A systematic literature review of health consumer

RESEARCH Open Access

A systematic literature review of healthconsumer attitudes towards secondary useand sharing of health administrative andclinical trial data: a focus on privacy, trust,and transparencyElizabeth Hutchings1* , Max Loomes2 , Phyllis Butow2,3,4 and Frances M. Boyle1,5

Abstract

We aimed to synthesise data on issues related to stakeholder perceptions of privacy, trust, and transparency in use ofsecondary data. A systematic literature review of healthcare consumer attitudes towards the secondary use and sharingof health administrative and clinical trial data was conducted. EMBASE/MEDLINE, Cochrane Library, PubMed, CINAHL,Informit Health Collection, PROSPERO Database of Systematic Reviews, PsycINFO, and ProQuest databases weresearched. Eligible articles included those reporting qualitative or quantitative original research and published in English.No restrictions were placed on publication dates, study design or disease setting. One author screened articles foreligibility, and two authors were involved in the full text review process. Data was extracted using a pre-piloted dataextraction template by one author and checked by another. Conflicts were resolved by consensus. Quality and biaswere assessed using the QualSyst criteria for qualitative and quantitative studies. This paper focuses on a subset of 35articles identified from the wider search which focus on issues of privacy, trust, and transparency. Studies included atotal of 56,365 respondents. Results of this systematic literature review indicate that while respondents identifiedadvantages in sharing health data, concerns relating to trust, transparency, and privacy remain. Organisations collectinghealth data and those who seek to share data or undertake secondary data analysis should continue to develop trust,transparency, and privacy with healthcare consumers through open dialogue and education. Consideration should begiven to these issues at all stages of data collection including the conception, design, and implementation phases.While individuals understand the benefits of health data sharing for research purposes, ensuring a balance betweenpublic benefit and individual privacy is essential. Researchers and those undertaking secondary data analysis need tobe cognisant of these key issues at all stages of their research. Systematic review registration: PROSPERO registrationnumber CRD42018110559 (update June 2020).

Keywords: Data sharing, Attitudes, Privacy, Trust, Transparency

© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate ifchanges were made. The images or other third party material in this article are included in the article's Creative Commonslicence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commonslicence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtainpermission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to thedata made available in this article, unless otherwise stated in a credit line to the data.

* Correspondence: [email protected] Clinical School, Faculty of Medicine, University of Sydney, NorthSydney, NSW, AustraliaFull list of author information is available at the end of the article

Hutchings et al. Systematic Reviews (2020) 9:235 https://doi.org/10.1186/s13643-020-01481-9

http://crossmark.crossref.org/dialog/?doi=10.1186/s13643-020-01481-9&domain=pdf

https://orcid.org/0000-0002-6030-954X

https://orcid.org/0000-0003-1042-0968

https://orcid.org/0000-0003-3562-6954

https://orcid.org/0000-0003-3798-1570

http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/publicdomain/zero/1.0/

mailto:[email protected]

BackgroundHealthcare provides an opportune setting for increaseddata sharing and secondary data analysis. Secondary dataanalysis of existing data originally collected for other pur-poses [1] can provide insights into real-world clinical prac-tice [2] and generate new clinical evidence [3]. There aremany forms of data collected during an individual’s inter-actions with health services, including administrative andclinical trial data which are the focus of this review.Administrative data are data originally collected for ad-ministrative and billing purposes [4], but have the capacityto be used to identify systemic issues and service gaps andused to inform improved health resourcing. Clinical trialsare expensive and take an approximately 17 years tocomplete, and less than 14% of the evidence is translatedinto practice [5]. Given the low rates of evidence beingtranslated into practice, it can be suggested that the sec-ondary use of this data has greater importance. The sec-ondary analysis of clinical trial data can further advancethe medical community’s understanding of diseases andpotentially limit the expenditure of funds on already testedhypotheses.Increased access to data for secondary use is complex

and continues to attract strong debate within the healthand scientific communities as well as the general public.While researchers are now being encouraged to increasedata accessibility for secondary research [6, 7], a range ofstakeholder-perceived barriers and concerns remain, in-cluding issues such as trust, transparency, and privacy[8, 9]. Despite the impact of these issues on willingnessto share data, there is a lack of synthesis of stakeholderviews to guide policy and practice.This paper presents the results of a subset of articles

identified in our systematic literature review and focuseson healthcare consumer concerns relating to privacy,trust, and transparency in the setting health administra-tive and clinical trial data reuse.

MethodsThis systematic literature review presents the results of asubset of articles identified in a larger review of articlesaddressing data sharing and was undertaken in accord-ance with the PRISMA statement for systematic reviewsand meta-analysis [10]. The protocol was prospectivelyregistered on PROSPERO (www.crd.york.ac.uk/PROS-PERO, CRD42018110559; updated June 2020).The following databases were searched: EMBASE/

MEDLINE, Cochrane Library, PubMed, CINAHL, Infor-mit Health Collection, PROSPERO Database of System-atic Reviews, PsycINFO, and ProQuest. The search wasconducted on 24 June 2020. No date restrictions wereplaced on the search; key search terms are listed inTable 1.

Our original goal was to focus on attitudes towardsdata reuse by breast cancer patients. However, due to apaucity of studies targeting this group, we re-ran thesearch without this limitation and present the results ofall disease settings and noted specific cases where breastcancer or any cancers were included. Breast cancer is adisease that impacts older individuals; therefore, respon-dents under the age of 18 years were excluded from thisanalysis, as were attitudes towards biobanking and gen-etic research.We noted that increasingly the delineation between

data collected for administrative purposes and otherforms of electronic documentation such as electronichealth records (EHR) (or other terms for these) becomesless clear. These records can contain both administrativeand clinical data. Where possible, EHRs were excludedfrom this literature review; however, we acknowledgethat the lack of separation has made this a grey area.Papers were considered eligible if they were published

in English in a peer-reviewed journal; reported originalresearch, either qualitative or quantitative with any studydesign, related to data sharing in any disease setting; andincluded subjects over 18 years of age. Reference list andhand searching was undertaken to identify additional pa-pers. Systematic literature reviews were included in thewider search but were not included in the results. Paperswere considered ineligible if they focused on electronichealth records (including other terms for these), healthinformation exchanges, biobanking and genetics, andwere review articles, opinion pieces, articles, letters, edi-torials or non-peer-reviewed theses from masters anddoctoral research. Duplicates were removed and titleand abstract and full text screening were undertakenusing the Cochrane systematic literature reviewprogramme ‘Covidence’ [11]. One author screened arti-cles for eligibility and two authors were involved in thefull text review process; conflicts were resolved byconsensus.

Table 1 Example search strategy

PubMed

1 ((data sharing) OR (data link*) OR (secondary data analysis) OR(data reuse) OR (data mining))

2 ((real world data) OR (clinical trial) OR (medical record*) OR (patientrecord*) OR (routine data) OR (administrative data))

3 attitud* OR view* OR opinion* OR perspective* OR satisfaction)

4 (patient* OR consumer*)

5 (doctor* OR clinician OR oncologist OR specialist*)

6 (Researcher* OR scientist* OR (data custodian*))

7 4 or 5 or 6

8 1 and 2 and 3

9 1 and 2 and 3 and 7

*Search includes ‘wildcards’ or truncation

Hutchings et al. Systematic Reviews (2020) 9:235 Page 2 of 41

http://www.crd.york.ac.uk/PROSPERO

http://www.crd.york.ac.uk/PROSPERO

Quality and bias were assessed at a study level usingthe QualSyst system for quantitative and qualitativestudies as described by Kmet et al [12]; this is a validatedtool and can be used to assess both qualitative andquantitative studies. No modifications were made to theQualSyst criteria prior to use. Quality and bias assess-ment was undertaken independently by two authors;conflicts were resolved by consensus. A maximum scoreof 20 is assigned to articles of high quality and low bias;the final QualSyst score is a proportion of the total, witha possible score ranging from 0.0 to 1.0 [12].Data extraction was undertaken by one author using a

pre-piloted form in Microsoft Office Excel; a second au-thor confirmed the data extraction. Conflicts were re-solved by consensus. Data points included author,country and year of study, study design and method-ology, health setting, and key themes and results. Whereavailable, detailed information on research participantswas extracted including age, sex, employment status,highest level of education, and health status.Quantitative data were summarised using descriptive

statistics. Synthesis of qualitative findings used a meta-aggregative approach, in accordance with guidelinesfrom Lockwood et al [13]. The main themes of eachqualitative study were first identified and then combined,if relevant, into categories of commonality. Using a con-stant comparative approach, higher-order themes andsubthemes were developed. Quantitative data relevant toeach theme were then incorporated. Using a frameworkanalysis approach as described by Gale et al [14], theperspectives of different groups towards data sharingwere identified. Where differences occurred, they arehighlighted in the results. Similarly, where systematicdifferences according to other characteristics (such asage or sex) occurred, these are highlighted.

ResultsThis search identified 10,499 articles, of which 323underwent full text screening; 75 articles met the inclu-sion criteria for the larger review. The PRISMA diagramis presented in Fig. 1. This article presents a subset ofthe results of the wider search which explores attitudesof health consumers towards privacy, trust, and trans-parency. The results relating to attitudes towards datasharing and reuse by researchers and healthcare profes-sionals, and attitudes towards consent in the context ofdata sharing and reuse by healthcare consumers are pre-sented in subsequent publications.A subset of 35 [15–49] of the 75 articles addressed

issues relating to privacy [15–49], trust [16–18, 21,23, 24, 26, 28–30, 32–37, 39–42, 44–46, 48], and/ortransparency [15–17, 26, 30, 32, 33, 37, 40, 42–44,48] and are included in this analysis (Fig. 1 and

Table 2). A total of 56,365 respondents were includedin the studies.

Study design, location, clinical focus, and studypopulationsQualitative research methodologies included face-to-faceinterviews and/or focus groups [32–34, 36–38, 49].Other designs included surveys [16–21, 23–29, 35, 39,41, 44] and combinations of deliberative sessions withsurveys [15, 40, 45, 46] and focus groups and interviews[43]. Two studies used a citizens’ jury model [47, 48]and another was a nested cohort within a randomisedcontrolled trial [22]. Studies were conducted in severalcountries; a breakdown by country is presented inTable 3.Most articles focused on the general public’s atti-

tudes towards secondary data usage, particularly ingeneral medicine [18, 22, 25–30, 34, 37–39, 43–45,48, 49], but also national cancer databases [31, 41],clinical trials [21, 32], fertility [33], pharmaco-epidemiological [47], and epidemiological [42] re-search. Other studies focused on health consumers’attitudes to secondary data usage in individuals: at-tending US Veterans Affairs (VA) facilities [40] or re-cently discharged from tertiary care [15], or witharthritis and other chronic conditions [36]. Otherswere in the setting of human immunodeficiency virus(HIV) [49], breast cancer (BC), colon cancer (CC)[35], or heterogeneous cancers [19, 20], acquired im-mune deficiency syndrome (AIDS), or multiple scler-osis (MS), or mental health concerns [23], presentingwith rare diseases [16, 17], in adults or parents ofchildren with cystic fibrosis (CF), sickle cell disease(SCD), or diabetes mellitus (DM) [35], or in adultswith potentially stigmatising health conditions (DM,hypertension, chronic depression, alcoholism, HIV,BC, or lung cancer) [46].The majority of articles discussed general attitudes to-

wards health data linkage and secondary use [16, 22, 27,30, 37, 39, 43, 46], linking health administrative data toclinical trial data [20] or clinical trial data reuse [21, 32],linking administrative data to survey data [38], access tomedical records [15, 19, 23, 25, 26, 28, 29, 34, 35, 40, 45,47, 48], statistical databases [49], research registries [17,18, 31, 33, 36, 41], and health data for epidemiologicalresearch [42]. Privacy as sociotechnical capital [24] andcommercial access to health data [44] were consideredin one article each.

