Publications

While clozapine is the most effective antipsychotic treatment for treatment-resistant schizophrenia, it remains underutilized across the United States, warranting a more comprehensive understanding of variation in use at the county level, as well as characterization of existing prescribing patterns. Here, we examined both Medicaid and Medicare databases to (1) characterize temporal and geographic variation in clozapine prescribing and, (2) identify patient-level characteristics associated with clozapine use. We included Medicaid and Fee for Service Medicare data in the state of Pennsylvania from January 1, 2013, through December 31, 2019. We focused on individuals with continuous enrollment, schizophrenia diagnosis, and multiple antipsychotic trials. Geographic variation was examined across counties of Pennsylvania. Regression models were constructed to determine demographic and clinical characteristics associated with clozapine use. Out of 8,255 individuals who may benefit from clozapine, 642 received treatment. We observed high medication burden, overall, including multiple antipsychotic trials. We also identified variation in clozapine use across regions in Pennsylvania with a disproportionate number of prescribers in urban areas and several counties with no identified clozapine prescribers. Finally, demographic, and clinical determinants of clozapine use were observed including less use in people identified as non-Hispanic Black, Hispanic, or with a substance use disorder. In addition, greater medical comorbidity was associated with increased clozapine use. Our work leveraged both Medicaid and Medicare data to characterize and surveil clozapine prescribing. Our findings support efforts monitor disparities and opportunities for the optimization of clozapine within municipalities to enhance clinical outcomes.

BACKGROUND: Hospitalizations involving opioid use disorder (OUD) are increasing. Medications for opioid use disorder (MOUD) reduce mortality and acute care utilization. Hospitalization is a reachable moment for initiating MOUD and arranging for ongoing MOUD engagement following hospital discharge. Despite existing quality metrics for MOUD initiation and engagement, few hospitals provide hospital based opioid treatment (HBOT). This protocol describes a cluster-randomized hybrid type-2 implementation study comparing low-intensity and high-intensity implementation support strategies to help community hospitals implement HBOT.

METHODS: Four state implementation hubs with expertise in initiating HBOT programs will provide implementation support to 24 community hospitals (6 hospitals/hub) interested in starting HBOT. Community hospitals will be randomized to 24-months of either a low-intensity intervention (distribution of an HBOT best-practice manual, a lecture series based on the manual, referral to publicly available resources, and on-demand technical assistance) or a high-intensity intervention (the low-intensity intervention plus funding for a hospital HBOT champion and regular practice facilitation sessions with an expert hub). The primary efficacy outcome, adapted from the National Committee on Quality Assurance, is the proportion of patients engaged in MOUD 34-days following hospital discharge. Secondary and exploratory outcomes include acute care utilization, non-fatal overdose, death, MOUD engagement at various time points, hospital length of stay, and discharges against medical advice. Primary, secondary, and exploratory outcomes will be derived from state Medicaid data. Implementation outcomes, barriers, and facilitators are assessed via longitudinal surveys, qualitative interviews, practice facilitation contact logs, and HBOT sustainability metrics. We hypothesize that the proportion of patients receiving care at hospitals randomized to the high-intensity arm will have greater MOUD engagement following hospital discharge.

DISCUSSION: Initiation of MOUD during hospitalization improves MOUD engagement post hospitalization. Few studies, however, have tested different implementation strategies on HBOT uptake, outcome, and sustainability and only one to date has tested implementation of a specific type of HBOT (addiction consultation services). This cluster-randomized study comparing different intensities of HBOT implementation support will inform hospitals and policymakers in identifying effective strategies for promoting HBOT dissemination and adoption in community hospitals.

TRIAL REGISTRATION: NCT04921787.

IMPORTANCE: Medicare began paying for medications for opioid use disorder (MOUD) at opioid treatment programs (OTPs) that dispense methadone and other MOUD in January 2020. There has been little research describing the response to this payment change and whether it resulted in more patients receiving MOUD or just a shift in who pays for this care.

OBJECTIVE: To describe how many and which Medicare beneficiaries receive care from OTPs and how this compares to those receiving MOUD in other settings.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study included all patients receiving MOUD care identified in 2019-2022 100% US Medicare Parts B and D claims. Patients receiving care in an OTP who were dually insured with Medicare and Medicaid in the 2019-2020 Transformed Medicaid Statistical Information System were also included.

