I am a physician-epidemiologist and PhD candidate in Epidemiology at UTHealth Houston, specializing in real-world evidence, health outcomes research, and observational health data analytics. I use epidemiologic design and reproducible computational methods to transform claims, EHR, survey, clinical, and population health data into evidence for healthcare and research decision-making.

My portfolio demonstrates SAS-based claims cohort construction, leakage-aware biomedical machine learning in R, survey-weighted epidemiology, cancer outcomes research, program evaluation, and research informatics.

• Real-World Evidence & Health Outcomes Research — healthcare utilization, treatment patterns, clinical and population outcomes, disparities, and decision-focused evidence generation
• Observational Health Data Analytics — claims, EHR, survey, and population-health data; cohort construction; observational study design; statistical modeling; and reproducible analysis
• Oncology & Infectious-Disease Research — cancer outcomes and disparities, healthcare access, infectious-disease surveillance, outbreak response, and community interventions

Real-World Evidence & Cohort Analytics
SAS • SQL • Administrative Claims • EHR • Cohort Construction • Claims-Based Phenotyping • Healthcare Utilization • Cost & ROI Analysis

Epidemiologic & Outcomes Research
Observational Study Design • Regression Analysis • Survey-Weighted Analysis • Disease Surveillance • Program Evaluation • Health Outcomes Analysis

Predictive Modeling & Reproducible Research
R • tidyverse • tidymodels • Machine Learning • Grouped Cross-Validation • Model Validation • Reproducible Analytical Workflows

Research Data, Programming & Communication
Stata • Python/pandas • Excel • SQLite • SEER • BRFSS • HINTS • MEPS • NSCH • Biopython/PubMed • Plotly • Research Dashboards

GIS & Mixed Methods
ArcGIS • NVivo • Geospatial Analysis • Thematic Analysis • Codebook Development • Data Visualization

From Administrative Healthcare Data to Reproducible Analytic Cohorts in SAS

Built a reproducible SAS healthcare claims analytics framework using six complementary datasets spanning 1.8+ million enrollment, inpatient, professional, pharmacy, and hospital-discharge records. Transformed raw administrative data into analysis-ready cohorts across multiple clinical and health-services use cases while preserving transparent eligibility, denominator, coding, and data-governance logic.

Methods: Cohort Construction • Claims-Based Phenotyping • Person-Time Denominators • Healthcare Utilization • Medication Exposure • Resource-Use Analysis • ICD-9/ICD-10 Portability
Tools: SAS • PROC SQL • DATA Step • ODS Graphics
Data: CMS DE-SynPUF • Pharmacy Claims • Texas THCIC Inpatient & Facility Data

Biopharma Relevance: Demonstrates foundational real-world data methods for defining observable populations, constructing clinical phenotypes, characterizing medication exposure, and measuring healthcare utilization and resource use.

View the full Claims to Cohorts repository

Leakage-Aware Machine Learning for Parkinson's Classification from Acoustic Voice Biomarkers

Rebuilt a Parkinson's disease classification workflow in R/tidymodels using repeated voice recordings from 252 participants. Implemented participant-grouped train/test splitting and repeated grouped cross-validation to prevent repeated-measures leakage, then compared six machine-learning classifiers under a common validation framework.

Methods: Machine Learning • Grouped Cross-Validation • Predictive Modeling • Model Validation • Permutation Importance
Tools: R • tidymodels • DALEX
Data: UCI Parkinson's Disease Classification Dataset

Biopharma Relevance: Demonstrates how validation design can alter apparent biomedical model performance and why participant independence matters when evaluating evidence generated from repeated clinical measurements.

View the full VoiceMark PD repository