Value-based Consent Model: A Design Thinking Approach for

Enabling Informed Consent in Medical Data Research

Simon Geller

, Sebastian Müller

, Simon Scheider

, Christiane Woopen

2,3 d

and Sven Meister

1,4 e

Healthcare, Fraunhofer Institute of Software and Systems Engineering, Emil-Figge-Straße 91, 44227 Dortmund, Germany

ceres, University of Cologne, Albertus-Magnus-Platz, 50923 Cologne, Germany

Center for Life Ethics, University Bonn, Schaumburg-Lippe-Str. 7, 53113 Bonn, Germany

Department of Health/School of Medicine, Witten/Herdecke University, Alfred-Herrhausen-Straße 50, 58455 Witten,

Germany

Keywords: Consent Model, Electronic Health Records, Big Data, Autonomy, Moral Values, Medical Research, Privacy,

Meta Consent.

Abstract: Due to new technological innovations, the increase in lifestyle products, and the digitalisation of healthcare

the volume of personal health data is constantly growing. However, in order to use, re-use, and link

personalised health data and, thus, unlock their potential benefits in health research, the authors of the data

need to voluntarily give their informed consent. That is a major challenge to health data research, because the

classic informed consent process requires the immense administrative burden to ask for consent, every time

personal health data is accessed. In this paper we argue that all alternative consent models that have been

developed to tackle this problem, either do not reduce administrative burdens significantly or do not conform

to the informed consent ideal. That is why we used the design thinking approach to develop an alternative

consent model that we call the value-based consent model. This model has the potential to reduce

administrative burdens while empowering research subjects to autonomously translate their values into

consent decisions.

1 INTRODUCTION

In medical data research an informed consent (IC) is

a process in which research subjects are given

information about specific studies and then

voluntarily permits research agents to access and use

their health data like x-ray images, clinical history,

drug prescriptions, and much more to conduct the

study. By informing research subjects in advance

about research objectives, potential advantages and

disadvantages, funding, and other relevant factors,

and by granting them the right to revoke their consent

at any time, the autonomy of subjects in medical

research is structurally protected (Kleinig, J., 2009).

However, the rise of big data in medical research

comes along with two challenges for the IC process.

First, it is impossible to inform subjects about the

https://orcid.org/0000-0002-6342-851X

https://orcid.org/0000-0001-9281-1246

https://orcid.org/0000-0003-4704-2833

https://orcid.org/0000-0002-7148-6808

https://orcid.org/0000-0003-0522-986X

scope, the methods, and the risks of research projects

that are going to access health data in the far future.

Second, to transfer the classic notion of IC into the

medical big data context, researchers would have to

repeat the consent process for each new data use and

each one of the research subjects would have to

repeatedly give their consent. It is easy to see that

with increased quantities of research requests both the

educational and the administrative burden on the part

of the researchers as well as the consent burden on the

part of the research subjects becomes impractical

(Ruyter et al., 2010; Mittelstadt and Floridi, 2016). To

face these challenges, a number of digital alternatives

to the analogue IC model have been introduced

recently (e.g., Helgesson, 2012; Kaye et al., 2015;

Ploug and Holm 2016). All of these models try to find

an equilibrium between the data subjects’ autonomy

Geller, S., Müller, S., Scheider, S., Woopen, C. and Meister, S.

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research.

DOI: 10.5220/0010828000003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 5: HEALTHINF, pages 81-92

ISBN: 978-989-758-552-4; ISSN: 2184-4305

and potential data efficiency benefits by introducing

various interpretations of the research scope that

subjects are able to consent to. They also try to

consider the various national and international laws

concerning medical data privacy like the EU GDPR

or the US HIPAA. Whether these IC models are able

to adequately preserve the research subjects’

autonomy is discussed critically and repeatedly

challenged (e.g., Caulfield and Kaye, 2009; Cheung,

2018; Manson, 2020).

In this paper, we will use the design thinking

method to analyse current IC approaches in the digital

health data research context and come up with a

suitable alternative IC model. First, we analyse the IC

models most commonly discussed in literature and

identify both normative and technical benefits and

risks for researchers and research subjects. Since

some of the identified risks may jeopardise the

research subjects’ autonomy in real live consent

situations, we come up with a new consent model that

we will call value-based consent model and that is

built on the Danish meta consent model by Ploug and

Holm, (2016), the idea of a cascade consent model

that was proposed by the German Ethics Council

(Deutscher Ethikrat, 2017), and the matrix model,

introduced by Christiane Woopen (2020). Its

innovation is the possibility for research subjects to

express consent preferences for different types of

research categories such as research scope, research

agent, funding, and many more in advance and, at the

same time, to manually introduce exceptions from

this setting. We call these exceptions dynamic

categories. They can be introduced in cases, where

settings in the meta consent conflict with each other.

This modification prevents automated consent

decisions that do not match a subjects' personal

values. Additionally, we promote the subjects’ right

to withdraw from its research participation by

introducing the possibility to transparently oversee all

consent decisions already made for past, present, and

future research studies in a so-called consent history

and to opt out of single studies before they start. This

way, the value-based consent model promotes value-

based consent decisions while supressing machine-

based consent decisions.

2 STATE OF THE ART

To understand what informed consent actually is and

how its elements can be justified, it is best to take a

closer look at the status of humans as self-determined

moral beings. This status is often justified by the

human capability to autonomously set and pursue

moral ends (Kant, 2011). In this context, Autonomy

is not so much a single momentum of purposeful

action as it is a process that comprises the capability

to, first, create and reflect personal desires, motives

and ideals, second, to form and change concrete

behavioural intentions and put those intentions into

actions and, third, to assess the foreseeable

consequences of actions (Woopen and Müller, 2021).

To ensure that subjects in medical research are

respected in their nature as morally autonomous

beings the IC must consider all three parts of the self-

determination process. It must enable people to form

and reflect their own preferences and desires

(Frankfurt, 1971). That includes the ability to project

individual changes in preferences and ostensible

choice inconsistencies into the IC architecture.

