Publications

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.

OBJECTIVE: The study objective was to examine associations between the use of biologic response modifiers (BRMs), corticosteroids, and oral small molecules (OSMs) and subsequent coccidioidomycosis infection risk among US Medicare beneficiaries with rheumatic or autoimmune diseases.

METHODS: This retrospective cohort study used US 2011 to 2016 Medicare claims data. We identified geographic areas with endemic coccidioidomycosis (≥25 cases per 10,000 beneficiaries). Among beneficiaries having any rheumatic/autoimmune diseases, we identified those initiating BRMs, corticosteroids, and OSMs. Based on refill days supplied, we created time-varying exposure variables for BRMs, corticosteroids, and OSMs with a 90-day lag period after drug cessation. We examined BRMs, corticosteroids, and OSMs and subsequent coccidioidomycosis infection risk using multivariable Cox proportional hazard regression.

RESULTS: Among 135,237 beneficiaries (mean age: 67.8 years; White race: 83.1%; Black race: 3.6%), 5,065 had rheumatic or autoimmune diseases, of which 107 individuals were diagnosed with coccidioidomycosis during the study period (6.1 per 1,000 person-years). Increased risk of coccidioidomycosis was observed among beneficiaries prescribed any BRMs (17.7 per 1,000 person-years; adjusted hazard ratio [aHR] 3.94; 95% confidence interval [CI] 1.18-13.16), followed by individuals treated with only corticosteroids (12.2 per 1,000 person-years; aHR 2.29; 95% CI 1.05-5.03) compared to those treated with only OSMs (4.2 per 1,000 person-years). The rate of those treated with only OSMs was the same as that of beneficiaries without these medications.

CONCLUSION: Incidence of coccidioidomycosis was low among 2011 to 2016 Medicare beneficiaries with rheumatic or autoimmune diseases. BRM and corticosteroid users may have higher risks of coccidioidomycosis compared to nonusers, warranting consideration of screening for patients on BRMs and corticosteroids in coccidioidomycosis endemic areas.

PURPOSE: To examine the association between prescription opioid use trajectories and risk of opioid use disorder (OUD) or overdose among nonmetastatic breast cancer survivors by treatment type.

METHODS: This retrospective cohort study included female nonmetastatic breast cancer survivors with at least 1 opioid prescription fill in 2010-2019 Surveillance, Epidemiology and End Results linked Medicare data. Opioid mean daily morphine milligram equivalents (MME) calculated within 1.5 years after initiating active breast cancer therapy. Group-based trajectory models identified distinct opioid use trajectory patterns. Risk of time to first OUD/overdose event within 1 year after the trajectory period was calculated for distinct trajectory groups using Cox proportional hazards models. Analyses were stratified by treatment type.

RESULTS: Four opioid use trajectories were identified for each treatment group. For 38,030 survivors with systemic endocrine therapy, 3 trajectories were associated with increased OUD/overdose risk compared with early discontinuation: minimal dose (< 5 MME; adjusted hazard ratio [aHR] = 1.73 [95% CI  1.43-2.09]), very low dose (5-25 MME; 2.67 [2.05-3.48]), and moderate dose (51-90 MME; 6.20 [4.69-8.19]). For 9477 survivors with adjuvant chemotherapy, low-dose opioid use was associated with higher OUD/overdose risk (aHR = 7.33 [95% CI 2.52-21.31]) compared with early discontinuation. For 3513 survivors with neoadjuvant chemotherapy, the differences in OUD/OD risks across the 4 trajectories were not significant.

CONCLUSIONS: Among Medicare nonmetastatic breast cancer survivors receiving systemic endocrine therapy or adjuvant chemotherapy, compared with early discontinuation, low-dose or moderate-dose opioid use were associated with six- to sevenfold higher OUD/overdose risk. Breast cancer survivors at high-risk of OUD/overdose may benefit from targeted interventions (e.g., pain clinic referral).

BACKGROUND: Endocrine therapy is the mainstay treatment for breast cancer (BC) to reduce BC recurrence risk. During the first year of endocrine therapy use, nearly 30% of BC survivors are nonadherent, which may increase BC recurrence risk. This study is to examine the association between endocrine therapy adherence trajectories and BC recurrence risk in nonmetastatic BC survivors.

