The Hospital for Sick Children (SickKids)
Neonatal Intensive Care Unit
Health Care Services
5001 to 10000
● Machine Learning
● 400M ES documents
● 30-40 properties in each ES document mapping
● 15B data elements imported
● 40 import asynchronous jobs (run time: 5 m to 30 h)
● Jobs combined paper mapping, electronic file, and a database dump
Affiliated with the University of Toronto, The Hospital for Sick Children (SickKids) is Canada’s most research-intensive hospital and largest centre dedicated to the improvement of children’s health in the country. By staffing professionals from all disciplines within both health care and research, SickKids provides the best in complex and specialized care through creating scientific and clinical advancements, sharing knowledge and expertise, as well as championing the development of accessible, comprehensive and sustainable child health systems.
SickKids decommissioned an electronic medical chart application (CIMS) which ran in the Neonatal, Pediatric and Cardiac Intensive Care Units since 1999.
The hospital now requires the migration of billions of records from CIMS to a new system that will allow physicians and researchers to find, examine and analyze this complex data.
CIMS data was stored in multiple file sources, including flat text files, mapping text files, b-tree file structures, and oracle database data extracts with over 1000 tables.
Billions of records stored over 20 years for over 50,000 patients needed to be reindexed in a new system with novel requirements such as sub-second searching, data analytics, reporting and metrics.
Manual cross mapping between text files, b-tree records, and database extract:
• 1000’s of text files
• 50,000+ Patients
• 400,000,000+ patines row data
• 2,000,000,000+ electronic chart inputs
Only single VM was available to process and import all data. TC ES exports had to fine tune the ES to make sure the system would import data without running out of memory which was a challenge to accomplish.
Given the complexity of the challenge Translucent broke up the project in to 4 phases:
• TC Consultants carefully mapped out the whole system in 40 + excel import files that combine layouts, labels and groups from a flat file, complex cross functional lookups between text files and db lookups
• These import files were the bases for system import
• TC engineers created asynchronous importers that executed up to 30 hours
• The importer used the mappings and DB data dump from the old system to construct ES documents
• Careful testing was done by TC team to match the old system with the new system
4. Searching and Analytics
• SK team is now using the system for searching patients for data lookups
• SK Research team is using the system to chart, graph and data mind the system
• Kibana is one of tools used for dashboards
A new Electronic Medical Chart Platform based on data transformation, mapping and migration of billions of charts inputs from legacy electronic medical chart application (CIMS). New electronic charts allow physicians and researchers to find 20 years of historical patient data with sub second searching, analytics and reporting.