Furthermore, the IC must take the bounded

rationality (Schlaile et al., 2018) of human beings

into account. Due to cognitive and physical

limitations as well as constraints in terms of time,

economic and social resources, the human ability to

assess all foreseeable consequences of one’s actions

is per se limited. Nonetheless, the quality and quantity

of information as well as its medium of

communication should enable research subjects to

make self-determined decisions. That includes the

necessity to receiving detailed information about

studies subjects are asked to participate in and the

structural ability to easily act on that very information

by consenting to, withdrawing from, or rejecting

research requests. An IC doesn’t serve its purpose if

these criteria are not met and subjects are not able to

form and change their preferences, retrieve relevant

information, and change their consent decisions.

Especially in the digital research context, the last

issue is most relevant. Jürgen Habermas identifies a

threat to autonomy in the pragmatic tendency to

replace complex decision-making processes with

mere technical processes. In such cases, it is no longer

the autonomous citizens within a society who define

social meaning through their own decisions and

discourses, but the few people who develop a

technology that makes important decisions for the

people (Habermas, 2004). Decisions about which

research projects are covered by a broad consent and

which studies are rejected due to hierarchical meta

structures, are part of that problem. Three criteria can

be derived from the discussion so far: an IC model

that transfers the ethical and legal reasons to conduct

the IC into the realm of medical health data research

can be considered to be adequate when research

subjects (i) are informed about crucial characteristics

of a research project before their participation, (ii) are

informed about foreseeable personal and social

HEALTHINF 2022 - 15th International Conference on Health Informatics

consequences of their research participation as well

as the limitations to that assessment, (iii) can easily

incorporate personal moral developments and social

changes in their decision-making process and adjust

their consent accordingly.

2.1 Specific Consent Model

With specific consent, subjects authorise research

agents to access a well-defined set of data for one

specific research purpose. In contrast to the confusing

perception given by the phrase ‘data donation’, which

is often used in such contexts, participants do not lose

their right to authorise or deny data access. They can

withdraw their consent at any time, even after

research has been conducted. Moreover, any data

access and use beyond the specific research purpose

needs new consent. The specific consent model

enables potential research subjects to use the specific

consent type in a digital format (Ploug and Holm,

2016). The information process can be provided by an

interface and support subjects in many ways. For

example, information can be made available in many

languages, its transfer can be supported audio-

visually, and the research subjects’ understanding of

the research in question can be checked with the help

of interactive elements (De Sutter et al., 2020). It is

also important to stress that in this model, information

can be altered to fit the subjects’ needs (Ploug and

Holm, 2016). Apart from that, the specific consent

model does not allow researchers to collect contact

data for follow-up studies. Some researchers argue

that this characteristic can cause tremendous

administrative burdens (Helgesson, 2012). Thus, a

consent model that enables subjects to use specific

consent only would demand a lot of efforts from

potential research subjects since it is plausible to

belief that the number of specific consent requests for

medical data research will increase significantly in

the near future (Mittelstadt and Floridi, 2016). In

consideration of the overwhelming numbers of

consent requests, we assume that study drop-outs and

blanked rejections have to be perceived as likely.

2.2 Broad Consent Model

In contrast, the broad consent can be introduced to

widen the research purpose subjects can consent to.

The definition of what an adequate scope of research

should look like is by no means fixed, but open for

debate. It may, for example, refer to the development

of different image recognition algorithms only and

authorise access to skin cancer images exclusively.

But it may also include a consent to grant access to all

related files in the EHR for the purpose of cancer

research in general. As a result, the broad consent

model reduces the number of research requests and,

therefore, the administrative expenses in correlation

to the extension of the research scope (Manson,

2020). However, a serious issue with the broad

consent model is that at the time the consent is given,

the quantity and quality of future studies conducted

with this data is unknown. The problem is that it is

not possible to inform subjects about the research

scope, the personal benefits and risks, the research

agents, or the funding of future research studies

whose design may not even have been invented today

(Caulfield and Kaye, 2009; Caplan, 2009).

Consequently, subjects cannot know, how their

involvement in research may affect future societies and

their future self. Moreover, not every broad consent

attaches an expiring date to the consents which means

that subjects can consent to a broad use of their data

and when research methods or governance policies

changes in the future, the consent will still hold (Ploug

and Holm, 2020). Attached to this problem is the issue,

that it is inconvenient to make use of the right to

withdraw consent, if subjects do not know in which

research projects their data is being used (Ploug und

Holm, 2016). Advocates of a broad consent model

argue that subjects can be well informed about the fact

that they do not have all relevant information on future

research projects (Taupitz and Weigel, 2012). As long

as transparent governance structures are put in place,

they belief that the broad consent model meet the IC

ideal (Manson, 2020).

2.3 Open Consent Model

The open consent model was implemented in the

Harvard Personal Genome Project, in which subjects

were able to consent to the public release of their

genome data after passing a very detailed test about

the properties of genes, the research areas genes are

used in, and all possible disadvantages that might

come along with public data use. The complexity of

the test gives the impression that the few who are able

to pass it are sufficiently informed to make an

autonomous decision (Angrist, 2009). The open or

blanked consent model enables subjects to grant

everyone the access to a specific set of health data

without any limitations regarding access time and

frequency, data use, or agency (Wendler, 2013).

Thus, the administrative burden of obtaining re-

consent is minimal. Because it is difficult to imagine

that regular data subjects are able to reach a level of

enlightenment where they can overview the most

important effects their open consent might have on

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research

their personal live and the society, most experts

disqualify this model for broad social application

(Cheung, 2018).

