Publications

BACKGROUND: Opioids are a widely prescribed class of medication for pain management. However, they have variable efficacy and adverse effects among patients, due to the complex interplay between biological and clinical factors. Pharmacogenetic testing can be used to match patients' genetic profiles to individualize opioid therapy, improving pain relief and reducing the risk of adverse effects. Despite its potential, the pharmacogenetic testing uptake (use of pharmacogenetic testing) remains low due to a range of barriers at the patient, health care provider, infrastructure, and financial levels. Since testing typically involves a shared decision between the provider and patient, predicting the likelihood of a patient undergoing pharmacogenetic testing and understanding the factors influencing that decision can help optimize resource use and improve outcomes in pain management.

OBJECTIVE: This study aimed to develop machine learning (ML) models, identifying patients' likelihood of pharmacogenetic uptake based on their demographics, clinical variables, medication use, and social determinants of health.

METHODS: We used electronic health record data from a single center health care system to identify patients prescribed opioids. We extracted patients' demographics, clinical variables, medication use, and social determinants of health, and developed and validated ML models, including a neural network, logistic regression, random forest, extreme gradient boosting (XGB), naïve Bayes, and support vector machines for pharmacogenetic testing uptake prediction based on procedure codes. We performed 5-fold cross-validation and created an ensemble probability-based classifier using the best-performing ML models for pharmacogenetic testing uptake prediction. Various performance metrics, uptake stratification analysis, and feature importance analysis were used to evaluate the performance of the models.

RESULTS: The ensemble model using XGB and support vector machine-radial basis function classifiers had the highest C-statistics at 79.61%, followed by XGB (78.94%), and neural network (78.05%). While XGB was the best-performing model, the ensemble model achieved a high accuracy (32,699/48,528, 67.38%), recall (537/702, 76.50%), specificity (32,162/47,826, 67.25%), and negative predictive value (32,162/32,327, 99.49%). The uptake stratification analysis using the ensemble model indicated that it can effectively distinguish across uptake probability deciles, where those in the higher strata are more likely to undergo pharmacogenetic testing in the real world (320/4853, 6.59% in the highest decile compared to 6/4853, 0.12% in the lowest). Furthermore, Shapley Additive Explanations value analysis using the XGB model indicated age, hypertension, and household income as the most influential factors for pharmacogenetic testing uptake prediction.

CONCLUSIONS: The proposed ensemble model demonstrated a high performance in pharmacogenetic testing uptake prediction among patients using opioids for pain. This model can be used as a decision support tool, assisting clinicians in identifying patients' likelihood of pharmacogenetic testing uptake and guiding appropriate decision-making.

BACKGROUND: Transitions of care (ToC) programs are important for patient safety, but their implementation and success remain highly variable across US hospitals, particularly for patients with multimorbidity and health-related social needs (HRSNs). Hospitalists, as key decision-makers at discharge, encounter firsthand the factors that hinder the success of ToC programs.

OBJECTIVE: To explore hospitalists' perspectives on successes, shortcomings, and implementation barriers in ToC programs, particularly during transitions from hospital to community settings.

METHODS: Rapid qualitative study featuring virtual focus groups with participants from the Hospital Medicine Reengineering Network (HOMERuN). Data were analyzed using a mixed inductive-deductive framework to identify key themes.

RESULTS: Twenty-two individuals from 19 different organizations participated in focus groups. None of the organizations offered comprehensive ToC programs to all patients. Four major themes emerged: (1) Diagnosis-specific ToC programs are effective but contribute to care fragmentation, particularly for patients with multimorbidity; (2) postdischarge follow-up is hindered by limited appointment availability, insurance barriers, and geographic challenges; (3) ToC programs often fail to address patient preferences, HRSNs, and health literacy, and lack adequate resources and leadership support; (4) successful programs require institutional commitment, dedicated funding, interprofessional collaboration, and community engagement. Participants emphasized the need to prioritize patient-centered care over financial return on investment.

CONCLUSIONS: Current ToC programs are fragmented, undermining safe and equitable transitions. Addressing HRSNs, fostering leadership support, and prioritizing patient-centered care over short-term financial metrics are essential for improving ToC outcomes.

BACKGROUND: The Medicare Part D Medication Therapy Management (MTM) program uses comprehensive medication reviews (CMRs) and targeted medication reviews to address medication-related problems in older adults. One such problem is central nervous system (CNS)-active polypharmacy, which is associated with impaired cognition and falls in older adults. To date, little is known about the prevalence of CNS-active polypharmacy among MTM enrollees and how they might differ from MTM enrollees without CNS-active polypharmacy; such insights would be helpful to MTM programs developing interventions to reduce CNS-active polypharmacy.

OBJECTIVE: To (1) estimate the prevalence of MTM enrollees with CNS-active polypharmacy in a nationwide cohort and (2) compare patient characteristics and MTM service use of MTM enrollees with CNS-active polypharmacy vs without.