Study qualityResults of the quality assessment are provided in Table 2.QualSyst [12] scores ranged from 0.5 to 1.0 (possiblerange 0.0 to 1.0). While none were blinded studies, mostprovided clear information on respondent selection and


data analysis methods and used justifiable study designsand methodologies. No key themes stood out for studieswhich received poorer judgements. No data were fromrandomised studies, with the highest level of evidencefrom a nested cohort study. Other data were obtainedfrom lower-quality studies such as surveys andinterviews.

ThemesTrustA total of 12,794 respondents provide a view on trust;results were from surveys, questionnaires, focus groups,and interviews. One study was a nested cohort in a ran-domised control trial and two used a citizens’ jurymodel. Participants emphasised that organisations mustdevelop, maintain, and promote high levels of patienttrust [24, 26, 42]. Developing this trust can be achievedthrough the maintenance of confidential records and byproviding information on how the individual’s informa-tion is used and by whom [40]. The importance of trust

in health organisations, clinicians, and university re-searchers was also noted [18, 26, 29, 40, 42], althoughgenerally respondents trusted that organisations wouldkeep their data private and confidential and that thiswould not be intentionally violated [40]. If a personalconnection with the research team is established, then itis easier for individuals to form a trusting relationship[23]. The highest levels of trust was placed in the doctor[16, 26, 46], the National Health Service (NHS) [26], andhospitals [16, 46], while the lowest trust was in commer-cial organisations [26], pharmaceutical companies andinsurance companies [46], or for-profit organisations[16]. An individual’s trust in an organisation was a deter-minant of what level of control they preferred over theirdata [40] and their willingness to participate in research[42], with trust overcoming concerns about privacy andconfidentiality [49]. Where an organisation shows clearand relevant connections between their research and theinformation contained in the records, respondentstrusted that the organisations will maintain the data

Fig. 1 PRISMA flow diagram


Table 2 Included studies

Author,location,datacollectiondate

Methodology,sampling, analysis

Healthcondition/setting

No. ofparticipants(N)

Participantdemographicsn (%)

Key themes(alphabetical)

Outcomes, result(s) QualSystScore

Qualitative

[15], Britain,October toDecember2004

Questionnaire, resultsexpressed inpercentages

Recentlydischargedmedical andsurgical patients/ hospital

166 Not reported PrivacyTransparency

Privacy• UK law currentlyallow for data use ifthe data is necessaryand the risk toprivacy isproportionate.

Transparency• Public education onthe secondary use ofdata promotestransparency.

0.9

[16],Europe,March toMay 2018

Survey, descriptivestatistics, and chi-square test forindependence

Rare diseases/patients andtheir familieswith rarediseases

2013 Age, years2 (0), 15–1770 (3), 18–24293 (15), 25–34852 (42), 35–49644 (32), 50–64152 (8), ≥65Sex, male473 (23)Location1775 (88), EU238 (12), non-EUDiagnostic status1909 (95), diagnosed104 (5), undiagnosedNumber of raredisease (n = 1909)1664 (87), 1174 (9), 244 (2), 313 (1), 414 (1), ≥5

PrivacyTransparencyTrust

PrivacyWillingness to sharedata is underscored bythe need to respectthe individual’s privacywishes and choicesregarding the use oftheir data.TransparencyImprovedcommunication willincrease transparencyin research, particularlyin the setting ofshared data.Trust• Not-for-profit stake-holders (89% formedical doctors, 79%researchers fromnon-profit organisa-tions, 77% for patientorganisations, 69%for healthcare profes-sionals other thanmedical doctors) isconsiderably higherthan trust in for-profitstakeholders.

• Participants withhigher levels ofeducation tend totrust governmentand institutions fromtheir country more(53%) than thosewho finished schoolearlier (44% amongthose who finishedschool before 20years old).

Distrust• 45% are in favourand 50% areopposed to sharingdata with researchersfrom thepharmaceuticalindustry.

• 16% are in favour

0.95


Table 2 Included studies (Continued)








and 80% areopposed data withinsurance companies

Older respondents areless likely to trust theprivate sector: 57% ofrespondents under 25trust researchers frompharmaceuticalindustry compared to36% for respondentsover 65 and only 28%compared to 9% forinsurance companies.

[17],Europe,2012

Questionnaire, resultsexpressed inpercentages

Leukodystrophy/Leukodystrophypatients andfamily

195 (149family, 46patients)

Age, years121 (62), 40–64Country130 (66.7), France24 (12.3), Italy9 (4.6), Belgium6 (3.1), Spain26 (13.3), Germany


Privacy• Ethics committeesprotect ‘patients’rights and privacy’.

Transparency• Information andtransparency areneeded on databasegovernance.

• Patient organisationsshould help with thedevelopment ofclinical trials.

• Transparency neededon the use of data,data storage and thelength of dataaccessibility.

• Information onsecondary use shouldbe provided at initialconsent.

• Pharmaceuticalindustry access todata was supportedby some, iftransparent.

Trust• High level of trust inresearchers using thedatabase.

• Patient’s/families hadgreat hope and trustin the developmentof database research.

0.85

[18],Finland,notreported

Survey, descriptivestatistics

General medical/general public

418 Age, years44 (10.5), ≤3047 (11.2), 31–4057 (13.6), 41–5088 (21.1), 51–60126 (30.1), 61–7049 (11.7), >707 (1.7), missingEducation74 (17.7), PS44 (10.5), SS165 (39.5), HS80 (19.1), University ofApplied Sciences or

PrivacyTrust

Privacy• Privacy protectionwas the mostcommon concern forregister-basedresearch.

• 80% consideredprotection of theirhealth informationimportant or veryimportant. 56% hadpositive attitudetowards their owninformation being

0.85










bachelor’s degree52 (12.4), Master’sdegree or higher3 (0.7), missing

used for research.• Sex, education, andhealth status did notaffect this viewsignificantly.

• Studies of majorpublic health impactare more importantthan individualprivacy protection—25% somewhatdisagreed and 12%strongly disagreed.

• Respondents in goodhealth more likely toagree than thosewith poor health.

• >70 years lessconcerned withprivacy than youngerage groups; 92% (31–40 years) statedprivacy of healthinformation was veryimportant orimportant.

Trust• 56% had a positive orvery positive attitudeabout theirinformation beingused researchpurposes, 31% wereneutral suggestingtrust in researchorganisations anddata protectionpractices.

• Younger participants,despite higherconcern for privacy,trust researchers andwere willing to lettheir information beused for research.

[19], Japan,October toNovember1995

Survey, Fisher’s exacttest and Mantel-Haenszel chi-square

Cancer/outpatients, andinpatients

293 Age, years11 (3.8), ≤2930 (10.2), 30–3949 (16.7), 40–4971, (24.2), 50–5981 (27.6), 60–6948 (16.4), ≥70Sex, male115 (39)

Privacy Privacy• 82.6% did not mindtheir own clinicaldata being used forhealthcare andtreatment skillsimprovement underthe condition thatprivacy would not beviolated.

• Younger malesseemed mostpositive to the use ofdata to improvehealthcare andtreatment skills.

• No significantdifference inresponses among

0.5










three sources ofpatients.

[20],Canada,August2014 andMay 2015

Survey, descriptivestatistics and Student’st tests andnonparametric tests

Cancer/outpatient clinic

569 Age, years59, medianCancer type109 (19.2), breast86 (15.1),gastrointestinal83 (14.6) genitourinary70 (12.3), thoracic73 (12.8), hematologic73 (12.8), head andneck63 (11.1),gynaecologic12 (2.1), otherClinical trialparticipation, yes183 (32.2)Education346 (60.8), university,college, professional39 (6.9), vocational,technical, diploma169 (29.7), elementary,HS15 (2.6), prefer not toanswer or missingSex, male234 (41)

Privacy Privacy• Willingness toprovide healthinformation forresearch.

• Cultural differencesbetween large citiesand smallercommunities mayalter an individual’slevel of concernabout privacy andconfidentiality.

• Linkage of clinicaltrial data withadministrativedatabases requiresconsideration ofethics and regulatoryprinciples andprocesses to ensureprivacy.

0.75

[21], USA,notreported

Structured survey,univariate statistics,logistic regressionmodels

Clinical trial data/interventionalclinical trialparticipants

771 Age, years (n = 762)63 (8.3), <25177 (23.2), 25–44286 (37.5), 45–64236 (31), ≥65Sex, male382 (51.1)Ethnicity (n = 768)518 (67.4), white113 (14.7), Black orAfrican American51 (6.6), AmericanIndian or AlaskanNative25 (3.3), Asian61 (7.9), OtherEducation (n = 752)40 (5.3), < HS125 (16.6), HS206 (27.4) somecollege238 (31.6), college143 (19), graduatedegree

PrivacyTrust

Privacy93% of respondentswere very orsomewhat likely toallow their data to beshared with universityresearchers82% were very orsomewhat likely toshare their data withresearchers from for-profit companies.• This willingness didnot change of thepurpose for the useof data, except foruse in litigation.

Provided adequatesafeguards were inplace, most werewilling to share theirdata for research.Trust• Some believed thatsharing this data mayreduce participationrates in clinicalresearch (37%), wereconcerned that itwould be used formarketing purposes(34%), stolen (30%),used to discriminate

0.9










(22%), or for profit(20%).

• Respondents had lesstrust inpharmaceuticalcompanies (18%trusted a great dealor a lot) andinsurance companies(15%) compared touniversities (63%).

• 8% were unwilling toshare with for-profitcompanies, 92%would be very orsomewhat willing toshare with thesecompanies.

• Multivariablemodelling found thatthose who believedthat the negativeaspects of datasharing outweighedthe positives, wassignificantly higher inthose who felt thatother people couldgenerally not betrusted, thoseconcerned aboutbeing re-identified, orthose concernedabout informationtheft.

• Those with a collegedegree were alsoassociated withfeeling the negativeaspects outweighedthe benefits.

• Low levels of trustresulted in lesswillingness to sharedata with researchersfrom both not-for-profit and for-profitcompanies.

[22], HongKong, notreported

RCT nested within acohort, chi-square test,multivariable logisticregression, likelihoodratio test

General medical/subsample of theFAMILY cohort

1200 Age, years94 (7.8), 18–29197 (16.4), 30–44423 (35.3), 45–59307 (25.6), 60–74179 (14.9), ≥75Education436 (36.3), primary587 (48.9), secondary177 (14.8), tertiarySex, male456 (38)

Privacy Privacy• Requesting HKIDsignificantly reducedconsent for datalinkage among adultsaged 18–44 years(9.9%) compared toother age groupsreflecting privacyconcerns.

• 60–74 years weremost likely to agreeto data linkage(40.7%).

0.95

[23], Survey, descriptive AIDS, MS, mental 235 Age, years Privacy Privacy 0.85










Canada,November2003

statistics, multiplelogistic regression

health /outpatients

68 (28.9), 20–39129 (54.9), 40–5934 (14.5), ≥604 (1.7), unknownEducation39 (16.6), <grade 1235 (14.9), grade 12156 (66.4), attended/finished post-secondary5 (2.1), no answerPrevious experiencewith medicalresearch, yes127 (54)Sex, male86 (36.6)

Trust • Privacy of health andfinancial records ofmost concern torespondents.

• To maintain privacy,consent to accessmedical recordsshould be sought.

• Concerns includedharm due to othersaccessing record;information beingavailable to ‘manypeople’.

• Privacy andautonomy are ofimportance.

