Peter C. Laussen MB.BS., FANZCA, FCICMChief, Department Critical Care Medicine

David and Stacey Chair in Pediatric Critical CareThe Hospital for Sick Children, Toronto

Professor of Anesthesia, University of Toronto

The Triumph of Clinical Integration

Disclosures

• Co-developer: Tracking, Trajectory and Trigger tool (T3)

• Medical advisor: Etiometry LLC, Boston MA

• Co-Developer: www.RiskyBusiness.events

“The triumph can’t be had without the struggle”

Wilma Rudolph; 3 gold medals, Rome Olympics 1960

Humans, Technology & Data Science

Clinical Integration - Precision Critical Care: Understanding the problems and how we work

Sources of Truth

Practice variability: Humans & systems

Decision making is a Dynamic process:

Quadrants for Adapting to Uncertainty™

Know your “space”: Dynamic, changes constantly

Precision Critical Care: Understanding the problems and how we work

Sources of Truth

Practice variabilityPhysiologic variability

Education & AI

New tools for learning and transferring information…..

What has been the clinical impact?

Estimated <0.1% in routine clinical use

Business Intelligence: Improve operational & safety-related decision support

• Monitoring of key quality & performance indicators

Precise population-based care:

• Data-driven clinical care and management throughout the patient journey

Individualized patient care:

• Diagnosis, trajectory and prognosis, & decision support

Big data in healthcare

Explore Data Model Data(e.g. Machine

Learning)

Translate

Results

Use Results in

Clinical Care

Get Data

Data CommonsData commons: facilitate access and organization of various structured and

unstructured data and enable real-time algorithms and analysis

Data

Data

Data

Data

Data

All patients (ownership)

All dataPermanently

Clinical useResearch

Training / labelling

InteractiveScalablePlatform

Usable:Augment

decision making

Laboratory Information

System

T3 Production Software

Test T3 Production Software

T3 User Interface Web-

Browser

AtriumDBPhysiological

DatabaseViNES

Data FlowTest T3 User

Interface Web-Browser

Infusion / other data

feeds

5 second Data

CCCU / PICU(42 Beds)

Waveform / 1 Second Data

Ventilator / other data

5 Second Data and Labs

AtriumDBAnalytics

Engine

AtriumDBWeb User Interface

AtriumDBAPI

Via EMR in HL7

Admit Discharge Database

Via Oracle Database

T3 Production

Server

T3 Analytics

Server

ViNES Server

HSC EMRServers

T3 Risk Analytics Software

ADT via Oracle

Database

Serial Connection to Philips Monitors

Ethernet Connection

Philips Gateway

Server

5 Second Data HL7

GatewayServers

Server Role

T3 Production • Short-Term Data Storage

• Web Interface Hosting

T3 Analytics • Long-Term Data Storage

• Analytics Engine

T3 Staging • Testing / Evaluation of New

T3 Software

ViNES Server • Device bridge and data

aggregator for waveform and

1 second metric data

T3 Staging Server

Server Role

Philips

Gateway

Server

• Hosts HL7 5s device metric

data feed

HSC EMR

Servers

• Provides HL7 feed of patient

lab information

• Provides ADT database

access

Lower frequency data (5 second)

High frequency data

HPC4Health

Sick Kids

Research Institute

High Speed Private Physical Connection

Server Role

AtriumDB • Permanent storage of device

metric and waveform storage

• Analytics Engine

• Programming Interface

• Web User Interface Hosting

HPC4Heatlh • Large Scale Compute Capability

• Secure Access to Data for

External Collaborators

Volume of Water

~200,000 ft3 /sec

Volume of Physiologic Data

~200,000 bytes/sec

Understanding the PHENOTYPE and risk for INJURY……

Eytan D, Goodwin AJ, Greer R, Guerguerian AM, Mazwi M, Laussen PC.

Distributions and behavior of vital signs in critically ill children by admission

diagnosis. Pediatric Critical Care Medicine. 2018 Feb;19(2):115-124.

Eytan D, Goodwin AJ, Greer R, Guerguerian AM, Laussen PC. Heart rate

and blood pressure centile curves and distributions by age of hospitalized

critically ill children. Frontiers in Pediatrics. 2017 Mar;17(5):52.

Predicting a PHYSIOLOGIC STATE……..

Prediction of EVENTS……..

Sepsis

"Atrial Fibrillation Classification Using Step-By-Step Machine

Learning“.

Goodfellow, Sebastian; Goodwin, Andrew; Greer, Robert; Laussen,

Peter; Mazwi, Mjaye; Eytan, Danny. Biomedical Physics & Engineering Express

Signal processing…….

More subtle phenomena……

• “Hidden variables”- things not easily measured directly at bedside (variability measures, SVR, oxygenation parameters, autoregulation)

Hering-Traube Mayer waves

We need……• Multi-modal and multi-site Physiologic Databank

• Data management platform:

- 4R’s: relational, reliable, robust and retrievable (beyond the 4 V’s)

• Linkages established with categorical registries and other sources of data

• APIs written => Distributive computing ready

• Governance structure, shared, open source (democratized)

• Knowledge translation and implementation

Pediatric cardiac critical care well placed to realize this

Sana TonekaboniBSc PhD (c)

Expanded teams and expertise

Michael BrudnoPhD

Senior ScientistGenetics and Genome Biology

Anna Goldenberg PhD

Senior ScientistResearch & Vector Institutes

Anusha JegatheeswarenMD, PhD

CV surgeon

Asad HabibB.Eng, PhD

Senior project manager

Patricia TrbovichPhD

HumanEra &IBBME program

University of Toronto

www.laussenlabs.ca