HL7® FHIR® Applications Roundtable Big Data on FHIR® HL7® FHIR® Applications Roundtable Pawan Jindal MD, MS(HI) Founder & President, Darena Solutions pjindal@darenasolutions.com
Agenda Big Data Overview Big Data in Healthcare Big Data on FHIR® Implementing a Big Data Solution
What Is Big Data? Big Data is often described as extremely large data sets that can’t be managed with conventional technologies. The challenge is not just about the size of the data, but the latency involved in leveraging the data for any meaningful purpose. The technologies used to implement a “Big Data Solution” utilize distributed computing and parallel processing techniques to help reduce perceived and actual latency in a scalable and cost-effective manner.
“Perceived Latency” With Big Data Batch(bulk) Data Batch Processing Machine Learning Models App (EHR) Report and Research Queries
Distributed Computing
Column Oriented Database diagnosis identifier period reason Encounter bodySite identifier code severity stage Condition gender identifier name birthDate Patient
Column Oriented Database Storage diagnosis Identifier : 1 period reason Encounter Identifier Property Name Value 1 period {value} reason diagnosis
Why Column Oriented? Schema independent Add columns as needed Compare with relational database Add columns as needed How many? Space usage optimization We are talking “Big Data”
Column Oriented Database Technologies
Why Cassandra? Decentralized architecture – no single point of failure Strong JSON support Secondary Indexes SQL like Query Language
Who uses Cassandra? https://www.slideshare.net/semLiveEnv/cassandra-for-mission-critical-data
Don’t Reinvent the Wheel Big Data on FHIR® Don’t Reinvent the Wheel
Defining Requirements Requirements for Distributed Computing Leverage FHIR® “Resources” Support Multiple FHIR Versions
FHIR in Column Oriented Database diagnosis identifier period reason Encounter Identifier Property Name Value 1 period {value} reason diagnosis 1 resourceType Encounter fhirVersion 3.1 tin 123456 npi
FHIR DSTU 2 FHIR DSTU 2 Mapping Layer Mapping Layer FHIR STU3 FHIR STU3
ndjson EHR 1/a 1/b Provider OAuth OAuth REST API REST API 5 6 1/c EHR FHIR API 4 2 7 CMS Blue Button 1/d 3 Apps IoMT 1/e
Use Case MIPS Reporting
Patient Bundle Patient Bundle Patient Medication Encounter Procedure Observation Medication Procedure Other Resources Patient Bundle Entry
Identifying FHIR Resources US Core Implementation Guide HQMF & QRDA-I Analysis Registry Measures Definitions
https://www. mymipsscore https://www.mymipsscore.com/mips-blog/2017/6/26/can-mips-on-fhir-burn-qrda-and-ccda
One more thing..
“Perceived Latency” With Big Data Batch(bulk) Data Batch Processing Machine Learning Models App Report and Research Queries
Machine Learning Models Patient Provider Batch(bulk) Data Real Time Data Batch Layer Speedy Layer Machine Learning Models Serving Layer Streaming Analytics Real Time Alerts Queries Report and Research
BIG
Founder & President, Darena Solutions Thank You Pawan Jindal MD, MS(HI) Founder & President, Darena Solutions pjindal@darenasolutions.com