D3.3 - Development of an architecture blueprint, including a specification of the central and local infrastructure components

1 A new, in e disciplina y, mul i-s akeholde Eu opean ne wo k o imp o e heal hca e se ices, esea ch and ou comes o Adolescen s and Young Adul s wi h cance . STRONG-AYA – No. 101057482 – D3.3 2 Deli e able Repo WP3 – In as uc u e and in e ope abili y Deli e able D3.3 De elopmen o an a chi ec u e bluep in , including a speci ica ion o he cen al and local in as uc u e componen s Due da e o deli e able: 30/09/2023 Ac ual submission da e: 29/09/2023 P ojec : STRONG-AYA Lead Con ibu o Leona d Wee, UNIMAAS Email [email p o ec ed] O he Con ibu o s And e Dekke , UNIMAAS Gijs Geleijnse, IKNL Ba an Beusekom, IKNL Joshi Hogenboom, UNIMAAS Ananya Choudhu y, UNIMAAS Emails [email p o ec ed]; [email p o ec ed]; [email p o ec ed]; [email p o ec ed]; [email p o ec ed] Due da e 30 09 2023 Deli e y da e 29 09 2023 Deli e able ype R Dissemina ion le el SEN Desc ip ion o Wo k Ve sion Da e STRONG-AYA – No. 101057482 – D3.3 3 V2.0 was e iewed by ex e nal commi ee and some mino addi ions by way o explana ion was eques ed. This is now added and esubmi ed. V2.1 30/09/2024 V 0.2 Amended ough d a 13-03-2023; inco po a ed some commen s ecei ed, adjus ed documen lay ou . Majo e ision o simpli y o gene al (non-specialis ) eade . Ma ked ce ain sec ions o alignmen wi h o he documen s and hen e en ually emo ed om his documen V2.0 13/03/2023 V1.0 10/12/2022 Desc ip ion: A chi ec u e bluep in will guide he de elopmen o local, egional, na ional and Eu opean in as uc u e componen s de eloped i e a i ely h oughou he p ojec Summa y (max ½ page) The a chi ec u al bluep in o he STRONG AYA in as uc u e ou lines he echnical in as uc u e equi emen s o build he undamen al ounda ion o he STRONG AYA ede a ed lea ning ecosys em. I should be ead in conjunc ion wi h he da a managemen plan (D3.1), he business a chi ec u e (D2.1), he Code o Conduc (D2.2), and he ope a ional ecosys em plans a he na ional and pan-Eu opean le els (D4.4, D4.6, D4.7) STRONG-AYA – No. 101057482 – D3.3 4 1 Table o Con en s SUMMARY (MAX ½ PAGE) .......................................................................................................................................... 3 1 TABLE OF CONTENTS .......................................................................................................................................... 4 2 DEFINITIONS....................................................................................................................................................... 5 3 ABBREVIATIONS ................................................................................................................................................. 6 4 PROJECT INTRODUCTION ................................................................................................................................... 7 PROJECT BACKGROUND ........................................................................................................................................................ 7 5 DELIVERABLE INTRODUCTION ............................................................................................................................ 9 5.1.1 Disclaime ................................................................................................................................................... 10 5.1.2 Complemen a y documen a ion ................................................................................................................. 10 6 FEDERATED DATA ECOSYSTEMS – LOCAL/CENTRE LEVEL...................................................................................11 TECHNICAL REQUIREMENT 1: DATA MANAGEMENT INSIDE PARTNER INSTITUTION .......................................................................... 11 TECHNICAL REQUIREMENT 2: TRUSTED COMPUTING ENVIRONMENT FOR RESEARCH ....................................................................... 13 TECHNICAL REQUIREMENT 3: MOVING EXTRACTED DATA INTO RESEARCH ENVIRONMENT ................................................................ 14 TECHNICAL REQUIREMENT 4: UPDATING OF DATA OVER TIME .................................................................................................... 14 TECHNICAL REQUIREMENT 5: PATIENT REPORTED OUTCOMES DATA COLLECTION ........................................................................... 14 7 FAIR DATA PRINCIPLES ......................................................................................................................................15 TECHNICAL REQUIREMENT 6: EXECUTING DOCKER CONTAINERIZED SOFTWARE .............................................................................. 