← Back to Portfolio
PhD Candidate · RWE & HEOR

Research &
Projects

7 projects spanning causal inference, health economics, EHR analytics, AI tools, and program evaluation — from dissertation to consulting.

01 · Dissertation 02 · Periodontal RWE 03 · Dental Study 04 · HAI Analytics 05 · Healthy Start 06 · Pediatric HCT 07 · AI Tools

01 · PhD Dissertation

Sequential Target Trial Emulation
for Acute Decompensated Heart Failure

📌 Lead Investigator ✅ Accepted · SER 2026

The central question of my dissertation is methodological and consequential: can we rigorously emulate a randomized controlled trial from observational electronic health record data in acutely ill inpatients — where treatment assignment is deeply confounded by indication?

Using the framework of target trial emulation, I am building a sequential analysis approach to evaluate time-varying treatment strategies in patients hospitalized with acute decompensated heart failure. The real-world EHR environment presents compounding challenges: treatments change with clinical trajectory, patients drop out non-randomly, and the very sickness that drives treatment also drives outcomes.

My approach draws on sequential treatment strategies, carefully constructed eligibility criteria that mirror a hypothetical trial's enrollment logic, and causal inference methods designed for time-varying exposures. The goal is to produce estimates of treatment effectiveness that are both internally valid and directly relevant to the clinical decision points practitioners actually face.

The broader implication: this framework, if validated, creates a replicable template for evidence generation in populations that cannot easily be randomized — with direct applications to RWE strategy in pharma drug development, post-market surveillance, and payer coverage decisions.

Full methodology and findings are in active development and being prepared for manuscript submission. Details available upon request for legitimate research discussion.

Setting
Inpatient EHR data — real-world treatment decisions in acutely ill patients
Core Methods
Target trial emulation · Sequential treatment strategies · Time-varying confounding · Causal inference from longitudinal EHR data
Key Challenge
Treatment is confounded by indication and clinical trajectory — the sickness that drives treatment also drives outcomes
Relevance to Industry
Direct application to RWE strategy in pharma, post-market surveillance, payer evidence generation, and clinical guideline development
Status
✅ Accepted presentation SER 2026 · Manuscript in preparation · PhD milestone complete
02

HEOR · RWE · Causal Inference

NIH R01 — Periodontal Care in Pregnancy ($1.5M)

Co-investigator and lead analyst on a 3-institution NIH-funded study generating real-world evidence on the cost-effectiveness of periodontal treatment in pregnant women. This project spans four interconnected analyses — each producing standalone evidence with direct payer and clinical relevance.

  • Applied G-estimation and marginal structural models to estimate causal effect of preventive periodontal care on periodontitis incidence 24% risk reduction
  • Built 50-state Medicaid fee schedule weighted cost models to estimate per-capita treatment burden and project economic savings ~$6k/capita/yr
  • Defined the first clinical responder endpoint for periodontitis treatment — a consensus definition with direct industry utility for trial design and regulatory submissions
  • Developing a 5–10 year tooth-loss prediction model using BigMouth national database — risk modeling across patient subgroups
  • Presented at SER 2025 · Manuscript in preparation for high-impact epidemiology journal
$6k
Projected savings/capita/year
24%
Risk reduction in periodontitis
03

Epidemiology · National Population Study

Dental Utilization — ADA National Study

A national cross-sectional study in collaboration with the American Dental Association examining dental care utilization patterns among childbearing-aged women with self-reported systemic conditions, using the ADA's national patient-reported outcomes dataset.

  • Designed cross-sectional study examining dental care utilization among women with systemic comorbidities across national ADA patient-reported outcomes data
  • Applied descriptive and inferential statistics in R/SAS to characterize utilization patterns by condition type, demographics, and insurance status
  • Identified key disparities in preventive dental care access in a population with elevated periodontal risk
  • 🏆 Received Certificate of Excellence in Oral Health Research — presented at APHA 2025
04

EHR Analytics · Consulting

HAI Predictive Analytics — Lexington Medical

A data-driven consulting engagement using EHR data from 250,000+ patient records to identify drivers of hospital-acquired infections (HAIs), length of stay, and clinical outcomes — and deliver an actionable prevention strategy.

  • Built survival models and segmented logistic regression using EHR data to evaluate length of stay and mortality
  • Identified high-risk patient clusters in ICU and med-surg transition settings — the most vulnerable windows for HAI acquisition
  • Root-cause analysis mapped care pathway inefficiencies across admission types and unit workflows
  • Delivered strategic plan projecting 15% reduction in HAI incidence through workflow redesign, automated alerts, and targeted prevention protocols
15%
Projected HAI reduction
05

Program Evaluation · Maternal Health

HRSA Healthy Start — Program Evaluation ($5M)

Evaluating a federally-funded maternal and child health program at PRISMA Health — leveraging real-world longitudinal data to assess maternal and perinatal outcomes across 12,000+ participants.

  • Designed and managed REDCap longitudinal data systems for 5,000+ participants across pregnancy and postpartum periods
  • Led QA/QI initiatives: improved data accuracy to 95%, reduced errors by 30%+, fully digitized data collection
  • Supervised community health workers and health educators; digital migration saved 2 hrs/day per staff member
  • Conducted R/SAS statistical analyses on maternal and perinatal outcomes; delivered insights improving program scalability
  • Ensured 100% on-time federal performance report submissions — every reporting period
  • Youngest presenter at HRSA National All-Grantee Meeting 2025
12k+
Program participants
95%
Data accuracy achieved
06

Survival Analysis · Multi-Center Registry

Post-HCT Survival — Pediatric Aplastic Anemia

Multi-center registry analysis examining conditioning regimen effectiveness and survival outcomes in children undergoing allogeneic hematopoietic cell transplantation (HCT) for severe aplastic anemia (SAA) — using competing-risk models to separate relapse from transplant-related mortality.

  • Applied competing-risk survival models to disentangle relapse vs. transplant-related mortality — a methodologically critical distinction for this patient population
  • Analyzed multi-center pediatric HCT registry/cohort data across several conditioning regimens
  • Key finding: Cyclophosphamide identified as optimal conditioning regimen — evidence with direct implication for future trial protocol design
  • Presented at SPER 2025 — Society for Pediatric and Perinatal Epidemiologic Research
07

AI · Data Science · Python

AI & Data Science Tools Portfolio

Building scalable tools that amplify and automate analytic workflows — from clinical NLP to strategy backtesting engines.

📈
AI-Enabled Strategy Backtester
Python · Pandas · NumPy · OOP Architecture
Modular backtesting engine for multi-timeframe trading strategies. Features scenario modeling, simulation design, and forecasting — built with a clean OOP architecture for extensibility.
⚡ Directly applicable to clinical trial simulation & resource modeling
🔬
Automated Univariate Analysis App
Python · Flask · SciPy · Matplotlib · Plotly
Full-stack web application that automates the full EDA and regression modeling pipeline: dataset upload → descriptive stats → visualizations → models → odds ratios. Designed to replicate biostatistician workflows at scale.
⚡ Replicates biostatistician workflows in a scalable, deployable tool
🤖
AI-Assisted Clinical Notes & Research Tool
Python · spaCy · LLMs · Oracle GenAI · Vector Search
NLP engine for clinical text mining and literature synthesis. Combines LLM pipelines, clinical NLP, and vector search to extract structured information from unstructured clinical text and research corpora.
⚡ EHR data extraction · Unstructured data analysis · Evidence synthesis