2.4 Dynamic Consent Model

The dynamic consent model as well as the meta

consent model try to mediate the extremes of the

former consent models. The dynamic consent enables

subjects to repeatedly give specific consents to

researchers to use their personal health data for

medical studies (Steinsbekk et al., 2013). While the

subjects’ data is stored permanently and does not

need to be deleted after each study, subjects can

actively oversee all research in which their data has

been used. Because the data is stored permanently,

subjects can be selected for research based on special

attributes like clinical history, blood type, social

media use, and so on. Empirical data indicates that

people are more likely to grant their consent to

research projects if research requests are managed

with a dynamic rather than a broad consent model

(Stoeklé et al., 2019). While advocates argue that the

administrative burden of the dynamic consent model

is likely to be smaller than in the specific consent

model, critics doubt that, because the program

constantly sends requests to potential subjects

followed by a long waiting period for responses

(Manson, 2019). In addition, Ploug and Holm assume

that people who are often confronted with consent

requests will stop reading the consent information and

give their consent or refusal out of habit. The authors

call that phenomenon “consent fatigue” (Ploug and

Holm, 2016). To the best of our knowledge this

phenomenon has not yet been empirically proven.

2.5 Meta Consent Model

The meta consent model can be thought of as a filter

program. It gives subjects the opportunity to set their

preferences regarding the study categories “type of

consent” (blanked refusal, broad consent, blanked

consent, specific consent), “type of data” (e.g., EHR,

gene material, tissue etc.), and “research context”

(e.g., commercial or non-commercial research,

funding situation, national or international research)

(Ploug and Holm, 2015; 2016). Blanked refusal

means that subjects refuse to consent to a given

research category in general. Subjects might, for

example, deny commercial agents access to their data.

By entering their preference settings in the meta

consent form, subjects can choose how they are going

to be asked for consent in the future. Now, if the same

subjects prefer to support all research concerning

cancer, they can give broad consent to cancer

research. This way, a study request on skin cancer

research that wants to use skin images to train a skin

cancer recognition software can do so without asking

for specific consent. The same way, a research project

on lung cancer can use the subjects’ EHR. As the

example suggests, in the meta consent model subjects

can choose alternative consent types for different

research categories. Unfortunately, there are cases

where consent choices on the meta level contradict

each other. If a for-profit organisation wants to do

cancer research, it is not obvious how the meta

consent form of the subjects in the example above can

generate a consent. For these cases, Ploug and Holm

introduce a prioritisation of consent decisions that

automatically solves technical inconsistencies.

Blanked refusal is prioritised over specific consent

over broad consent over blanked consent. For the

case above, the blanked refusal to private businesses

is weighted higher than the broad consent to support

cancer research. The meta consent model has

empirically been proven to gain trust among Danish

research subjects (Ploug and Holm, 2017).

There are some problems with the meta consent

model as well. First, the meta consent model has a

higher administrative burden than the broad consent

model because the system needs to send consent

requests constantly and waits for individual answers

(Manson, 2020). Second, human preferences are not

as ordered and consistent as the automated conflict

solution suggests. Preferences and desires do not need

to be complete or transitive to acknowledge the moral

autonomy of research subjects (Sunstein, 1996).

Subjects might, for example, not consent to value

blanked refusal over broad consent in a specific case.

Referring to the introduction of this article, we belief

this momentum to be particularly problematic

because consent decisions are actively delegated to an

automated mechanism that is not controlled by the

research subjects. Finally, it is important to note that

a meta-consent model can be shaped with a variety of

meta criteria to choose from and priority rules to

govern conflicts. Ploug and Holm introduced only

one of many ways to design such a model.

3 RESEARCH DESIGN

We applied a design thinking approach as

methodological framework to our research. The

method of design thinking is increasingly applied in

various scientific domains, particularly information

systems research. It is characterizable as a systematic

approach to find solutions for complex issues with the

HEALTHINF 2022 - 15th International Conference on Health Informatics

aid of multidisciplinary researchers (Wylant, 2008;

Plattner et al., 2011; Wölbling et al., 2012). In this

context, a common model by HPI School of Design

Thinking outlines an iterative process encompassing

six dedicated steps. These steps are passed iteratively

in multiple loops and carried out in a sequential order

while allowing to return to previous steps within an

iteration. The original process model consists of the

steps ‘understand’, ‘observe’, ‘define the point of

view’, ‘ideate’, ‘prototype’, and ‘test’ (HPI School of

Design Thinking, n.d.). For our research design,

however, we performed several adaptations. Similar

to the proposed framework, we defined six steps and

maintained both the sequential order including the

possibility of backward stepping within an iteration

and the iterative nature of the overall methodological

process. Likewise, our research methodology consists

of the steps: (1) awareness building, (2) knowledge

base development, (3) formation of opinion, (4)

ideation, (5) conceptualization, and (6) validation.

Our adapted iterative approach contains an embedded

loop from the last to the first step to ensure agility of

the research process.

Figure 1: Design thinking Steps.

We carried out the steps depicted in Figure 1

iteratively with a project team of 12 researchers, who

contributed their expertise from the research fields

ethics, law, economics, social sciences, information

technology, healthcare, price and service

management, education, and media research. In the

first step, awareness building, each researcher

developed an own understanding for phenomena

related to a given domain under investigation. In the

second step, each researcher relied on individual and

self-selected methods for knowledge base

development, typically a literature analysis, to acquire

comprehensive information about the phenomenon

under investigation. Naturally, this step differed

among researchers since it is subject to the

researchers’ expertise in terms of the domains

relevant for phenomena elicitations. The creation of a

knowledge base allowed each researcher the

formation of an opinion based on own expertise

established over the time of the project. Such

expertise was used by the researchers in the phase of

ideation to generate novel ideas for consent models.

In conceptualization, these ideas were concretized in

several group meetings, five workshops, and weekly

small group meetings over a period of nine month. As

a result of this step, the first two authors of this article

developed first blueprints for consent models,

typically emphasizing certain aspects, e.g., legal or

ethical issues. The final phase of each iteration, called

validation, was carried out in plenums with all

researchers of the project team. Such plenums took

place according to fixed schedules and terminated an

iteration. All researchers presented and shared their

ideas on state of art consent models and the consent

model development. Subsequently, (dis-) advantages,

problems, and opportunities were discussed as well as

potential (dis-) similarities among model proposals.