16 TECHNICAL REQUIREMENT 7: DEFINING METADATA FOR MAPPING TO THE COS ............................................................................ 16 TECHNICAL REQUIREMENT 8: ADDING HARMONIZED COS SCHEMA AS METADATA ......................................................................... 17 8 FEDERATED ANALYTICS AND FEDERATED LEARNING .........................................................................................17 TECHNICAL REQUIREMENT 9: DATA VISUALIZATION DASHBOARDS ............................................................................................... 18 TECHNICAL REQUIREMENT 10: FEDERATED LEARNING OF STATISTICAL MODELS.............................................................................. 18 TECHNICAL REQUIREMENT 11: PROTECTIONS AGAINST RUNNING UNAUTHORIZED CONTAINERS ........................................................ 18 9 SECURE WEB INFRASTRUCTURE – THE PAN-EUROPEAN ECOSYSTEM ................................................................19 TECHNICAL REQUIREMENT 12: SOFTWARE INFRASTRUCTURE VANTAGE6 ..................................................................................... 19 TECHNICAL REQUIREMENT 13: USE OF THIRD PARTY SECURE AGGREGATION TO ENHANCE SAFETY ..................................................... 19 TECHNICAL REQUIREMENT 14: SERVICES SUPPORTED BY SUBCONTRACTOR ................................................................................... 19 APPENDIX: FAIR DATA ...............................................................................................................................................21 STRONG-AYA – No. 101057482 – D3.3 5 2 De ini ions STRONG AYA conso ium membe s a e e e ed o as ollowing wi hin his ex : 1. NKI-AVL – S ich ing he Nede lands Kanke Ins i uu – An oni an Leeuwenhoek Ziekenhuis (NL) 2. YCE – You h Cance Eu ope (RO) 3. INT – Fondazione IRCCS Ins i u o Nazionale dei Tumo i (IT) 4. FFUND – FFUND BV (NL) 5. CLB – Cen e de Lu e Con e le Cance Leon Be a d (FR) 6. ECO – Eu opean Cance O ganisa ion (BE) 7. UNIMAAS – Uni e si ei Maas ich (NL) 8. IKNL – S ich ing In eg aal Kanke cen um Nede land (NL) 9. EORTC – Eu opean O ganisa ion o Resea ch and T ea men o Cance AISBL (BE) 10. IGR – Ins i u Gus a e Roussy (FR) 11. MSCNRIO – Na odowy Ins y u Onkologii im. Ma ii Sklodowskiej-Cu ie – Pans wowy Ins y u Badawczy (Ma ie Sklodowska-Cu ie Na ional Resea ch Ins i u e o Oncology) (PL) 12. UOM – Uni e si y o Manches e (UK) 13. UOL – Uni e si y o Leeds (UK) 14. LTHT – Leeds Teaching Hospi als Na ional Heal h Se ice T us (UL) 15. SOUTHAMPTON – Uni e si y o Sou hamp on (UK)  G an Ag eemen (including i s annexes and amendmen s): he ag eemen signed be ween he bene icia ies o he HORIZON Resea ch and Inno a ions Ac ions (he ea e e e ed o as Ho izon) and he Eu opean Heal h and Digi al Execu i e Agency (he ea e e e ed o as HADEA) o he unde aking o he STRONG AYA p ojec (G an Ag eemen no. 101057482).  Bene icia y: Signa o ies o he G an Ag eemen  Associa ed Pa ne : En i ies which pa icipa e in he ac ion bu wi hou he igh o cha ge cos s o claim con ibu ions.  P ojec : he sum o all ac i i ies ca ied ou in he amewo k o he G an Ag eemen .  Conso ium: he STRONG AYA conso ium, including all he a o emen ioned pa ne s.  Conso ium Ag eemen : The ag eemen made be ween STRONG AYA membe s o he implemen a ion and execu ion o he ac ion ou lined in he G an Ag eemen . The ag eemen shall no a ec he pa ies’ obliga ions o HADEA on behal o he Eu opean Union, and/o o one ano he a ising om he G an Ag eemen . STRONG-AYA – No. 101057482 – D3.3 6 3 Abb e ia ions Ac onym/Abb e ia ion Meaning HCP Heal h Ca e P o ide PRO Pa ien Repo ed Ou come PROM Pa ien Repo ed Ou come Measu e COS Co e Ou come Se WP Wo k Package WPL Wo k Package Lead(s) WP1 Wo k Package 1 (De elopmen Co e Ou come Se AYA wi h cance & da a collec ion) WP2 Wo k Package 2 (Go e nance, Da a Secu i y and E hics) WP3 Wo k Package 3 (In as uc u e and In e ope abili y) WP4 Wo k Package 4 (Ope a ion o STRONG AYA ecosys ems, s akeholde and pa ien in ol emen , dissemina ion, exploi a ion, communica ion) WP5 Wo k Package 5 (Scien i ic coo dina ion and p ojec managemen ) KPI Key Pe o mance Indica o OA Open Access PAB Pa ien Ad iso y Boa d EC Eu opean Commission HADEA Eu opean Heal h and Digi al Execu i e Agency SC S ee ing Commi ee MT Managemen Team DMP Da a Managemen Plan STRONG-AYA – No. 101057482 – D3.3 7 4 P ojec In oduc ion P ojec backg ound Cance a adolescen and young adul (AYA) age is a e, al hough 4-6 imes mo e equen han paedia ic cance (i.e. p epubescen pe iod). Howe e , his a i y does no e lec he signi ican pe sonal and socie al cos s o cance in his popula ion, as e lec ed in he po en ial yea s o li e los o sa ed, he dec eased p oduc i