008), WHO essential medicine status (essential mOR, 2.2 [95% CI, 1.3-5.2] vs nonessential mOR, 4.6 [95% CI, 3.2-6.7]; P = .02), and for brand-name vs generic status (brand-name mOR, 8.1 [95% CI, 4.0-16.0] vs generic mOR, 2.4 [95% CI, 1.7-3.6]; P = .002).

CONCLUSIONS AND RELEVANCE: In this national cross-sectional study, supply chain issues associated with drug shortages increased at the beginning of the COVID-19 pandemic. Ongoing policy work is needed to protect US drug supplies from future shocks and to prioritize clinically valuable drugs at greatest shortage risk.

IMPORTANCE: Drug shortages are a persistent public health issue that increased during the COVID-19 pandemic. Both the US and Canada follow similar regulatory standards and require reporting of drug-related supply chain issues that may result in shortages. However, it is unknown what proportion are associated with meaningful shortages (defined by a significant decrease in drug supply) and whether differences exist between Canada and the US.

OBJECTIVE: To compare how frequently reports of drug-related supply chain issues in the US vs Canada were associated with drug shortages.

DESIGN, SETTING, AND PARTICIPANTS: Longitudinal cross-sectional study conducted from January 2023 to March 2024 using drug-related reports of supply chain issues from 2017 to 2021 that were less than 180 days apart in Canada and the US. Shortages were assessed using data from the IQVIA Multinational Integrated Data Analysis database, comprising 89% of US and 100% of Canadian drug purchases.

EXPOSURE: Country (Canada vs US), timing of report issuance (before vs after the COVID-19 pandemic), and characteristics of the supply chain prior to the reports of drug-related supply chain issues (including World Health Organization essential medicine status, Health Canada tier 3 medicine [moderate risk classification], whether there was sole-source manufacturing of the drug, the formulation, the price per unit, ≥20 years since drug approval, and the number of therapeutic alternatives).

MAIN OUTCOMES AND MEASURES: A drug shortage (a decrease of ≥33% in monthly purchased standardized drug units) within 12 months, relative to the average units purchased during the 6 months prior to the report of supply chain issues to a US or Canadian reporting system.

RESULTS: Among the 104 drug-related reports of supply chain issues in both countries, 49.0% (95% CI, 39.3%-59.7%) were associated with drug shortages in the US vs 34.0% (95% CI, 25.0%-45.0%) in Canada (adjusted hazard ratio [HR], 0.53 [95% CI, 0.36-0.79]). The lower risk of drug shortages in Canada vs the US was consistent before the COVID-19 pandemic (adjusted HR, 0.47 [95% CI, 0.30-0.75]) and after the pandemic (adjusted HR, 0.31 [95% CI, 0.15-0.66]). After combining reports of supply chain issues in both countries, the shortage risk was double for sole-sourced drugs (adjusted HR, 2.58 [95% CI, 1.57-4.24]) and nearly half for Canadian tier 3 medicines (moderate risk) (adjusted HR, 0.56 [95% CI, 0.32-0.98]).

CONCLUSIONS AND RELEVANCE: Drug-related reports of supply chain issues were 40% less likely to result in meaningful drug shortages in Canada compared with the US. These findings highlight the need for international cooperation between countries to curb the effects of drug shortages and improve resiliency of the supply chain for drugs.

BACKGROUND: Over half of veterans enrolled in the Veterans Health Administration (VA) are also enrolled in Medicare, potentially increasing their opportunity to receive low-value health services within and outside VA.

OBJECTIVES: To characterize the use and cost of low-value services delivered to dually enrolled veterans from VA and Medicare.

DESIGN: Retrospective cross-sectional.

PARTICIPANTS: Veterans enrolled in VA and fee-for-service Medicare (FY 2017-2018).

MAIN MEASURES: We used VA and Medicare administrative data to identify 29 low-value services across 6 established domains: cancer screening, diagnostic/preventive testing, preoperative testing, imaging, cardiovascular testing, and surgery. We determined the count of low-value services per 100 veterans delivered in VA and Medicare in FY 2018 overall, by domain, and by individual service. We applied standardized estimates to determine each service's cost.

KEY RESULTS: Among 1.6 million dually enrolled veterans, the mean age was 73, 97% were men, and 77% were non-Hispanic White. Overall, 63.2 low-value services per 100 veterans were delivered, affecting 32% of veterans; 22.9 services per 100 veterans were delivered in VA and 40.3 services per 100 veterans were delivered in Medicare. The total cost was $226.3 million (M), of which $62.6 M was spent in VA and $163.7 M in Medicare. The most common low-value service was prostate-specific antigen testing at 17.3 per 100 veterans (VA 55.9%, Medicare 44.1%). The costliest low-value service was percutaneous coronary intervention (VA $10.1 M, Medicare $32.8 M).

CONCLUSIONS: Nearly 1 in 3 dually enrolled veterans received a low-value service in FY18, with twice as many low-value services delivered in Medicare vs VA. Interventions to reduce low-value services for veterans should consider their substantial use of such services in Medicare.

BACKGROUND: Oral anticoagulation reduces stroke risk for patients with atrial fibrillation (AF). Prior research demonstrates lower anticoagulant prescribing in Black than in White individuals but few studies have examined racial differences in facility-level anticoagulant prescribing for AF.

OBJECTIVE: To assess variation in anticoagulant initiation by race within Veterans Health Administration (VA) facilities.

DESIGN: Retrospective cohort study.

PARTICIPANTS: Black and White patients enrolled in the VA with incident AF from 2020 through 2021.

MAIN MEASURES: The primary outcome was rate of any anticoagulant initiation (i.e., warfarin or direct oral anticoagulant [DOAC]) or any DOAC therapy within 90 days of an AF diagnosis, overall and for Black and White patients at each facility. We also estimated the adjusted Black-White risk difference.

KEY RESULTS: In 82 VA facilities serving 26,832 Black and White patients, overall unadjusted rates of any anticoagulant therapy ranged from 56.8 to 87.1% across facilities; the corresponding ranges for Black and White patients were 47.6 to 91.3% and 58.2 to 87.1%, respectively. Overall unadjusted rates of DOAC therapy ranged from 55.1 to 85.5% by facility; ranges for Black and White patients were 42.8 to 86.9% and 56.4 to 85.5%, respectively. The adjusted risk difference between Black and White patients ranged from - 29.9 (95% CI, - 54.9 to - 4.8) to 14.2 (95% CI, - 9.1 to 25.0) across facilities for any anticoagulant therapy and from - 28.8 (95% CI, - 58.3 to 0.8) to 15.0 (95% CI, - 8.0 to 38.1) for DOAC therapy. For any anticoagulant therapy there were 3 facilities where prescribing was statistically higher in White than Black patients; for DOAC therapy there were 5 such facilities.

CONCLUSIONS: In a national cohort of patients with AF, we observed large facility-level variation and adjusted risk differences in any anticoagulant and DOAC initiation, overall and by race. These findings represent a target for local quality improvement in AF care.