Healthcare Data Engineering — I make clinical quality methodology production-ready
I started in writing and editing, spent nearly a decade in mixed-methods health services research at UW-Madison starting in 2009, and moved into data engineering when I realized the tools I needed to study complex healthcare systems didn't exist yet. The thread across all of it is an interest in complex systems, not numbers for their own sake. The numbers and tools are just the best available instrument for whatever system I'm studying. Right now that means Medicare Star Ratings, HEDIS pipelines, and data integration across claims and eligibility sources. I build with Python and SQL and I care most about analytics that ship as products, not reports. I continue to engage with research questions at the edges of the field — my writing focuses on methodology problems in healthcare data that don't have clean answers yet.
Years Experience
Organizations
Blog Posts
Technical Skills
Publications
Presentations
Right now I’m at Baltimore Health Analytics building a Medicare Advantage analytics platform, where I lead the data engineering and methodology. I’m focused on the places where CMS Technical Notes and production reality diverge — which turns out to be most of them.
52 weeks of writing activity
Technical documentation and literature review supporting the ordinal logistic regression model used in the interactive Medicare Star Rating predictor on this site.
ECDS adoption is not just a data modernization effort. It introduces measurable distribution risk into the Stars ecosystem.
Concurrency sounds like “do more at once.” In real data systems it’s mostly coordination under constraint, and in healthcare analytics the constraints include trust, auditability, and deadlines.
How version control and pull requests transformed healthcare analytics from chaotic file management into transparent, collaborative systems that save lives.
Standalone, in-browser dashboard implemented as a single HTML file using Chart.js 4.4.1, PapaParse 5.4.1, jsPDF 2.5.1, and chartjs-plugin-datalabels to visualize local CSV data and generate PDF reports with no server dependency and no data leaving the user's machine.
Analytics used in case study at OCHIN, Inc. examining healthfinch's Charlie Practice Automation Platform implementation across multiple community health centers. Focus on workflow optimization and return on investment achieved through data-driven decision making.
Analyzed organization-wide care delivery changes using interrupted time series analysis on clinic panel data. Measured impact of change initiatives with segmentation and regression modeling.
Redesigned primary care patient survey for UW Health Patient Relations/Resources department. Conducted pretesting with patient interviews to ensure survey validity and usability improvements.
Conducted research studying cancer prevention screening rates in patients with co-morbid severe mental illness using logistic regression and the Elixhauser Comorbidity Index.
Full-stack MVP built with FastAPI, PostgreSQL, and Redis/Celery that uses O*NET occupation and skill data with logistic-regression calibration to surface job-transition paths — originally motivated by healthcare workforce shortages and the question of which adjacent roles clinical and administrative staff can reskill into. Generates Ready Now, Trainable, and Long-Term Reskill recommendations with gap analysis.
This model runs entirely in your browser. No data leaves your machine. It demonstrates how Medicare Advantage health plans can estimate their Star Rating from quality measure inputs.
Adjust sliders to see guidance.
Synthetic coefficients calibrated to CMS 2025 measure weights and MA-PD star distribution. Not trained on contract-level data. Does not include case-mix (CAI) adjustment. Full methodology →
Pilot experimental curriculum trial evaluating approaches to broadening medical students' exposure to the range of illness experiences, with a focus on depression education.
Led comprehensive mixed-methods study on how environmental design influences team interactions in family medicine clinics. Secured $18,000 grant, conducted 120 hours of observations, facilitated 9 focus groups with 40 participants.
Analysis of cost patterns and savings achievement across Medicare Shared Savings Program ACOs, finding that higher-baseline-cost organizations were more likely to achieve shared savings.
Methodological contribution on gathering diverse community opinions to inform health system interventions, presented at a national implementation science conference.
Qualitative study examining how primary care teams use and optimize EHR systems, identifying key barriers and facilitators to effective adoption.
Conference proceedings presenting preliminary findings on EHR optimization challenges in primary care settings.
17 presentations across conferences, workshops, and seminars
Since 2018, my methodology work has shifted to long-form technical writing — see my blog for current work on Star Ratings methodology, ordinal regression, and HEDIS implementation.
Baltimore Health Analytics
The person who makes sure the methodology is right before it ships to health plans.
Health Catalyst
Owned the data platform for a clinical quality product used by health systems nationwide.
healthfinch
Promoted to lead the analytics function and tie it directly to revenue.
healthfinch
First analytics hire — built the reporting infrastructure from zero.
University of Wisconsin-Madison, Department of Family Medicine and Community Health
Nine years across the UW School of Medicine, starting as a research specialist and advancing to lead statistician on published, federally-funded work.
Sustainable Clarity
Joanna Laucirica, PMP
Director, Customer Operations, Health Catalyst
Jessica McCay
Director of Customer Success, Health Catalyst
Jake St. Germain
Software Developer, Health Catalyst
Emily Rodriguez
Professional Services Team Lead, Lunit (formerly Health Catalyst)
University of Wisconsin-Madison
University of Wisconsin-Madison
University of Wisconsin-Madison
Most production work lives in private repos. Public repos reflect personal projects, research tools, and open learning work.
Regression work from UW Madison biostatistics/SAS coursework
MVP web application that uses O*NET occupation skill data to help users discover job transition opportunities based on current skills and experiences
Narrative, single-page briefing wall that explains how Stars / HEDIS operations are shifting by state
Interested in connecting? Reach out through any of the channels below.
name | role | domain | tools -------------+--------------------+-----------------------------------+---------------------- Zaher Karp | Lead Data Engineer | Healthcare Payer / Regulated Data | SQL, Python, dbt, AWS (1 row)