DV4RDA Page 1
Adding Da a Visi a ion language o an IRB p o ocol
Templa e Ve sion:
Templa e Da e:
1.1c
No embe 24, 2025
P epa ed by: Pa icia Buendia, Seonyoung Kim, DV4RDA p ojec o he EOSC-
Fu u e/RDA A i icial In elligence and Da a Visi a ion Wo king G oup (AIDV WG), Resea ch
Da a Alliance (RDA)
DOI: 10.15497/RDA00142
DESCRIPTION
Adding Da a Visi a ion (DV) language o an IRB p o ocol equi es placing he desc ip ion wi hin
he sec ions ha add ess da a handling, p i acy p o ec ions, secu i y, and ex e nal da a
access. Al hough IRB p o ocol empla es a y ac oss ins i u ions (e.g., HRP-503 o biomedical
s udies, HRP-583 o social/beha io al s udies, HRP-593 o d ug/de ice ials), all empla es
sha e common co e sec ions whe e DV language can be inco po a ed.
This documen p o ides adap able DV language designed o i he mos common IRB p o ocol
sec ions, ega dless o he speci ic empla e used.
HOW TO USE THIS DOCUMENT
This guidance applies o h ee ca ego ies o esea ch scena ios whe e DV may be ele an :
1. Single-ins i u ion p ima y s udies wi h mixed access le els
2. Mul i-ins i u ion p ima y s udies
3. Ex e nal seconda y euse o sensi i e da a
Depending on he scena io, DV language may appea in di e en IRB sec ions. The sec ions
p o ided below include op ional DV pa ag aphs ha can be added o omi ed depending on
whe he DV is used o in e nal analysis, c oss-ins i u ion collabo a ion, o ex e nal seconda y
access.
IMPORTANT CONSIDERATIONS FOR THE IRB
When inse ing DV language in o an IRB p o ocol, ensu e ha he desc ip ion is:
●Clea : A oid o e ly echnical language. IRBs mus unde s and how DV p ese es
con iden iali y.
●Speci ic: S a e whe e he da a is s o ed, who has access o ha s o age, wha
sys em/pla o m is used o da a isi a ion, how he isi ing ools a e
au ho ized/con olled, and how ou pu s a e de-iden i ied
● Complian : Ensu e alignmen wi h HIPAA, GDPR, ins i u ional secu i y policies, da a-use
limi a ions s a ed in consen o ms
DV4RDA Page 2
●Risk-Awa e: DV should be desc ibed as a isk-mi iga ing s a egy ha educes da a
mo emen , duplica ion, and unau ho ized access.
USE CASES
DV language bene i s IRB p o ocols in h ee use cases:
1. Single-Ins i u ion P ima y S udies wi h Mixed Access Le els
Al hough many single-ins i u ion p ima y s udies g an ull da a access o all membe s o
he in e nal esea ch eam, his is no always app op ia e o pe mi ed. In p ac ice,
p ima y s udy eams o en include indi iduals in di e en oles who should no ha e ull
isibili y in o he aw, iden i iable da ase . In hese si ua ions, Da a Visi a ion (DV) can
se e as a secu e p ima y wo k low o suppo ole-app op ia e analy ics while
main aining egula o y and ins i u ional compliance.
When DV is Rele an in a Single-Ins i u ion P ima y S udy
DV may be used wi hin a single ins i u ion when he esea ch eam includes membe s
who ha e limi ed o es ic ed access o iden i iable da a, such as:
●Con ac o s, ex e nal endo s, o so wa e enginee s: These indi iduals may
suppo da a p ocessing, pla o m de elopmen , quali y con ol, o analy ic
pipelines, bu a e no ins i u ionally au ho ized o iew iden i iable da a. DV
enables hem o execu e code o pe o m sys em-le el asks wi hou exposing
aw da a ou side he secu e en i onmen .
●S uden s, ainees, pos docs, o junio pe sonnel: Some eam membe s may
con ibu e o analysis o explo a o y wo k bu a e no app o ed (o do no need)
o see he ull da ase . DV suppo s educa ional o analy ical engagemen
h ough con olled, p i acy-p ese ing ou pu s.
●Remo e o hyb id wo ke s: Fo compliance o secu i y easons, ce ain
ins i u ions es ic he s o age o handling o sensi i e da a on pe sonal o o -si e
de ices. DV elimina es he need o download o ans e da a o emo e sys ems
by keeping all compu a ion wi hin he secu e ins i u ional en i onmen .
