Hi, I am Akshay Bhat. I am a PhD Candidate in Information Science at Cornell University specifically Cornell Tech in New York City.
For last four years I have been working on building a large scale healthcare data analysis platform similar to ATLAS but on top of HCUP and State claims databases.I plan on publishing/sharing it with other researcher soon. This platform called Computational Healthcare allows researchers & physicians to query aggregated statistics from millions of pre-computed results through a search engine style interface. Additionally Computational Healthcare contains a library that makes it easy to analyze, aggregate and build predictive models. It offer out-of box compatibility with several publicly available claims databases such as Texas Inpatient Public Data file (Direct download from Texas website following click through agreement, 11 Million visits), HCUP Nationwide Readmission Database (Direct download after payment, 14 Million visits), and HCUP SID/SASD/SEDD databases (particularly 146 Million visits from NY, FL, CA). Computational Healthcare is written in modern software engineering stack and uses Python, Django, Postgres, levelDB, Celery/RabbitMQ, Protocol Buffers, Tensorflow, Docker etc. The goal is to enable an extremely seamless experience for researcher by allowing studies to be run with zero configuration (only copy files obtained/downloaded from Government in to specified folders) while maintaining privacy using privacy policies when performing aggregation. I joined this forum to learn about OHDSI efforts and integrating its data model with the one used by Computational Healthcare.
While building Computational Healthcare I had to solve similar set problems, with some of them solved in different manner. E.g. rather than using CMS synthetic data, Texas Public Data files are used. The data model is patient / visit objects (to better match HCUP) and is specified using Protocol Buffer objects. Additionally Computational Healthcare contains aggregation primitives which allow worry-free computation of aggregate statistics on "bags of patients/visits". The eventual goal of Computational Healthcare is to serve as a replacement for HCUPNet.
More information about Computational Healthcare including a demo can be found on: www.computationalhealthcare.com
I am also interested in (cyber-) security, computer vision, machine learning and have experience interning at Dropbox inc. & WebMD inc.