METHODS: Cross-sectional, observational study of 2019-2021 Medicare 5% fee-for-service data linked to 2020-2021 MTM data. Patient characteristics and MTM use were compared between MTM enrollees with and without CNS-active polypharmacy.

RESULTS: Among 38,733 MTM enrollees in 2021, 4,144 (10.7%) experienced CNS-active polypharmacy. Compared with those without CNS-active polypharmacy, the CNS-active polypharmacy cohort was more likely to be male (72% vs 55%, standardized mean difference [SMD] = 34.7%), be dually enrolled in Medicaid (42% vs 24%, SMD = 41.1%), and have greater comorbidity burden (Charlson Comorbidity Score of 6.0 vs 5.0, SMD = 13.2%). The CNS-active polypharmacy cohort also had more prior-year health care utilization, such as being more likely to have an inpatient stay (37.5% vs 29.0%, SMD = 18.1%) or emergency department visit (53.7% vs 43.0%, SMD = 21.6%), as well as number of outpatient visits (7.0 vs 5.0, SMD = 20.8%) and number of unique prescription drugs (21.0 vs 15.0, SMD = 91.9%). The number of targeted medication reviews received was greater in the CNS-active polypharmacy cohort, but a lower proportion (35% vs 39%) participated in a CMR.

CONCLUSIONS: More than 1 in 10 MTM enrollees experience CNS-active polypharmacy, which is higher than the general Medicare fee-for-service beneficiary population. MTM enrollees with CNS-active polypharmacy are more likely to be male, dually enrolled in Medicaid, and with greater comorbidity burden and prior-year use of health care and medications, suggesting that interventions for this population may need to account for additional clinical and socioeconomic needs. Despite being at greater risk of adverse drug events including impaired cognition and falls, just over one-third of MTM enrollees with CNS-active polypharmacy participate in a CMR, suggesting opportunity for more targeted outreach and intervention by MTM programs, Part D plan sponsors, and Centers for Medicare & Medicaid Services.

BACKGROUND: Lecanemab, an anti-amyloid monoclonal antibody for Alzheimer's disease, was approved for use in the Veteran's Health Administration (VHA) in 2023. Little is known about the real-world use and outcomes of lecanemab.

METHODS: This retrospective cohort study included Veterans who initiated lecanemab in the VHA between October 2023 and September 2024. Data on demographics, healthcare utilization, medication exposures, and neuroimaging were extracted from the VA Corporate Data Warehouse. Chart review was used to collect information on dementia diagnosis, Montreal Cognitive Assessments (MoCA), and Apolipoprotein E (APOE) genotype. We examined medication adherence, brain MRIs, 6-month MoCAs, amyloid-related imaging abnormalities (ARIA), and healthcare utilization from initiation to study end (12/31/24).

RESULTS: Overall, 32 Veterans (mean [SD] age 75 [6] years, 100% male, 97% white, 84% urban-dwelling) initiated lecanemab; 53% had mild cognitive impairment and 47% mild dementia. The baseline mean MoCA score was 21 (SD 3). Half were heterozygous for APOE ε4 and half non-carriers. The median days of follow-up was 172 (range 62-349), and the median number of lecanemab infusions received was 11 (2-24). Brain MRI follow-up occurred regularly, with all eligible patients receiving an MRI between the 4th and 5th infusion, 92% (22/24) between the 6th and 7th infusions, and 92% (11/12) between the 13th and 14th infusions. However, of the 28 patients with 6 months follow-up, only 9 (32%) had a MoCA completed and the mean change in MoCA was -0.2 (SD 4.2). During follow-up, 25% of patients had imaging abnormalities: 2 (6.2%) experienced acute stroke and 6 (18.8%) experienced ARIA (3 ARIA-edema, 3 ARIA-hemorrhage, and 1 both). Three patients (9.4%) stopped lecanemab for ≥30 days by the end of the study period. Reasons for discontinuation included ARIA, side effects, and patient preference. During follow-up, 0 patients died, 3 (9.4%) were hospitalized within VHA, and 11 (34.4%) had VHA emergency department or urgent care visits.

CONCLUSIONS: In the first year lecanemab was available in VHA, the few patients initiated on treatment were mostly white, male, and urban residents. The finding that 25% of patients experienced ARIA or stroke underscores the importance of monitoring the lecanemab safety and effectiveness long-term.

INTRODUCTION: Medicaid provides access to care for low-income patients facing life-threatening illnesses who are cared for in intensive care units (ICUs). Despite the growth of Medicaid coverage with the Affordable Care Act, little is known about Medicaid's role in critical care for the Medicaid Expansion population, adults ages 19-64.

METHODS: Using hospital discharge data from Virginia, we examined payer composition between 2016 and 2023 and analyzed 2023 demographic and clinical data for adults ages 19-64.