i y and quali y-o -li e due o he impac o he disease du ing o ma i e yea s, and he long- e m complica ions o disabili ies 1 . AYAs wi h cance o m a unique g oup; hey ace age-speci ic issues (e.g. In e ili y, unemploymen , inancial p oblems) and dec eased quali y o li e due o cance and i s ea men . Unlike dedica ed heal hca e and ials o paedia ic cance pa ien s, AYA-speci ic heal hca e se ices a e sca ce and a y ac oss Eu ope. AYAs who a e a he co e o socie y and economy need access o age- app op ia e and high-quali y heal hca e. De ining AYAs wi h cance as 15 o 39 yea s a ini ial cance diagnosis 2 , hei annual cance incidence is 42.2/100.000, wi h 156.431 cases in Eu ope and 1.231.007 cases wo ldwide epo ed in 2018 ( oge he 6.8% o all cance s) 3 . Popula ion-based da a om 27 Eu opean coun ies suppo s ha AYAs ha e lowe su i al han child en bu highe han adul s a ec ed by cance . Ad ances in cance ea men ha e led o inc eased su i al a es o AYAs wi h cance , imp o ing by 82% o all cance s be ween 1990 and 2007 4 . Howe e , su i al imp o emen in AYA is mo e challenging han o child en and olde cance su i o s, which migh be due o he ac ha AYA ha e he highes absolu e excess isk o second p ima y malignan neoplasms 5 . AYA ace some dis inc challenges gi en ha hey do no belong o nei he paedia ic no adul oncology g oups. Cha ac e is ic ea u es o his popula ion g oup include unique spec um o cance ypes, di e en umou biology, unique complex psychological needs, dis inc la e sequelae, including impai ed e ili y, and pallia i e ca e. These ai s imply ha clinical managemen , ea men , diagnosis, psychological suppo will need o be designed and de eloped o AYA’s speci ic needs. Fo example, AYAs diagnosed wi h b eas and p os a e ca cinomas ha e wo se su i al han olde pa ien s because o he biological di e ences be ween hem, highligh ing he need o a ge sc eening me hodologies, ea men and policies o hei needs 6 . AYAs wi h cance also ace signi ican psychological challenges, including subs ance abuse, men al heal h issues, suicidal idea ions and inc eased emo ional bu den om cance and cance - ela ed mo bidi y. Finally, ailo ing cance ca e o AYA’s needs is di icul and due o many di e en complex ac o s including he low a e o pa icipa ion by AYAs in clinical ials and cance esea ch 7 . Despi e he inc easing awa eness and a g owing body o he scien i ic li e a u e, hese unique issues emain o be ully ecognised and add essed by he Eu opean heal h sys ems and AYAs wi h cance a e equen ly unde se ed. In pa , his may ha e esul ed om he adi ional dicho omy be ween he in eg a ed paedia ic (“pa ien / amily-cen ed”) ca e se ices e sus dispe sed (“disease- cen e ed”) adul oncology 1 S oneham SJ. AYA su i o ship: The nex challenge. Cance 2020; 126: 2116-2119. 2 Adolescen and Young Adul Oncology Re iew G oup. Closing he gap: Resea ch and Ca e Impe a i es o Adolescen s and Young Adul s wi h Cance . Na ional Ins i u e o Heal h, Na ional Cance Ins i u e, and Li es ong Young Aul Alliance: Be hesda, MD, USA, 2006. 3 T ama A, Bo a L, S elia o a-Fouche E. Cance Bu den in Adolescen s and Young Adul s: A Re iew o Epidemiological E idence. Cance J 2018; 24: 256-266 4 T ama A, Bo a L, Foschi R e al. Su i al o Eu opean adolescen s and young adul s diagnosed wi h cance in 2000-07: popula ion-based da a om EUROCARE- 5. Lance Oncol 2016; 17: 896-906. 5 Keegan THM, Bleye A, Rosenbe g AS e al. Second P ima y Malignan Neoplasms and Su i al in Adolescen and Young Adul Cance Su i o s. JAMA Oncol 2017; 3: 1554-1557. 6 S a k D, Bielack S, B ugie es L e al. Teenage s and young adul s wi h cance in Eu ope: om na ional p og ammes o a Eu opean in eg a ed coo dina ed p ojec . Eu opean Jou nal o Cance Ca e, 25(3), 419–427. 7 Hayashi RJ. Adolescen and young adul cance su i o ship: The new on ie o in es iga ion. Cance 2019; 125: 1976-1978. STRONG-AYA – No. 101057482 – D3.3 8 se ices 8 , 9 , 10 . Up o hal o AYAs wi h cance epo unme in o ma ional and se ice needs, impac ing hei di ec (su i al a es) and indi ec (long e m e ec s and men al heal h) eco e y o pa icipa e in socie y 11 . Fu he mo e, aligned wi h his ba ie is he low a e o heal h ca e u ilisa ion by AYAs, especially p ima y ca e, gi en hei challenges o sus ain heal h insu ance co e age 12 . Acco ding o a su ey conduc ed h ough he ne wo ks o he AYA Wo king G oup o he Eu opean Socie y o Medical Oncology (ESMO) and he Eu opean Socie y o Paedia ic Oncology (SIOP Eu ope), 67% o p ac i ione s do no ha e access o specialised cen es o AYA wi h cance , 67% had no access o a specialis cance se ice o la e e ec s managemen and 38% had no access o e ili y specialis s. Unde -p o ision and inequali y o AYA cance ca e is common ac oss Eu ope and especially in he Eas e n and