• There was not awillingness to tradeloss of privacy forpublic good.

Trust• Easier to trustresearchers when theindividual is involvedand meets someoneinvolved in theresearch.

• Education is neededto foster trust inhealth researchprocesses and toprovide moreinformation onsocietal perspectivesof personal healthdata.

[24], USA,August toNovember2014

Health InformationNational Trends Survey(HINTS), multivariateregression analyses


3212 Age, years, mean53.48Sex, male1317 (41)

PrivacyTrust

Privacy• Three key facets ofhealth privacy: capitalawareness of privacy,attitude towards theimportance of privacyand data sharing, andconfidence in theability to maintainprivacy.

• Positive relationshipbetween privacycapital andengagement.

• Development ofhealth privacy capitalis susceptible tosociodemographicdisparities.

• Higher levels ofeducation andincome associatedwith higher levels ofhealth privacy capital.

• Awareness andconfidence were alsorelated to the social

0.9










capital index.• Females and olderpeople were morelikely to be aware ofhealth information.

• Those with higherlevels of healthprivacy capital weremore likely to reporthigher levels ofdigital interactionwith healthprofessionals.

• Higher levels ofincome were foundto do consistentlybetter in healthoutcomes, and theytended to be moreactively engaged indigital interactionwith healthprofessionals.

Trust• Privacy awareness asa predictor seems toplay an enabling rolefor those who lackbasic social networksbecause it helpsthem fully trust andefficiently navigatethe health system totheir benefit.

• Promoting trust inhealth service isimportant.

[25],Europe,August toNovember2013

Survey, descriptiveanalysis and choicemodelling analysis,Westin’s methodology


20,882 Age, years2163 (10.4), 18–243512 (16.8), 25–343719 (17.8), 35–443763 (18), 45–543735 (17.9), 55–643990 (19.1), ≥65Health status, self-rated12,823 (61.4), good orvery goodSex, male9960 (47.7)

Privacy Privacy• 48.9% to 60.6% ofrespondentsexpressed concernsabout different levelsof access to healthdata.

• 38.4% of respondentsagreed thathealthcare providersare currentlysuccessful inproviding effectivedata security.

Sweden, Slovenia, andDenmark had thehighest proportionsrespondents notconcerned aboutprivacy; highestconference was inLithuanianrespondents.• Anonymised sharingof information withacademic researcherswas not preferred,

1










thus pointingtowards a preferencefor individual-levelbenefits over broaderpopulation-levelbenefits.

• Respondents wereaverse to sharinghealth informationwith third parties,such as insuranceproviders andpharmaceuticalcompanies.

[26],NorthernIreland,SeptembertoDecember2015

NILT survey, univariateand multivariateanalyses


1202 Results of weighteddemographicsAge, years144 (12), 18–24175 (14.6), 25–34172 (14.3), 35–44214 (17.8), 45–54180 (15.0), 55–64310 (25.8), ≥657 (0.6), not answered/refusedEducation224 (18.6), noqualification555 (46.2), school level369 (30.7), graduatelevel54 (4.5), notanswered/ refusedSex, male559 (46.5)


Privacy• 85% agreed datashould be used forresearch whichbenefits society ifanonymous andprivacy weremaintained.

• 95% in favour ofsharing data withinthe health service,over two thirds werein favour of healthinformation beingshared to improveaccess to servicesprovided by othergovernmentdepartments.

• Not willing to sharedata with ‘justanyone’:

• 16% happy foruniversity researchersto use data; 39% didnot want commercialaccess to their data;20% did not carewho used their data

• 74% did not mindhow their data wasused if anonymised;46% believed thateven if the data werede-identified you canstill identify theperson.

• 83% believed thatthe right to privacyshould be respectedover everything else.

• A significant minorityshowed limitedsupport for datasharing.

Transparency• Minimumrequirements fortrust: transparency,knowledge about

0.9










what information isheld and what it isused for, and datasecurity.

• Transparentcommunicationshould help improvetrust.

TrustTrust in keepinformation secureand use appropriately:• 91% GP surgery; 86%NHS; 73%governmentdepartments; 72%academic researchers;51% charities; 41%commercialorganisations.

• 5% of respondentsdid not trust any ofthe organisations.

• 42% had ever hadconcerns about howany of thoseorganisations usedthe information theykept.

• Low levels of trustwere associated witha need for consent touse data.

• Support for datasharing based on:trust in organisations,data protectionmeasures, and publicbenefit.

Levels of trust differedbetween:• Sex – females –generally moretrusting oforganisations.

• Age – ≥55 years lesstrusting ofcommercialorganisations while25–44 years weremore trusting ofacademic researchers.

• Trust included theability to keep datasecure, accuraterecords, capacity tochange or deleteincorrect data, dataused to discriminateagainst individual.

• 50% believed thatcommercialorganisations shouldbe required to










provide more datasafeguards,compared to otherorganisations.

[27], UnitedKingdom,2009

BHPS Wave 18,multivariate bivariateprobit models

Generalmedicine/BHPSWave 18participants

6433 Not reported Privacy Privacy• Attitudes to privacyare related to viewson consenting todata linkage.

• Privacy concernswere the strongestnegative predictor forconsent.

Trust• General trust inothers positivelyaffect consent.

0.65

[28], NewZealand,notreported

Survey, chi-squaretests


203 Age, years106 (56), 18–3469 (37), 31–6014 (7), ≥61Sex, male61 (32)

PrivacyTrust

Privacy• Respondents morewilling to sharehealth information ifidentity removed.

• 60% concernedabout sharing evenanonymousinformation withpeople other thanhealth professionals.

Trust• If an individual’sprivacy requirementsare not met it maylead to a reduction oftrust.

0.7

[29],Canada,March toApril 2005

Survey, responsefrequencies


1230 Age, years480 (39), 18–39504 (41), 40–59246 (20), ≥ 60Education406 (33), HS or less172 (14), some post-secondary492 (40), completedpost-secondary123 (10), post-graduate or profes-sional degreeSex, male554 (45)

PrivacyTrust

Privacy• Protection of privacyof their personalinformation was veryimportant (74%) torespondents.

• 56% increasedconcern aboutprivacy in the past5 years.

• 68% agreed with thestatement: ‘researchthat could bebeneficial to people’shealth is moreimportant thanprotecting people’sprivacy’.

Trust• High levels of trust(26–35%) in disease-based foundations,hospitals, universityresearchers, and datacollectionorganisations.

• High levels of distrustin the insuranceindustry (42%), drug

1










companies (28%),and government(27%).

Qualitative

[30], UnitedKingdom,notreported

Interviews, thematicanalysis wasundertaken using theFramework approach

General medical/individualsincluded in theALSPAC birthcohort study

55 Age, years12 (21.8), 1735 (63.6), 188 (14.5), 19Education7 (12.7), at university25 (45.5), A-levels8 (14.5), GCSE’s12 (12.8), other3 (5.45), noneHealth status, self-reported9 (16.4), disability/long-term illness46 (83.6), no disability/long-term illnessSex, male24 (43.6)


Privacy• Data linkage bestpractice providesenough privacyprotection.

• Threat to privacy wasa potential harm.

• Data anonymisationwas a solution toprivacy concerns.

Transparency• Data linkageprocesses should beclear and transparent.

Trust• A lack of trust abouthow data would beused.

• Trust was associatedwith the need forconsent.

• Important thatresearch does notundermine anindividual’s trust inresearch.

0.95

[31],England,Wales, andScotland(UK), Marchto April2008

Face to faceinterviews, adjustedproportions

National cancerdatabase/general public

2872 Age, years1315 (46), 16–44997 (35), 45–64564 (20), ≥65Education542 (19), Degree orhigher1496 (52), Belowdegree837 (29), NoqualificationsEver had cancer? No2701 (94)Sex, male1319 (46)

Privacy Privacy• 81% would supportlegislation tounderpin theNational CancerRegistry.

• 95% did not believethat a letter fromtheir primary caretrust inviting them toa cancer screeningtest was an invasionof privacy.

• 80% did not considerthe inclusion of theirpostcode or nameand address in theregistry, or thereceipt of a letterinviting them to takepart in a researchstudy, was aninvasion of theirprivacy.

• 2% saw all threescenarios as aninvasion of privacy.

• Small but significantdifferences werefound according tocountry, ethnicity,socioeconomicstatus, housing

0.95










tenure, and theexperience of cancerin the immediatefamily.

• The majority did notconsider theconfidential use ofpersonal, identifiableinformation by theNational CancerRegistry for thepurposes of publichealth research andsurveillance to be aninvasion of privacy.

• The confidential useof identifiable healthinformation forresearch withoutindividuals’ consenthas not damaged thepublic’s trust so far.

[32],Belgium,February2017

Interviews, deductiveanalysis usingQUAGOL

Reuse of clinicaltrial samples anddata/ clinical trialparticipants

16 Age, years35–79, mean 62,median 64Sex, male7 (43.75)Education10 (62.5), highereducation6 (37.5), college oruniversityEthnicity15 (93.5), Belgium1 (6.25) PolishCancer types4 (25), colorectal3 (18.75), ovarian1 (6.25), gastric andlung1 (6.25), colorectal andlung2 (12.5), pancreatic2 (12.5), gastric1 (6.25),cholangiocarcinoma1 (6.25), unreported


Only results about datasharing are reportedData was seen byparticipants to be asimilar resource totissue samples;however, this positionis not supportedlegally where thesamples are notconsidered the same.Privacy2 (6.25) explicitlyspecified thatinformation providedto other researchgroups would becoded, illustrating awish to protect theirprivacy.• If this is protected,then there was awillingness to sharedata for research.

Participants supportedmedical research andthe reuse of data foraltruistic reasons. Mostinterviewed believedthat the data can andshould be reused.However, theparticipants shouldlittle interest in thespecific purpose forreuse.• 2 (12.5) expressed aduty to contribute toscience and wantedthe maximumpotential extractedfrom their data.

0.9










The privacy andsecurity of systems toexchange data isessential.TransparencySome participantswanted to beinformed when theirdata was being usedas this increasesresearch transparency.Participants weremostly positive aboutdigitisation in healthas the use of thesetools may increasecontrol andtransparency and allowfor greaterparticipation inresearch.TrustParticipants weregenerally trusting ofthe initial researchteam.• 1 (6.25) participantnoted that furtherresearch wasacceptable, if it ‘stayswithin the oncologyarea’.

Responses showedthat participantstrusted their cliniciansto use the datacorrectly in theirdisease setting.General trust in thehealth services andclinicians.14 () trusted the ethicscommittees to be thedecision maker aboutaccess to data.Distrust• 2 (12.5) participantdistrusted ‘unknown’researchers,particularly thepotential for datamisuse and security.

1 (6.25) participantexpressed his distrustin electronic systemsgenerally as theybecome alternatives totraditional provision ofcare.

[33],England,September2015 toDecember

Interviews and onlinesurvey, thematicanalysis

HumanFertilisation andEmbryologyAuthorityregistry/fertility

60 (20,interview, 40,online survey)

Interview populationAge, years36 median, 30–46rangeEthnicity


Interview population14 (70) agreed toshare data2 (10) refused to sharetheir data

0.9










2017 clinic attendees 16 (80, British whiteSex, male5 (25)Occupation14 (70), managerial orprofessional2 (10), intermediate3 (15) routine ormanual1 (5), student

3 (15) were unsureabout sharing data1 (5) agreed anddisagreed with datasharing at differenttimesOnline survey32 (80), agreed toshare data4 (10), refused to sharetheir data2 (5), were unsureabout sharing data2 (5), agreed anddisagreed with datasharing at differenttimesPrivacyDifferences in sharingwere seen byprofessional status,with those inintermediate ormanual or routine jobsbeing more likely toshare data (interviewparticipants only).Sharing data for thegreater good wasimportant to some,however othersbelieved that it maycause harm (fraud,identity theft, targetedfor marketing).• Documentationneeds to clearly saythat the data willonly be used forresearch and notmarketing.

