Publications

BACKGROUND: One of the largest-ever retail drug shortages began in 2018 when several angiotensin II receptor blockers (ARBs) for treating hypertension, heart failure, and chronic kidney disease-valsartan, losartan, and irbesartan-were recalled for carcinogenic impurities. The long-term consequences of the ARB shortages and whether certain groups experienced more adverse outcomes is unknown.

OBJECTIVE: To evaluate changes in adherence and health outcomes after ARB recalls and to identify patients who experienced greater changes in access and adverse clinical outcomes.

METHODS: Using an integrated claims and electronic health record dataset and a difference-in-differences design, we evaluated changes in the proportion of days covered (PDC) for ARBs and similar drugs (angiotensin-converting enzyme inhibitors [ACE-Is]), uncontrolled blood pressure, major cardiovascular event (MACE)-related acute care visits, and all-cause ambulatory care visits in the 12 months before vs 18 months after recalls for valsartan, losartan, and irbesartan users vs patients taking similar, nonrecalled drugs (ACE-Is, nonrecalled ARBs). Triple-difference models characterized heterogeneous associations by pre-recall patient demographic (race, ethnicity, age), clinical (baseline indication, mental health conditions), and adherence variables.

RESULTS: Adjusting for pre-recall patient characteristics, we observed no significant changes in PDC for ARBs and ACE-Is (combined), uncontrolled blood pressure, or ambulatory care visits among 86,507 recalled ARB users vs 123,583 comparison drug users in the 18 months after the recalls. Following the recalls, medication switches increased on average by an additional 2.08 percentage points (p.p.) per quarter (95% CI = 2.01-2.15) for recalled ARB vs comparison drug users, a 195.9% relative increase. We observed the most switches in the 90-day period immediately after valsartan's recall (difference-in-difference: 9.48 p.p.; 95% CI = 9.36-9.59; relative change = 892%). Cumulatively, 55.2% of valsartan, 7.6% of losartan, and 18.9% of irbesartan users switched medications after 18 months. We observed an increase in the proportion of recalled ARB vs comparison patients who experienced medication gaps exceeding 30 days (1.13 p.p. per quarter on average; 95% CI = 0.97-1.30), which was most apparent after approximately 15 months (5 quarters). Although MACE-related acute care visits did not change in the quarter (90 days) immediately after valsartan's recall, we observed an increase of 1.40 additional visits per 1,000 recalled ARB vs comparison drug patients in each subsequent quarter, a 9.3% relative increase. Results were similar across most subgroups.

CONCLUSIONS: The 2018 ARB recalls were associated with immediate changes in antihypertension medication use. Many patients transitioned to alternative medications. Although overall impacts on clinical outcomes were minimal and not statistically significant, small increases in medication gaps and MACE-related acute care visits among some patients occurred after more than 1 year. The ARB recalls may have been associated with fewer adverse events than other recent shortages owing to the widespread availability of alternative treatments in the same or similar drug class.

IMPORTANCE: Monoclonal antibodies (mAbs) targeting calcitonin gene-related peptide (CGRP) or its receptor (anti-CGRP mAbs) offer effective migraine-specific preventive treatment. However, concerns exist about their potential cardiovascular risks due to CGRP blockade.

OBJECTIVE: To compare the incidence of cardiovascular disease (CVD) between Medicare beneficiaries with migraine who initiated anti-CGRP-mAbs vs onabotulinumtoxinA in the US.

DESIGN, SETTING, AND PARTICIPANTS: This retrospective, sequential cohort study was conducted among a nationally representative population-based sample of Medicare claims from May 2018 through December 2020. Data analysis was performed from August to December 2023. This study included fee-for-service Medicare beneficiaries aged 18 years or older with migraine who initiated either anti-CGRP mAbs or onabotulinumtoxinA. Beneficiaries who had a history of myocardial infarction (MI), stroke, cluster headache, malignant cancer, or hospice service within a 1-year baseline period prior to treatment initiation were excluded. To minimize channeling bias from new drug introductions and time-related bias due to the COVID-19 pandemic, 5 cohorts were established, representing sequential 6-month calendar intervals based on the initial prescription or date of index anti-CGRP mAbs or onabotulinumtoxinA use.

EXPOSURE: Anti-CGRP mAbs vs onabotulinumtoxinA.