METHODS: This retrospective cohort study included Medicare beneficiaries in the United States (US) with incident nonmetastatic BC followed by endocrine therapy initiation in 2010-2019 US Surveillance, Epidemiology, and End Results linked Medicare data. We calculated monthly fill-based proportion of days covered in the first year of endocrine therapy. We applied group-based trajectory models to identify distinct endocrine therapy adherence patterns. After the end of the first-year endocrine therapy trajectory measurement period, we estimated the risk of time to first treated BC recurrence within 4 years using Cox proportional hazards models.

RESULTS: We identified 5 trajectories of adherence to endocrine therapy in BC Stages 0-I subgroup (n = 28,042) and in Stages II-III subgroup (n = 7781). A trajectory of discontinuation before 6 months accounted for 7.0% in Stages 0-I and 5.8% in Stages II-III subgroups, and this trajectory was associated with an increased treated BC recurrence risk compared to nearly perfect adherence (Stages 0-I: adjusted hazard [aHR] = 1.84, 95% CI = 1.46-2.33; Stages II-III: aHR = 1.38, 95% CI = 1.07-1.77).

CONCLUSIONS: Nearly 7% of BC survivors who discontinued before completing 6 months of treatment was associated with an increased treated BC recurrence risk compared to those with nearly perfect adherence among Medicare nonmetastatic BC survivors.

OBJECTIVE: We aimed to estimate Coccidioides serologic screening rates before initiation of biologic disease-modifying antirheumatic drugs including tofacitinib (b/tsDMARDs), conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), and/or noninhaled corticosteroids.

METHODS: This retrospective cohort study used 2011 to 2016 US Medicare claims data and included beneficiaries with rheumatic or autoimmune disease residing in regions within Arizona, California, and Texas endemic for Coccidioides spp. with ≥1 prescription for a b/tsDMARD, csDMARD, and/or noninhaled corticosteroid. We estimated prior-year serologic screening incidence before initiating b/tsDMARDs, csDMARD, and/or noninhaled corticosteroid.

RESULTS: During 2012 to 2016, 4,331 beneficiaries filled 64,049 prescriptions for b/tsDMARDs, csDMARDs, and noninhaled corticosteroids. Arizona's estimated screening rate was 20.1% (95% confidence interval [95% CI] 14.5-25.7) in the year before prescription initiation for b/tsDMARDs, 8.1% (95% CI 6.5-9.7) before csDMARDs, and 6.9% (95% CI: 5.6-8.2) before corticosteroids. Screening rates for b/tsDMARDs (2.8%, 95% CI 0.0-6.7), csDMARDs (1.0%, 95% CI 0.0-2.0), and corticosteroids (0.8%, 95% CI: 0.4-1.1) were negligible in California and undetected in Texas. Adjusted screening rate before prescription for b/tsDMARDs in Arizona increased from 14.5% (95% CI 7.5-21.5) in 2012 to 26.7% (95% CI 17.6-35.8) in 2016. Rheumatologists prescribing b/tsDMARDs in Arizona screened more than other providers (20.9% [95% CI 13.9-27.9] vs 12.9% [95% CI 5.9-20.0]).

CONCLUSION: Coccidioides serologic screening rates among Medicare beneficiaries with rheumatic/autoimmune diseases on b/tsDMARDs, csDMARDs, and noninhaled corticosteroids was low in Coccidioides spp.-US endemic regions between 2012 and 2016. Alignment of screening recommendations and clinical practice is needed.

OBJECTIVE: To estimate the therapeutic inertia prevalence for patients with type 2 diabetes, develop and validate a machine learning model predicting therapeutic inertia, and determine the added predictive value of area-level social determinants of health (SDOH).

METHODS: This prognostic study with a retrospective cohort design used OneFlorida data (linked electronic health records (EHRs) from 1240 practices/clinics in Florida). The study cohort included adults (aged ≥ 18) with type 2 diabetes, HbA1C ≥ 7% (53 mmol/mol), ≥one ambulatory visit, and ≥one antihyperglycemic medication prescribed (excluded patients prescribed insulin before HbA1C). The outcome was therapeutic inertia, defined as absence of treatment intensification within six months after HbA1C ≥ 7% (53 mmol/mol). The predictors were patient, provider, and healthcare system factors. Machine learning methods included gradient boosting machines (GBM), random forests (RF), elastic net (EN), and least absolute shrinkage and selection operator (LASSO). The DeLong test compared the discriminative ability (represented by C-statistics) between models.