EXPOSURE: Receiving MOUD care in an OTP.

MAIN OUTCOMES AND MEASURES: Comparisons of 2022 beneficiaries treated in OTPs vs other non-OTP settings in 2022.

RESULTS: The share of Medicare beneficiaries treated by OTPs rose steadily from 4 per 10 000 (14 160 beneficiaries) in January 2020 to 7 per 10 000 (25 596 beneficiaries) in August 2020, then plateaued through December 2022; of 38 870 patients (23% ≥66 years; 35% female) treated at an OTP in 2022, 96% received methadone. Patients in OTPs, compared to those receiving MOUD in other settings, were more likely be 65 years and younger (65% vs 62%; P < .001), less likely to be White (72% vs 82%; P < .001), and more likely to be an urban resident (86% vs 74%; P < .001). When Medicare OTP coverage began, there was no associated drop in the number of dually insured patients with Medicaid with an OTP claim. Of the 1854 OTPs, 1115 (60%) billed Medicare in 2022, with the share billing Medicare ranging from 13% to 100% across states.

CONCLUSIONS AND RELEVANCE: This study showed that since the initiation of Medicare OTP coverage in 2020, there has been a rapid increase in the number of Medicare beneficiaries with claims for OTP services for MOUD, and most OTPs have begun billing Medicare. Patients in OTPs were more likely to be urban residents and members of racial or ethnic minority groups than the patients receiving other forms of MOUD.

BACKGROUND: Permanent supportive housing (PSH) programs, which have grown over the last decade, have been associated with changes in health care utilization and spending. However, little is known about the impact of such programs on use of prescription drugs critical for managing chronic diseases prevalent among those with unstable housing.

OBJECTIVE: To evaluate the effects of PSH on medication utilization and adherence among Medicaid enrollees in Pennsylvania.

DESIGN: Difference-in-differences study comparing medication utilization and adherence between PSH participants and a matched comparison cohort from 7 to 18 months before PSH entry to 12 months post PSH entry.

SUBJECTS: Pennsylvania Medicaid enrollees (n = 1375) who entered PSH during 2011-2016, and a propensity-matched comparison cohort of 5405 enrollees experiencing housing instability who did not receive PSH but received other housing services indicative of episodic or chronic homelessness (e.g., emergency shelter stays).

MAIN MEASURES: Proportion with prescription fill, mean proportion of days covered (PDC), and percent adherent (PDC ≥ 80%) for antidepressants, antipsychotics, anti-asthmatics, and diabetes medications.

KEY RESULTS: The PSH cohort saw a 4.77% (95% CI 2.87% to 6.67%) relative increase in the proportion filling any prescription, compared to the comparison cohort. Percent adherent among antidepressant users in the PSH cohort rose 7.41% (95% CI 0.26% to 14.57%) compared to the comparison cohort. While utilization increased in the other medication classes among the PSH cohort, differences from the comparison cohort were not statistically significant.

CONCLUSIONS: PSH participation is associated with increases in filling prescription medications overall and improved adherence to antidepressant medications. These results can inform state and federal policy to increase PSH placement among Medicaid enrollees experiencing homelessness.

BACKGROUND AND OBJECTIVES: Fall injuries are prevalent in older adults, yet whether higher spending occurs after nonfracture (NFFI) and fracture is unknown. We examined whether incident fall injuries, including NFFI and fractures, were associated with higher Medicare spending in 12 months after incident events in older adults.

RESEARCH DESIGN AND METHODS: The Health, Aging, and Body Composition Study included 1 595 community-dwelling adults (53% women, 37% Black; 76.7 ± 2.9 years) with linked Medicare Fee-For-Service (FFS) claims at 2000/01 exam. Incident outpatient and inpatient fall injuries (N = 448) from 2000/01 exam to December 31, 2008 were identified using the first claim with a nonfracture injury diagnosis code with a fall E-code, or a fracture diagnosis code with/without an E-code. Up to 3 participants without fall injuries (N = 1 147) were matched on nonfall events to 448 participants in the fall injury month. We calculated the change in monthly FFS spending in 12 months before versus after index events in both groups. Generalized linear regression with centered outcomes and gamma distributions examined the association of prepost expenditure changes with fall injuries (including NFFI and fractures) adjusting for related covariates.