MAIN OUTCOMES AND MEASURES: The primary outcome was time to first MI or stroke. Secondary outcomes included hypertensive crisis, peripheral revascularization, and Raynaud phenomenon. The inverse probability of treatment-weighted Cox proportional hazards models were used to compare outcomes between the 2 treatment groups.

RESULTS: Among 266 848 eligible patients with migraine, 5153 patients initiated anti-CGRP mAbs (mean [SD] age, 57.8 [14.0] years; 4308 female patients [83.6%]) and 4000 patients initiated onabotulinumtoxinA (mean [SD] age, 61.9 [13.7] years; 3353 female patients [83.8%]). Use of anti-CGRP mAbs was not associated with an increased risk of composite CVD events (adjusted hazard ratio [aHR], 0.88; 95% CI, 0.44-1.77), hypertensive crisis (aHR, 0.46; 95% CI, 0.14-1.55), peripheral revascularization (aHR, 1.50; 95% CI, 0.48-4.73), or Raynaud phenomenon (aHR, 0.75; 95% CI, 0.45-1.24) compared with onabotulinumtoxinA. Subgroup analyses by age group and presence of established non-MI or stroke CVD showed similar findings.

CONCLUSIONS AND RELEVANCE: In this cohort study, despite initial concerns regarding the cardiovascular effects of CGRP blockade, anti-CGRP mAbs were not associated with an increased risk of CVD compared with onabotulinumtoxinA among adult Medicare beneficiaries with migraine, who were predominantly older adults or individuals with disability. Future studies with longer follow-up periods and in other populations are needed to confirm these findings.

OBJECTIVES: To highlight the often overlooked role of user interface (UI) design in mitigating bias in artificial intelligence (AI)-based clinical decision support (CDS).

MATERIALS AND METHODS: This perspective paper discusses the interdependency between AI-based algorithm development and UI design and proposes strategies for increasing the safety and efficacy of CDS.

RESULTS: The role of design in biasing user behavior is well documented in behavioral economics and other disciplines. We offer an example of how UI designs play a role in how bias manifests in our machine learning-based CDS development.

DISCUSSION: Much discussion on bias in AI revolves around data quality and algorithm design; less attention is given to how UI design can exacerbate or mitigate limitations of AI-based applications.

CONCLUSION: This work highlights important considerations including the role of UI design in reinforcing/mitigating bias, human factors methods for identifying issues before an application is released, and risk communication strategies.

BACKGROUND: Ischemic stroke is a leading cause of death and disability. Society guidelines recommend pharmacotherapies for secondary stroke prevention. However, the role of sex differences in prescription and adherence to guideline-directed medical therapies (GDMT) after ischemic stroke remains understudied. The aim of this study was to examine sex differences in prescription patterns and adherence to GDMT at 1 year after ischemic stroke in a cohort of commercially insured patients.

METHODS: Using the Truven Health MarketScan database from 2016 to 2020, we identified patients admitted with ischemic stroke. GDMT was defined as any statin, antihypertensive agents, or oral anticoagulant prescription within 30 days after discharge. Medication adherence was estimated using the proportion of days covered at 1 year. The proportion of days covered <0.80 was used to define nonadherence. A multivariable model adjusting for covariates was performed to identify the factors associated with nonadherence at 1 year. This analysis was restricted to new users of GDMT.

RESULTS: Among 155 220 patients admitted with acute ischemic stroke during the study period, 15 919 met the inclusion criteria. The mean age was 55.7 years, and 8218 (51.7%) were women. Women were less likely to be prescribed statins (58.0% versus 71.8%) and antihypertensive agents (27.7% versus 41.8%). In this subset of patients with atrial flutter/fibrillation, women were also less likely to be prescribed oral anticoagulants (41.2% versus 45.0%). Women were more likely to be nonadherent (ie, proportion of days covered <0.80) to statins (47.3% versus 41.6%; P<0.0001), antihypertensives (33.3% versus 32.2%; P=0.005), and the combination of both (49.6% versus 45.0%; P=0.003). On multivariable analysis, women were likely to be nonadherent to statins and antihypertensive agents at 1 year (odds ratio, 1.23 [95% CI, 1.08-1.41]).

