Juan P Brito

Has grant

Mayo Clinic Minnesota
Country flagUnited States

Research Interests

ApplyKite TwinLearns as you apply

Make your next move.

Prepare your application

Review drafts before you send them.

Explore related searches

Recent Grants

Grant: Open

Comparative Effectiveness and Safety of Osteoporosis Drug Therapies

Open Date: 2022-09-15

Close Date: 2027-06-30

Grant: Open

De-implementation of inappropriate thyroid ultrasound

Open Date: 2022-07-14

Close Date: 2027-06-30

Articles (22)

Year:

2026

Year:

2026

BACKGROUND Named Entity Recognition (NER), a key task in extracting structured information from unstructured text, such as names, dates, and locations. Traditional systems that apply NER to speech rely on a pipeline of Automatic Speech Recognition (ASR) and text-based NER, which can suffer from error propagation and performance degradation. There is a growing need to understand the capabilities of more advanced end-to-end (E2E) models that process speech directly, especially in the context of unscripted, human-to-human conversations relevant to clinical settings. OBJECTIVE To systematically evaluate the performance, architectures, data sources, and methodological innovations of E2E spoken NER models applied to unscripted, spontaneous human-to-human conversations. METHODS Following PRISMA 2020, we searched MEDLINE, Scopus, IEEE Xplore, arXiv, and other databases from inception, up to June 13, 2025. Studies were eligible if they applied E2E NER models to audio or audiovisual recordings of unscripted conversations between at least two people. Pipeline-only, text-only, scripted, and human–machine interaction studies were excluded. Data on conversational corpora, annotation schemas, model architectures, training strategies, and NER and ASR performance metrics were extracted and synthesized. RESULTS Of 1,094 records found, five studies met inclusion criteria. Four evaluated models on the SLUE-VoxPopuli benchmark, while one focused on spontaneous telephone and contact-center dialogues. All E2E approaches used self-supervised speech encoders (wav2vec 2.0, HuBERT, WavLM), and four studies employed the character-level Connectionist Temporal Classification (CTC) loss, and language-model–guided decoding. On SLUE-VoxPopuli, the strongest E2E systems achieved up to 77.2% NER F1 (ROVER ensemble) with 8.6% WER, surpassing competitive pipeline baselines. However, performance declined sharply on real-world conversational corpora, with F1 falling to 30.6% on noisy cross-domain telephone speech. E2E models required large amounts of external automatic speech recognition (ASR) and spoken language understanding (SLU) pretraining data to match or exceed pipelines, highlighting sensitivity to noise, domain shift, and limited in-domain labels. CONCLUSIONS E2E models leveraging state-of-the-art self-supervised learning encoders, contextual modeling, and semi-supervised learning represent the forefront of spoken NER. While they hold significant promise, their performance on noisy, real-world conversational data highlights a critical gap that must be addressed before they can be reliably deployed in challenging domains like healthcare. Future progress will depend on the development of domain-specific datasets and models robust to the complexities of spontaneous human interaction. Registration: PROSPERO (CRD42024627286).

Year:

2026

Objective To evaluate the effectiveness of multiple decision aid strategies in promoting high quality shared decision making for prevention of stroke in patients with non-valvular atrial fibrillation. Design Cluster randomized controlled trial. Setting Six academic medical centers in the United States. Participants Patient participants were aged ≥18 with a diagnosis of non-valvular atrial fibrillation, at risk for stroke (CHA 2 DS 2 -VASc ≥1 for men, ≥2 for women), and scheduled for a clinical appointment to discuss stroke prevention strategies. Participating clinicians were those who manage stroke prevention strategies for participating patients. Intervention Patients were randomized to use a patient decision aid or usual care; clinicians were randomized to use an encounter decision aid or usual care with all participating patients. Main outcome measures Primary outcome measures were quality of shared decision making measured by OPTION12, knowledge of atrial fibrillation and its management, and decisional conflict. Results 1117 participants across six sites were included in the analysis. Compared with usual care, the combined use of both the patient decision aid and the encounter decision aid improved the quality of shared decision making (adjusted mean difference 12.1 (95% confidence interval (CI) 8.0 to 16.2; P<0.001), improved patients’ knowledge (odds ratio 1.68 (95% CI 1.35 to 2.09; P<0.001), and reduced patients’ decisional conflict (adjusted mean difference −6.3 (95% CI −9.6 to −3.1; P<0.001). Statistically significant improvements were also observed with the encounter decision aid alone versus usual care for all three outcomes and with the patient decision aid alone versus usual care for quality of shared decision making and knowledge. No important differences were observed in treatment choices for stroke prevention or in participants’ satisfaction. No statistically significant difference in the length of visit across study groups was detected. Conclusion Patients who received any decision aid (encounter decision aid, patient decision aid, or both) had lower decisional conflict, better shared decision making, and greater knowledge than those receiving no decision aid, except for the effect of the patient decision aid on decisional conflict, which did not reach statistical significance. The study establishes that use of either pre-visit or in-visit decision aids individually or in combination is advantageous compared with usual care. Trial registration ClinicalTrials.gov NCT04357288 .