Bene i s o Using DV Wi hin a Single-Ins i u ion S udy
Using DV as an in e nal wo k low can:
● P e en unnecessa y c ea ion o duplica e da ase s ac oss de ices and s o age
loca ions (a common bu unde - ecognized secu i y isk)
● Reduce he need o g an b oad access p i ileges beyond wha is jus i ied o
each ole
● Enhance audi abili y and compliance wi h IRB-app o ed access limi a ions
● S eamline collabo a ion by allowing con olled analysis wi hou eloca ing da a
DV4RDA Page 3
2. Mul i-Ins i u ion P ima y S udies
Mul i-ins i u ion p ima y s udies in ol e esea ch eams dis ibu ed ac oss wo o mo e
o ganiza ions, o en wi h di e en echnical in as uc u es, secu i y con ols, da a access
policies, and egula o y equi emen s. In hese se ings, Da a Visi a ion (DV) can se e
as a p ima y mechanism o secu e, coo dina ed da a analysis ac oss ins i u ions wi hou
mo ing o duplica ing sensi i e da ase s.
Why DV is Valuable in Mul i-Ins i u ion S udies
When mul iple ins i u ions collabo a e, i is a ely app op ia e o all pa ne s o ecei e
ull-access copies o he aw da a. Challenges include:
● Da a T ans e Res ic ions: Sensi i e o egula ed da a (e.g., PHI, genomic
da a, beha io al da a) may no be legally o con ac ually ans e able ac oss
ins i u ions due o egula o y equi emen s.
● Role-based access equi emen s: Mo ing da a be ween ins i u ions c ea es
mul iple uncon olled copies, inc easing he isk o di e gence, co up ion, o
unau ho ized use.
● He e ogeneous secu i y en i onmen s: No all ins i u ions mee he same
secu i y s anda ds (e.g., HIPAA, NIST 800-171), making i unsa e o dis ibu e he
aw da ase widely.
DV add esses hese challenges by enabling compu a ion o occu whe e he da a eside,
while collabo a o s in e ac only wi h con olled in e aces and app o ed ou pu s.
Common Scena ios Whe e DV Suppo s Mul i-Ins i u ion S udies
DV is pa icula ly e ec i e o :
● C oss-si e analysis in conso ia, including pooled model aining o
ede a ed-s yle wo k lows
● Collabo a o s con ibu ing s a is ical code o models wi hou needing aw da a
● Ex e nal analys s o me hodologis s who equi e access o pa e ns, esul s, o
ends bu no iden i ie s
● Cloud-based o pla o m-based pipelines whe e pa ne ins i u ions ha e limi ed
access pe missions
Bene i s o Using DV in Mul i-Ins i u ion S udies
DV o e s se e al ad an ages in in e -ins i u ional collabo a ion:
● Remo es he need o ans e , eplica e, o ha monize sensi i e da a ac oss
o ganiza ions
● En o ces consis en , cen alized secu i y and access con ols
● Simpli ies compliance wi h di e ing o con lic ing ins i u ional policies
● Suppo s ep oducibili y by ensu ing all analyses ope a e on he same
au ho i a i e da a sou ce
● P o ides audi logs and access con ols o c oss-ins i u ion in e ac ions
DV4RDA Page 4
3. Ex e nal Seconda y Reuse o Sensi i e Da a (DV as an Al e na i e o DUA)
Ex e nal seconda y euse in ol es esea che s ou side he o iginal s udy eam
eques ing access o da a collec ed du ing he p ima y s udy. T adi ionally, hese
eques s a e handled h ough a Da a Use Ag eemen (DUA) ha allows he seconda y
esea che o ecei e a copy o he da ase . Howe e , DUA-based access o en p esen s
signi ican legal, e hical, and adminis a i e ba ie s – especially o sensi i e o
po en ially e-iden i iable da a. DV o e s a secu e al e na i e ha enables meaning ul
analysis wi hou ans e ing aw da a ou side he o igina ing ins i u ion.
Why DV Is Valuable o Ex e nal Seconda y Reuse
The DUA ou e can place a high bu den on bo h he o igina ing ins i u ion and he
eques ing esea che due o:
● Legal and con ac ual cons ain s: DUAs equi e mul i-le el e iew (legal,
compliance, da a go e nance), p olonged nego ia ion, and assessmen o he
ecei ing ins i u ion’s secu i y en i onmen .
● E hical limi a ions: Pa icipan consen may es ic edis ibu ion o iden i iable,
genomic, o sensi i e da a. T ans e ing a da ase can aise compliance
conce ns.