CONCLUSIONS: In this real-world analysis of commercially insured patients with ischemic stroke, women were less likely initiated on GDMT within 30 days after discharge. Women were more likely to be nonadherent to statins and antihypertensive agents at 1 year. Future efforts and novel interventions are needed to understand the reasons and minimize these disparities.

Predicting major bleeding in nonvalvular atrial fibrillation (AF) patients on direct oral anticoagulants (DOACs) is crucial for personalized care. Alternatives like left atrial appendage closure devices lower stroke risk with fewer nonprocedural bleeds. This study compares machine learning (ML) models with conventional bleeding risk scores (HAS-BLED, ORBIT, and ATRIA) for predicting bleeding events requiring hospitalization in AF patients on DOACs at their index cardiologist visit. This retrospective cohort study used electronic health records from 2010 to 2022 at the University of Pittsburgh Medical Center. It included 24,468 nonvalvular AF patients (age ≥18) on DOACs, excluding those with prior significant bleeding or warfarin use. The primary outcome was hospitalization for bleeding within one year, with follow-up at one, two, and five years. ML algorithms (logistic regression, classification trees, random forest, XGBoost, k-nearest neighbor, naïve Bayes) were compared for performance. Of 24,468 patients, 553 (2.3%) had bleeding within one year, 829 (3.5%) within two years, and 1,292 (5.8%) within five years. ML models outperformed HAS-BLED, ATRIA, and ORBIT in 1-year predictions. The random forest model achieved an AUC of 0.76 (0.70 to 0.81), G-Mean of 0.67, and net reclassification index of 0.14 compared to HAS-BLED's AUC of 0.57 (p < 0.001). ML models showed superior results across all timepoints and for hemorrhagic stroke. SHAP analysis identified new risk factors, including BMI, cholesterol profile, and insurance type. In conclusion, ML models demonstrated improved performance to conventional bleeding risk scores and uncovered novel risk factors, offering potential for more personalized bleeding risk assessment in AF patients on DOACs.

BACKGROUND: Racial disparities exist in access to health care and management of multiple health conditions including chronic pain; however, racial disparities in pre- and postoperative pain management in lower extremity amputation are not well-studied. Our objective was to examine the association between different racial and ethnic groups and prescription opioid and other analgesics use before and after lower extremity amputation. We hypothesize prescription opioid and other analgesic use among Black, Hispanic, and Native American US Medicare beneficiaries undergoing lower extremity amputations will be lower compared to White US Medicare beneficiaries.

METHODS: This retrospective cohort study included a 5% national sample of all Medicare beneficiaries from 2011 to 2015 and 15% national sample of fee-for-service Medicare beneficiaries from 2016 to 2018 undergoing nontraumatic, lower extremity amputations. The exposure of interest was racial and ethnic group membership (ie, Black, Hispanic, Native American, White, and others-with others being the combination of the categories Asian and other) as provided in Medicare claims data. Using multivariable generalized estimating equations with a logistic link to account for repeated measurements over time, we estimated the odds of prescription opioid use within 6 months before and after lower extremity amputation across different racial and ethnic groups separately, adjusting for sociodemographic and health status factors (eg, Elixhauser index). Adjusted odds ratios (aORs) and 95% confidence intervals (95% CI) were reported.

RESULTS: Among 16,068 eligible beneficiaries who underwent major and minor amputations (mean age = 65.1 ± 12.7 years; female = 36.1%), 10,107 (62.9%) were White, 3462 (21.5%) were Black, 1959 (12.2%) were Hispanic, 247 (1.5%) were Native American, and 151 (2.9%) were beneficiaries of other races. During the 6 months before lower extremity amputation, Hispanic beneficiaries (aOR, 0.71, 95% CI, 0.65-0.78) and beneficiaries of other races (aOR, 0.60, 95% CI, 0.47-0.76) had significantly lower odds of using prescription opioids compared to White beneficiaries. Similarly, Hispanic beneficiaries (aOR, 0.78, 95% CI, 0.71-0.84) and beneficiaries of other races (aOR, 0.63, 95% CI, 0.51-0.78) were associated with lower odds of opioid use in the 6 months after amputation compared to White beneficiaries.