In a new iteration, the first two authors tried to refine

their consent model based on the feedback received

by the group. Naturally, this involved further

awareness building and an extension of the

knowledge base. Our findings generated through this

process allowed us to systematically compare the

models, which is presented in section 4. Among

others, this systematic comparison has led to the

further development of Ploug and Holms meta

consent model into a new meta consent model

variation that is the value-based consent model.

4 COMPARISON OF THE

CONSENT MODELS

As we have discussed in part 2 there are many issues

with the current IC models that can either severely

limit Big Data research or critically diminish the

autonomy of research subjects. The specific consent

model incorporates the fundamental elements of the

classic notion of IC the best. But, compared to the

alternatives, it carries the highest burdens for

researchers and research subjects. The broad consent

model and the open consent model both reduce or

minimise these burdens at the expenses of the

research subjects’ autonomy. Autonomy can be

restricted by inconvenient refusal options and by

static broad consent types for an unknown number of

studies with an unknown quality. The dynamic

consent and the meta consent model try to smoothen

this gradual autonomy issues by introducing more

convenient choice architectures. Unfortunately, the

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research

Table 1: Benefits vs. risks and burdens of contemporary consent models.

Consent Model Benefit Risks and

urdens

Specific Consent -genuine implementation of the IC ideal

(Ploug and Holm, 2016; De Sutter et al., 2020)

-huge financial and administrative burden

(Helgesson, 2012; Manson, 2019)

-burden of being informed numerous times

(Steinsbekk et al., 2013)

-Risk of stud

dro

outs

(

Steins

ekk et al., 2013

)

Broad Consent -reduced administrative burden compared to

specific consent (Manson, 2020)

-subjects are not informed in detail about the

studies they consent to (Caulfield and Kaye, 2009;

Caplan, 2009)

-model may scare potential research subjects away

Blanked/ Open

Consent

-minimal administrative burden

(Angrist, 2009)

-does not comply with IC ideal in broad public

settings (Cheung, 2018)

ossibilities to intervene are minimal

Dynamic

Consent

-simplified way to contact and re-contact

research subjects (Steinsbekk et al., 2013)

-reduced administrative burden compared to

pecific consent (Steinsbekk et al., 2013)

-the ‘Re-Consent’ option might scare potential

research subjects off (Steinsbekk et al., 2013)

Meta Consent -model produces procedurally consistent

consent decisions (Ploug and Holm, 2016)

-subjects can express preferences comparatively

accurate (Ploug and Holm, 2016)

-subjects might give consent to studies that they

prefer not to consent to (Ploug and Holm, 2016)

-higher administrative burden compared to the

broad consent model (Manson, 2019)

first one is still burdensome in administration and

potentially foster a consent fatigue. The latter might

technically generate consents that do not reflect the

informed decisions of the subjects. Table 1

summarise the benefits and burdens of all consent

models as discussed before. This table relates to the

present analysis and is not exhaustive.

Based on the comparison above, we belief that a

meta consent model approach is the best solution, so

far, to realise the IC ideal in the context of medical

data research. It enables research subjects to affect the

frequency and type of research requests they receive

and it also acknowledges their ability to form and

express personal preferences concerning research

categories like research objectives, research agents,

and funding in general.

However, we recognise that the limitations of

Ploug and Holms model jeopardise the IC ideal, for it

creates the option to generate technical consents that,

eventually, do not correspond with the subjects’

actual intentions. It does so by providing only

technical solutions to resolve conflicting consent

types, by allowing subjects to choose between a few

categories within the meta consent form only, and by

not facilitating the subjects’ ability to identify and

subsequently correct potentially erroneous consents.

To overcome those issues and to come closer to the

IC ideal without increasing administrative burdens

disproportionately, we propose an extended version

of the meta consent model that we will call value-

based consent model.

5 THE VALUE-BASED CONSENT

MODEL

So far, it has become evident that an exclusive

specific consent model is too demanding to be used in

a medical data research setting and that all the other

consent models sacrifice important elements of the IC

ideal in their efforts to make medical research more

efficient. The meta consent model is least affected by

this critique. It struggles primarily with the scope of

preferences subjects can choose from and the

technical solution to conflicts of preference

incoherencies. To overcome these problems, we

propose to adapt the meta consent model by

introducing additional consent and refusal options

that enables research subjects to translate their values

into a fine-grained preference matrix that better

reflect their preferences on how to be approached for

IC requests.

The idea to adopt the meta consent model in a way

that better matches the EU General Data Protection

Regulation(GDPR) as well as the autonomous

decision-making process of research subjects have

already been elaborated in political statements

elsewhere (Deutscher Ethikrat, 2017;

Datenethikkomission, 2019; Woopen, 2020). We also

introduce a learning strategy – dynamic categories –

that enables the value-based consent model to favour

individual values over technical choice consistencies.

The interplay between both mechanisms, the adapted

version of the meta consent form that we call value-

HEALTHINF 2022 - 15th International Conference on Health Informatics

based consent structure, and the dynamic categories,

enable subjects to autonomously translate their

personal values into preferences and IC decisions and

to adjust those decisions at any time.