Sou he n-Eas . Fu he mo e, he Wo king G oup also epo ed an absence o ou come measu es o moni o ing and e alua ing AYA cance ca e p og ams and con ol 13 . Among he ecommended u u e s eps, i has been iden i ied ha one o he mos impo an con ibu ions o AYA esea ch would be o pool da a (e.g. pa ien - epo ed ou comes, clinical and ea men da a) ac oss ins i u ions and coun ies and c ea e la ge coho s o esea che s o Add ess he bu den o cance in AYA 14 . The e is a lack o da a s anda diza ion, da a in e ope abili y and (p ospec i e) collec ion o ou comes o ele ance o AYAs wi h cance . The STRONG-AYA p ojec aims o ackle he unde ep esen a ion o AYA’s expe iences and ou comes when na iga ing he heal hca e sys em and in clinical ca e by de eloping na ional in as uc u es o ou come da a managemen and clinical decision-making wi hin a pan-Eu opean ecosys em and es ablishing communica ion eedback o AYAs wi h cance and he heal hca e sys ems. This will be key o imp o ing heal hca e se ices, esea ch, ou comes and policies o AYAs and o ul ima ely be e cance ca e o his pa ien g oup. To his aim, STRONG-AYA b ings oge he an in e na ional mul i-disciplina y conso ium ac oss se en Eu opean coun ies, led by he Ne he lands Cance Ins i u e (NKI) and composed o academic esea ch o ganisa ions (Eu opean O ganisa ion o Resea ch and T ea men o Cance (EORTC), Uni e si y o Sou hamp on, Uni e si y o Leeds, Uni e si y o Manches e , Maas ich Uni e si y, Ne he lands Comp ehensi e Cance O ganisa ion (IKNL)), clinical pa ne s (I alian Na ional Tumou Ins i u e, Léon Bé a d Cen e, Gus a e Roussy Ins i u e, Ma ia Sklodowska-Cu ie Na ional Resea ch Ins i u e o Oncology, he Leeds Teaching Hospi als Na ional Heal h Se ice T us ), s akeholde and pa ien o ganisa ions (You h Cance Eu ope, Eu opean Cance O ganisa ion) and a consul ing company (FFUND). Building on p e ious ini ia i es, a STRONG-AYA da a ecosys em will be se up o alue-based ca e, esea ch and policy o AYA wi h cance by: 1. De eloping a Co e Ou come Se (COS) speci ically o AYAs wi h cance , ia a pa icipa i e consensus p ocess de ining mos impo an aspec s o hose di ec ly a ec ed by AYA cance , including pa ien s and heal hca e p o essionals. 8 Fe a i A, S a k D, Pecca o i FA e al. Adolescen s and young adul s (AYA) wi h cance : a posi ion pape om he AYA Wo king G oup o he Eu opean Socie y o Medical Oncology (ESMO) and he Eu opean Socie y o Paedia ic Oncology (SIOPE). ESMO Open 2021; 6: 100096. 9 Osbo n M, Johnson R, Thompson K e al. Models o ca e o adolescen and young adul cance p og ams. Pedia Blood Cance 2019; 66: e27991. 10 Fa dell JE, Pa e son P, Wake ield CE e al. A Na a i e Re iew o Models o Ca e o Adolescen s and Young Adul s wi h Cance : Ba ie s and Recommenda ions. J Adolesc Young Adul Oncol 2018; 7: 148-152. 11 Keegan TH, Lich ensz ajn DY, Ka o I e al. Unme adolescen and young adul cance su i o s in o ma ion and se ice needs: a popula ion-based cance egis y s udy. J Cance Su i 2012; 6: 239-250. 12 Hayashi RJ. Adolescen and young adul cance su i o ship: The new on ie o in es iga ion. Cance 2019; 125: 1976-1978. 13 Salous os E, S a k D, Michailidou K e al. Repo on ESMO/SIOPE Eu opean Landscape p ojec key esul s: Mapping he s a us and needs in AYA cance ca e. La e-b eaking and de e ed publica ion abs ac s public heal h 2017; 28, 5: V643. 14 Smi h AW, Seibel NL, Lewis DR e al. Nex s eps o adolescen and young adul oncology wo kshop: An upda e on p og ess and ecommenda ions o he u u e. Cance 2016; 122: 988-999. STRONG-AYA – No. 101057482 – D3.3 9 2. Implemen ing he COS ac oss se e al na ional Eu opean heal hca e sys ems. Da a will be collec ed a local le el and will hen be included in a da a in eg a ion pla o m. An o e all ecosys em amewo k o da a analy ics and ou pu will be c ea ed suppo ing ede a ed analyses and he c ea ion o epo s ac oss clinical and pa ien - epo ed da a and making na ional eposi o ies o (pa ien - epo ed) heal h da a, a ailable o indi idual pa ien s, pa ien o ganisa ions, egula o y au ho i ies, as well as he pa ien s’ heal h ca e p o ide s o in o m clinical decision-making. The i e esul ing na ional ecosys ems will be connec ed o each o he in o he pan-Eu opean ecosys em using a ede a ed app oach also u ilizing he o e all ecosys em amewo k. 3. Dissemina ing he COS o a wide ange o local as well as pan-Eu opean s akeholde s, in pa icula by de eloping analy ical ools o p ocess and p esen pa ien ou come da a and es ablish eedback loops ha in o m pa ien s and clinicians. Implemen ing hese p ojec objec i es a e i e wo k packages wi hin he STRONG-AYA p ojec :  WP1: De elopmen Co e Ou come Se AYA wi h cance and Da a Collec ion (Lead: SOUTHAMPTON)  WP2: Go e nance, Da a Secu i y and E hics (Lead: EORTC)  WP3: In as uc u e and In e ope abili y (Lead: UNIMAAS)  WP4: Ope a ion o STRONG-AYA ecosys ems, S akeholde and Pa ien in ol emen , Dissemina ion, Exploi a ion, Communica ion (Lead: E.C.O.)  