Publications

OBJECTIVES: In 2022, the FDA issued a drug safety communication based on case studies that transmucosal buprenorphine, a medication for opioid use disorder (MOUD), may contribute to dental disease. We sought to assess longitudinal dental care utilization patterns among patients with opioid use disorder (OUD) in the 18 months before and 36 months after MOUD initiation.

METHODS: Using data from the Veterans Affairs Corporate Data Warehouse (2003-2020), we created a cohort of patients coded for OUD and prescribed MOUD. Outcomes included preventive and therapeutic dental visits and oral infections within 18 months before and up to 36 months after index MOUD. We used unadjusted Poisson models to estimate incidence per 1000 patients by MOUD. We performed analyses with and without interval censoring for edentulism, death, or a 60-day gap in MOUD.

RESULTS: Among 49,675 eligible patients, 21,551 received methadone, 17,759 transmucosal buprenorphine, 8993 oral naltrexone, and 1372 injectable naltrexone. Median (IQR) days on treatment varied by drug: methadone 95 (52, 257), buprenorphine 309 (102, 968), oral naltrexone 49 (30, 101), injectable naltrexone 28 (28, 85). We observed an immediate increase in dental visits from a baseline range of 135-144 visits/1000/6-month period to 223-686 visits/1000/6-month period after initiating any MOUD. Patterns were similar by MOUD agent and formulation. Results were similar in analyses with and without interval censoring.

CONCLUSIONS: Both preventive and therapeutic dental utilization increased immediately following initiation with MOUD. Future observational studies of the effects of MOUD on adverse dental outcomes should account for confounding due to health-seeking behavior.

BACKGROUND: Opioid use disorder (OUD) remains a critical public health crisis in the United States. Despite widespread policy and clinical interventions, early identification of individuals at risk for developing OUD remains challenging due to limitations in traditional screening approaches and a lack of individualized risk stratification methods. Machine learning (ML) methods offer an opportunity to develop timely, high-performing, and explainable predictive models that can enhance OUD prevention strategies in clinical settings.

OBJECTIVE: This study aims to develop and validate an ML model using electronic health record (EHR) data to predict the 3-month risk of incident OUD among adults initiating opioid therapy and to stratify patients into clinically actionable risk groups.

METHODS: This prognostic modeling study used 2017-2022 OneFlorida+ EHR data to develop and validate ML algorithms predicting 3-month incident OUD risk. We included 182,083 adults (≥18 y) without cancer, overdose, or OUD or hospice history who received ≥1 outpatient, noninjectable opioid prescription. Using 183 predictors measured in sequential 3-month intervals, we developed an elastic net, least absolute shrinkage and selection operator, gradient boosting machine (GBM), and random forest models on randomly split training, testing, and validation sets. Model performance was assessed using C-statistics, predictive values, and number needed to evaluate, with patients stratified into risk deciles for clinical applicability. Model explainability was assessed using Shapley additive explanations, and fairness was evaluated using standard metrics. We externally validated the best-performing model using an independent cohort from the 2018-2020 UPMC (formerly University of Pittsburgh Medical Center) health system.

RESULTS: In the validation sample (n=60,694), GBM (C-statistics=0.879, 95% CI 0.874-0.884) and elastic net (C-statistics=0.872, 95% CI 0.867-0.877) outperformed least absolute shrinkage and selection operator (C-statistics=0.846, 95% CI 0.840-0.851) and random forest (C-statistics=0.798, 95% CI 0.792-0.804), with GBM model requiring the fewest predictors (n=75) for predicting 3-month incident OUD. Using the GBM algorithm to predict the subsequent 3-month OUD risk, the top decile subgroup had a positive predictive value of 3.26%, a negative predictive value of 99.8%, and a number needed to evaluate of 31. The top decile (n=6696) captured  68% of patients with OUD. Shapley additive explanations analysis identified age, number of outpatient visits, history of back and other pain conditions, comorbidity burden, and opioid prescribing patterns as the strongest predictors of incident OUD. Fairness assessment showed an acceptable false negative rate parity across race, age, and sex. In external validation on the UPMC cohort, the GBM model maintained good discrimination (C-statistics=0.756, 95% CI 0.750-0.762) and effective risk stratification.

CONCLUSIONS: An ML algorithm predicting incident OUD derived from OneFlorida+ EHR data performed well in external validation with data using UPMC. The algorithm might be valuable for incident OUD risk prediction and stratification across health systems, with potential to inform early intervention.

BACKGROUND: Concerns about the influence of the pharmaceutical industry on medical education, ranging from education of students to professional development, have led professional societies to recommend regulation of interactions between industry and medical schools. The objective of this study was to evaluate conflict of interest (COI) policies at medical schools in 2023 compared to 2014.