5.1 Research Categories

For the possibility to use broader consent types such

as broad consent and blanked refusal, as it is common

in the original meta consent model, we propose an

adjustment of the meta consent form (Ploug and

Holm, 2016). In doing so, we also define the term

research category, which can denote single research

fields as well as other characteristics of medical

studies such as ‘research objective’, ‘type of data’ and

‘type of context’. All of those are part of the meta

consent model by Ploug and Holm (2016). Because

every study has a research objective, a source of

funding, research agents like hospitals or research

groups, and so on, it can always be characterised with

a specific set of research categories. For example, a

study that seeks the data of research subjects to

develop an image recognition algorithm for skin

cancer can be characterised by its objective (skin

cancer research), the type of data that is being

processed (images of benign/malignant moles), and

contextual parameters (e.g., funded by the Ministry of

Health). Each research category is composed of

many subcategories. For example, the objective ‘skin

cancer’ research is a subcategory of ‘cancer research’

which has other subcategories as well like ‘neoplasms

of digestive organs’ or ‘neoplasm of breast’. The

category ‘skin cancer’ may in turn have further

subcategories like ‘melanoma’ or ‘basal-cell

carcinoma’. All study requests can be represented as

sets of research categories. For example, a study

dealing with the analysis of skin images with a

repeated data query conducted over 5 years in the

field of skin cancer research on behalf of the public

could look like this:

Study (x) = {…, Virology:false, SkinCancer:true,

DataType-SkinImages:true,

MultipleDataRetrieval:true,

ResearchAgend-PrivateCompany:false,

…}

Now, for a fine-grained adaptation of the broad

consent and its counterpart the broad refusal,

research subjects are able to consent to studies that

correspond to research categories on different super

and sub-levels. If broad consent is given in one

research category, the consent affects all

subcategories if no other consent decision has been

made. In contrast to the classic meta consent model,

however, subjects have the option of manually

consent or refuse to individual subcategories. For

example, research subjects may choose to give their

consent to all research projects that want to use their

Figure 2: Representation of the linkage between studies and research categories.

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research

data for ‘cancer research’. By giving broad consent to

this research category, subjects also consent to all its

associated subcategories. If some subjects like to

make other decisions in certain subcategories like

‘skin cancer’ research, they can select another option

in those subcategories. Due to the inheritance of

consent decisions to respective subcategories the

fine-grained preferences of the research subjects can

be represented in a tree structure and all its

components.

Figure 2 shows the linkage of different studies to

their research categories in a tree structure. The

blanked consent option enables subjects to consent to

any research request regardless of the research

category it belongs to. At this time this option is

legally prohibited in the EU and only listed for the

sake of completeness.

What level of differentiation of medical research

areas, data types, funding models and more can

commonly be understood in an IC process and,

therefore, what exact set of research categories are

needed to improve the IC process for research

subjects cannot be determined theoretically. We will

come back to this issue in the limitations.

5.2 Additional Refusal Types

In order to enable subjects to translate their values

into a new fine-grained adaptation of the old meta

consent form, this subsection explicates some refusal

types that have already been in use in other consent

models.

5.2.1 Specific, Broad and Blanked Refusal

One new way for subjects to communicate their

preferences in the value-based consent model is to

choose a refusal to certain research categories and its

subcategories. The specific refusal that is already

implied in the specific consent model allows subjects

to refuse to participate in specific studies and in

studies that share certain research categories. For

example, if subjects are not comfortable to authorise

the research team of Google to access their health data

in one particular case, they can use the specific

refusal. If the same subjects are uncomfortable with

Google in general using their data for research, they

can make a broad refusal for Google to use their data

in any further studies. Analogously, to the broad

consent, the broad refusal affects multiple studies

belonging to that research category at once.

Likewise, the blanked refusal expresses the

preference to deny data access to any kind of medical

data research, whatsoever. This refusal type differs

from the blanked refusal option that is known to the

meta consent model in that it refers to the entire

preference tree (Section 5.3).

5.2.2 Legal Obligation for Data Processing

For reasons of transparency, it could be useful to

communicate legal acts of obliged health data access.

For example, when the Centre for Disease Control

and Prevention accesses its citizens EHR files to

count new COVID-19 cases and evaluate counter

measures, it is entitled to do so by law. Noteworthy,

communicating legal obligations for data processing

does not embody a novel governmental power, but

rather makes such operations more transparent.

5.3 Value-based Consent Structure

As in Ploug and Holm (2016) original meta consent

form, in the value-based consent model subjects have

the option to express their consent preferences for

each research category and subcategory. In contrast

we present the adapted version in a hierarchical tree

structure which simplifies the representation of

multidimensional research categories and allows

subjects to combine different consent and refusal

types within these dimensions.

Figure 3: Preference Tree.

Figure 3 shows that some research categories are

described as ‘open’. The term ‘open’ simply means

that the subjects have not made a consent decision in

this category yet. If, at the time a research agent sends

out a study’s request, all the relating research

categories are still ‘open’, a specific consent request

will be forwarded to the subjects. Whereby, the

subjects are made aware that they could set a consent

decision in the value-based consent structure. A

specific consent option means, that requests should be

forwarded directly to the research subjects as specific

consent requests, whereas the broad consent option

reflects the broad consent decision to grant a consent

HEALTHINF 2022 - 15th International Conference on Health Informatics

to all corresponding research requests. In addition, we

introduce a prioritisation of consent decisions for

subcategories and supercategories in which consent

and refusal decisions in a subcategory override the

corresponding supercategory. For example, research

subjects may choose to give broad refusal to the

research category ‘cancer research’ which will mark

all subcategories like ‘skin cancer research’ as

refusal. Now, if subjects like to set exceptions, they

can choose to give a different consent option to

specific subcategories. This makes it possible to

express global decisions such as ‘All except these’ as

well as ‘None except these’ through the consent tree.

The opportunity to add exceptions to a broad refusal

option is actively suppressed by the prioritisation

logic of the meta consent model by Ploug and Holm

(2016).

5.4 Dynamic Categories

In the meta-consent form as well as in the value-based

consent structure it is very likely that two or more

preferences conflict with each other on a regular

basis. For the skin cancer research case, such a

conflict may arise for subjects who have given a

broad consent to cancer research but broad refusal to

commercial research agents. Now, if Google wants to

conduct skin cancer research, it is not clear whether

the consent or the refusal should be more important.