WP5: Scien i ic Coo dina ion and P ojec Managemen (Lead: NKI) STRONG-AYA will enable AYA ca e and esea ch o bene i om collec ion and pooling o pa ien -cen e ed da a and collabo a ion among all s akeholde s: pa ien s, heal hca e p o essionals, scien is s, and policymake s. Mo e widely, he p ojec will le e age he ne wo k o in e es ed o ganisa ions and ne wo ks es ablished unde STRONG-AYA o long- e m s eng hened p omo ion o he necessa y implemen a ion o specialis AYA cance se ices ac oss Eu ope. This will ul ima ely b ing no el insigh s in o AYA cance ca e, esea ch and policy, con ibu ing o he long- e m imp o emen o ou comes o people wi h AYA cance . 5 Deli e able in oduc ion The STRONG-AYA conso ium belie es ha inno a i e and pa adigm-b eaking ways o accessing da a and sha ing knowledge ac oss ins i u ional siloes is u gen ly needed o imp o e heal hca e o his unique g oup o people. The key pilla s o he p ojec a e : (1) consensus (Delphi) de elopmen o a Co e Ou comes Se i.e. COS; (2) in as uc u e o collec ing, managing and enhancing AYA da a among pa icipa ing coun ies; (3) dissemina e ou comes and da a analysis ools a he pan-Eu opean le el o imp o e pa e ns o ca e o AYAs. A ede a ed design o da a managemen and da a analysis was chosen as a p omising app oach o add ess agmen a ion o da a, o add ess adop ion ba ie s associa ed wi h cen alized pa ien da a eposi o ies, and o be mo e easily able o suppo scale-up wi h nume ous u u e pa ne s. Da a managemen ha complies wi h FAIR (Findable-Accessible-In e ope able-Reusable) da a p inciples is a co ne s one o ou echnical in as uc u e. The FAIR p inciples will acco dingly be applied a pa ne , na ional and pan-Eu opean le els o da a, and subsequen ly digi al a e ac s gene a ed (e.g., so wa e, publica ions, epo s) om he p ojec wo k. STRONG-AYA – No. 101057482 – D3.3 16 Seman ic in e ope abili y means ha he p ecise meaning o he da a is p ese ed and unde s ood h oughou all ansac ions be ween coope a ing pa ne s. In STRONG-AYA, we will be suppo ing he open wo ld pa adigm ha allows us o wo k wi h s uc u ed da a ex ac ion om each pa ne in any o ma and any e minology. We will add a FAIR-based implemen a ion o seman ic in e ope abili y on op o he AYA da a, in o de o make ou da ase s mu ually unde s andable o each o he . An illus a ion o how we place a seman ically in e ope able anno a ions “laye ” on op o he da a can be iewed he e (app oxima ely 20 minu es ideo play ime). STRONG-AYA pa ne s a e ecommended o iew he online ideo, o ge an idea o wha he gene al idea looks like. Technical Requi emen 6: Execu ing Docke con aine ized so wa e STRONG-AYA pa ne s should use a Docke applica ion om UNIMAAS con aining a basic se o ools ha assis s wi h making s uc u ed da a mo e FAIR. This Docke applica ion will be s a ed in he us ed STRONG- AYA i ual machine whe e he ex ac ed COS da a is al eady being s o ed. This applica ion includes a G aphical Use In e ace (GUI). A) Via he GUI, he local PI o he da a can selec which da abase(s) o p ocess in o a non- abula o ma known as Resou ce Desc ip o F amewo k (RDF). This RDF con ains he pa ien -le el con en s o he PI’s da abase. B) The GUI hem p omp s he local PI o ma ch which ields in he own da a (p esumably in he PI’s na i e language and in hei own coding sys em) co esponds o which expec ed COS i ems. C) The e is ex a space o addi ional desc ip ions in ee ex (in English please) whe e he local PI can desc ibe hei ields mo e ully, o de ine addi ional da a ields ha hey hink is impo an bu migh no be immedia ely ma ching wi h he cu en e sion o COS. D) The local PIs da abase schema layou and i s codebook (bu no he da a con en s) will be sa ed as a W3C Web On ology Language (OWL) ile. The ma ched labels and he PI’s own addi ional ex no es will also be sa ed in he OWL ile. E) The RDF ile (da a con en s) and he OWL (local schema and codebook) will be a chi ed in hei own us ed i ual machine as a “g aph da abase”. F) The local PI, wi h he assis ance o hei ins i u ional eam, is ee o inspec he OWL iles o con i m ha i con ains schema, codebook and ma ched COS i ems, bu i does no con ain any indi idual pa ien da a. Technical Requi emen 7: De ining me ada a o mapping o he COS Only an OWL ile om he p ocedu e in Tech Requi emen 6 will be p i a ely and secu ely sha ed (e.g. by secu ed ile ans e ools) by he pa ne PI o he echnology pa ne UNIMAAS. No e ha he OWL ile shall con ain p o ec ed in o ma ion abou he da abase o ganiza ion and ad hoc schema, bu i does no con ain any indi idually-iden i iable pa ien in o ma ion. UNIMAAS da