• Concerned aboutharms not just tothem, but also theirchildren.

• Not sharing is amechanism toprotect themselves.Stigma around IVFtreatments.

• HFEA data is requiredby legislation;however, someexpressed concernthat that therequirement tocollect this type ofdata is not requiredin fertile couples.

TransparencyMany were not surewhat they had alreadyagreed to in respect tosharing data.










Variation in theunderstanding ofanonymisation andidentifiable data.TrustMany already thoughtthat data was alreadyshared for researchpurposes.Trust in partners,clinics, hospitals, andwider institutions wasnoted.Misconception aboutthe data being shared,some believed that itwas NHS data which isgenerally seen as atrustworthy datacustodian.• This trust was notuniversal as somehad lots trust afterrecent data beachesby this organisation.

Trust in the fertilityclinic was reduced asthey were seen asbusinesses.

[34],Scotland(UK), Mayto June2009

Focus groups,thematic analysis


19 Age, years1 (5), <6015 (79), 60–743 (16), ≥75Numbers taken partin medical research6 (32)Numbers withchronic healthcondition13 (68)Numbers withloyalty cards15 (79)Sex, male6 (32)

PrivacyTrust

Privacy• Self-worth was articu-lated as a question ofrights (i.e. ‘Isn’t therea right to privacythough?’).

• Beliefs that anindividual has a‘natural right toprivacy’ werebalanced with agenuinecommitment tosupporting medicalresearch.

• AS belief that theinstitutions andresearchers to keeptheir medical recordsprivate andconfidential, wasassociated with adecreasedrequirement forconsent.

Trust• Respondents acutelyaware of previousinformation securitybreaches, howevercontinued to displaya high level of trustin the organisationsstaff.

• Trust was highest for

0.85










clinicians involved ina person’s care butwas extended toresearchers moregenerally.

• Concerns aboutcontrol and accessand a generalisedscepticism andmistrust of thegovernment andlarge commercial andinsuranceorganisations.

[35], USA,March1996 toFebruary2000

A single structuredinterview, Pearson’schi-square test for in-dependence, logisticregression models(binary responses) andordinal logisticregression

General medical/individuals withgenetic orchronic healthcondition* (or afamily memberwith condition)

602 DiseaseParticipants wereequally divided(approximately) intothe six diseases. BCand CC eachcomprised of 50participants with apersonal history of thedisease and 50 with afamily member withthe disease.

PrivacyTrust

Privacy• 2% thoughtcomputerised recordswere a negative dueto privacy violations.

• Those who agreedwith anonymisedcomputer records,believed that itwould advancemedical researchwithout violatingprivacy.

Trust• Some did not trustthat computerdatabases set upanonymously forresearch would besecure.

• If individuals trust theentity asking to usethe data, they maybetter trust that therecords will remainconfidential and willbe used forworthwhile purposes.

• Important that theintegrity of medicalcare and researchenterprise ismaintained, andpatients trust theirphysicians andmedical institutions.

• Institutions have aresponsibility to taketrust seriously, andnever assume asimple right toconduct researchwith private,identifiable data.

1

(28),Australia,February toDecember2006

Focus groups,thematic analysisSurvey, chi-square testof independence


723 (23, focusgroup700, survey)

Age, years (surveypopulation)35 (5), 18–19138 (19.7), 20–34141 (20), 35–44

Privacy Note: Results fromsurvey populationPrivacy• 98% supportive ofgeneral medical

0.95










208 (29.7), 45–59178 (25.4), ≥60 yearsEducation (surveypopulation)66 (9.4), not finishedHS159 (22.7), finished HSbut no HSC131 (18.7), finished HSand HSC17 (2.4), sometechnical orcommercial/ TAFE24 (3.4), finishedtechnical orcommercial/ TAFE13 (1.8), someuniversity/ C.A.E.85 (12.1), tertiarydiploma15 (2.1), now atUniversity/ C.A.E.145 (20.7), university/C.A.E. degree40 (5.7), post-graduatedegree

research; 73% wouldbe happy to sharetheir health data forresearch; 12% wouldnot share their data;14% were not sure.

• 37% concernedabout sharing theirhealth informationfor research; 33%were not concerned.

• The youngest andoldest age groupswere less concernedabout privacy (51.4%and 53.4%respectively)compared to theother age groups(range between 71%and 76%).

• Those with poorerhealth were lessconcerned aboutprivacy compared tothose in good health(63.8% vs 67.7%). Ofthose who did notcomment on healthstatus 93.3% hadconcerns aboutprivacy.

• Even with additionalsecurity measuresand anonymiseddata, 25% remainedconcerned aboutprivacy.

• Those not employedwere less privacyconcerned (60%)than respondentswho were employed(range 66–74%).

• Most concerns aboutdata sharing relatedto: sexuallytransmitted diseases;issues of abortion,infertility; familymedical history/genetic disorders;drug/alcoholincidents; mentalillness; list of previousoperations/procedures/dates;and currentmedications.

• For cancer: 12% werevery concerned,15.7% wereconcerned, 9.7%were slightly










concerned, while62.3% were notconcerned.

• Most want to beasked for theirpermission beforetheir healthinformation is usedfor any purposeother than medicaltreatment (92%), andthey would like toknow theorganisation anddetails of theresearch beforeallowing the use oftheir health records(83%).

• Many not aware thatremoving namesfrom medical recordsdoes not guaranteeconfidentiality.

• Insurances thatsecurity measuresprotectconfidentiality ofpersonal healthinformation are inplace are important,as are reassurancesthat this will be doneconsistently andreliably.


Focus groups,grounded theoryapproach

Arthritis andother chronicconditions/hospital

23 Age, years, mean59 (SD ±13), range36–84Sex, male5 (21)Chronic Illness10 (43.5)

PrivacyTrust

Privacy• Confidentiality wasnoted as a concerngenerally for patients.

Trust• Trust, distrust, andconfidentiality wereinfluential and aconsideration inpatients’ viewstowards researchregistry participation.

• A trusting doctor-patient relationshipmight be an import-ant factor influencingregistry participation.

0.95


Focus groups,emergent contentanalysis

Generalmedicine/general health

30 Age, years1 (3), 18–304 (13, 31–404 (13), 41–508 (27), 51–604 (13), 61–706 (20), 71–801 (3), ≥80Sex, male14 (47)Education


Privacy• Some respondentswere concernedabout the loss ofprivacy.

• Respondentspreferred that theirelectronic data beaccessed over paper-based records.

• Concerns regarding

1










11 (37), some HS7 (23), HS7 (23), some college3 (10), collegeEthnicity4 (13), white5 (17), black20 (67), Latino2 (7), other

privacy violationsrevealing highlypersonal informationand third-party ac-cess to data wereexpressed.

• If no permission touse data sought,some considered itan invasion of theirprivacy.

• Using the data tosupport researchrecruitment was seena violation of privacy.

TransparencyDevelopment oftransparent policiesand practices willcontinue to supportthe secondary use ofhealth data.Trust• There is apsychologicalcomponent ofuncertainty andmistrust.

• Lack of trustregardingvolunteering forresearch.

• Participants wantedto have face to facecontact withresearchers duringthe recruitmentprocess to increasetrustworthiness.

• Researchorganisations andinstitutional reviewboards need todevelop outreachprogrammes toengage communitiesand build trust.

[38], USA,2006 and2008

Interview (telephoneand enhanced face toface), multilevelrandom effects logisticregression

Generalmedicine/healthand retirementstudy

6384 Age, by birth cohort747 (11.7), <1923428 (6.7), 1923–19303543 (55.5), 1931–1941792 (12.4), 1942–1947875 (13.7), 1948–1953Sex, male2522 (39.5)Ethnicity5235 (82.0), white875 (13.7), black275 (4.3), otherEducation1481 (23.2), 0–11 years2190 (34.3), 12 years1334 (20.9), 13–15

Privacy Privacy• Concerns regardingprivacy andconfidentiality werereflected in rates ofconsent to datalinkage (LR 43.48,p < .01)

• As the rate ofconcerns regardingprivacy andconfidentialityincreases, lower ratesof consent are seen.

• Where respondentsrefused to answerquestions regarding

1










years1379 (21.6), ≥16 years

their finances,consent for datalinkage decreases.

The hypothesis thatprivacy andconfidentialityconcerns arenegatively related toconsent wassupported.

[39],Australia,notreported

Interviews, frameworkapproach

Generalmedicine/general public

26 Age, yearsBetween 24 and 41Education3 (12), ≤ Year 126 (23), TAFE16 (62), tertiary1 (4), post-graduateSex, male6 (23)

PrivacyTrust

Privacy• The protection ofprivacy was reflectedin the need forconsent, withoutwhich privacy mightbe breached.

• Some believed de-identified informationdoes not breachprivacy.

• Many not concernedabout use ofidentifiable data tocreate the linkagekey. This mayindicate that viewson privacy arechanging.

• Conflicting values,notably betweenprivacy protectionand public benefits.

Trust• Concerns aboutidentifiability whenthe linkage beingdone by researchers;not a lack of trust inresearchers. Someconcerns about thepotential for a personknown to them tofind out privateinformation.

• Lack of trust thatinformation wouldremain anonymous.

0.9

Mixed methods

[40], USA,November2003 toJune 2004

Deliberative sessionsand surveysQuantitative - chi-square for categoricaldata and ANOVA forcontinuous variablesQualitative-contentanalysis

General medical/veterans

217 Characteristics ofdeliberators providedAge, years—mean(SD)65 (12)Education80 (37), BS or BA orhigherSex, male206 (95)


Note: Results fromdeliberators providedPrivacy• Compared touniversity researchers(75%), participantswere less inclined togive permission for alocal hospital to usetheir medical recordsfor a preventivehealth programme(61%) and even fewer(51%) were inclined

0.9










to give permission toa drug company formarketing purposes.

• Concern regardingthe protection ofprivacy, particularly ifenough was beingdone to protect theirprivacy.

• Many conformablewith using theirrecords for research iftheir informationstayed within the VA.

• Participants wantedto know what datawas being used andfor what purposes;particularly how thisinformation would beused to help otherveterans.

• 75% were unawarethat in the USmedical recordscould be usedwithout permission.

• Concerns regardingaccess to informationabout stigmatisingconditions wereraised, particularlyHIV/ AIDS or mentalhealth illnesses.

• Concerned about thepotential of their databeing sold.

Transparency• Even those with ahigh level of trust inthe VA, want to befully informed abouthow their recordswere being used forresearch and anyconclusions from thisresearch.

• Ensuring medicalrecords are beingkept private andconfidential in a waythat patients can seeis important.

Trust• At follow up survey(4–6 weeks afterbaseline): 32%believed that medicalresearchers at a VAhospital wouldalways keep theirinformationconfidential andprivate (change from










37% at baseline);compared to 8% forhealth insurancecompany researchersand 9% forpharmaceuticalcompany researchers.

• Respondents whotrust the VA werelikely to recommenda less-stringentprocess for obtainingconsent.