Year:

2025

Year:

2025

Year:

2025

Importance The increasing use of glucagon-like peptide-1 receptor agonists (GLP-1RA) demands a better understanding of their association with thyroid cancer. Objective To estimate the risk of incident thyroid cancer among adults with type 2 diabetes being treated with GLP-1RA vs other common glucose-lowering medications. Design, Setting, and Participants This was a prespecified secondary analysis of a target trial emulation of a comparative effectiveness study using claims data for enrollees in commercial, Medicare Advantage, and Medicare fee-for-service plans across the US. Eligible participants were adults with type 2 diabetes at moderate risk for cardiovascular disease and without history of thyroid cancer who had newly filled prescriptions for GLP-1RA, sodium-glucose cotransporter 2 inhibitor (SGLT2i), dipeptidyl peptidase-4 inhibitor (DPP4i), or sulfonylurea from January 1, 2014, to December 31, 2021. Data were analyzed February 1 to October 31, 2024. Main Outcomes and Measures Overall and piecewise (<1, 1-2, and ≥2 years since treatment initiation) hazard ratios (HRs) for thyroid cancer with use of GLP-1RA vs the other 3 drug classes were estimated using inverse propensity score weighted Cox proportional hazards models. Modified intention-to-treat (mITT) (primary) and as-treated (sensitivity) analyses were performed. Results Of 351 913 patients (mean [SD] age, 65.3 [8.5] years; 173 391 [49.3%] females and 178 522 [50.7%] males), 41 112 started treatment with GLP-1RA; 76 093, with DPP4i; 43 499, with SGLT2i; and 191 209, with sulfonylurea therapy. The numbers of patients diagnosed with thyroid cancer were 69 (0.17%) in the GLP-1RA group, 172 (0.23%) in the DPP4i group, 72 (0.17%) in the SGLT2i group, and 381 (0.20%) in the sulfonylurea group. In the mITT analysis, GLP-1RA initiation was not significantly associated with increased overall risk for thyroid cancer compared to the other 3 diabetes drugs (HR, 1.24; 95% CI, 0.88-1.76). However, the risk for thyroid cancer was significantly higher within the first year after GLP-1RA initiation (HR, 1.85; 95% CI, 1.11-3.08) and was amplified in the overall as-treated analysis that censored patients when therapy was discontinued or another medication was added (HR, 2.07; 95% CI, 1.10-3.95). Conclusions and Relevance This secondary analysis of a target trial emulation of a comparative effectiveness study found that despite the low absolute risk of thyroid cancer among patients receiving GLP-1RA therapy, there was an increased risk of new thyroid cancer diagnoses within the first year of GLP-1RA initiation compared to 3 other diabetes drugs. This finding may have been due to enhanced early detection; therefore, further research is necessary to understand the underlying causes of this association.

Year:

2025

Year:

2025

Year:

2025

Objectives Diagnostic reconciliation is the collaborative process between patients and clinicians to create and reconcile evidence-based, feasible, and desirable care plans. However, the specific components of this process remain unclear. The objective of this study was to develop the first comprehensive framework to elucidate the diagnostic reconciliation process. Methods We followed a multi-step and iterative approach to develop the framework, including a focused systematic review of diagnostic conversations, quantitative evaluation of recordings of real-life clinical visits recordings, and stakeholder engagement (e.g., patients, clinicians, researchers). Results We identified 17 potential components to the process of diagnostic reconciliation through literature review and stakeholder engagement. After review of 56 clinical visits and further stakeholder engagement, we developed a final framework including four categories: 1) understanding the need for a test/referral, 2) logistics of test/referral scheduling, 3) test/referral information, and 4) test/referral results. Conclusions The proposed framework lays the foundation for evaluation and improvement of diagnostic conversations in practice. Clinicians can enhance patient-centered diagnosis by co-creating diagnostic plans of care in practice and using the components described in the novel diagnostic reconciliation framework.

Year:

2025

Year:

2025

Year:

2025

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Year:

2024

Collaborators (13)

Geoffrey Barnes

Associate Professor

University of Michigan

UNITED STATES
View details

Guillermo Umpierrez

Professor of Medicine

Emory University School of Medicine

UNITED STATES
View details

Rozalina McCoy

Associate Professor of Medicine

University of Maryland School of Medicine

UNITED STATES
View details

Daniel Witt

Professor and Chair

University of Utah

UNITED STATES
View details

Elizabeth Jackson

Professor

The University of Alabama at Birmingham School of Medicine

UNITED STATES
View details

Diana Barb

Associate Professor

University of Florida

UNITED STATES
View details

Javad Azadi

-

UNITED STATES
View details

Marie McDonnell

Section Chief, Diabetes

Brigham and Women's Hospital

UNITED STATES
View details

David Toro-Tobon

Mayo Scholar

Mayo Clinic

UNITED STATES
View details

Christina LaVecchia

Assistant Professor

University of Cincinnati

UNITED STATES
View details

Kenzie Cameron

Professor

Northwestern University Feinberg School of Medicine

UNITED STATES
View details

Elissa M. Ozanne

University of Utah

UNITED STATES
View details

Carma L. Bylund

Professor and Associate Chair

University of Florida

UNITED STATES
View details