a scien is s will wo k closely wi h each pa ne PI and wi h he leade o Wo k Package 1, o c ea e he cus omized anno a ions o ha ins i u ion, so ha hei da a is FAIR and sui ed o he needs o he STRONG-AYA p ojec . STRONG-AYA – No. 101057482 – D3.3 17 Technical Requi emen 8: Adding ha monized COS schema as me ada a Technical pa ne UNIMAAS will p o ide a cus omized anno a ion ile o each pa ne , so ha he local da abase schema will be au oma ically mapped wi h he o e -a ching COS and hus becomes in e ope able wi h he es o he conso ium da a. Each local PI shall keep his cus om anno a ion ile in hei own ins i u ion STRONG AYA de ice. 8 Fede a ed analy ics and ede a ed lea ning Ma hema ical me hods adjus ed o mul iple geog aphically-dispe sed da ase s will used o analyze da a and c ea e s a is ical models. We ha e es ed he alidi y o hese me hods, and can show ha he esul s o compu a ion a e he same as i we had analyzed all ha da a in one single loca ion. Fede a ed analy ics is conce ned wi h gene a ing coho summa ies and desc ip i e s a is ics on-demand om dispe sed da ase s (on-demand he e means he abili y o in e ac i ely and dynamically gene a e hem in nea - eal ime, such ha he summa y s a is ics do no need o be p e-calcula ed be o ehand o be a chi ed alongside he unde lying da a). Fede a ed lea ning is simila ly conce ned wi h i ing s a is ical models on demand om dispe sed da a (e.g., eg ession models, decision ees and machine lea ning models – including deep lea ning neu al ne wo ks). The STRONG-AYA p ojec suppo s alida ed and ce i ied ools o use s o pe o m hei own ede a ed analy ics and ede a ed lea ning. Fede a ed analy ics is somewha analogous o pe o ming a pooled s a is ical me a-analysis o indi idual esea ch s udies. Each s udy is sel -con ained and independen , bu i communica es ( h ough i s published manusc ip ) i s own agg ega ed s a is ical esul s o he e iewe , bu he e is no indi idual pa ien da a exchanged. The e iewe has o collec he agg ega ed esul s o each s udy, and hen come up wi h he pooled global s a is ics as hei me a-analysis. Fede a ed lea ning is somewha analogous o how a poli ical union unc ions. Fo example, he Eu opean Union is a conso ium consis ing o au onomous na ions, bu i is able o implemen policies on impo an global ma e s such as ade policy, h ough a ce ain p ocess o i e a ion un il e e yone con e ges on a consensus policy. In STRONG-AYA, a cen al agg ega o de ice se es an analogous unc ion as he au ho o he pooled me a-analysis o as he Eu opean Pa liamen . I has he ole o communica ing wi h nodes and inding a consensus among nodes (mo e abou he cen al agg ega o de ice in he sec ion on Secu e Web In as uc u e). No e : These anno a ions si on op o he ex ac ed COS da a as pe sis en “me ada a”. This me ada a will be needed when doing ede a ed analysis and ede a ed lea ning. Ou design allows mul iple anno a ions o simul aneously co-exis on op o he same da a ( o example, due o di e en u iliza ion cases o changes in e minology o e ime). The me ada a laye s keep he seman ic in e ope abili y o da a ac oss he STRONG-AYA conso ium, wi hou ha ing o edi any o he unde lying COS da a gene a ed by he ins i u ion. STRONG-AYA – No. 101057482 – D3.3 18 Technical Requi emen 9: Da a isualiza ion dashboa ds Fede a ed da a analy ics will be suppo ed in he o m o in e ac i e da a explo a ion “dashboa ds” which displays summa y s a is ical da a, no any indi idual da a. The desi ed layou and summa ies displayed will be cus omized o he use and/o some p e-de ined dashboa d layou s will be a ailable o he use o choose om. A e he use logs in and au hen ica es, he dashboa d will be accessible o ha use (assuming hey ha e he co ec le els o pe mission) as a web in e ace on hei own compu e de ices. Impo an : The in as uc u e (see sec ion on Secu e Web In as uc u e) shall ha e buil -in p o ec ion agains e ie ing s a is ical summa ies om any pa ne ins i u ion i hei local summa y s a is ic con ains less han 10 subjec s. Technical Requi emen 10: Fede a ed lea ning o s a is ical models Fede a ed lea ning o s a is ical models will be suppo ed, so ha app o ed use s can i widely known s a is ical models, o example logis ic eg ession, wi hou displaying indi idual pa ien s da a. The desi ed models and he co a ia es o be i ed will be cus omized o he use and/o some p e-de ined modelling modules will be p o ided o he use o choose om. A e he use logs in and au hen ica es, he a ailable models will be accessible o ha use (assuming hey ha e he co ec le els o pe mission) as a web in e ace on hei own compu e de ices. Impo an : The in as uc u e (see sec ion on Secu e Web In as uc u e) shall ha e buil -in p o ec ion agains making s a is ical models om any pa ne ins i u ion i hei