[41],England,June toJuly 2016

Surveys andinterviews, notdescribed

Cancer registry/cancer patientsand non-cancerpatients, cancer

2033 (1033with cancer,1000 generalpublic)

Age, years, cancergroup31 (3), 18–34155 (15), 35–54847 (82), ≥55Age, years, generalpublic290 (29), 18–34350 (35), 35–54350 (35), ≥55Cancer status,cancer group only186 (18), localised/stable31 (3), advanced671 (65), remission/cancer freeCancer type, cancergroup only52 (5), bladder93 (9), bowel/colorectal258 (25), breast134 (13), prostate62 (6), cervical/womb155 (15), skin300 (29), all othersFamily or friend whohas/had cancer,general publicgroup only640 (64), yesSex, male475 (46), cancer group490 (49), generalpublic

PrivacyTrust

Privacy• Privacy concerns:12% in the cancergroup and 9% in thegeneral public group.

• Practical aspects ofprivacy: identifiableinformation beingincluded (6% in thecancer group; 2% inthe general publicgroup) and concernover third partieshaving access to data(6% in the cancergroup and 2% ofgeneral public).

• Respondents withcancer who opposethe current methodof collection, but thatsupport collection ofcancer data ingeneral were mostlybeing put off by thelack of consent andlack of informationrather than privacyconcerns.

• The trade-off be-tween data collectionfor improved cancerservices and treat-ment versus no datacollection to maintainprivacy and security;support for thistrade-off was highacross both groups.

Trust• Concern regardingtrust in organisationsto maintain datasecurity was noted ina small group ofrespondents.

0.6

[42],Australia,notreported

Focus groups andsemi-structured inter-views, open codingand NVivo analysis

Epidemiologicalresearch/generalpublic andexpert

45 (calculatedbased on thebelow)(4 focus

Not reported PrivacyTransparencyTrust

Privacy• Respondents wantedto know where theirnames and contact

0.9










stakeholders groups withgeneralpublic (4 to 8persons pergroup)2 focusgroups withAboriginaland TorresStrait Islanderpeoples (4 to8 persons pergroup)5 peoplefrom diverseculturalbackgrounds20 expertstakeholders)

details were sourcedfor studies.

Transparency• Wanted to know thesource of funds forresearch, indicating adesire fortransparency and theability to judge themerits and motivesfor the research, toknow who hasvested interests in itsfindings.

• Concern about a lackof transparency fromresearch undertakenby pharmaceuticalcompanies (forexample notpublishing adverseresults).

Trust• Prepared toparticipate inepidemiologicalresearch, particularlyif it is conducted bya trusted publicinstitution(government healthdepartments,charities, universities).

• Trust is critical to theconduct of researchand it is importantthat researchinstitutions do notact in ways thatbetray trust.

• Widespreadcommunity distrustof researchconducted orsponsored bypharmaceuticalcompanies andprivate companies.

• Pharmaceuticalcompanies wererepeatedly singledout by participants –concerns regardingtheir motivationswere raised (i.e.profits).

[43],England,April toMay 2013

Focus groups andinterviews notdescribed


50 Age, years18–70

PrivacyTransparency

Note: Telephoneinterviewees weredefined as the ‘pro-privacy group’ and wererespondents who wereespecially cautiousabout releasingpersonal data and

0.9










actively taking somemeasures to protectagainst doing so. Thefocus grouprespondents wererecruited as owners ofproducts such as aloyalty store card.Privacy• Respondents fromthe telephoneinterview group weremore fearful of ‘whatifs’ and were cautiousin releasing personaldata because theyworked in industriesresponsible forhandling personaldata (e.g. banking,insurance, sellingdatabases).

• Concerns regardingthe potential for datato be lost, stolen,hacked, or leaked,and shared withoutconsent.

• Invasion of privacy,with a sense of BigBrother watching;incorrect orinaccurate datacollection, whichwould be hard tocorrect and undo;potentialdiscrimination (e.g.data falling into thehands of anemployer).

• Strong feeling thatpersonal health dataare confidential,private, and sensitive,and should not beshared outsidesecure, authorisedbodies, and especiallynot with privatecompanies such asemployers, insuranceproviders and drugmanufacturers.

• Mental health datawas sometimesregarded asparticularly personaland sensitive.

• Some concernregarding thepotential for futurediscrimination beingintroduced, or was










already beingpractised, within theNHS (e.g. low priorityon waiting list; beingrefused treatmentuntil lifestyle changesare made).

• Concerns regardingpotentialprivatisation; healthdata may bevulnerable to outsideparties.

• Possibility ofindividualidentification was acause for concern.

• Fears about research/clinical trial data werelow and related toanonymity being lostand possibleunwanted mediaattention.

• Data linkage wasmore acceptable ataggregate levelrather than individuallevel.

Transparency• Clarity, transparency,and reassurancerequired whenaddressing issues oflinking and use ofpersonal data.

[44], GreatBritain,NovembertoDecember2015

Deliberativeworkshops and face toface interview, notreported

Commercialaccess to healthdata/generalpublic, doctors,individuals withchronic or raredisease

2263 (246focus groups2017interviews)

Not reported PrivacyTransparencyTrust

Privacy• Clear benefit both toindividuals and towider society was theonly ‘good’ rationalefor breaching privacy.

• Different data typescame have differentprivacy expectations.

• New ways ofcollecting andsharing data maygive rise toconflictingexpectations arounddata privacy.

• Concepts of privacyare in flux.

• Privacy should be thedefault position andthat any sharingwould infringe thatprivacy.

• Data anonymisationwas a potentialsafeguard.

• Permission for acommercial access

0.95










was noted as aconcern somerespondents feel thatthey are somehowlosing their privacy.

• Respondents withpessimistic dystopianoutlooks were notonly concernedabout the potentialnegative effect onthemselves but fearlarge-scale negativeimpact for all society.

Transparency• Underlying concernsfor equity,transparency, andindependent scrutinyof research by bodiesfree from vestedinterest. Restrictingthird-party access,and transparency insharing results andpublishing analysis.

• Transparency wasuniversally important.

Trust• Motivations ofcommercialcompanies werequestioned, and theprivate sector wasmistrusted in general.

• When benefits areperceived (and theorganisation istrusted) allparticipants acceptedcommercial access tohealth data inprinciple.

• The nature of thesafeguards is not asimportant as the trustthat comes withknowing there is asafeguard in place.

• Education onaggregation andanonymisation isrequired to buildpublic trust.

• 20% of respondentssaid commercialorganisations cannotbe trusted to storethe data safely, 18%say that profit shouldnot be made fromtheir NHS healthdata, even if thereare potential societal










and health benefits;13% fear that acommercialorganisation mightsell the data on toanother commercialorganisation.

• Some lackedunderstanding of, ortrust in,anonymisation, andalso did not knowhow data is actuallyheld in the healthservice.

• Access to data byinsurance companyto adjust premiumswas met withuniversal disapproval.

• If respondents didnot trust theorganisationconducting theresearch, they calledinto question theultimate publicbenefit.

• Only allowing namedindividuals within anorganisation toaccess health datahas had little traction– the individual’smay not betrustworthy.

• Some have limitedtrust in commercialorganisationsundertaking researchand not only fear nopublic benefit butallowing access tohealth data willcreate new publicharms.

[45],Canada,notreported

Public dialogues andsurvey, ANOVA andMANOVA


98 Age, years37 (38), 20–3935 (36), 40–5926 (27), ≥60Education26 (26), HS or less18 (18.7), some post-secondary42 (42.7), completedpost-secondary12 (12.5), post-graduate or profes-sional degreeSex, male40 (40.8)

PrivacyTrust

Privacy• Providing informationabout research allowsindividuals to feel likethey are contributingwhile respectingprivacy.

• Some respondentswere less privacyconcerned.

Trust• Accountable systemsfor managing datacan increasetrustworthiness.

0.5

[46],Canada,

Survey and focusgroups, regression

Generalmedicine//DM,

1780 (n = 1137)Age, mean

PrivacyTrust

Completed survey n =403

0.9










November2006 toJuly 2007andSeptember2007

analysis usinggeneralised estimatingequations

HT, chronicdepression,alcoholism, HIV,BC, LC, andgeneral public.

54 yearsSex, male765 (43)Education(33), HS or lessSelf-describedhealth587 (33), poor or verypoor712 (40), fair481 (27), good orexcellent

PrivacyRespondents whowere more privacysensitive were lessinclined to participatein the study.Disclosure concernsdiffered across healthconditions.Respondents whocompleted the surveyvia telephone wereless privacy concernedcompared to thosewho completed itonline.TrustParticipants trustedthat their informationwould be usedappropriately.Higher levels of trustwere placed in theindividual (doctors)(64%), as well ashospitals (43%); thiswas higher than foruniversity researchers(28%).Lowest rates of trustwere recorded for:pharmaceuticalcompanies (9%),provincialgovernments (9%), andinsurance industries(6%).Research undertakenfor profit or linking ofincome, education oroccupation had lowerrates of consent.Trust is reduced in theevent of high-profiledata breaches.

Other

[47], NewZealand,notreported

Citizens’ jury, notapplicable

Pharmaco-epidemiologicalresearch/generalpublic

13 Age, years18–65 years (7 (54)were 45)Sex, male6 (46)

Privacy Privacy• Given appropriateprivacy safeguards,an informed publicdoes not alwaysplace personalprivacy abovesocietal benefits.

• A balance betweenthe public interestand medicine safetyis important.

• The use and linkageof medicalinformation forresearch on medicinesafety is warranted

1










given the existingprotections and theminimum ofidentifiableinformation.

The jury wascomfortable with thesmall loss of privacy tosupport public goodand safety…

[48], UnitedKingdom,notreported

Citizens’ jury, notapplicable


34 Age, years8 (23.5), 18–2910 (29.4), 30–4410 (29.4), 45–596 (17.6), ≥60Sex, male17 (50)Education13 (38.2), noqualification11 (32.4),apprenticeship orother qualification10 (29.4), degree levelor above


Privacy• Public benefit was akey justification foraccess.

• Where data was usedfor public benefit(such as improvedmedical care andtreatments, improvedpublic health, ormanagement ofpublic funds) andorganisations made aclear and compellingcase for why theyneed these patientrecords, accessshould be granted.

• An individual’s rightto privacy should notprevent research thatcan benefit thegeneral public.

Privacy statements:a. We should share allthe data we canbecause it benefits theservices and me—aslong as I can opt out ifI chooseb. We should notshare data as the risksto people’s privacyand security outweighthe benefits19 (55.9), agree morewith a than b.11 (32.4), agree morewith b than with a.4 (11.8), agree equallywith both or don’tagree with either ordon’t know.Transparency• Transparency aboutdata use and accesswas important.

• Involving individualsin the decisionregarding data usewould increasetransparency.

Trust• Important to show a

0.8


appropriately [48]. Ensuring researchers act in the pa-tient’s best interest and clearly and transparently disclos-ing the research being undertaken also built trust [40].Respondents were generally trusting of the original re-search team and they trusted that they would use theirdata appropriately [32]. In a study about the use of fertil-ity data, many respondents believed that registry datawas already used for research purposes thus showing anestablished trust in the clinic, hospital, and wider healthinstitutions [33].The ability to maintain data security, privacy, confi-

dentiality, and accurate records, change or delete incor-rect data, and ensure that data would not be used todiscriminate against an individual, all contributed tolevels of trust [26]. Granting access to a small number ofnamed individuals was not seen as a solution to resolv-ing privacy concerns, as these individuals themselvesmay not be trustworthy [44]. Any research undertakenusing secondary data analysis must not undermine or









clear, relevantconnection betweenthe research questionand the informationcontained in therecords. Someorganisations had atrack record ofprotecting data andrecords and could betrusted to maintaincontrol of data.

• Some believed thatorganisations couldnot be trusted tomaintain recordsappropriately.

• Organisations mayuse the data toexploit or manipulateindividuals orpopulations or mightmanipulate the datato support their ownagenda.

• Providing moreinformation on datause does not increasepublic trust.

