I’m a clinician-trained epidemiologist and healthcare data analyst who turns surveillance, clinical, observational, regulatory, and geospatial data into clear, reproducible evidence for public-health decisions.
Explore my complete portfolio · LinkedIn · Email
Open to healthcare data analyst, epidemiologist, biostatistician, clinical data analyst, and related opportunities with employers able to support future U.S. work authorization.
Decision question: How much time elapsed between firm initiation of a medical-device recall and publication in an FDA Enforcement Report?
- Analyzed 11,005 recall records initiated during 2022–2025
- Built a reproducible openFDA acquisition pipeline, typed DuckDB SQL layer, Python analysis, and automated QA
- Found median initiation-to-report times of 48 days for Class I, 50 for Class II, and 62 for Class III records
- Distinguished a statistically detectable class difference from its small practical effect size (epsilon-squared = 0.011)
- Added a substantive SAS validation program with explicit rank, tie-correction, effect-size, reporting, and cross-software comparison steps
Tools: SAS, Python, DuckDB SQL, openFDA, nonparametric statistics, regulatory data, QA/QC
View: Repository · Portfolio case study
Decision question: Which age groups carried the greatest cumulative hospitalization burden from COVID-19, influenza, and RSV during the 2025–26 respiratory season?
- Analyzed 53 weeks of CDC RESP-NET observed cumulative rates across three viruses and 11 age groups
- Found the highest combined rates among adults aged 85+ and infants younger than 1 year
- Showed that RSV accounted for 71% of the infant rate while influenza dominated most older age groups
- Reconciled the component series to CDC’s separately published combined series
Tools: SAS, Python, CDC RESP-NET, surveillance epidemiology, QA/QC
View: Repository · Portfolio case study
Decision question: Where do publicly reported emergency-department time measures indicate operational benchmarking priorities?
- Analyzed CMS Care Compare data for 4,658 facilities and six ED measures
- Used DuckDB SQL for transformation, facility joins, and explicit QA checks
- Added a substantive SAS workflow for import, validation, grouped analysis, and reporting
- Found a 69-minute median difference between very-high- and low-volume ED groups (95% bootstrap CI: 65–75)
Tools: SAS, Python, DuckDB SQL, CMS Care Compare, bootstrap inference, QA/QC
View: Repository · Dashboard
Decision question: How much laboratory-defined CKD is detectable among U.S. adults with diagnosed diabetes, and how often do affected adults report awareness?
- Integrated five CDC/NCHS NHANES files with survey weights, strata, and PSUs
- Derived race-free CKD-EPI 2021 eGFR and combined it with urine albumin-creatinine ratio
- Built a complete SAS survey workflow with an independently executed Python validation analysis
- Estimated analytic CKD in 39.3% of adults with diagnosed diabetes and an 82.0% awareness gap among those with diabetes and CKD
Tools: SAS, Python, NHANES, complex-survey analysis, CKD-EPI 2021, QA/QC
View: Repository · Portfolio case study
Decision question: Do staffing capacity and nursing-staff turnover align with low CMS overall ratings after facility and market context are considered?
- Analyzed 14,690 active nursing homes in CMS Care Compare data
- Fit a multivariable logistic model adjusting for occupancy, case mix, bed count, ownership, and state
- Found a 54.7% low-rating share in the lowest adjusted-staffing quartile versus 23.6% in the highest
- Kept interpretation explicitly associational rather than causal
Tools: R, DuckDB SQL, Python, CMS Care Compare, logistic regression, QA/QC
View: Repository · Portfolio case study
Decision question: Where do high population-adjusted incidence and high operational case burden point to different surveillance priorities?
- Analyzed CDC’s 2025 provisional pertussis report across 52 reporting jurisdictions
- Built a typed DuckDB SQL workflow with validation, ranking, and threshold flags
- Prepared a QGIS-ready state join with explicit handling of New York State and New York City
- Reconciled jurisdiction records exactly to the reported national total of 28,783 cases
Tools: DuckDB SQL, QGIS, infectious-disease surveillance, QA/QC
View: Repository · Dashboard
- Analysis: SAS, R, Python, SQL, SPSS, Epi Info
- Visualization and GIS: ArcGIS, QGIS, Tableau, Power BI
- Methods: epidemiologic study design, regression, complex-survey analysis, propensity-score methods, nonparametric statistics, reproducible analysis, QA/QC
- Data environments: infectious-disease surveillance, clinical trials, observational research, regulatory data, REDCap, NEDSS
- First-author study of juvenile idiopathic arthritis, anxiety, and depression in 192,019 U.S. children and adolescents (2025)
- Co-author of a machine-learning study of juvenile idiopathic arthritis screening in 223,195 U.S. children, published in Frontiers in Public Health (2025)
- Co-author of a propensity-score-weighted study on asthma and juvenile idiopathic arthritis in Pediatric Rheumatology (2026)
I prioritize transparent assumptions, reproducible workflows, careful data-quality checks, appropriate interpretation, and communication that distinguishes association from causation. All portfolio projects use public data and exclude PHI, confidential employer information, and proprietary Texas DSHS material.