● Secu i y isks: Once da a lea e he o iginal en i onmen , he da a owne loses
con ol o e s o age, downs eam sha ing, and long e m-s ewa dship.
● Adminis a i e wo kload and delays: DUA p ocessing is o en slow,
esou ce-in ensi e and inconsis en ac oss ins i u ions.
DV a oids hese issues by allowing ex e nal use s o conduc analysis wi hou ecei ing
a physical copy o he da a.
Common Scena ios Whe e DV Suppo s Seconda y Reuse
DV is pa icula ly e ec i e when ex e nal esea che s:
● Need access o pa e ns, agg ega e ends, o easibili y-le el esul s
● Analyze highly sensi i e da a ha canno be edis ibu ed
● Mus comply wi h consen language ha limi s b oad da a sha ing
● Requi e explo a o y wo k be o e a possible o mal da a ans e
● Do no ha e an app o ed secu e en i onmen o s o e sensi i e da a
Bene i s o Using DV o Seconda y Reuse
Using DV as an access pa hway o e s se e al ad an ages:
● No da a ans e : elimina es legal and echnical e iew ied o da ase expo
● No p oli e a ion o copies: educes go e nance, acking, and compliance bu den
● Suppo s es ic i e consen : enables analysis compa ible wi h limi ed-sha ing
language
● G ea e secu i y: access is audi able, ime-limi ed, and igh ly con olled
● Lowe adminis a i e ba ie : o en as e app o al compa ed o DUAs
● Risk educ ion: minimizes ins i u ional liabili y by keeping da a in one secu e
en i onmen
DV4RDA Page 5
PLACEMENT OF DV LANGUAGE IN IRB PROTOCOLS
Ac oss common IRB empla es (HRP-503, HRP-583, HRP-593), he mos app op ia e sec ions
o include DV language a e:
● Da a Managemen and Secu i y
● P i acy and Con iden iali y P o ec ions
● Seconda y Use, Fu u e Use, o Da a Sha ing (when p esen )
● Da a Analysis Plan (i DV a ec s how analysis is pe o med)
● In o med Consen (i DV a ec s how pa icipan da a mo e o a e p o ec ed)
No all s udies will equi e DV language. The IRB language p o ided in la e sec ions is op ional
and should be included only when DV meaning ully a ec s how da a a e collec ed, analyzed,
sha ed, o p o ec ed.
PROPOSED ADDITIONS TO IRB PROTOCOL TEMPLATE
The ollowing sec ions p o ide op ional DV language ha can be inse ed in o an IRB p o ocol
when applicable.
1. Sec ion: Da a Collec ion & Managemen
● Whe e o add: Wi hin he desc ip ion o how da a will be collec ed, s o ed, and accessed
o analysis.
● S anda d P o ocol Language: Pa icipan da a collec ed o his s udy, including
[speci y da a ypes], will be s o ed secu ely a [loca ion] in acco dance wi h ins i u ional
policies and all applicable egula ions.
● Op ional DV Language (include only i DV is used): I his s udy uses Da a Visi a ion
(DV) o in e nal analysis, mul i-ins i u ion collabo a ion, o ex e nal seconda y access,
sensi i e pa icipan da a will emain wi hin he secu e compu ing en i onmen whe e
hey a e o iginally s o ed (e.g., ins i u ion-managed secu e esea ch se e s o an
app o ed da a encla e). Access o his en i onmen is es ic ed o au ho ized pe sonnel
as de ined in he IRB-app o ed s udy oles.
Ra he han ans e ing o downloading aw da a, app o ed analy ical code o models
a e b ough o he da a’s loca ion using a secu e DV pla o m (e.g., FAIRlyz). All code
execu ion is pe mission-con olled, logged, and e iewed. Only agg ega ed, de-iden i ied
ou pu s ha mee ins i u ional p i acy h esholds and comply wi h HIPAA, GDPR (i
applicable), and pa icipan consen es ic ions will be accessible ou side he secu e
en i onmen . No iden i iable o indi idual-le el da a will be expo ed, copied, o s o ed
locally by use s.
DV4RDA Page 6
2. Sec ion: Con iden iali y & P i acy
● Whe e o add: Wi hin he desc ip ion o how he s udy p o ec s pa icipan p i acy.
● S anda d P o ocol Language: P o ec ing pa icipan p i acy is essen ial. Iden i iable
da a will be limi ed o au ho ized pe sonnel, and secu e s o age sys ems will be used o
main ain con iden iali y.