RESULTS: Monthly spending increased after versus before fall injuries (USD$2 261 vs $981), nonfracture (N = 105; USD$2 083 vs $1 277), and fracture (N = 343; USD$2 315 vs $890) injuries (all p < .0001). However, after adjusting for covariates in final models, fall injuries were not significantly associated with larger increases in spending/month versus nonfall events (differential increase: USD$399.58 [95% CI: -USD$44.95 to $844.11]). Fracture prepost change in monthly spending was similar versus NFFI (differential increase: USD$471.93 [95% CI: -USD$21.17 to $965.02]).

DISCUSSION AND IMPLICATIONS: Although substantial increases occurred after injuries, with fracture and NFFI increasing similarly, changes in monthly spending after fall injury were not different compared to nonfall events. Our results contribute to the understanding of subsequent spending after fall injury that may inform further research on fall injury-related health care spending.

BACKGROUND: Opioid use disorder contributes to maternal morbidity and mortality in the United States. Little is known about how the patterns of buprenorphine dose and duration throughout pregnancy may affect neonatal and postpartum outcomes.

OBJECTIVE: To determine the associations between trajectories of buprenorphine utilization and dose during pregnancy on maternal and neonatal health outcomes.

STUDY DESIGN: Retrospective cohort study among 2925 pregnant persons with opioid use disorder, followed from the estimated start date of pregnancy through 90 days after delivery. We used administrative healthcare data from Medicaid-enrolled individuals to assess buprenorphine dose and use and maternal (postpartum buprenorphine continuation and overdose) and neonatal (low birthweight, neonatal abstinence syndrome (NAS)) outcomes. Group-based trajectory modelling was used to identify trajectories of buprenorphine dose and use during pregnancy. Weighted multivariable logistic regression assessed the association between buprenorphine trajectories and outcomes.

RESULTS: We identified 8 trajectories of buprenorphine utilization and dose during pregnancy. Regression analyses found that high doses of buprenorphine and a longer duration of buprenorphine use during pregnancy was associated with higher odds of postpartum buprenorphine continuation and reduced rates of overdose. Higher doses and longer duration of buprenorphine treatment were not associated with an increase in NAS or term low birth weight, relative to moderate or low doses or shorter treatment duration.

CONCLUSION: A longer duration and higher dose of buprenorphine treatment during pregnancy were associated with improved odds of postpartum buprenorphine continuation and were not associated with adverse neonatal outcomes.

IMPORTANCE: Individuals with opioid use disorder (OUD) and criminal justice system involvement experience high rates of overdose death. Historical data point to limited use of medications for opioid use disorder (MOUD) in criminal justice system-referred treatment for OUD as playing a role. However, how MOUD use among those referred to treatment by the criminal justice system has changed relative to other referral sources over time is still unclear, as well as how it varies across states.

OBJECTIVE: To examine disparities in the use of MOUD between individuals referred to treatment by the criminal justice system compared to other referral sources over time.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study included admissions to specialty substance use treatment facilities for OUD in the national Treatment Episodes Dataset-Admissions from 2014 to 2021. Logistic regression models were used to examine trends in the probability of MOUD use among individuals with and without criminal justice referrals for OUD treatment, as well as any differential trends by state. The data were analyzed from September 2023 to August 2024.

MAIN OUTCOME AND MEASURE: The main outcome was the probability that treatment for individuals with OUD included MOUD.

RESULTS: A total of 3 235 445 admissions were analyzed in the study data. Among individuals referred to OUD treatment by the criminal justice system, the probability that treatment included MOUD increased by 3.42 percentage points (pp) (95% CI, 3.37 pp to 3.47 pp) annually from 2014 to 2021. This was faster than the increase in the probability of MOUD use for noncriminal justice-referred admissions (2.49 pp [95% CI, 2.46 pp to 2.51 pp) and reduced, but did not eliminate, disparities in MOUD use between individuals with and without criminal justice system-referred treatment. In 2021, only 33.6% of individuals in criminal justice system-referred treatment received MOUD, 15.6 pp lower than for individuals referred to treatment by other sources. Trends in the probability of MOUD use varied substantially for individuals in criminal justice system-referred treatment across states, but very few experienced enough growth to eliminate this disparity.