AIDS acquired immune deficiency syndrome, ALSPAC Avon Longitudinal Study of Parents and Children, BC breast cancer, BHPS British Household Panel Survey,CAE Centre for Adult Education, CATI computer assisted telephone interviewing, CF cystic fibrosis, CC colon cancer, DM diabetes mellitus, ED emergencydepartment, EU European Union, GCSES General Certificate of Secondary Education, GED general educational development, HFEA Human Fertilisation andEmbryology Authority, HIV human immunodeficiency virus, HKID Hong Kong Identity Card number, HS high school, HSC high school certificate, HT hypertension,LC lung cancer; LR likelihood ratio, MS multiple sclerosis, NHI National Health Index, NHS National Health Service, NILT Northern Island Life and Times, NZ NewZealand, SCD sickle cell disease, PS primary school, QUAGOL Qualitative Analysis Guide of Leuven, SS secondary school, TAFE Technical and Further Education, VAVeterans Affairs, UK United Kingdom, USA United States of America*Adults or parents of children with CF, or adults or parents of children with SCD, or adults or parents of children with DM, or adults with HIV, or adults with BC, oradults with CC

Table 3 Studies by country

Country study undertaken (inalphabetical order)

Number ofstudies

Reference

Australia 3 [39, 42, 49]

Belgium 1 [32]

Canada 5 [20, 23, 29, 45, 46]

England 1 [33]

Europe 3 [16, 17, 25]

Finland 1 [18]

Hong Kong 1 [22]

Great Britain 1 [15]

Japan 1 [19]

New Zealand 2 [28, 47]

Northern Ireland 1 [26]

United Kingdom 8 [27, 30, 31, 34, 41,43, 44, 48]

United States of America 7 [21, 24, 35–38, 40]


compromise an individual’s trust in medical research[30]. The level of respondents’ education influencedtheir view of trust, with a higher level of education beingmore trusting of their government and research institu-tions compared to those who finished their educationearlier [16]. In the setting of fertility, most respondentswere willing to share their data, suggesting trust in theorganisation and registry [33].

DistrustIn contrast, the theme of distrust was noted in severalarticles representing a total of 6830 respondents and in-cluded data from questionnaires, surveys, focus groups,and a citizens’ jury. A general distrust in the health sys-tem, research, and sharing of health information [30, 36]was noted, with some respondents not trusting any or-ganisation with their data [26] or the organisational cap-acity to maintain records appropriately [48]. While therewas a desire to support the use of anonymised healthdata for research purposes, concerns regarding trust inthe systems and data security remained [34, 40]. Theprovision of information on the source of research fund-ing [40, 42] and data management systems [45] can in-crease transparency and trust, but providing moreinformation on data use does not necessarily increasepublic trust [48]. In a study from the UK, some individ-uals with a ‘pessimistic dystopian’ mindset had limitedtrust in commercial organisations accessing health data,believing it would create new harms [44], with somesuggesting that organisations may use the data inappro-priately (exploit or manipulate individuals or populationsor might manipulate the data to support their ownagenda) [48]. Access to information by pharmaceuticalcompanies and insurance agencies had lower levels ofsupport, suggesting a distrust in these organisations.Older respondents (≥ 65 years of age) showed less trustin these organisations compared to younger respondents(≤ 25 years of age) [16]. Respondents who believed thatdata sharing had more negative than positive effectswere more likely to have a college education [21]. Gen-erally, these respondents believe that people could notbe trusted and were concerned about data reidentifica-tion and information theft [21]. These low levels of trustwere associated with a decreased willingness to sharedata with both for-profit and non-profit organisationsalike [21]. Sharing data with an ‘unknown’ researcherwas also associated with distrust; further, some believedthat the increased digitisation of healthcare would leadto a decrease in the traditional provision of care [32]. Inthe setting of fertility, respondents’ levels of trust de-creased given some respondents saw them as a business;it is essential that the information provided to peopleclearly state the purpose of data reuse and should notethat it would not be used for purposes such as marketing

[33]. Lucero et al. noted that there is a psychologicalcomponent to uncertainty and mistrust. This leads to adistrust in volunteering for research and the need for or-ganisations and ethics review boards to engage withcommunities to build trust [37]. To decrease distrust, re-spondents wanted to have face-to-face contact with re-searchers during a study’s recruitment process [37].

Privacy and confidentiality: differences according todemographic and health characteristicsA total of 44,366 respondents provide a view on privacyand confidentiality. Responses were obtained throughsurveys, deliberative workshops, dialogues and inter-views, and questionnaires.

General concerns about privacy Concerns about priv-acy and confidentiality were one reason for not sharinghealth data [16, 28, 29, 36, 37, 40, 44]. One study notedthat the respondents’ concerns about privacy had in-creased over the past 5 years [29]. Concerns about thesense of ‘big brother’ and the potential for data to beused to discriminate [43] were expressed, with someconsumers expressing a belief in the natural right (notdependent on law or custom) to privacy [34]. Wheresafeguards were in place to protect the data, most re-spondents in one study were willing to share their data,irrespective of the proposed data use [21, 32], except inthe setting of litigation [21].

Demographic characteristics The inclusion of an indi-vidual’s postcode, name or address, and receiving a letterinviting them to participate in research from the cancerregistry was not considered to be a breach of privacy[31]. In a study of UK respondents, no substantial differ-ences in privacy concerns were found according to sexor age; however, small but significant variations werenoted by factors such as education, ethnicity, socioeco-nomic status, and an experience of cancer in the imme-diate family [31].In other studies, the relationship between age and sex

and concerns regarding trust and privacy were contra-dictory. Younger respondents expressed higher levels oftrust in researchers and were more willing to let theirdata be used for research, but they also had high levelsof privacy concerns [18]. Conversely, other studies notedthat older respondents were more likely to agree to datalinkage [22], while respondents aged 18 to 19 years andover 60 years had lower levels of privacy concerns com-pared to other groups [49].Levels of concern about privacy were also influenced

by the respondent’s level of education and employment.Those with commercial or technical qualifications hadmore concerns regarding privacy compared to all othereducation groups and those with a post-graduate degree


had the fewest privacy concerns [49]. One article consid-ered privacy in the context of sociotechnical capital,composed of awareness of privacy, attitudes towards theimportance of privacy and data sharing, and confidencein the ability to maintain privacy [24]. Individuals withhigher levels of education and income had higher ratesof health privacy capital [24]. In a second study, respon-dents who were employed in manual, routine, or inter-mediate work were more likely to share their datacompared to those in professional roles [33]. Respon-dents employed by a government organisation weremore concerned about privacy [49]. Respondents whodid not respond to finance questions had lower rates ofconsent for data linkage [38]. Other influences on priv-acy included social networks [24]. In one study, differ-ences in rates of privacy concern between those whoanswered the survey online compared to those who an-swered via telephone were noted; those who answeredby telephone were less privacy concerned [46]. Some dif-ferences in privacy concerns were noted by country. Astudy of European respondents found those based inSweden, Slovenia, and Denmark were less concernedabout privacy concerned compared to respondents fromLithuania [25].

Health status Health status also impacted privacy con-cerns. Respondents in good health were more likely toagree to the use of data in healthcare registries com-pared to those in poor health [18]. Nevertheless, in thesetting of no additional digital security measures (re-stricted access, etc.) being applied, individuals with poorhealth were less concerned about privacy compared tothose in good health [49].

Sensitivity of dataA total of 3347 respondents provided a view on sensi-tivity of data. Individuals may consider some forms ofmedical data to be more sensitive than others. Datarelated to sexually transmitted diseases, family med-ical history including genetic disorders, drug and alco-hol use [49], and mental illness [43, 49] raised themost privacy concerns, particularly the possibility ofinappropriate data access [23, 49]. A UK report notedthat ‘new ways of collecting and sharing data, undernew circumstances, can give rise to conflicting expec-tations around data privacy’ [44] and that differenttypes of data came with different privacy expectations[44]. In one study, respondents believed that data wasa similar resource as tissue samples, suggesting thatdata is equally as sensitive as biospecimens [32]. Astudy of respondents who were seeking fertility ser-vices found that while there was a willingness toshare data, they were concerned about the potentialfor the data to cause harm (potential for stigma), not

only to them but also their children [33]. Further,some respondents were concerned that the data col-lected on them, while required by legislation, was notcollected on fertile couples [33].

Control of dataA total of 6859 respondents provided a view on controlof data; results were obtained from surveys, a nestedcohort, focus groups, and a citizens’ jury. Individuals’desire to maintain some control over their health datawas evident across studies, with many seeing this as keyto transparency [40, 48]. Respondents were selectiveabout those with whom they were willing to share theirdata. Respondents to two UK studies preferred theirdata to stay within the NHS [34, 43], with some believ-ing that once data left the organisation control wouldbe lost [34]. Health data access by private/commercialorganisations [21, 26, 37, 41–44], pharmaceutical com-panies [21, 42, 43], and health insurance companies[21, 43, 44], all were seen as inappropriate. Respondentswere concerned that insurance companies would usehealth data to adjust premiums which was consideredinappropriate and without clear public benefit [44]. Notallowing third parties to access their data was based ona distrust of these organisations [42, 44], perceived lackof transparency from research conducted by pharma-ceutical companies [42], concern about the companies’motivations (e.g. profit, marketing) [40, 42, 44], doubtsabout data security, distrust in their capacity to put so-ciety before profit, and a belief that a commercial or-ganisation may on-sell their data [44]. In one study,some respondents indicated they would prefer that re-search not be undertaken if it required allowing com-mercial access to health data; however, mostrespondents wanted third-party access to health data ifdisallowing this resulted in research not being under-taken [44]. In contrast, other studies found respondentswere happy to allow pharmaceutical company access totheir registry data if it is undertaken in a transparentmanner [17], and where consent is sought, respondentsin a second study accepted the principle of commercialaccess to health data [44]. Respondents to one surveysuggested that increased sharing may be used for mar-keting purposes, stolen, used for profit, or to discrimin-ate against an individual [21]. Further, somerespondents wanted to be informed when their datawas being used [32]. Digitisation of health data wasseen by some as a mechanism to increase control andtransparency over their data, and increased participa-tion in research [32]. Some believed that consent wasrequired to access records for research, or to identifypotential research participants, without which it was aviolation of their privacy [37].


Benefit to societyThe importance of research and its benefit to societywas noted as important in several studies with a total of7006 respondents. It was noted that society’s views onprivacy may be changing, creating conflicting values be-tween privacy protection and public benefit [39]. Gener-ally, respondents were positive about sharing their datafor research [18, 20, 49]. In some circumstances, societalbenefit may outweigh concerns regarding privacy [29,44, 47]; further, research using health data should have asocietal impact and not be undertaken just for academicreasons [40]. Ensuring transparency about the publicbenefit of research and sharing of results and analysis ata study’s conclusion [44] were important. Where datawas used for public benefit, such as improved medicalcare and treatments, improved public health, or manage-ment of public funds, and organisations made a clearand compelling case for access to the data, access shouldbe granted as it could potentially benefit both the indi-vidual and the health service [48]. Public benefit wasseen as a justification for access to health data and an in-dividual’s right to privacy should not prevent researchthat could benefit the general public [48]. Altruism wasalso noted as reason to support health research usingexisting data, with some wanting their data to be used to‘maximum potential’ [32]. In the setting of fertility data,sharing for the greater good was important to some [33];however, this was not universal as some believed that itincreased the risk of harm (fraud, identity theft, targetedmarketing) [33] and that the premise that public benefitoutweighs privacy concerns was not supported [18]. Inone survey, some respondents valued maintaining indi-vidual control over their data more than societal benefitand respondents expressed a lack of willingness to tradeloss of privacy for public good [23]. This was echoed ina second study where the majority of respondents be-lieved that the right to privacy should be respectedover all else; however, respondents also believed thatif the data were made anonymous and privacy wasmaintained, data should be used for research thatbenefits society [26].