local model con ains less han 10 subjec s. Technical Requi emen 11: P o ec ions agains unning unau ho ized con aine s The s a is ical summa ies and s a is ical models will be packaged as sel -con ained “mic o”-so wa e applica ions called Docke con aine s. These docke con aine s execu e exclusi ely inside he pa ne ’s us ed STRONG-AYA compu ing de ice. A) All p og amming language code o he ede a ed analysis and he ede a ed lea ning Docke applica ions shall be made open sou ce and open access on a publicly eadable so wa e eposi o y (i.e. Gi Hub). This is o make any s a is ical me hods used on COS da a o be ully anspa en and ully audi able by anyone. B) Only Docke applica ions ha a e digi ally signed, independen ly es ed and app o ed as ee om malicious con en s a e allowed o be dis ibu ed h ough he STRONG-AYA ne wo k and execu ed on he pa ne s’ de ices. C) Technology pa ne s IKNL and UNIMAAS will p o ide S anda d Ope a ing P ocedu es o de eloping, ce i ying, es ing and app o ing Docke applica ions o he STRONG-AYA conso ium o app o al and en o cemen . No e : The pu pose o en o cing a minimum h eshold o 10 is o make i g ea ly mo e di icul o e-iden i y any pa icula human subjec om he summa y s a is ics o he modelling esul s alone. I any pa ne ’s da ase con ains less han 10 subjec s o any ede a ed s a is ic o ede a ed model, his pa ne ’s en i e da a shall be epo ed as i “missing in i s en i e y” o he pu pose o ha summa y s a is ics o ha pa icula model. This h eshold can be aised o lowe ed, bu only a he in as uc u e-wide le el and only wi h he ag eemen o he whole conso ium. STRONG-AYA – No. 101057482 – D3.3 19 D) App o ed Docke applica ions o ede a ed analysis/lea ning wi hin he STRONG-AYA conso ium shall be placed in a p i a e p o ec ed Docke applica ions eposi o y (a.k.a. “STRONG- AYA Algo i hm S o e”). 9 Secu e web in as uc u e – The pan-Eu opean ecosys em To ca y ou he asks equi ed in he STRONG-AYA p ojec along he equi emen s se ou in his documen , he conso ium needs a so wa e echnology ha is us ed, well-main ained and p o en o ha e been success ully used o da e among clinical ins i u ions o he pu pose o clinically- ela ed o epidemiologically- ela ed esea ch. Technical Requi emen 12: So wa e in as uc u e Van age6 STRONG-AYA shall use he open-sou ce Van age6 (p iVAcy p ese iNg ede aTed leA ninG in as uc u E o Secu e Insigh eXchange). Technology pa ne IKNL o e sees he so wa e managemen o he STRONG-AYA e sion(s) o he Van age6 code. S anda d Ope a ing P ocedu es shall be p o ided o he STRONG-AYA conso ium o ins alling and upda ing he equi ed so wa e. Technology pa ne IKNL publishes and upda es a So wa e Managemen Plan a egula in e als, as s a ed by he Ne he lands eScience Cen e guidelines on sus ainable so wa e, on behal o he STRONG-AYA conso ium. This So wa e Managemen Plan shall be open access and accessible ia a publicly eadable web page managed by he conso ium. Technical Requi emen 13: Use o hi d pa y secu e agg ega ion o enhance sa e y STRONG-AYA shall implemen a “ us ed hi d-pa y secu e agg ega o ” me hodology o ede a ed analy ics and ede a ed lea ning based on Van age6 so wa e in as uc u e. Technology pa ne s IKNL and UNIMAAS shall ake so wa e design s eps o ensu e ha : a) The e shall be no di ec pa ne - o-pa ne web ne wo k connec ions pe mi ed by he Van age6 in as uc u e. b) Only signed, es ed, ce i ied and app o ed Docke applica ions aken om one designa ed “STRONG-AYA Algo i hm S o e” will be pe mi ed o be sen ou and execu ed by he de ices o he pa ne s. c) The Van age6 so wa e shall ha e ac i e in-buil p o ec ions ha p e en ansmission o any summa y s a is ic o modelling esul om an ins i u ion de ice ha has less han 10 subjec s (subjec o change by uling o he conso ium) o a pa icula summa y s a is ic o pa icula model i ing. d) E en hough no indi idual pa ien da a will be ansmi ed, none heless pa ne s IKNL and UNIMAAS shall ha e indus y-g ade web message enc yp ion u ned on du ing ede a ed analysis and ede a ed lea ning p ocesses unning on he ne wo k. Technical Requi emen 14: Se ices suppo ed by subcon ac o As s a ed in he o iginal p ojec plan submi ed o he unding body, echnology pa ne Medical Da a Wo ks has been sub-con ac ed by UNIMAAS o hos a “ us ed hi d-pa y secu e agg ega o ” machine on hei p o ec ed cloud se ices o use by he STRONG-AYA conso ium. Medical Da a Wo ks will o e see he egis a ion o ins i u ions, use s and esea che s o access he his secu e web in as uc u e. STRONG-AYA – No. 101057482 – D3.3 20 Medical Da a Wo ks u he ensu es ha ne wo k connec i i y, use au hen ica ion, web applica ions and o he p ocesses ha pass ia he secu e cen al agg ega o con o ms o he Secu ed Hype - ex T ans e P o ocol (HTTPS) web indus y s anda ds. Figu