Recently, we took a look at the top 100,000 providers from the PokitDok Marketplace and viewed our database's structure according to each provider's primary and secondary specialty associations. We utilize Gephi for high level graph database visualizations that we can then use for further exploration if we see interesting or unusual sub-graphs. With our top 100,000 providers, we identified current consumer trends in healthcare ecommerce and identified useful applications for future data products. Let's take a deep dive into two of our high resolution graph images below.
Top 100K PokitDok Providers Clustered by Specialty
This graph is a visual representation of the top 100,000 most active providers from the PokitDok Marketplace, including relationships to their medical specialties and individual practice locations. The ‘most active providers’ are defined as a combination of those who access the system and/or are searched for by consumers. Using Gephi, we are able to easily import graphml files that we are able to extract from our graph database (Titan) with merely a few command line actions. We exported our graph database to a graphml file and uploaded it into Gephi to create the image above.
Yellow vertices represent specialties, like internal medicine and behavioral health, and are generally the largest and most visible due to the large number of connections (edges) they have. Other smaller vertices that may not be as visible in the images are Providers (red) and individual practice locations (blue). Edges connecting yellow specialty vertices to red provider vertices are denoted with the yellow-orange-red gradient edges. Similarly, connections between providers and individual practice locations, which are not easily visible in the large scale image, follow a gradient from red to blue. There are vertices in the graph with colors we have not accounted for, but for our purposes we will focus only on the three aforementioned vertex types: specialty (yellow), provider (red), and individual practice locations (blue). Each sub-graph cluster indicates a network of providers who share common specialties. The six largest clusters represent, from left to right, the following specialties: pharmaceuticals, alternative medicine, allied health and rehab, internal medicine, behavioral health, and dentistry.
This sprawling structure visualizes the primary and secondary relationships across our internal specialty ontology. Across these clusters, it is interesting to observe the structural differences between the largest subgraph, internal medicine, and the others. Specifically, internal medicine encompasses a broad area of medical care including, but not limited to: surgery, radiology, anesthesiology, oncology, etc. As such, the subgraph containing these specialties lacks distinguishing structure when compared to the subgraph for a a more specialized area of healthcare, like Behavioral Health. Additionally, visual data representations like this offer insight into provider specialty trends by location. These can then be used to proactively distribute medical talent based both on oversaturation of certain practice types, and undersaturation of others.
Most notably, this image yields insight into current consumer trends for healthcare ecommerce. To date, this image demonstrates that consumers are most comfortable searching for, scheduling with, and purchasing medical procedures from the doctors represented in the six main areas of specialties above: pharmaceuticals, alternative medicine, allied health and rehab, internal medicine, behavioral health, and dentistry.
Behavioral Health Cluster
This graph, made with Gephi, shows a magnified version of the behavioral health provider subgraph from the PokitDok Marketplace. The yellow vertices represent specialties under the behavioral health umbrella; the most prominent specialty vertex is for Behavioral Health. Providers are shown in red, while individual practice locations are represented in blue.
The large yellow Behavioral Health vertex is the parent category for this field and consequently, it has the most connections. In other words, all of the providers in this data set identify themselves as behavioral health providers. The next largest vertices in this image also represent specialities, which are, in descending order: counselor, social work, marriage and family therapy, and psychology - all of which are sub-specialties of Behavioral Health in our curated PokitDok specialty ontology.
The differences between the yellow specialties are interesting, especially when compared with the location in which physicians with the specialties are practicing (shown in blue). These juxtapositions offer insight into the way providers choose their secondary specialties, if they do, and in which locations they choose to practice. Data sets like this offer transparency and greater clarity not only into the number of doctors in certain specialties, but also where certain kinds of care are lacking. We can use this information to suggest a better plan of care and to allocate talent accordingly.
- Snake_Byte #35: Building Custom Slack Bots with Python - Mimic Your Coworkers - August 11, 2017
- Snake_Byte #14: Writing Made Easy - January 6, 2017
- Top 100K PokitDok Providers - November 30, 2015
- Visualizing Medicare Physician Co-Occurrence and Payment Data Geographically - Part 2 - July 14, 2015
- Visualizing Medicare Physician Co-Occurrence & Payment Data Geographically - Part 1 - July 9, 2015
- National Provider Data Revealed - A Choropleth Map Analysis - October 9, 2014
- AWS Presents: Managing a Healthcare System with Blockchain - June 28, 2017
- DocRank: Computer Science Capstone Project with the College of Charleston - June 5, 2017
- Notes from the Field: Consensus 2017 - May 31, 2017
- Snake_Byte #28: Term x Document Matrices via Sklearn's DictVectorizer - April 21, 2017
- DC Blockchain Codeathon - April 11, 2017
- Snake_Bytes #5: Vector Distances - October 12, 2016
- TinkerPop and Titan from a Python State of Mind: A Tutorial via Docker - December 15, 2015
- Online Recommendation Systems: Getting Personal with Gremthon - Part 2 - April 14, 2015
- Online Recommendation Systems: What it Really Means to be 'Somebody Like You' - Part 1 - April 7, 2015
- The top 3 things we love most about the PokitDok HealthGraph - January 27, 2015