Views about specific data sharing scenariosDigitisation of health records

Data linkage A total of 700 respondents provided aview on health data linkage. An Australian study identi-fied concerns relating to confidentiality during the datalinkage process, specifically the possibility that the indi-vidual making the linkage may know the person and findout confidential information about them, although thiswas not universal [39]. The use of de-identified data wasnot seen as a breach of privacy [39] and that the currentdata linkage best practice provides sufficient privacy

protection [30]. Transparency about process and datausage was an important factor in individuals’ decisions toallow data linkage [43]. The importance of ensuring priv-acy when undertaking the linkage of clinical trial data tohealth administrative data was seen as important [20]. Asurvey of UK respondents found that they were not con-cerned about health record linkage as long as the data wasused to increase health knowledge, consistency betweenhealth services, and administrative efficiency [43].

Registries and patient-provided data A total of 25,814respondents provide a view on registries and patient-provided data. Studies indicated that use of health infor-mation exchanges and digital health platforms can im-prove care [25]. While these positives were noted, viewson use of these technologies varied widely [36].While respondents agreed with the principles of elec-

tronic data sharing, they desired transparency and amechanism for independent scrutiny of data access anduse [44]. Some respondents indicated a preference formore electronic health data sharing [25].There was a high level of trust in using data from dis-

ease registries [17] and a willingness to share data; in acancer setting, only a small number of respondents wereopposed to data collection [41]. The most common con-cern about registry-based research was the protection ofprivacy [18]. One mechanism suggested to ensure max-imum transparency in registry research was to involvepatient organisations in the development of clinical trialsand registries [17].

Strategies to address privacy and trust concernsData securityA total of 25,052 respondents provided a view on datasecurity. Many health organisations already have well-established protocols to ensure patient privacy. Trans-parent information about data security and protectionmeasures is important to maintain trust [26, 32] withsome suggesting that systems to protect confidentialityshould be more secure for shared data than those usedin usual medical practice [40]. Some respondents wereparticularly concerned about unauthorised access totheir data [40] particularly the chance for data to be lost,stolen, or ‘hacked’, or shared without consent [43], al-though trust in researchers and health care providers tomaintain data security [25] remained. Some respondentsin one study were happy to ‘trade-off’ any potential riskto privacy and security for benefits, like improved treat-ment and services [41]. Breaches in data security signifi-cantly reduced the levels of trust in an organisation tokeep the individual’s health data private [33, 46]. Inter-estingly in one study, respondents were more willing toallow access to their electronic records, compared topaper-based records [37]


The role of legal and ethics bodies in protecting privacyA total of 4219 respondents provided a view on the rolelegal and ethical bodies have in protecting privacy. Theuse of laws, regulations, and policies to protect an indi-vidual’s privacy in the UK [31], the USA [40], andAustralia [49] were noted. Without developing an under-standing of individual privacy concerns and perceptionsof privacy, King et al. note that it will be ‘impossible toprovide adequate law as well as effective technical solu-tions for protecting privacy’ [49]. In the UK, laws allowfor data use if the risk to privacy is proportionate [15];the NHS Code of Practice on Confidentiality establishesrules for the protection of privacy [31]. In relation tonew laws to improve data collection, one study notedthat 81% of respondents (N = 2335) would support a lawmaking a cancer registry statutory in the UK [31]. In theUSA, mechanisms such as the National Institute ofHealth (NIH) requirement for manuscripts from NIH-funded research to be made publicly available were con-sidered beneficial in fostering public accountability andtrust [40]. Further, the US Health Insurance Portabilityand Accountability Act of 1996 (HIPAA) establishes anational standard for the protection of health informa-tion [40]. However, some believe that concerns aboutprivacy are not fully addressed by HIPAA, which treatsall health data, except psychotherapy, the same [40]. In astudy, some respondents were unaware that under somecircumstances their medical data could be used withouttheir permission [40]. Respondents in two studies advo-cated for clear and consistently applied penalties for in-dividuals who breach privacy, such as job termination,paying fines, and/or going to jail [40]; measures such asthis may increase perceptions of trust and accountability[40]. The role of ethics and institutional review boardsin protecting privacy was noted in two articles [17, 40].Respondents supported the role of ethics committees tomanage access to health data and trusted their decisions[32]. It is important that health consumers recognise therole of these bodies in regulating access to data for re-search [40] and in protecting patient rights [17]. Finally,the development of clear policies and procedures willallow for more support for the secondary use of data,while increasing transparency for the healthcare con-sumer [37].

AnonymisationA total of 5302 respondents provided a view on dataanonymisation. Data anonymisation was central to an in-dividual’s decision to share health data for research orhealth and service improvement programmes [26, 28, 40,49]. There was a lack of understanding between theterms anonymisation and identifiable data [33]. In onestudy, many respondents were in favour of anonymousdatabases for research, noting it was beneficial and

would advance medical research without impacting ontheir privacy [35]. In the setting of appropriate privacy,confidentiality frameworks, and ethical oversight, Parkinand Paul note that an informed public are more likely tobe receptive to research using potentially identifiablehealth information [47].Even when data is de-identified, some respondents

remained concerned in the setting of extra security mea-sures and data anonymisation [49], which were not seenas safeguards [44]. Some respondents believed that even ifdata had identifying features removed it was not com-pletely de-identified [26] and were concerned about shar-ing de-identified data with non-healthcare professionals[28]. Respondents were asked about their preferences foreither a computer system or human programmer to an-onymise (extract and link) data; some expressed concernabout the potential for identification of individuals andnoted a need for trust in the people undertaking thesetasks [34]. While respondents recognised the capacity ofcomputers to undertake the anonymisation process, theysuggested they would not trust a completely computerisedsystem citing concerns about data infrastructure and dataaccuracy [34].

Communication and educationA total of 8511 respondents provided a view on the im-portance of communication and the role of education inpromoting data sharing. Providing increased informationabout data use and research more generally allowed indi-viduals to feel their privacy is being maintained while con-tributing to health research with societal benefits [45].Providing education was also seen as a mechanism to im-prove transparency [15, 16, 33] and trust [23, 26]. Specificinformation on how and when the data will be used [15,45], and knowing how and where their contact detailswere sourced [42] were all important to individuals. Infor-mation on the data aggregation and anonymisation pro-cesses [44], and the systems used to protect data [40],should be provided. In a UK cancer registry study, cancerpatients opposed to current data collection processes weremore concerned about lack of information about theregistry and consent processes than privacy [41]. Informa-tion and education about database governance, includingdata storage, length of data accessibility, and use of data,should be clear at the time of consent, particularly for dataheld in disease registries [17].

ConsentMost of the articles included in this subset discussed theconnection between issues of privacy, trust, and trans-parency, and consent. Specific issues of consent in rela-tion to secondary data use and sharing are discussed in aseparate publication. Broadly, seeking consent for thesecondary analysis of health data, either anonymised or


potentially identifiable, was seen as a way to build trust,respect, and transparency [26, 30] and address an indi-vidual’s privacy concerns [23, 27, 39].

DiscussionThis systematic literature review highlights the ongoingcomplexity associated with secondary data analysis andlinking health data. Data gaps identified included a pau-city of information specifically related to our primaryarea of interest, and the attitudes of breast cancer pa-tients towards the secondary use and sharing of healthadministrative and clinical trial data. Interestingly, giventhe high rate of cancer more generally in society, thispopulation was underrepresented in the results.While respondents believed that the principles of

data sharing were sound, significant concerns regard-ing privacy, information security, trust, and transpar-ency remain. Further, the diversity of attitudestowards privacy suggests that there is little clarity onwhat predicts an individual’s attitudes towards priv-acy, highlighting an area for future study. Many re-spondents supported the use of health data for socialbenefit; however, this was not universal. The literatureunderscores the importance of communication be-tween those who collect data or act as data custo-dians and health consumers. Health consumersshould be provided clear information on how theirdata privacy will be maintained, how the data will besecured, and how access to their data will be regu-lated. Providing increased information to health con-sumers about how, when, and where their health datamay be used, and with whom it may be shared, is es-sential in the development and maintenance of trans-parent data sharing systems and policies. Concernsrelating to privacy and the misuse of data may be, inpart, mitigated by increased education of health con-sumers regarding their national privacy laws and reg-ulations. Providing information on penalties forbreaches of privacy and how an individual’s healthdata can and cannot be used is important. This mayreduce some specific concerns regarding inappropriateuse of data, ‘big brother’ sentiments, and any percep-tions of discrimination based on data. While not spe-cifically discussed in the articles, it is important tonote that as the use of artificial intelligence increasesin healthcare, ensuring penalties for discriminationbased on data analysis will become more complex.Examples of discriminatory algorithms, in society andin healthcare have been highlighted by researchers[50–52], and these need to be closely examined andtested as our reliance on data-driven healthcare in-creases. Further, health consumers need to be pro-vided information about how any research isundertaken including anonymisation and aggregation

processes, and the requirement for ethics committeeoversight.Our results suggest that trust is an important compo-

nent in the discussion regarding the secondary analysisof health data: trust in the organisations, clinicians, andinfrastructures used to maintain data. Onora O’Neill haswritten extensively on issues of trust in a modern societyand in health and argues that despite the sentimentexpressed by some that trust more generally in societyhas decreased, it has not; rather the culture of suspicionhas increased [53]. Therefore, it is essential that organi-sations wishing to undertake secondary analysis on theirdatasets need to develop trust between themselves andhealth consumers.

LimitationsThe papers included in this study were limited to thoseindexed on major databases; some literature on this topicmay have been excluded if it was not identified during the‘grey’ literature and hand searching phases. As the searchwas restricted to English language publications, some rele-vant literature may have been excluded from the search.Given the initial focus of this research being attitudes to-wards data sharing and reuse in breast cancer, individualsunder 18 years of age were excluded from the analysis. Afinal limitation of this research is that much of the datawas from research methods (surveys, interviews) that arenot considered to be level 1 evidence; however, a rando-mised controlled trial methodology is not necessarily ap-propriate to this research subject.

ImplicationsResults of this systematic literature review indicate thatwhile respondents identified advantages in health infor-mation data sharing, including post-market medicationsurveillance and the potential to decrease medical errors,concerns relating to trust, transparency, and the protec-tion of privacy remain. Additional work is therefore re-quired within these areas during the conception, design,and implementation phases of any health data sharingprogrammes to ensure the balance between public bene-fit and individual privacy is maintained.

ConclusionThe literature confirms that while consumers under-stand the benefits of health data sharing for researchpurposes, issues of trust, transparency, and privacy re-main central to acceptance of health data sharing pol-icies and programmes in the general community.Researchers and those undertaking secondary data ana-lysis should work with consumer organisations to ensureconsumer concerns are addressed.


AcknowledgementsThe authors would like to thank Ms Ngaire Pettit-Young, Information First,Sydney, NSW, Australia, for her assistance in developing the search strategy.

Authors’ contributionsEH, PB, and FB were responsible for developing the study concept and thedevelopment of the protocol. EH and ML were responsible for the dataextraction and data analysis. FB and PB supervised this research. All authorsparticipated in interpreting the findings and contributed to the intellectualcontent of the manuscript. All authors have read and approved themanuscript.

FundingThis project was supported by the Sydney Vital, Translational CancerResearch, through a Cancer Institute NSW competitive grant. The viewsexpressed herein are those of the authors and are not necessarily those ofthe Cancer Institute NSW. FB is supported in her academic role by theFriends of the Mater Foundation.

Availability of data and materialsAll data generated or analysed during this study are included in thispublished article.