The issue of conflicting consent types not only occurs

in cases in which broad consent and broad refusal

conflict with each other, but also in those where

specific consent preferences and broad consent

decisions preferences contradict each other. To solve

this issue Ploug and Holm introduce a prioritisation

schema, as described in section 2.5. According to

their schema, the leaner decision is preferred and

applied as soon as two preference definitions in the

meta-consent form contradict each other, whereby the

refusal is given the highest priority:

Blanked Refusal > Specific Consent > Broad

Consent > Blanked Consent

In line with this prioritization logic, the refusal

towards commercial research agents like Google

would be of a higher priority than the cancer research

consent and Googles cancer research request would

be denied. This prioritisation does not correspond to

the IC ideal, since the subjects cannot decide which

of the respective preference they would favour in a

conflict. On a more technical level, the scheme will

likely generate refusal decisions for research studies

in which subjects would have liked to participate. To

eliminate this problem, we replace the static

prioritization logic with what we call dynamic

categories.

Dynamic categories enable subjects to

individually decide how to deal with conflicting

preference choices. In cases in which consent

decisions conflict with each other, the affected

research request triggers a request for specific

consent. Subjects are notified that the request has

been generated by conflicting preference choices in

the value-based consent structure. After reading the

consent information and approving or rejecting the

specific consent request, subjects are given the option

to give a broad consent or a broad refusal to all future

study requests that trigger the very same conflict.

That means that any case of conflict enables subjects

to create exceptional broad consent or broad refusal

rules which is the dynamic category. For the Google

cancer study case, subjects are notified about the

conflicting preferences and can choose, either to treat

every conflict of this kind as specific consent request

or to generate a new dynamic category that solves the

conflict for all studies that share the same research

categories. In the first case, a different study on skin

cancer that is conducted by Google would be brought

forward as a new specific consent request. In the

second case and in dependence to the previous

decision, subjects either consent or refuse their

participation in forthcoming cancer studies that are

conducted by commercial agents like Google, Pfizer,

or Nestle without having to go through a new IC

process. For the dynamic categories, only broad

consent and refusal options are available, since a

specific consent choice would generate the same

effect as the absence of the dynamic category.

Since dynamic categories can only be generated

when consent decisions conflict in the value-based

consent structure, it is evident that they are deleted

when all underlying contradictions cease to exist. As

soon as subjects alter the structure in a way that the

conflicts that establish a dynamic category are

removed, the subjects are being notified. In addition,

subjects can change their consent or refusal decisions

or delete the dynamic category at any time.

There are two ways in which research categories

can overlap. First, there are conflict-free overlaps

which are overlaps of any number of research

categories for which the same consent or refusal type

has been chosen and any number of research

categories for which the decision is still open.

Second, there are overlaps that are characterised by

conflicting consent decisions. The following four

conflict cases are conceivable:

1. Broad or blanked consent with broad or blanked

refusal

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research

2. Broad or blanked consent with specific consent

preference

3. Broad or blanked refusal with specific consent

preference

4. Broad or blanked consent with broad or blanked

refusal with specific consent preference.

The procedure to solve those conflicts is always

the same. A specific consent request is sent to the

subjects who then can give a specific consent or

refusal and create a dynamic category by choosing a

broad consent or refusal decision for all cases alike.

5.5 Opt out and Reconsideration

Options

We introduced an opt out and a reconsideration

opportunity to the value-based consent model, that

enables research subjects to revoke or change their

consent decisions up to the point of data retrieval. We

belief this option to be necessary to effectively

exercise the right to withdrawal consent in a fast-

moving research field that is medical data research.

For this purpose, a consent history is available that

entails a list of all research requests that have been

given a consent or a refusal to. Subjects can use their

consent history to modify specific consent decisions

as well as the specific consent or refusal decisions that

derived through broad consent and broad refusal

settings. For example, research subjects that decide to

give broad consent to commercial research agents

and to skin cancer research at first, might, at a later

time, read the list of all research studies that they have

given their consent to and realise that they consented

to participate in the research of a company that they

would rather not have given consent to. In the value-

based consent model the subjects can opt out in such

situations without further ado as long as the data has

not been retrieved jet. The other way around it is also

possible that subjects have given a broad refusal to

commercial agents, but by looking on the list of

refused research studies they may realise that they

want to support a specific skin cancer study by

Google. As long as this study is not due, the subjects

can alter their former broad consent decision in such

cases by using the option to reconsider the study for

the specific consent process. It is also possible to opt

out of a specific consent that is characterised by

multiple data retrievals over a fixed period of time. In

the consent history a specific consent of such kind is

tagged prominently so that research subjects can

easily distinguish it from other consent types. To

revoke a consent, subjects can simply opt out of the

data retrieval that is scheduled next.

The opt out and the reconsider option is a

transparency feature that communicates the relation

between subjects and all relevant research studies

openly. The oversight of the consent history and, thus,

the impact of broad consent and broad refusal

options on specific research requests promotes the

competence of subjects to translate their values into

consent preferences adequately. In all alternative

models that offer broad consent options, subjects may

face the problem of not being able to imagine, which

specific research projects may serve an abstract

research objection, which specific agents are affected

by a broad consent decision or which type of data

might be risky to share in a given research context.

With this transparent structure, the value-based

consent model ensures that subjects will become

increasingly better at understanding research

categories. A remarkable side effect of the opt out

feature is, that it excludes the original notion of a

blanked consent type. Even the broadest consent

decision like the decision to give consent to all

research objectives, research agents, and all other

research categories, whatsoever, does not equal the

classic blanked consent decision because subjects are

always able to change their decision based on

transparently communicated research activities and

opt out of individual research projects.

6 DISCUSSION

In this article we used the design thinking approach

to develop the value-based consent model. The value-

based consent model enables consent requests for the

use of personal health data in medical research

projects to be answered via the value-based consent

structure and additional dynamic categories. In this

process, the research categories of each study are

matched with the selected consent preferences of each

subject. In contrast to the original meta consent form

by Ploug and Holm, the new structure enables

subjects to choose two additional refusal options:

specific refusal and broad refusal. We believe that

representing the super and sub research categories in

a tree structure (see Figure 2) is also more accessible

than the matrix notation used by Ploug and Holm. The

dynamic categories, then, transfer the power to decide

how to deal with conflicting consent preferences from

a technical static prioritization logic to the research

subjects themselves. By adding a transparent consent

history that can be used to audit all former and current

consent decisions and to opt out or reconsider consent

choices, subjects are also able to increase their

HEALTHINF 2022 - 15th International Conference on Health Informatics

medical and digital literacy and adjust their consent

choices to their personal values.

