PPADS: Physician-PArent Decision-Support for NICU Dr. Monique Frize, O.C., P. Eng., FIEEE Systems and Computer Engineering (SCE) , Carleton University Erika Bariciak, MD Children’s Hospital of Eastern Ontario (CHEO) Jeff Gilchrist, PhD Adjunct professor, SCE, Carleton University Frize Bariciak Gilchrist Medinfo 2013
19
Embed
PPADS: Physician- PArent Decision-Support for NICU
PPADS: Physician- PArent Decision-Support for NICU. Dr. Monique Frize , O.C., P. Eng., FIEEE Systems and Computer Engineering (SCE) , Carleton University Erika Bariciak , MD Children’s Hospital of Eastern Ontario (CHEO) Jeff Gilchrist, PhD Adjunct professor, SCE, Carleton University. - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Frize Bariciak Gilchrist Medinfo 2013
PPADS: Physician-PArent Decision-Support for NICU
Dr. Monique Frize, O.C., P. Eng., FIEEESystems and Computer Engineering (SCE) , Carleton University
Erika Bariciak, MDChildren’s Hospital of Eastern Ontario (CHEO)
Jeff Gilchrist, PhDAdjunct professor, SCE, Carleton University
Frize Bariciak Gilchrist Medinfo 2013
Content
• Objectives (CDR, ANN, PPADS)
• Methodology
• Results
• Conclusion and Future Work
Frize Bariciak Gilchrist Medinfo 2013
Objectives• Collect data from NICU patients in real-time
and store in Clinical Data Repository• Adapt our outcome estimation system to
processing real-time data• Develop PPADS: A decision-support module
for physicians and a module for parents.
Frize Bariciak Gilchrist Medinfo 2013
Clinical Data Repository Design
• Database uses open source MySQL and Entity-Attribute-Value format + time (EAVT).
• Confidential information separated from data for research.
• Study ID is used to associate patient ID with research data in an anonymous way
Frize Bariciak Gilchrist Medinfo 2013
Clinical Data Repository
Hospital Network
ADTSystem
HL7HL7
HL7
Clinical DataRepositoryHL7
MIRTH
MySQL Database
PrivatePatient
Data
De-IdentifiedMedical Data
Lab SystemPatient Monitors
Frize Bariciak Gilchrist Medinfo 2013
Outcome estimations
• Decision Trees and Artificial Neural Network
• Estimating Mortality
• Using real time data and summary data.
Frize Bariciak Gilchrist Medinfo 2013
Methodology
• 1. Evaluation of CDR: storage, speed, complexityand pilot survey at hospital (MeMeA 2013)
• 2. Outcome estimationsCollected 85 million data points from 634 infants in NICU 2010 and 2011Admission (12 hrs), 24 hrs, 48 hrs5x2 cross validation approach Pre-processing and ANN analysis
• 3. Developed PPADS: Physician module and Parent module
Frize Bariciak Gilchrist Medinfo 2013
Results--CDR
• CDR was better performing than the single EAV and the Multi-data type in terms of storage space, speed and complexity of query.
• Pilot test at CHEO with over 75 researchers and physicians: very positive results.
Frize Bariciak Gilchrist Medinfo 2013
Results– Outcome estimation
• Best results were obtained with data collected at 48 hours after admission.
• Mortality estimations: Specificity of 99 %; sensitivity of 63%;
PPV of 73%; and NPV of 98%... which meet our clinician’s expectations.
Frize Bariciak Gilchrist Medinfo 2013
Results-- PADS
PPADS: Identified criteria and mode of operation Examined standards, applied IPDAS Usability 8 parents and 5 neonatologists with very positive results.