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Competing interestsEH, ML, PB, and FB declare that they have no competing interests.

Author details1Northern Clinical School, Faculty of Medicine, University of Sydney, NorthSydney, NSW, Australia. 2Department of Psychology, The University ofSydney, Sydney, NSW, Australia. 3Centre for Medical Psychology &Evidence-Based Decision-Making (CeMPED), Sydney, Australia.4Psycho-Oncology Co-Operative Research Group (PoCoG), The University ofSydney, Sydney, NSW, Australia. 5Patricia Ritchie Centre for Cancer Care andResearch, Mater Hospital, North Sydney, Sydney, Australia.

Received: 19 April 2020 Accepted: 14 September 2020

References1. Glaser BG. Retreading research materials: the use of secondary analysis by

the independent researcher. Am Behav Sci. 1963;6(10):11–4.2. Carrato A, Falcone A, Ducreux M, Valle JW, Parnaby A, Djazouli K, et al. A

systematic review of the burden of pancreatic cancer in Europe: real-worldimpact on survival, quality of life and costs. J Gastrointestinal Cancer. 2015;46(3):201–11.

3. Khozin S, Blumenthal GM, Pazdur R. Real-world data for clinical evidencegeneration in oncology. JNCI J Nat Cancer Inst. 2017;109(11).

4. Cadarette SM, Wong L. An introduction to health care administrative data.Can J Hosp Pharm. 2015;68(3):232–7.

5. Balas EA, Boren SA. Managing clinical knowledge for health careimprovement. Yearbook of Medical Informatics. 2000.

6. National Health and Medical Resarch Council (NHMRC). Open access policyCanberra: Australian Government; 2018 [updated November 2018. Availablefrom: https://www.nhmrc.gov.au/about-us/resources/open-access-policy].

7. National Health and Medical Resarch Council (NHMRC). National Statementon Ethical Conduct in Human Research (2007) - updated 2018 Canberra:Australian Government 2018 [Available from: https://www.nhmrc.gov.au/about-us/publications/national-statement-ethical-conduct-human-research-2007-updated-2018#block-views-block-file-attachments-content-block-1].

8. Kostkova P, Brewer H, de Lusignan S, Fottrell E, Goldacre B, Hart G, et al.Who owns the data? Open data for healthcare. Front Public Health. 2016;4.

9. Esmaeilzadeh P. The impacts of the perceived transparency of privacypolicies and trust in providers for building trust in health informationexchange: empirical study. JMIR Med Inform. 2019;7(4):e14050.

10. Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Preferred reportingitems for systematic reviews and meta-analyses: the PRISMA statement.PLoS Med. 2009;6(7):e1000097-e.

11. Veritas Health Innovation. Covidence systematic review software.Melbourne: Cochrane Collaboration; 2018.

12. Kmet LM, Cook LS, Lee RC. Standard quality assessment criteria forevaluating primary research papers from a variety of fields. Edmonton:Alberta Heritage Foundation for Medical Research (AHFMR; 2004.

13. Lockwood C, Munn Z, Porritt K. Qualitative research synthesis:methodological guidance for systematic reviewers utilizing meta-aggregation. Int J Evid Based Healthcare. 2015;13(3):179–87.

14. Gale NK, Heath G, Cameron E, Rashid S, Redwood S. Using the frameworkmethod for the analysis of qualitative data in multi-disciplinary healthresearch. BMC Med Res Methodol. 2013;13(1):117.

15. Campbell B, Thomson H, Slater J, Coward C, Wyatt K, Sweeney K. Extractinginformation from hospital records: what patients think about consent. BMJQual Saf. 2007;16(6):404–8.

16. Courbier S, Dimond R, Bros-Facer V. Share and protect our health data: anevidence based approach to rare disease patients’ perspectives on datasharing and data protection-quantitative survey and recommendations.Orphanet J Rare Dis. 2019;14(1):175.

17. Darquy S, Moutel G, Lapointe A-S, D'Audiffret D, Champagnat J, Guerroui S,et al. Patient/family views on data sharing in rare diseases: study in theEuropean LeukoTreat project. Eur J Hum Genet. 2016;24(3):338.

18. Eloranta K, Auvinen A. Population attitudes towards research use of health careregistries: a population-based survey in Finland. BMC Med Ethics. 2015;16:48.

19. Hamajima N, Tajima K. Patients' views on reference to clinical data. JEpidemiol. 1997;7(1):17–9.

20. Hay AE, Leung YW, Pater JL, Brown MC, Bell E, Howell D, et al. Linkage ofclinical trial and administrative data: a survey of cancer patient preferences.Curr Oncol. 2017;24(3):161.

21. Mello MM, Lieou V, Goodman SN. Clinical trial participants’ views of the risksand benefits of data sharing. N Engl J Med. 2018;378(23):2202–11.

22. Ni MY, Li TK, Hui RWH, McDowell I, Leung GM. Requesting a uniquepersonal identifier or providing a souvenir incentive did not affect overallconsent to health record linkage: evidence from an RCT nested within acohort. J Clin Epidemiol. 2017;84:142–9.

23. Page SA, Mitchell I. Patients' opinions on privacy, consent and thedisclosure of health information for medical research. Chronic Dis Can. 2006;27(2):60–7.

24. Park YJ, Chung JE. Health privacy as sociotechnical capital. Comput HumBehav. 2017;76:227–36.

25. Patil S, Lu H, Saunders CL, Potoglou D, Robinson N. Public preferences forelectronic health data storage, access, and sharing—evidence from a pan-European survey. J Am Med Inform Assoc. 2016;23(6):1096–106.

26. Robinson G, Dolk H, Given J, Karnell K, Gorman EN. Public attitudes to datasharing in Northern Ireland. Northern Ireland: Administrative Data ResearchCentre; 2016.

27. Sala E, Burton J, Knies G. Correlates of obtaining informed consent to datalinkage: respondent, interview, and interviewer characteristics. SociolMethods Res. 2012;41(3):414–39.

28. Whiddett R, Hunter I, Engelbrecht J, Handy J. Patients’ attitudes towardssharing their health information. Int J Med Inform. 2006;75(7):530–41.

29. Willison DJ, Schwartz L, Abelson J, Charles C, Swinton M, Northrup D, et al.Alternatives to project-specific consent for access to personal informationfor health research: what is the opinion of the Canadian public? J Am MedInform Assoc. 2007;14(6):706–12.

30. Audrey S, Brown L, Campbell R, Boyd A, Macleod J. Young people’s viewsabout consenting to data linkage: findings from the PEARL qualitative study.BMC Med Res Methodol. 2016;16:34.

31. Barrett G, Cassell JA, Peacock JL, Coleman MP. National survey of Britishpublic’s views on use of identifiable medical data by the National CancerRegistry. Br Med J. 2006;332(7549):1068–72.

32. Broes S, Verbaanderd C, Casteels M, Lacombe D, Huys I. Sharing of clinicaltrial data and samples: the cancer patient perspective. Front Med. 2020;7:33.

33. Carson C, Hinton L, Kurinczuk J, Quigley M. ‘I haven’t met them, I don’t haveany trust in them. It just feels like a big unknown’: A qualitative studyexploring the determinants of consent to use Human Fertilisation andEmbryology Authority registry data in research. BMJ Open. 2019;9(5):e026469.


https://www.nhmrc.gov.au/about-us/resources/open-access-policy

https://www.nhmrc.gov.au/about-us/publications/national-statement-ethical-conduct-human-research-2007-updated-2018#block-views-block-file-attachments-content-block-1



34. Haddow G, Bruce A, Sathanandam S, Wyatt JC. ‘Nothing is really safe’: afocus group study on the processes of anonymizing and sharing of healthdata for research purposes. J Eval Clin Pract. 2011;17(6):1140–6.

35. Kass NE, Natowicz MR, Hull SC, Faden RR, Plantinga L, Gostin LO, et al. Theuse of medical records in research: what do patients want? J Law MedEthics. 2003;31(3):429–33.

36. Lee SB, Zak A, Iversen MD, Polletta VL, Shadick NA, Solomon DH.Participation in clinical research registries: a focus group study examiningviews from patients with arthritis and other chronic illnesses. Arthritis CareRes. 2016;68(7):974–80.

37. Lucero RJ, Kearney J, Cortes Y, Arcia A, Appelbaum P, Fernandez RL, et al.Benefits and risks in secondary use of digitized clinical data: views ofcommunity members living in a predominantly ethnic minority urbanneighborhood. AJOB Empirical Bioethics. 2015;6(2):12–22.

38. Sakshaug JW, Couper MP, Ofstedal MB, Weir DR. Linking survey andadministrative records: mechaisms of consent. Sociol Methods Res. 2012;41(4):535–69.

39. Xafis V. The acceptability of conducting data linkage research withoutobtaining consent: lay people’s views and justifications. BMC Med Ethics.2015;16(1):79.

40. Damschroder LJ, Pritts JL, Neblo MA, Kalarickal RJ, Creswell JW, Hayward RA.Patients, privacy and trust: Patients’ willingness to allow researchers toaccess their medical records. Soc Sci Med. 2007;64(1):223–35.

41. Macmillan Cancer S, Cancer Research UK, Ipsos M. Perceptions of the cancerregistry: attitudes towards and awareness of cancer data collection. London:Cancer Research UK; 2016.

42. Slegers C, Zion D, Glass D, Kelsall H, Fritschi L, Brown N, et al. Why dopeople participate in epidemiological research? J Bioethical Inquiry. 2015;12(2):227–37.

43. Wellcome Trust, C. M. Insight. Summary report of qualitative research intopublic attitudes to personal data and linking personal data. London:Wellcome Trust; 2013.

44. Wellcome Trust, Mori I. The one-way mirror: public attitudes to commercialaccess to health data. London: Wellcome Trust; 2016.

45. Willison DJ, Swinton M, Schwartz L, Abelson J, Charles C, Northrup D, et al.Alternatives to project-specific consent for access to personal informationfor health research: insights from a public dialogue. BMC Med Ethics. 2008;9(1):18.

46. Willison DJ, Steeves V, Charles C, Schwartz L, Ranford J, Agarwal G, et al.Consent for use of personal information for health research: do people withpotentially stigmatizing health conditions and the general public differ intheir opinions? BMC Med Ethics. 2009;10:10.

47. Parkin L, Paul C. Public good, personal privacy: a citizens’ deliberation aboutusing medical information for pharmacoepidemiological research. JEpidemiol Community Health. 2011;65(2):150–6.

48. Tully MP, Bozentko K, Clement S, Hunn A, Hassan L, Norris R, et al.Investigating the extent to which patients should control access to patientrecords for research: a deliberative process using citizens’ juries. J MedInternet Res. 2018;20(3).

49. King T, Brankovic L, Gillard P. Perspectives of Australian adults aboutprotecting the privacy of their health information in statistical databases. IntJ Med Inform. 2012;81(4):279–89.

50. Hao K. Facebook’s ad-serving algorithm discriminates by gender and race.MIT Technology Review [Internet]. 2019. Available from: https://www.technologyreview.com/2019/04/05/1175/facebook-algorithm-discriminates-ai-bias/.

51. O'Neil C. Weapons of math of destruction: how big data increasesinequality and threatens democracy. London: Allen Lane; 2016.

52. Parikh RB, Teeple S, Navathe AS. Addressing bias in artificial intelligence inhealth care. JAMA. 2019;322(24):2377–8.

53. O'Neill O. A question of trust; the BBC Reith Lectures 2002. Cambridge:Cambridge University Press; 2002.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.


https://www.technologyreview.com/2019/04/05/1175/facebook-algorithm-discriminates-ai-bias/



Documents

A systematic literature review of health consumer