OPN Healthcare

SOFTWARE BUILDING AND DATA VISUALIZATION

SCOPE OF PROJECT

Geographic Analytics and Patient Modeling

Timeline: September 2024 - January 2025

What We Did:  Developed a modular data visualization web app, conducted geographic data analysis.

Tools: Python, Streamlit, ML Modeling, Web Scraping, Data Analytics/Visualization frameworks

“We offered two projects to UCLA Data Science Union and they executed both unique projects with professionalism and technical aptitude. The process from start to finish was flexible yet thorough which was greatly appreciated given the diversity of skills required for both projects. OPN is excited to implement what they have worked on into our own analysis and applications to help us continue to improve and deliver value-based care”

Goal of Project

  • Present patient distributions and travel distances in an interactive dashboard

  • Identify capacity bottlenecks and underserved regions for potential new facilities

  • Analyze monthly claims trends and repeat vs. single-visit patient load

  • Provide easy-to-use filtering, distance computations, and key metrics for stakeholders

Guiding Questions

  • How can we best map patients’ travel patterns to each facility?

  • Which distance computation method (Haversine, OSMnx, Google Maps) is most appropriate?

  • How can we visualize capacity vs. demand over time?

  • Can external public health data enrich our analysis of patient clusters?

Description of Project

 OPN Healthcare is a management services organization focused on cancer care, and they work with oncology providers to develop systems for value-based cancer care in the community. They had an untapped resource – geo-coded patient claims data – and asked DSU to uncover findings to better serve their partners and the community at large.

Our Findings

Phase 1: Data Ingestion & Prep

  • Ingested and cleaned patient claims CSV data

  • Set up columns for latitude/longitude, dates of service, line-of-business categories

  • Implemented filters for date ranges, line-of-business, and minimum claims

Phase 2: Distance Computation & Visualization

  • Deployed multiple distance methods (Haversine, OSMnx, Google Maps API)

  • Used Folium to create interactive maps highlighting distance thresholds

  • Built histograms, bar charts, and line charts for patient visits and monthly trends

Phase 3: Facility Load Analysis

  • Aggregated patient data by facility address and city

  • Computed repeat vs. single-visit patients

  • Identified high-load regions and potential inefficiencies in facility location

Phase 4: Recommendations & Expansion Strategy

  • Proposed adding facility capacity metrics for load balancing

  • Suggested advanced clustering (DBSCAN, k-means) to discover underserved areas

  • Documented best practices for real-time data updates and API integrations

Our Impact

  • Real-Time Insights: The Streamlit app consolidates data and visualizations in one place, enabling OPN to respond more quickly to changing patient trends

  • Cost Savings: Identifying high-travel distances can guide decisions on where new facilities might reduce travel time and improve patient satisfaction

  • Efficiency: Filtered dashboards and on-demand distance computations streamline internal analyses that once took days to complete

Reflection

While the web app enables rapid data exploration, incorporating facility capacity, staff levels, and real-time data pipelines can further enhance decision-making. Our prototype lays the groundwork for ongoing improvements in quality cancer care management.