Citation: Harborth, D.; Pape, S.;
McKenzie, L.T. Why Individuals Do
(Not) Use Contact Tracing Apps: A
Health Belief Model Perspective on
the German Corona-Warn-App.
Healthcare 2023,11, 583. https://
Academic Editor: Daniele Giansanti
Received: 28 December 2022
Revised: 8 February 2023
Accepted: 9 February 2023
Published: 15 February 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
Why Individuals Do (Not) Use Contact Tracing Apps: A Health
Belief Model Perspective on the German Corona-Warn-App
David Harborth * , Sebastian Pape and Lukas Tom McKenzie
Faculty of Economics and Business, Goethe University Frankfurt, 60326 Frankfurt am Main, Germany
The World Health Organization declared the emergence of the novel coronavirus (SARS-
CoV-2) in January 2020. To trace infection chains, Germany launched its smartphone contact tracing
app, the “Corona-Warn-App” (CWA), in June 2020. In order to be successful as a tool for ﬁghting
the pandemic, a high adoption rate is required in the population. We analyse the respective factors
inﬂuencing app adoption based on the health belief model (HBM) with a cross-sectional online study
including 1752 participants from Germany. The study was conducted with a certiﬁed panel provider
from the end of December 2020 to January 2021. This model is primarily known from evaluations of
medical treatments, such as breast cancer screenings, but it was rarely applied in prior work for a
health-related information system such as the CWA. Our results indicate that intrinsic and extrinsic
motivation to use the CWA are the strongest drivers of app use. In contrast, technical barriers, privacy
concerns and lower income are the main inhibitors. Our ﬁndings contribute to the literature on
the adoption of contact tracing apps by questioning actual users and non-users of the CWA, and
we provide valuable insights for policymakers regarding inﬂuences of adoption and potential user
groups of disease prevention technologies in times of pandemics.
contact tracing apps; SARS-CoV-2; COVID-19; health belief model; user adoption of
health-related technologies; quantitative user study
Shortly after a cluster of pneumonia cases of unknown aetiology was detected in
Wuhan (China) in late 2019, the World Health Organization (WHO) declared the emergence
of the novel coronavirus (SARS-CoV-2/COVID-19) in January 2020 [
]. In an effort
to trace coronavirus infection chains, Germany launched an open-source smartphone
contact tracing app (“Corona-Warn-App”) in June 2020. The Corona-Warn-App (CWA)
informs users who were exposed to a person later tested positive on the basis of Exposure
Notification APIs [
]. The CWA works independently and, in particular, does not rely
on the health offices, which has the advantage that it can warn people even if the health
offices are overloaded. The app is under active development, and since our survey, new
features such as information on COVID-19 incident rates, the possibility for check-ins
and a vaccination certification and test management were added. For the CWA to be
successful, a high proportion of the German population has to adopt the technology
and register themselves once they have been declared infected with COVID-19. Besides
heavy advertisements from the German government, the development of the CWA in
Germany was accompanied by public discussions about its usefulness, i.e., the claim that
more than 50% of the population needed to use it to be effective [
], and data privacy
issues. The discussion on data privacy issues started with the question of whether the
CWA should facilitate a centralised or decentralised architecture [
]. A decentralised
approach was chosen, resulting in ongoing claims by politicians that privacy would hinder
a beneficial use of the CWA [
] while privacy experts pointed out that data protection is
not a hindrance and necessary for a high adoption rate [
]. However, a high press coverage
Healthcare 2023,11, 583. https://doi.org/10.3390/healthcare11040583 https://www.mdpi.com/journal/healthcare
Healthcare 2023,11, 583 2 of 14
of these discussions could alter individuals’ perceptions, who usually are neither privacy
experts nor epidemiologists [
]. On that ground, we investigate which variables have led
German citizens to use or not use the CWA. We address this question with the theoretical
lens of the Health Belief Model (HBM), which was originally developed in the 1950s
and remains one of the most widely applied conceptual frameworks for health-related
behaviours, especially disease prevention behaviours [
]. The HBM represents a profound
and established model for analysing subjective expectations and barriers regarding the
disease prevention behaviour (use of CWA) which is a function of the subjective value of
an outcome as well as the subjective probability or expectation, that a particular action
will achieve that outcome. However, since the original HBM did not consider infectious
diseases, we had to adapt several constructs.
While the HBM has been applied to COVID-Tracing apps in earlier studies [
of them were conducted before the actual introduction of an app and worked by describing
potential properties of the upcoming app and asking for the behavioural intention to use it.