CONCLUSIONS: Among fee-for-service Medicare beneficiaries, Hispanic and other (eg, Asian) fee-for-service Medicare beneficiaries had lower odds of prescription opioid use than their White counterparts before and after nontraumatic, lower extremity amputations. Efforts to determine the underlying reasons are needed to ensure equitable health care access.

AIMS: To develop machine-learning algorithms for predicting the risk of a hospitalization or emergency department (ED) visit for opioid use disorder (OUD) (i.e. OUD acute events) in Pennsylvania Medicaid enrollees in the Opioid Use Disorder Centers of Excellence (COE) program and to evaluate the fairness of model performance across racial groups.

METHODS: We studied 20 983 United States Medicaid enrollees aged 18 years or older who had COE visits between April 2019 and March 2021. We applied multivariate logistic regression, least absolute shrinkage and selection operator models, random forests, and eXtreme Gradient Boosting (XGB), to predict OUD acute events following the initial COE visit. Our models included predictors at the system, patient, and regional levels. We assessed model performance using multiple metrics by racial groups. Individuals were divided into a low, medium and high-risk group based on predicted risk scores.

RESULTS: The training (n = 13 990) and testing (n = 6993) samples displayed similar characteristics (mean age 38.1 ± 9.3 years, 58% male, 80% White enrollees) with 4% experiencing OUD acute events at baseline. XGB demonstrated the best prediction performance (C-statistic = 76.6% [95% confidence interval = 75.6%-77.7%] vs. 72.8%-74.7% for other methods). At the balanced cutoff, XGB achieved a sensitivity of 68.2%, specificity of 70.0%, and positive predictive value of 8.3%. The XGB model classified the testing sample into high-risk (6%), medium-risk (30%), and low-risk (63%) groups. In the high-risk group, 40.7% had OUD acute events vs. 16.5% and 5.0% in the medium- and low-risk groups. The high- and medium-risk groups captured 44% and 26% of individuals with OUD events. The XGB model exhibited lower false negative rates and higher false positive rates in racial/ethnic minority groups than White enrollees.

CONCLUSIONS: New machine-learning algorithms perform well to predict risks of opioid use disorder (OUD) acute care use among United States Medicaid enrollees and improve fairness of prediction across racial and ethnic groups compared with previous OUD-related models.

BACKGROUND: Narcolepsy is linked to adverse cardiovascular disease (CVD) outcomes, but few studies have examined its associations with subclinical CVD, including in children. We assessed the relationship between narcolepsy and subclinical CVD outcomes, including hypertension, hyperlipidemia, diabetes, and nonalcoholic fatty liver disease/nonalcoholic steatohepatitis.

METHODS AND RESULTS: We conducted a retrospective cohort study using MarketScan Commercial and Medicare Supplemental databases from January 1, 2005 to December 31, 2021. Patients included N=22 293 diagnosed with narcolepsy (NT1 and NT2) and N=63 709 propensity-score-matched without. Patients with narcolepsy were identified as those with ≥2 outpatient insurance claims for narcolepsy (type 1 or type 2) within a 1-year interval with 1 claim being nondiagnostic. Main outcomes were diagnosis of hypertension, hyperlipidemia, diabetes, and nonalcoholic fatty liver disease/nonalcoholic steatohepatitis following index date, as well as a composite measure for CVD and major adverse cardiovascular events. Compared with propensity-score-matched patients without narcolepsy, patients with narcolepsy had an increased risk for hypertension (hazard ratio [HR], 1.40 [95% CI, 1.34-1.47]), hyperlipidemia (HR, 1.41 [95% CI, 1.35-1.47]), diabetes (HR, 1.50 [95% CI, 1.38-1.64), nonalcoholic fatty liver disease/nonalcoholic steatohepatitis (HR, 1.48 [95% CI, 1.28-1.73]), CVD composite (HR,1.61 [95% CI, 1.35-1.47]), and major adverse cardiovascular events (HR,1.69 [95% CI, 1.43-2.00]). Results remained significant following adjustment for narcolepsy medications including stimulants, wake-promoting agents, and oxybates. Results stratified by age groups showed similar findings, including heightened risk for those <25 years old.

CONCLUSIONS: Narcolepsy is associated with greater risk of subclinical CVD even in patients as early as childhood. Detection of these outcomes early in the course of narcolepsy could help reduce the burden of adverse cardiovascular events later in life.