All these changes establish a digital IC process

that converges with the ideal presented at the

beginning. (i) By informing subjects continuously

and transparently about past, current and future

studies in a learning friendly environment, the value-

based consent model supports the subject’s digital

and medical literacy, thus, enables them to build their

consent decisions on the latest insides and educated

assessments. (ii) Based on the easily accessible and

comprehensive information provided by the value-

based consent model and with access to the advice of

medically trained personnel who always needs to

accompany consent processes subjects are

empowered to understand the risks and consequences

of research participation and the limits of that

assessment. Finally, (iii) the differentiated choice

architecture, the simple to use intervention options,

and the consent history give subjects the opportunity

to incorporate newly gained experiences and personal

value shifts in the IC decision-making process.

However, there is also the issue of the

administrative burdens of digital IC models for

secondary use of research data. We have referred to

empirical studies that indicate that research subjects

are less likely to engage in medical data research that

is governed by broad or open consent models. In

contrast subjects place more trust in IC processes that

are governed by a dynamic consent model or a meta

consent model. Since the value-based consent model

is comparatively more transparent and does not

patronise subjects with priority rules, we assume that

subjects would prefer it over other models. If this was

the case, the comparatively large number of subjects

that would potentially use the new IC model would

reduce the challenges of the recruitment process and,

therefore, the administrative burden. The hypothesis

needs to be proven empirically in future studies.

6.1 Limitations

The effectiveness and the usability of the value-based

consent model has not been evaluated empirically so

far. Since the meta consent form is less complex than

the value-based consent structure and since the new

model incorporates a number of new refusal options

empirical studies on the application of the meta

consent model that have been conducted so far cannot

substitute the gap. Another limitation is the

configuration of the value-based consent structure. In

their work, Ploug and Holm pointed out that the meta

consent form can be extended to contain more meta

categories to choose from. However, the number and

type of research categories it should contain in order

to allow subjects to make autonomous decisions

without being either overwhelming or patronised, has

not yet been determined. This issue needs to be

addressed in further empirical studies as well.

ACKNOWLEDGEMENT

This work is funded by German Federal Ministry of

Education and Research (BMBF) (grant 16SV8500,

‘ViCon - Virtual Consent Assistant for an informed

and sovereign consent’).

REFERENCES

Angrist, M. (2009). Eyes wide open: The personal genome

project, citizen science and veracity in informed consent.

Personalized Medicine, 6(6), 691–699.

https://doi.org/10.2217/pme.09.48

Benke, K., & Benke, G. (2018). Artificial intelligence and big

data in public health. International Journal of

Environmental Research and Public Health, 15(12).

https://doi.org/10.3390/ijerph15122796

Caplan, A. L. (2009). What No One Knows Cannot Hurt

You: The Limits of Informed Consent in the Emerging

World of Biobanking, in: Solbakk, J. H., Holm, S., &

Hofmann, B. (ed.) The Ethics of Research Biobanking

(2pp.5-32). Dordrecht: Springer. https://doi.org/

10.1007/978-0-387-93872-1

Caulfield, T., & Kaye, J. (2009). Broad consent in

biobanking: Reflections on seemingly insurmountable

dilemmas. Medical Law International, 10(2), 85–100.

https://doi.org/10.1177/096853320901000201

Cheung, A. S. Y. (2018). Moving beyond consent for citizen

science in big data health and medical research.

Northwestern Journal of Technology and Intellectual

Property, 16(1).

Datenethikkomission (2019). Gutachten der Datenethik-

kommission Available at: https://www.bmi.bund.de/

SharedDocs/downloads/DE/publikationen/themen/it-

digitalpolitik/gutachten-datenethikkommission.pdf;

jsessionid=9E1A40B6C43577B2234D8CF241C2241A.

2_cid373?__blob=publicationFile&v=6 [Accessed

September 9, 2021].

Deutscher Ethikrat (2017). Big Data und Gesundheit -

Datensouveränität als informationelle Freiheits-

gestaltung Available at: https://www.ethikrat.org/

fileadmin/Publikationen/Stellungnahmen/deutsch/stellu

ngnahme-big-data-und-gesundheit.pdf [Accessed

September 9, 2021].

Frankfurt, H. G. (1971). Freedom of the will and the concept

of a person. The Journal of Philosophy, 68(1), 5-20.

https://doi.org/10.2307/2024717

Habermas, J. (2004). Freiheit und determinismus. Deutsche

Zeitschrift Für Philosophie, 52(6), 871–890.

https://doi.org/10.1524/dzph.2004.52.6.871

Value-based Consent Model: A Design Thinking Approach for Enabling Informed Consent in Medical Data Research

Helgesson, G. (2012). In defense of broad consent.

Cambridge Quarterly of Healthcare Ethics, 21(1), 40–

50. https://doi.org/10.1017 /S096318011100048X

HPI School of Design Thinking (n.d.) The six phases of the

Design Thinking process. Available at:

https://hpi.de/en/school-of-design-thinking/design-

thinking/background/design-thinking-process.html

[Accessed: September 08, 2021].

Kant, I. (2011). Groundwork of the metaphysics of morals: A

German-English edition (M. J. Gregor, Trans.) (J.

Timmermann, Ed.). Cambridge: University Press.

https://doi.org/10.1017/CBO9780511973741

Kaye, J., Whitley, E. A., Lund D., Morrison, M., Teare, H.,

& Melham, K. (2015). Dynamic consent: A patient

interface for twenty-first century research networks.