● Op ional DV Language (include only i DV con ibu es o p i acy p o ec ions): The
s udy may use Da a Visi a ion (DV) as a p i acy-p ese ing s a egy o p e en
unnecessa y mo emen o duplica ion o iden i iable da a. Unde DV, aw da a emain in
he secu e compu ing en i onmen , and analyses occu locally wi hin ha en i onmen .
Only app o ed agg ega e o de-iden i ied esul s may lea e he secu e en i onmen .
Access o he DV sys em is logged, moni o ed, and es ic ed o au ho ized indi iduals.
No iden i iable da a lea es he secu e en i onmen a any ime. Resul s a e also
e iewed o ensu e hey do no unin en ionally e eal in o ma ion abou indi iduals.
3. Sec ion: Da a Analysis Plan
● Whe e o add: Wi hin he desc ip ion o analy ical me hods.
● Op ional DV Language (include only i DV a ec s analysis wo k lows): I he s udy
uses Da a Visi a ion (DV), analy ical code will be execu ed wi hin he DV pla o m, which
en o ces secu e, con olled access o he unde lying da a. S a is ical analyses,
compu a ional models, o machine lea ning wo k lows will un inside he secu e
en i onmen , ensu ing ha aw da a a e ne e expo ed o eplica ed. Only agg ega ed
esul s o de-iden i ied ou pu s ha mee disclosu e con ol equi emen s and comply
wi h HIPAA and ins i u ional policies will be e u ned o s udy in e p e a ion and
epo ing.
4. Sec ion: S o age o Da a and Specimens
● Whe e o add: Wi hin he desc ip ion o whe e da a a e s o ed and how long hey a e
e ained.
● Op ional DV Language (include only i DV is used): When DV is used, all analysis
occu s wi hin he secu e s o age en i onmen , educing he c ea ion o duplica e
da ase s and minimizing he isk o unau ho ized access h ough da a ans e .
DV4RDA Page 7
5. Sec ion: Seconda y Use, Fu u e Use, o Ex e nal Da a Sha ing
●Whe e o add: In he sec ion desc ibing how s udy da a may be sha ed wi h ex e nal
esea che s.
●Op ional DV Language ( ecommended o sensi i e da a): Ex e nal eques s o
access o s udy da a may be suppo ed h ough ei he (1) a adi ional Da a Use
Ag eemen (DUA), which pe mi s he ans e o a da ase o an ex e nal in es iga o , o
(2) a Da a Visi a ion (DV) pa hway ha allows ex e nal esea che s o un app o ed
analyses wi hin a secu e en i onmen wi hou ecei ing aw da a.
Unde he DV app oach, he da a emain wi hin he ins i u ion’s secu e en i onmen , and
ex e nal use s ecei e only agg ega e o de-iden i ied esul s. Be o e any esul s a e
eleased, he DV sys em e iews he ou pu s o ensu e hey do no con ain small cell
sizes, iden i iable combina ions, o o he in o ma ion ha could po en ially e eal
pa icipan iden i ies. Ou pu s ha do no mee he ins i u ion’s p i acy s anda ds a e no
eleased.
6. Sec ion: Pa icipan No i ica ion and In o med Consen
●Whe e o add: In he sec ion desc ibing how pa icipan s will be in o med abou da a
handling.
●Op ional DV Language (include only when DV a ec s consen language):
Pa icipan s may be in o med ha hei iden i iable da a will emain in a secu e
en i onmen and will no be ans e ed ou side ha en i onmen o analysis. I DV is
used, analyses occu only wi hin a secu e, access-con olled en i onmen , and only
de-iden i ied summa ies o agg ega e esul s a e sha ed ex e nally. This app oach
educes p i acy isks by minimizing da a mo emen and p e en ing he expo o
iden i iable in o ma ion.
Ci e as:
P Buendia, S Kim. “Adding Da a Visi a ion language o an IRB p o ocol.” DV4RDA P ojec o
he EOSC-Fu u e/RDA A i icial In elligence and Da a Visi a ion Wo king G oup. Resea ch
Da a Alliance. DOI: 10.15497/RDA00142 No embe 15, 2025.
Acknowledgemen s
This DV4RDA p ojec has ecei ed unding h ough RDA TIGER om he
Eu opean Union’s Ho izon Eu ope amewo k p og amme unde g an
ag eemen No. 101094406. Views and opinions exp essed a e howe e
hose o he au ho s only and do no necessa ily e lec hose o he
Eu opean Union o ins i u ions ep esen ed he e. Nei he he Eu opean
Union no he ins i u ions can be held esponsible o hem.