METHODS: This study used a cross-sectional design to evaluate the COI policies at the top 30 medical schools identified by US News and World Report rankings. The authors collected policies by survey and review of public websites, and assessed their quality across 15 domains informed by guidelines published by leading national organizations and previous PharmFree Scorecards. Each domain was graded on a 3-level scale derived from professional organization guidelines, which when totaled corresponded to the following letter grades: an "A" (score 38-45), "B" (32-37), "C" (25-31), and "I/F" (< 25). This study assessed industry payments to school leadership using the 2023 Open Payments database.

RESULTS: Eleven of thirty medical schools submitted COI policies, and the remainder were analyzed based on publicly available information. No school received an "A," 22 (73.3%) schools received a "B", 6 (20.0%) schools received a "C", and 2 (6.7%) schools received an "I/F". Most schools had model policies around COI enforcement (29/30, 96.7%), gift acceptance (25/30, 83.3%), and ghostwriting (24/30, 80.0%). No schools had model policies in limiting direct faculty payments. When comparing 2014 and 2023 Scorecards over the shared 14 domains, 14 (46.7%) schools had a decrease in score, 11 (36.7%) schools had an increase, and 5 (16.7%) schools had no change. Faculty at every school accepted industry payments, including 20 (16.7%) deans and 52 (19.3%) of clerkship directors.

CONCLUSIONS: Medical school COI policies remain less stringent than consensus recommendations; thus, renewed attention to policies and implementation is needed to ensure bias-free medical education.

AIMS: Urinary tract infections (UTIs) are common long-term complications in people with neurogenic bladder (NB). However, there are limited data on how UTIs impact different aspects of quality of life (QoL) in people with NB. Our objective was to understand UTI-related QoL impacts in Veterans with NB.

METHODS: Twenty-three Veterans with NB and UTI diagnoses in the prior year participated in focus groups to share their perceptions of and experiences with UTIs including QoL impacts. Transcripts were coded using inductive and deductive reasoning. A patient survey was also developed using items modified from existing surveys validated for people with NB and new items generated by the study team using the focus group data. Qualitative results on QoL impacts from focus groups were integrated with the quantitative survey data to provide a more comprehensive understanding of UTI-related QoL impacts in people with NB.

RESULTS: UTIs most significantly impacted daily activities, primarily by impairing mobility and restricting independence which led to more limited participation in social, family, and leisure activities. Psychological impacts were more prominent in the focus groups than the survey data.

CONCLUSIONS: Results suggest that people with NB may experience substantial QoL impacts from UTIs, and patient-centered interventions may be needed to decrease the impact of UTIs.

BACKGROUND: The 2023 Predicting Risk of Cardiovascular Disease Events equations estimate 30-year atherosclerotic cardiovascular disease (ASCVD) risk for adults aged 30 to 59 years to inform preventative treatment decisions. We aimed to characterize 30-year ASCVD risk in the eligible US population.

METHODS: We examined adults aged 30 to 59 without known ASCVD who participated in the National Health and Nutrition Examination Survey, 2017 to March 2020 cycle. Using survey weighting to generate nationally representative estimates with 95% CIs, we described 10-year and 30-year ASCVD risk and risk factor control. We then estimated the absolute risk reduction of statin use in populations at high 30-year risk (≥20%.

RESULTS: The cohort included 3229 participants without known ASCVD (mean [SD] age, 44.6 [8.8] years; 49.8% women), representative of 101.9 million (95% CI, 92.2-111.6) US adults. The mean estimated 10-year ASCVD risk was 2.0% (95% CI, 1.9%-2.1%), and the mean 30-year risk was 9.7% (95% CI, 9.4%-10.1%). Of the 9% of the population with high estimated 30-year ASCVD risk, 32.4% (95% CI, 24.0%-40.7%) reported statin use. Most adults with high 30-year ASCVD risk had multiple uncontrolled risk factors, including elevated blood pressure (70.8% [95% CI, 62.4%-79.2%]), obesity (59.9% [95% CI, 52.6%-67.2%]), and elevated total cholesterol (56.2% [95% CI, 45.5%-66.9%]). Expanding primary prevention statins to adults with high 30-year ASCVD risk would change recommendations for 2.5 million (95% CI, 1.9-3.2) adults not currently receiving statins, with an average number needed to treat over 10 years to prevent 1 ASCVD event of 78.3 (95% CI, 74.6-82.0).

CONCLUSIONS: Use of the Predicting Risk of Cardiovascular Disease Events 30-year ASCVD risk equations would identify a population of US adults with low 10-year but high 30-year risk who may warrant enhanced primary prevention strategies.