big data: impact on global health and clinical decision making
TRANSCRIPT
Impact ofBig DataonGlobalHealth and Clinical Decision Making
Professor Dr.BedirhanUstun
4V‘sofBigData
Volume• Dataquantity
Velocity• DataSpeed
Variety• DataTypes
Veracity• Messiness
Security
Smarter Healthcare Multi-channel
sales
Telecom
Manufacturing
Traffic Control Trading Analytics
SearchQuality
EverybreathyoutakeEverymoveyoumakeEverybondyoubreakEverystepyoutakeI'llbewatchingyou
Byte
Byte : one grain of rice
Kilobyte
Byte : one grain of riceKilobyte : cup of rice
Megabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of rice
Gigabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucks
Terabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container Ships
Petabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container ShipsPetabyte : Blankets Manhattan
One Byte Exabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container ShipsPetabyte : Blankets ManhattanExabyte : Blankets west coast states
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container ShipsPetabyte : Blankets ManhattanExabyt : Blankets west coast statesZettabyte : Fills the Pacific Ocean
Zettabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container ShipsPetabyte : Blankets ManhattanExabyte : Blankets west coast statesZettabyte : Fills the Pacific OceanYottabyte : An EARTH SIZE RICE BALL! Yottabyte
Byte : one grain of riceKilobyte : cup of riceMegabyte : 8 bags of riceGigabyte : 3 Semi trucksTerabyte : 2 Container ShipsPetabyte : Blankets ManhattanExabyte : Blankets west coast statesZettabyte : Fills the Pacific OceanYottabyte : A EARTH SIZE RICE BALL!
90%oftoday’sstoreddatawasgeneratedinjustthelasttwoyears.
Datageneration
MobileDevices
Readers/Scanners
Sciencefacilities
Microphones
Cameras
SocialMedia
Programs/Software
Market Size
Source:Wikibon Taming BigData
Europeprecisionmedicinemarketsize,byapplication,2013- 2023(USDMillion)
BigDataUseCases
BeyondtheHype… Hope ?
• BigDataisnot aFAD• YOUarealready usingit…• Itisheretostay• BigDatahasMinimalStructure
• BigDataIsusuallyRawData• ItisNOT likeatypicalRelationalDatabase
• BigDataisavailable - andLessExpensive• BigDataisnotcollectedforapurpose- hasnomap• Itisyourbusiness– yourtimeandmoneyisatwork
EndofPart1
Avoidingane-Tower of Babel onBigData
Millionsoftypesofdata- nolinkage
Linkingdatafromdifferentsources:xAPIs
CAPTURINGMEDICALDATA
GenealogyofICDà 1664
353years
38
ReportingofMortalityintheWorld
Information Paradox
0
100000000
200000000
300000000
400000000
500000000
600000000
700000000
800000000
1 2 3 4
YLLs
VR countries vs No VR
Burden of Mortality
1.hasBIGHOLES
WhyisthisSooooo important?
UK experience:16billionpounds and…
2. Doesnottalktoeachother
GIGO:GarbageIn
Out?
theinformationYOU-
₋ have isnot whatyouwant
₋want isnot whatyouneed
₋ need isnot whatyoucanhave
Finagle's LawofInformation
Computers areSTUPID
?Theycannotaskquestions
¿ Theymay– onlyifyouenablethem-
giveyouanswers.
PabloPicasso
KnowledgeRepresentationthe triad of things, thoughts and words(Ogden&Richards,1923)
APPLETERM
Ontology (philosophy)theOrganizationofRealityJ !!!
ü Ontology(computerscience)– theexplicit– operationaldescriptionof
theconceptualizationofadomain• Entities• Atributes• Values
• Anontologydefines:– acommonvocabulary– asharedunderstanding/exchange:
• amongpeople• amongsoftwareagents• betweenpeopleandsoftware
– toreusedata- information– tointroducestandardstoallow
interoperability
Whatis“NOntology”?
