Ca se study: Cha lle ng e s fa c e d b y E MI F in utilising - PowerPoint PPT Presentation

Ca se study: Cha lle ng e s fa c e d b y E MI F in utilising the OMOP CDM Jo ha n va n de r L e i E ra smus Me dic a l Ce nte r Ro tte rda m

Outline • Sc a ffo lding • E MI F a nd a CDM • E MI F a nd the OMOP CDM • Ong o ing a c tivitie s/ c ha lle ng e s E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 2

I nfo rma tio n • Synta c tic : “g ra mma r” • Se ma ntic : “me a ning ” • Pra g ma tic : “c o nse q ue nc e s” E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 3

Vie ws o n info rma tio n pra g ma tic se ma ntic synta c tic Regulatory context Analytical tools Mapped data CDM Source Data E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 4

Vie ws o n info rma tio n pra g ma tic se ma ntic synta c tic E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 5

CDM… • On the synta c tic le ve l • Multiple so lutio ns po ssib le • Mo de ls a re dyna mic • De b a te o fte n: se ma ntic a nd pra g ma tic • But tha t disc ussio n is o fte n inde pe nde nt o f a spe c ific mo de l E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 6

F ro m me dic a l info rma tic s pe rspe c tive : • T he q ue stio n whic h CDM to use is pro b a b ly no t the rig ht q ue stio n…… • I s a ‘ so c io -te c hnic a l’ c o nstruc t pra g ma tic • Ca n o nly b e unde rsto o d in se ma ntic c o nte xt synta c tic • De fine s ro le s a nd re spo nsib ilitie s E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 7

Pro je c t o ve rvie w ACADEMIC PARTNERS 37 14 E uro pe a n c o untrie s c o mb ining 57 pa rtne rs €56 millio n wo rth o f re so urc e s 3 pro je c ts in o ne 5 ye a r pro je c t (2013–2017) SME PART NE RS E F PIA PART NE RS 9 10 PAT IE NT ORGANISAT I ON 1 E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 8

Why is E MI F ne e de d? Po te ntia l a pplic a tio ns o f Re a l Wo rld Da ta Launch/ Discovery Development Post-Launch  Biomarke r disc ove ry  T  Analysis of tre atme nt rial de sign and fe asibility analysis pathways  Pre dic tive mode lling  E  Colle c tion of c linic al and le c tronic he alth re c ord  Dise ase insight ge ne ration (E HR)-fac ilitate d e c onomic e vide nc e (opportunity re c ruitme nt  Ongoing e ffic ie nc y and ide ntific ation)  Prospe c tive c ohort safe ty monitoring se le c tion E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 9

E MI F Se tting • Da ta fro m ve ry dive rse so urc e s • Po pula tio n b a se d • Ho spita l b a se d • Dise a se spe c ific c o ho rts • Bio b a nks • Dive rse da ta • Bro a d spe c trum o f re se a rc h q ue stio ns • Ove ra ll purpo se : fa c ilita te re -use o f da ta E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 10

E MI F a nd CDM Cha lle ng e s • Cle a r ne e d fo r a CDM • Bro a d spe c trum o f c o ding sc he me s, la ng ua g e s, a nd se tting s • Ne e d to sto re AL L so urc e da ta inc luding so urc e vo c a b ula rie s • Po ssib ility to e sc a pe / re fine to study-spe c ific so lutio ns • Re pro duc ib le re se a rc h: Ope n, T ra nspa re nt, So urc e da ta , Ma pping s, Ana lytic a l to o ls • F le xib ility in ro le tra nsfe r (e .g ., study c o o rdina to r) • Multiple te c hnic a l infra struc ture s E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 11

E MI F a nd OMOP CDM: why? • No silve r b ulle t… • b ut no t ye t a no the r mo de l !!! • Dive rsity o f the E U se tting : Suppo rt fo r Sta nda rdize d Vo c a b ula rie s • No t limite d to spe c ific a na lytic a l use c a se • Ope n so urc e • Multiple pla tfo rms • OHDSI • Ope n c o lla b o ra tive • Gro wing in E U E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 12

E MI F Da ta b a se s b e ing ma ppe d to the OMOP-CDM E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 13

E T L re q uire s multi-disc iplina ry te a m Medical knowledge CDM Knowledge (Local) Data Depending on knowledge preferences & available skills, EMIF can take on different roles ETL Development Project Coordination Local Database / Infrastructure management EMIF E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 14

T o o ls suppo rting the pro c e ss De fine De ve lo p Ac c e pta nc e Ana lyze Ma pping De plo y E T L Ma pping s T e st E T L T e sting / QA Da ta So urc e L o g ic (E T L ) github White rabbit Rabbit in a hat Achilles usagi E va lua te Infra struc ture E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 15

E MI F Intro d uc tio n No v 2017 16

Curre nt Cha lle ng e s: E T L The following factors were found to be most impactful on overall speed and quality of the ETL: 1. Source Database research readiness: The ‘quality’ of the input data structure – and the availability of internal knowledge on how the database is defined- are the primary driver of efficiency and quality of the CDM Mapping. 2. Strong project management : superior results in terms of quality and speed can be achieved when resources are allocated and active project management is executed. 3. Vocabulary mappings: establishing the vocabulary mappings is the most resource intensive step. It’s recommended to set realistic goals with associated timings (e.g. map the top 20% of lab tests, covering 80% of all occurrences). E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 17

E va lua tio n o f tra nsla tio n: Struc tura l Ma pping Did all my source data end up in the CDM? Prevalence CDM Lost in translation? Prevalence Source • Can be very good reason for differences: business rules assessment • Iterative process to optimize the ETL • No structural CDM limitations encountered so far

E va lua tio n o f tra nsla tio n: Vo c a b ula ry Ma pping IPCI Database Example • High data coverage. • Term coverage is further improved by extending the Standard Vocabularies, e.g. RxNorm-Extension to accommodate European Drug market E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 19

Curre nt E MI F CDM Ac tivitie s • Re plic a tio n o f e xisting E MI F Use Ca se (s) o n CDM • Co ntrib ute to vo c a b ula ry e xte nsio n • Co ntrib ute to to o l de ve lo pme nt • T ra ining o f sta ke ho lde rs in using the CDM a nd OHDSI T o o ls • I nitia te a nd pa rtic ipa te in OHDSI Ne two rk studie s E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 20

E xa mple : T re a tme nt Pa thwa y Study Hripcsak G. et al. Characterizing treatment pathways at scale using the E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 21 OHDSI network. PNAS 2016 113 (27) 7329-7336;

I PCI : T ype 2 Dia b e te s EMIF will run this study on CDM databases in Europe E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 22

F ina l Re ma rks • E MI F will inte nsify its pa rtic ipa tio n in the OHDSI ne two rk b y suppo rting the E uro pe a n OHDSI initia tive (www.o hdsi-e uro pe .o rg ) c o o rdina te d b y E ra smus MC • E MI F is in the pro c e ss o f a susta ina b ility a sse ssme nt to suppo rt a nd use the da ta ne two rk po st E MI F We b e lie ve tha t the a do ptio n o f the OMOP-CDM a nd the a c tive OHDSI c o mmunity will e na b le tra nspa re nt a nd re pro duc ib le re se a rc h a t a n unpre c e de nte d sc a le in E uro pe E uro pe a n Me d ic a l Info rma tio n F ra me wo rk De c 2017 23