European Journal of Human Genetics, 23(2), 141–146.

https://doi.org/10.1038/ejhg.2014.71

Kleinig, J. (2009). The nature of consent In F. Miller & A.

Wertheimer (ed.), The ethics of consent (pp. 3–22).

Oxford: University Press. https://doi.org/10.1093/

acprof:oso/9780195335149.003.0001

Manson, N. C. (2019). The biobank consent debate: Why

‚meta-consent‘ is not the solution? Journal of Medical

Ethics, 45(5), 291-294. http://dx.doi.org/10.1136/

medethics-2018-105007

Manson, N. C. (2020). The case against meta-consent: Not

only do Ploug and Holm not answer it, they make it even

stronger. Journal of Medical Ethics, 46(9), 627–628.

https://doi.org/10.1136/medethics-2019-105955

Mittelstadt, B. D., & Floridi, L. (2016). The ethics of big data:

Current and foreseeable issues in biomedical contexts.

Science and Engineering Ethics, 22(2), 303–341.

https://doi.org/10.1007 /s11948-015-9652-2

Plattner, H., Meinel, C., & Leifer, L. (2011). Design

Thinking: Understand - improve - apply. Understanding

Innovation. Berlin: Springer. http://sub-hh.ciando.com/

book/?bok_id=232558 https://doi.org/10.1007/978-3-

642-13757-0

Ploug, T., & Holm, S. (2015). Meta consent: A flexible and

autonomous way of obtaining informed consent for

secondary research. BMJ, 350, h2146. https://doi.org/

10.1136/bmj.h2146

Ploug, T., & Holm, S. (2016). Meta consent - a flexible

solution to the problem of secondary use of health data.

Bioethics, 30(9), 721–732. https://doi.org/10.1111/

bioe.12286

Ploug, T., & Holm, S. (2017). Eliciting meta consent for

future secondary research use of health data using a

smartphone application - a proof of concept study in the

danish population. BMC Medical Ethics, 18(1), 1–8.

https://doi.org/10.1186/s12910-017-0209-6

Ploug, T., & Holm, S. (2020). The ‚expiry problem‘ of broad

consent for biobank research - and why a meta consent

model solves it. Journal of Medical Ethics, 46(9), 629–

631. https://doi.org/10.1136/medethics-2020-106117

Ruyter, K. W., LÕuk, K., Jorqui, M., Kvalheim, V.,

Cekanauskaite, A., & Townend, D. (2010). From

research exemption to research norm: Recognising an

alternative to consent for large scale biobank research.

Medical Law International, 10(4), 287–313.

https://doi.org/10.1177/096853321001000403

Schlaile, M. P., Klein, K., & Böck, W. (2018). From bounded

morality to consumer social responsibility: A

transdisciplinary approach to socially responsible

consumption and its obstacles. Journal of Business

Ethics, 149(3), 561–588. https://doi.org/10.1007/

s10551-016-3096-8

Steinsbekk, K. S., Kåre Myskja, B., & Solberg, B. (2013).

Broad consent versus dynamic consent in biobank

research: Is passive participation an ethical problem?

European Journal of Human Genetics, 21(9), 897–902.

https://doi.org/10.1038/ejhg.2012.282

Stoeklé, H.‑C., Turrini, M., Charlier, P., Deleuze, J.‑F.,

Hervé, C., & Vogt, G. (2019). Genetic Data, Two-Sided

Markets and Dynamic Consent: United States Versus

France. Science and Engineering Ethics, 25(5), 1597–

1602. https://doi.org/10.1007/s11948-019-00085-4

Sunstein, C. R. (1996). Social Norms and Social Roles.

Columbia Law Review, 96(4), 903. https://doi.org/

10.2307/1123430

Sutter, E. de, Zaçe, D., Boccia, S., Di Pietro, M. L., Geerts,

D., Borry, P., & Huys, I. (2020). Implementation of

electronic informed consent in biomedical research and

stakeholders‘ perspectives: Systematic review. Journal

of Medical Internet Research, 22(10), e19129.

https://doi.org/10.2196/19129

Taupitz, J., & Weigel, J. (2012). The necessity of broad

consent and complementary regulations for the

protection of personal data in biobanks: What can we

learn from the german case? Public Health Genomics,

15(5), 263–271. https://doi.org/10.1159/000336604

Wendler, D. (2013). Broad versus blanket consent for

research with human biological samples. The Hastings

Center Report, 43(5), 3–4. https://doi.org/10.1002/

hast.200

Wölbling, A., Krämer, K., Buss, C. N., Dribbisch, K., LoBue,

P., & Taherivand, A. (2012). Design Thinking: An

Innovative Concept for Developing User-Centered

Software. In Maedche, A., Botzenhardt, A., & Neer, L.

(ed.), Management for Professionals. Software for

People (pp. 121–136). Berlin: Springer. https://doi.org/

10.1007/978-3-642-31371-4_7

Woopen C. (2020). Stellungnahme zum Gesetzesentwurf der

Bundesregierung "Entwurf eines Gesetzes zum Schutz

elektronischer Patientendaten in der

Telematikinfrastruktur. Deutscher Bundestag BT-

Drucksache 19/18793. Deutscher Bundestag. Available

at: https://www.bundestag.de/resource/blob/698068/

d2839c2cc2c1ec321bbb843df38362a3/19_14_0165-26-

_ESVe-Prof-Woopen_PDSG-data.pdf [Accessed

September 9, 2021].

Woopen, C., & Müller, S. (2021). Die Digitalisierung als

Herausforderung für die Bestimmung des Selbst im

Gesundheitswesen. In C. Piallat (ed.), Der Wert der

Digitalisierung: Gemeinwohl in der digitalen Welt

(pp. 123–146). Bielefeld: transcript Verlag.

Wylant, B. (2008). Design Thinking and the Experience of

Innovation. Design Issues, 24(2), 3–14.

https://doi.org/10.1162/desi.2008.24.2.3

HEALTHINF 2022 - 15th International Conference on Health Informatics