PlacingWHOClassificationsinHIS&IT
PopulationHealth• Births• Deaths• Diseases• Disability• Riskfactors
e-HealthRecordSystems
ICD
ICF
ICHI
Classifications
KRs
Terminologies
Clinical• DecisionSupport• Integrationofcare• Outcome• Safety
Administration• Scheduling• Resources• Billing
Reporting• Cost• Needs• Outcome
THECONTENTMODELAnyCategoryinICDisrepresentedby:
1.ICDConceptTitle1.1.FullySpecifiedName
2. ClassificationProperties2.1.Parents2.2Type2.3.UseandLinearization(s)
3.TextualDefinition(s)
4.Terms4.1.BaseIndexTerms4.2.InclusionTerms4.3.Exclusions
5.BodyStructureDescription5.1.BodySystem(s)5.2.BodyPart(s)[AnatomicalSite(s)]5.3.MorphologicalProperties
6.ManifestationProperties6.1.Signs&Symptoms6.2.Investigationfindings
7.CausalProperties7.1.EtiologyType7.2.CausalProperties- Agents7.3.CausalProperties- CausalMechanisms7.4.GenomicLinkages7.5.RiskFactors
8.TemporalProperties8.1.AgeofOccurrence&Occurrence Frequency8.2.DevelopmentCourse/Stage
9.SeverityofSubtypesProperties
10.FunctioningProperties10.1.ImpactonActivitiesandParticipation10.2.Contextualfactors10.3.Bodyfunctions
11.SpecificConditionProperties11.1BiologicalSex11.2.Life-CycleProperties
12. TreatmentProperties
13.DiagnosticCriteria
BigDataOrganizationZoom-inZoom-Out
Summery:Useontologytobridgedatasetsacrossdomains• Basictechnology
• Terms(classes/instances)definedinontologyareusedascommonvocabulary forsearchdata.
• IftheontologyhasmappingtoMultipleDBs,theusercansearchacrossthem.
• MotivationandIssue• CombinationsofmultipledatasetscouldbevaluableforBigDataAnalysis.• However,togetallcombinationsacrossmultipleBigDataisnotrealisticfortheirsize.
• Requestsbytheusersareverydifferentaccordingtotheirinterests.
• OntologyEngineeringforBigDatatoSolvetheissue• OntologyExplorationcontributetoobtainmeaningfulcombinations(=viewpoints)accordingto theusers’interests.
3. NeedsBig Intell igence
Knowledge
INPUTS
Big Data Science
OUTPUT
• Mechanisms
• Interventions
• Policies
• Statistics
• Aggregation
• Ontologies
• Data
• Information
ComputationalProcessing
RewritingICD using {SNOMED}exampleofDepressiveDisorderF32.0
A. Lowmood {41006004}
Lossofinterest {417523004 }
Lowenergy {248274002}
1. Appetite (decrease,increase) {64379006, 72405004}
2. Bodyweight (decrease,increase) {89362005, 8943002}
3. Sleep (decrease,increase){59050008, 77692006}
4. Psychomotor (decrease,increase){398991009, 47295007}
5. Libidoloss {8357008}
6. Lowselfesteem {286647002, 162220005}
7. Guilt,selfblame {7571003} 8. Thoughtsofdeath…
9. SuicideIdeation {102911000, 6471006}
B.
Grade 3 hypertension
Grade 2 hypertension
Grade 1 hypertension
Highnormal
normal
optimal
120 130 140 150 160 170 180
Systolicpressure
Diastolicpressure
172
102
110
105
100
95
90
85
80
KnowledgeRepresentation
62
Real Time Public HealthRule-based Aggregation @ Individual, Facility, Population levels
Public Health, Epi & Surveillance
Findings InterventionsEvents
Clinical Information
ReimbursementResource Management
BeyondSemanticInteroperabilityforHIS
• SearchusingConceptsaboveWords• HowmanypatientsdohavediabetesmellitustypeII?
• ExtractionofConcepts fromHealthRecords• AutomatedextractionofHbA1cresults ofselectedpatientswithDMtypeIIfromlabreportswithinlastyear
• StatisticalIndexonCommunity Collections• Calculationofcoveragegap fortreatmentneedfordiabetesmellitus
• ConceptNavigation acrossCollections• ComparisonofregionA withregionB etc
4. needsUSER Tools
Clinical Use Case: Exploration of Cough
Fever
386661006
COUGH
49727002
WET COUGHsputum
28743005
HemoptisiaBlood in Sputum
207069003
• X-ray : Tbc? • Culture
399208008
104184002
• Diagnosis: Tuberculosis 154283005A 15.0
• Treatment: DOTs { 324453004 }
ALGORITHMS
From www.research.vt.edu/.../images/Asymmetry.jpg
Informationasymmetry in HEALTH CARE
GARBAGE IN:GOLD OUT?
• …recognition?• …diagnosis?• …accuracyofdiagnosis?• …treatment?
— prescription?— compliance?
•…outcomes?• …patientsatisfaction?• … patientsafety?
IsHealth lessvaluablethanStockExchange?
- WhatdoyouthinkofBIGDATA?
- Ithinkitwouldbe agoodidea.