CONCLUSIONS AND RELEVANCE: The results of this cross-sectional study suggest that targeted efforts to address persistent disparities in MOUD use among those with OUD and criminal justice system involvement are needed to address the poor health outcomes experienced by this population.

AIMS: To assess diabetes-related complications, glycemic levels, and healthcare utilization 12 months after exposure to therapeutic inertia among patients with type 2 diabetes mellitus (T2D).

METHODS: This retrospective cohort study analyzed data from the OneFlorida Clinical Research Consortium (electronic health records from Florida practices/clinics). The cohort included adult patients (≥18 years old) with T2D who had an HbA1c≥7.0% (53 mmol/mol) recorded from January 1, 2014-September 30, 2019. Therapeutic inertia (exposed vs. not exposed) was evaluated during the six months following HbA1c≥7.0% (53 mmol/mol). The outcomes assessed during the 12-month follow-up period included diabetes-related complications (continuous Diabetes Complications and Severity Index (DCSI)), glycemic levels (continuous follow-up HbA1c lab), and healthcare utilization counts. We analyzed data using multivariable regression models, adjusting for covariates.

RESULTS: The cohort included 26,881 patients with T2D (58.94% White race, 49.72% female, and mean age of 58.82 (SD=13.09)). After adjusting for covariates, therapeutic inertia exposure was associated with lower DCSI (estimate=-0.14 (SE=0.03), p < 0.001), higher follow-up HbA1c (estimate=0.14 (SE=0.04), p < 0.001), and lower rates of ambulatory visits (rate ratio=0.79, 95% CI=0.75-0.82).

CONCLUSIONS: Findings communicate the clinical practice implications and public health implications for combating therapeutic inertia in diabetes care.

We used group-based trajectory models to identify 4 distinct trajectory patterns of adherence to preexposure prophylaxis (PrEP) among 20 696 users. Only 44.5% were consistently PrEP adherent, with younger age, being female, or having substance use disorder or depression associated with early discontinuation. Public health efforts are needed to improve PrEP adherence.

OBJECTIVE: To develop a natural language processing (NLP) package to extract social determinants of health (SDoH) from clinical narratives, examine the bias among race and gender groups, test the generalizability of extracting SDoH for different disease groups, and examine population-level extraction ratio.

METHODS: We developed SDoH corpora using clinical notes identified at the University of Florida (UF) Health. We systematically compared 7 transformer-based large language models (LLMs) and developed an open-source package - SODA (i.e., SOcial DeterminAnts) to facilitate SDoH extraction from clinical narratives. We examined the performance and potential bias of SODA for different race and gender groups, tested the generalizability of SODA using two disease domains including cancer and opioid use, and explored strategies for improvement. We applied SODA to extract 19 categories of SDoH from the breast (n = 7,971), lung (n = 11,804), and colorectal cancer (n = 6,240) cohorts to assess patient-level extraction ratio and examine the differences among race and gender groups.

RESULTS: We developed an SDoH corpus using 629 clinical notes of cancer patients with annotations of 13,193 SDoH concepts/attributes from 19 categories of SDoH, and another cross-disease validation corpus using 200 notes from opioid use patients with 4,342 SDoH concepts/attributes. We compared 7 transformer models and the GatorTron model achieved the best mean average strict/lenient F1 scores of 0.9122 and 0.9367 for SDoH concept extraction and 0.9584 and 0.9593 for linking attributes to SDoH concepts. There is a small performance gap (∼4%) between Males and Females, but a large performance gap (>16 %) among race groups. The performance dropped when we applied the cancer SDoH model to the opioid cohort; fine-tuning using a smaller opioid SDoH corpus improved the performance. The extraction ratio varied in the three cancer cohorts, in which 10 SDoH could be extracted from over 70 % of cancer patients, but 9 SDoH could be extracted from less than 70 % of cancer patients. Individuals from the White and Black groups have a higher extraction ratio than other minority race groups.

CONCLUSIONS: Our SODA package achieved good performance in extracting 19 categories of SDoH from clinical narratives. The SODA package with pre-trained transformer models is available at https://github.com/uf-hobi-informatics-lab/SODA_Docker.