e 2: Diag am explaining he gene al s eps in ol ed in ede a ed lea ning ( ede a ed analy ics wo ks in a e y simila way, bu i migh no equi e he s eps 2 and 3 o be epea ed mul iple imes). Image sou ce : The Open Da a Ins i u e “Fede a ed Lea ning” (C ea i e Commons A ibu ion Sha e-Alike 4.0 In e na ional License) To pa icipa e in a ede a ed ask, de ices con olled by each STRONG-AYA pa ne ins i u ion de ice needs o be ac i ely connec ed o a cen al coo dina ion de ice ha ac s as a s a is ical agg ega o . The agg ega o ells he connec ed de ices which compu a ion o execu e, e.g. calcula e a summa y s a is ic o i a eg ession model. Compu a ions ha a e e i ied and app o ed by he STRONG-AYA conso ium will be packaged as digi ally-ce i ied Docke applica ions. The nodes execu e he eques ed compu a ion as a Docke applica ion, only on he local da ase inside i s own i ual machine. The agg ega ed esul o he compu a ion (no he da a) is communica ed o he cen al agg ega o . The cen al agg ega o collec s he local esul s om each node, and by i sel compu es he global esul . When needed, e.g. i ing a eg ession model, he abo e p ocess has o i e a e and upda e many imes, un il a single con e gen global model eme ges. STRONG-AYA – No. 101057482 – D3.3 21 Appendix: FAIR da a In STRONG-AYA, we expec ha schema, de ini ions and e minologies a e qui e likely o change o e ime. Wi h igid da abase designs, he da abases gene ally need o be aken down and e-buil o accommoda e s uc u al changes. This will no be an op imal idea o STRONG-AYA, since we a e ully expec ing ou da ase s o g ow and ex end, as we lea n mo e and mo e wi h each o he . Also, we an icipa e ha a small numbe o p esen and u u e STRONG-AYA membe s migh no ha e a ully de eloped in e nal da a wo k low, combining clinical case epo o ms, elec onic heal h eco ds, ea men e i ica ions sys ems, e c. I would seem a p esen o be highly esou ce in ensi e o impose a ixed schema on a pa ne , unless hey jus happen o be al eady collec ing da a in ha o ma . Fu he mo e, p esen and u u e membe s o STRONG-AYA pa icipa e in many simul aneous p ojec s, hus i becomes unwieldy o main ain a snapsho o hei da a in many di e en schemas due o hei many di e en p ojec s. Figu e 3: Di e en da a collec ion pa adigms: wa ehouses, lakes and lakehouses (le o igh ) Simila i ies and di e ences among da a collec ion pa adigms a e illus a ing in he abo e diag am : (Le ) adi ional da a wa ehouse, (middle) uns uc u ed da a lake, and ( igh ) da a lakehouse. The unique possibili ies in a da a lakehouse a e : (1) o e ime, we can build da a collec ion in such a way as o use di e en ypes o da a, (2) by adding me ada a anno a ion and go e nance p ocessing on op o he da a i would be possible o make he unde lying da a FAIR, and suppo au oma ed e ie al om he da a, and (3) a ange o applica ions such as business epo s, da a analy ics and epidemiological modelling can be buil abo e he FAIR da a. The ins i u ional da a wo k lows inside a gi en STRONG-AYA pa ne will (na u ally) be unique and dis inc om wo k lows elsewhe e. Howe e , we can expec ha he e will be some high-le el kind o gene aliza ions o he wo k low. The igu e below is only an example o a possible gene aliza ion, acknowledging ha he e will be many di e en a ia ions o he heme. STRONG-AYA – No. 101057482 – D3.3 22 The p incipal da a sou ces in STRONG-AYA consis o a wide ange o mission-c i ical compu e sys ems ha collec da a and s o e da a (g ey box). These sys ems con ol ope a ions wi hin hospi als, egis ies o in e ace wi h pa ien s h ough some kind o elec onic su ey o ms. STRONG-AYA does no expec o y o use all o his da a, only he da a ha is pa o he CORE OUTCOMES SET (COS). The ask o he ins i u ion is o ge he necessa y pe missions and app o als o ex ac COS- ela ed da a om hese p ima y da a sys ems, and b ing he da a in o a p o ec ed esea ch en i onmen o he STRONG-AYA p ojec (o ange box). Along he way, he e will be asks equi ing he help o Da a P o ec ion o ice s, IT expe s, da abase expe s and in e nal e hics e iew commi ees o ge he da a. Inside he p o ec ed esea ch en i onmen o he STRONG-AYA p ojec (o ange box) we will as a conso ium ins all a se o so wa e ools and sa egua ds o (a) make he da a FAIR and mu ually in e -ope able wi hin he conso ium (b) s o e he me ada a anno a ions ha makes he da a FAIR, and (c) o add a connec i i y ga eway so ha summa y s a is ics and epidemiological models ha a e compu ed locally can be communica ed o a cen al agg ega ion se e , which will pu he many local esul s oge he in o a single global esul . No e ha da a s ays inside he p o ec ed esea ch en i onmen (o ange box) and no indi idual pa ien -le el in o ma ion is sen ou side o his en i onmen . Figu e 4 Da a low in STRONG AYA