RESULTS: Medicaid's share of ICU stays more than doubled from 2016 (14.1%) to 2023 (31.8%). While only 25.6% of Medicaid hospitalizations involved ICU care, these stays account disproportionately for charges (51.7%), hospital days (36.9%), and readmissions (32.3%). Common reasons for admission include potentially preventable conditions: sepsis, diabetes, heart failure, and alcohol use. Medicaid patients, despite being younger than their commercially insured counterparts, have more comorbidities (4+ comorbidities: 49.9% vs 38.9%) and are more likely to be readmitted in adjusted models (29.7% [95% Confidence Interval: 29.1-30.4] vs 24.3% [95% Confidence Interval: 23.6%-25.1%]).

CONCLUSIONS: This work demonstrates the crucial role of Medicaid as a payer for seriously ill adults and the need for increased attention by Medicaid programs to ICU patients before, during, and after hospitalization.

BACKGROUND AND AIMS: Comparative effectiveness research studies commonly restrict cohorts to individuals who initiate a medication and do not have evidence of prior treatment. This is particularly challenging in research on medications for opioid use disorder (MOUD) because of sporadic use or intermittent adherence. We examined the impact of different lookback windows and washout criteria to identify MOUD initiator cohorts on sample size, cohort characteristics, and misclassification of treatment initiation.

DESIGN AND SETTING: Cohort study using the Merative™ MarketScan® Multi-State Medicaid Database (2011-2022).

PARTICIPANTS: Medicaid-insured adults aged 18-64 with an MOUD prescription from 01/01/2022 to 12/31/2022 and a history of opioid use disorder (OUD) with at least 3 months of continuous enrollment.

MEASUREMENTS: We created treatment initiator cohorts with increasingly restrictive lookback windows for inclusion (6-, 12-, 24-, 36-months, all-available). During each lookback window, we required [1] continuous enrollment; [2] continuous enrollment and OUD diagnosis; or [3] continuous enrollment, OUD diagnosis, and no prior treatment with MOUD. We defined prior treatment with MOUD as: (a) ≥ 30 days use (less restrictive definition; allowed for some prior treatment); or (b) ≥ 1 day use (more restrictive definition; did not allow prior treatment). We quantified changes in cohort sample size, demographic characteristics, and proportion of prevalent use episodes misclassified as MOUD treatment initiation (gold standard: 36-month lookback window).

FINDINGS: We identified 103 794 eligible MOUD initiators (64.8% buprenorphine, 24.8% methadone, 10.4% naltrexone). Sample size of the cohorts decreased with increasingly restrictive lookback windows and washout criteria: [1] continuous enrollment (range, 96.9% for 6 months to 51.8% for 36 months); [2] continuous enrollment and less restrictive washout (range, 29.7% to 8.4%); and [3] continuous enrollment and more restrictive washout (range, 22.2% to 5.8%). All-available lookback performed similarly to a 12-month lookback. Longer lookback windows resulted in initiator cohorts with a greater proportion of individuals who were older, female, and of a minoritized race/ethnicity. The proportion of people with prevalent MOUD use misclassified as treatment initiation increased steadily with decreasing duration of lookback windows (24-, 12-, and 6-month); we observed misclassification among 16.1% to 49.2% of individuals (less restrictive washout), and 16.8% to 53.2% of individuals (more restrictive washout).

CONCLUSIONS: The choice of lookback window duration and washout criteria in research on medications for opioid use disorder (MOUD) presents tradeoffs between cohort sample size, demographic characteristics, and misclassification of treatment initiation. This study offers practical guidance for researchers planning to perform comparative studies in MOUD.

Background/Objectives: The Developing and Evaluating a Machine Learning Opioid Prediction & Risk-Stratification E-Platform (DEMONSTRATE) trial aims to assess the usability, acceptability, feasibility, and effectiveness of implementing a machine learning (ML)-based clinical decision support (CDS) tool-the Overdose Prevention Alert-which predicts a patient's risk of opioid overdose within three months. Methods: This single-arm study uses a pre-post implementation design with mixed-methods evaluation in 13 University of Florida Health, Gainesville, internal medicine and family medicine clinics. Eligible patients are aged ≥18 years, received an opioid prescription within the year prior to their upcoming primary care visit, are not receiving hospice care, do not have a malignant cancer diagnosis, and are identified by the ML algorithm as high risk for overdose. The Overdose Prevention Alert triggers when a primary care provider (PCP) signs an opioid order in electronic health records. We will evaluate effectiveness by comparing pre- and post-implementation outcomes using a composite patient-level measure defined by the presence of any of the following 6 favorable indicators: (1) evidence of naloxone access; (2) absence of opioid overdose diagnoses and naloxone administration; (3) absence of emergency department (ED) visits or hospitalizations due to opioid overdose or opioid use disorder (OUD); (4) absence of overlapping opioid and benzodiazepine use within a 7-day window; (5) absence of opioid use ≥50 morphine milligram equivalent daily average; (6) receipt of referrals to non-pharmacological pain management. Additional quantitative metrics will include alert penetration, usage patterns, and clinical actions taken. Usability and acceptability will be assessed using a 12-item questionnaire for PCPs and semi-structured interviews. Expected Results: The trial will provide insights into real-world ML-driven CDS implementation and inform future strategies to reduce opioid-related harm.