RESULTS: The cohort included 31,087 patients with type 2 diabetes (mean age = 58.89 (SD = 13.27) years, 50.50% male, 58.89% White). The therapeutic inertia prevalence was 39.80% among the 68,445 records. GBM outperformed (C-statistic from testing sample = 0.84, 95% CI = 0.83-0.84) RF (C-statistic = 0.80, 95% CI = 0.79-0.80), EN (C-statistic = 0.80, 95% CI = 0.80-0.81), and LASSO (C-statistic = 0.80, 95% CI = 0.80-0.81), p < 0.05. Area-level SDOH significantly increased the discriminative ability versus models without SDOH (C-statistic for GBM = 0.84, 95% CI = 0.84-0.85 vs. 0.84, 95% CI = 0.83-0.84), p < 0.05.

CONCLUSIONS: Using EHRs of patients with type 2 diabetes from a large state, machine learning predicted therapeutic inertia (prevalence = 40%). The model's ability to predict patients at high risk of therapeutic inertia is clinically applicable to diabetes care.

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.

Background: Chronic cough (CC), characterized as a cough lasting >8 weeks, is a common multi-factorial syndrome in the community, especially in older adults. Methods: Using a pre-existing algorithm to identify patients with CC within the 2011-2018 Medicare beneficiaries, we examined trends in gabapentinoid use through repeated cross-sectional analyses and identified distinct utilization trajectories using group-based trajectory modeling (GBTM) in a retrospective cohort study. Individuals without CC but with any respiratory conditions related to cough served as a comparator group. Results: Among patients with CC, gabapentinoid use increased from 18.6% in 2011 to 24.1% in 2018 (p = 0.002), with a similar upward trend observed in the non-CC cohort but with overall lower usage (14.7% to 18.4%; p < 0.001). Patients with CC had significantly higher burdens of respiratory and non-respiratory comorbidities, as well as greater healthcare service and medication use compared to the non-CC cohort. The GBTM analyses identified three distinct gabapentinoid utilization trajectories for CC and non-CC patients: no use (77.3% vs. 84.5%), low use (13.9% vs. 10.3%), and high use (8.8% vs. 5.2%). Conclusions: Future studies are needed to evaluate the safety and effectiveness of gabapentinoid use in patients with refractory or unexplained CC in real-world settings.

IMPORTANCE: One-third of older adults in the US have depression, often treated with psychotherapy and antidepressants. Previous studies suggesting an increased risk of falls and related injuries (FRI) associated with antidepressant use may be affected by confounding by indication or immortal time bias.

OBJECTIVE: To evaluate the association between FRI risk and first-line treatments in older adults with depression.

DESIGN, SETTING, AND PARTICIPANTS: This cohort study used a target trial emulation framework with a cloning-censoring-weighting approach with Medicare claims data from 2016 to 2019. Participants included fee-for-service beneficiaries aged 65 years or older with newly diagnosed depression. Data were analyzed from October 1, 2023, to March 31, 2024.

EXPOSURES: First-line depression treatments including psychotherapy, sertraline, escitalopram, citalopram, mirtazapine, duloxetine, trazodone, fluoxetine, bupropion, paroxetine, and venlafaxine.

MAIN OUTCOME AND MEASURE: One-year FRI rate, restricted mean survival time (RMST), and adjusted hazard ratio (aHR) with 95% CI.

RESULTS: Among 101 953 eligible beneficiaries (mean [SD] age, 76 [8] years), 63 344 (62.1%) were female, 7404 (7.3%) were Black individuals, and 81 856 (80.3%) were White individuals. Compared with the untreated group, psychotherapy use was not associated with FRI risk (aHR, 0.94 [95% CI, 0.82-1.17]), while other first-line antidepressants were associated with a decreased FRI risk (aHR ranged from 0.74 [95% CI, 0.59-0.89] for bupropion to 0.83 [95% CI, 0.67-0.98] for escitalopram). The FRI incidence ranged from 63 (95% CI, 53-75) per 1000 person-year for those treated with bupropion to 87 (95% CI, 83-90) per 1000 person-year for those who were untreated. The RMST ranged from 349 (95% CI, 346-350) days for those who were untreated to 353 (95% CI, 350-356) days for those treated with bupropion.

CONCLUSIONS AND RELEVANCE: In this cohort study of older Medicare beneficiaries with depression, first-line antidepressants were associated with a decreased FRI risk compared with untreated individuals. These findings provide valuable insights into their safety profiles, aiding clinicians in their consideration for treating depression in older adults.