Healthcare Data Projects & Case Studies

Explore the projects I've worked on throughout my career and the tools I used to solve them.


HIPAA De-Identification Pipeline

The Problem

Healthcare organizations need to analyze patient data for operational insights, but strict HIPAA regulations prevent the use of identifiable patient information (PHI/PII) outside of production systems.

What I Built

A reusable Python toolkit that automatically de-identifies patient data to meet HIPAA Safe Harbor standards.

The Result

Reduced de-identification time from hours to minutes with a production-ready, open-source toolkit now available to any healthcare organization.

Tools
PostgreSQL Python (pandas, Faker, pyodbc) Streamlit HIPAA Safe Harbor standards

Note: The demo app may take a few seconds to load if inactive. For production use with real patient data, please download the toolkit from GitHub and run it locally in your secure environment.


Provider Data Deduplication Case Study

The Problem

Duplicate physician records from inconsistent data entry were causing patients to appear in wrong provider queues, disrupting care coordination and creating operational bottlenecks.

What I Built

Identified the root cause of those duplicate records, quantified the scope of the problem, and developed both immediate workarounds and long-term solutions.

The Result

Restored accurate patient-provider matching for 1,500+ affected records and eliminated queue assignment failures that were delaying patient care.

Tools
SQL Server T-SQL Python Microsoft Excel data profiling gap analysis cross-functional stakeholder management

Patient Churn Prediction Model

The Problem

Healthcare practices need to identify at-risk patients and prioritize retention outreach, but predicting churn accurately and translating it into actionable business decisions isn't as simple as it sounds.

What I Built

A synthetic patient dataset generator and a full churn-prediction pipeline — EDA, Random Forest model with threshold tuning, feature importance validation across two models, and a revenue-at-risk layer to prioritize outreach.

The Result

Identified billing issues as the strongest churn driver, with Distance To Facility (Miles) and an Insurance Type of 'Self Pay' also contributing. Built a revenue-at-risk model surfacing $52K in monthly revenue tied to at-risk patients.

Tools
Python (pandas, scikit-learn, numpy) Random Forest statistical testing (chi-square, correlation) Jupyter

Additional Work

Slump Dog Sluggers

A Tableau Public dashboard using a PostgreSQL database that tracked individual Philadelphia Phillies’ batting performance during the 2025 season. Designed a multi-timeframe rolling average analysis (7-game, 14-game, 30-game windows) to try and test how big of a factor does age play in a player’s ability to bounce back from a slump.

View on GitHub View on Tableau Public

The $97.7M Hit Drought Tracker

A Streamlit dashboard set up to give a quick snapshot of the stats for the Phillies’ top 4 hitters during the 2026 season and testing how effective they are compared to the rest of the team. Also includes in memoriam, injured reserved, and restricted list trackers.

View the tracker on Streamlit

Workout Wednesday Challenges

Building technical skills through weekly Tableau visualization challenges that require creative problem-solving, dashboard design, and advanced techniques.

View my solutions on Tableau Public

IBM Data Visualization Course Final Project

Built two dashboards (sales performance and service operations tracking) using a provided auto sales dataset as the capstone for IBM’s Data Visualization course - part of my ongoing IBM Data Analyst Professional Certification. Earned 100% on the final evaluation.

View Certificate View Dashboards


All project code and documentation available on GitHub.