In contrast, our participants were sampled by users and non-users of the already existing
CWA in Germany. We conducted a representative user study with the help of a market
research institute with 1752 participants in Germany. We sampled participants evenly
distributed into these two groups (users of the app, N = 896 and non-users, N = 856). Thus,
we could utilise the actual use as the dependent variable. Furthermore, it has been shown
that, in particular, the older population and those in lower-income households might also
face technical barriers to using the app [
], which could not be considered in the previous
HBM studies. With our study design, we can investigate such factors in a more realistic
way. We investigate the effect of each (partially adapted) factor from the HBM on the
likelihood of using the CWA based on a logistic regression model. Our results contribute to
the theoretical IS knowledge base by establishing that the HBM is a useful theory for health
information systems and can also be used to explain the behaviour of people for infectious
diseases. Our results also contribute to the literature on contact tracing app adoption by
showing that perceptions of technical barriers and privacy concerns, as well as a lower
income, are the main inhibitors for adopting the CWA, while the intrinsic and extrinsic
motivations to use the app are the main drivers.
2. The Health Belief Model
The health belief model was developed through a set of independently investigated
research questions. It has been formulated in order to explain the widespread failure
of people to participate in case-finding programs to prevent and detect asymptomatic
]. The design and dissemination followed the work by a group of behavioural
psychologists in the U.S. Public Health Service between 1950 and 1960 at a time when
health professionals were alarmed by the fact that few people were getting screened for
tuberculosis (TB). The HBM states that health-related behaviours reflect both an individ-
ual’s level of fear of perceived health threats and the expected fear-reduction potential
of taking a recommended action [
]. Based on this notion, we will examine how the
perceived health threat of contracting SARS-CoV-2 will affect a person’s willingness to
use the CWA. Originally, the HBM contained five theoretical constructs, which all depict
the psychological explanation of whether and why people will take action to engage
in disease preventives [
]. These constructs are (1) perceived susceptibility (percep-
tions of the likelihood of getting a disease or condition), (2) perceived severity (beliefs
about the seriousness of the illness, condition, or unwanted outcome and the potential
consequences), (3) perceived barriers (beliefs about the influences that discourage adop-
tion of the promoted action or new behaviour), (4) perceived benefits (beliefs about the
positive consequences of adopting the promoted action) and finally (5) cues to action
(internal or external factors that could trigger health behaviour). The seminal work on
operationalisations for these constructs by Champion (1984) [
] provides the basis for
Healthcare 2023,11, 583 3 of 14
Medical studies have been using HBM for decades in order to measure constructs’
association with diseases and health behaviours, such as breast cancer [
], or vaccine
]. For example, studies on breast cancer find that a person’s adherence to
mammography is significantly associated with greater perceived susceptibility, lower barriers,
higher benefits, and cues in the form of recommendations from healthcare providers .
The HBM has also been applied to SARS-CoV-2 to explore adherence to precautionary
measures and preventive behaviours (e.g., face mask-wearing, social distancing) in several
countries, such as Ethiopia [
] or China [
]. Other studies analyse the acceptance of SARS-
CoV-2 vaccines in various populations, such as Malaysia [
]. Important predictors of a
deﬁnite intention to take a vaccine are high perceived beneﬁts and lower perceived barriers
to receiving the vaccine, and higher levels of perceived susceptibility to infection . An
important contribution of our study is the explicit focus on the form of the disease, namely
an infectious disease, for with such a disease, not only the person’s own behaviour and
health matters but also the behaviours and interests of others. Covering this aspect is one
of the key parts of adapting and developing HBM constructs and hypotheses in the latter
parts of this paper.
HBM in the Context of Contact Tracing Apps
Prior work on the HBM in the context of contact tracing apps can be divided into
two groups. The first group of work employs HBM-based items in their survey but did not
further examine the actual influence of these constructs on use [
]. The second group of
research uses similar constructs to those in the HBM but does not explicitly refer to the HBM
to evaluate the reasons for using the CWA in Germany [
]. For example,
Munzert et al.
(2021) find an increased risk of severe illness, which is relatable to the construct of perceived
severity in the HBM, to encourage a higher uptake rate of CWA. To the best of our knowledge,
there are only two studies which use the HBM in the context of tracing apps while focusing
on a specific country, such as Belgium [
] or France [
]. However, in contrast to our
study, which explains user behaviour, these studies focus on intentions or potential use.
Several articles in the IS literature show that there is a so-called intention-behaviour gap
which might introduce biases in the results of such studies, especially when dealing with
privacy-related judgements [
]. Thus, by having a dependent variable based on quotas,
our analysis advances the results of prior work as we can significantly reduce the bias due
to self-reported information and social desirability bias.
3. Research Hypotheses
Against the backdrop of the related work and prior literature on contact tracing apps
(see Section 2), we introduce our adapted HBM for the case of the German CWA in this
section. We needed to adjust the model at several points in order to address that we are
dealing with an infectious disease which also has immediate medical consequences on
other people (unlike, for example, cancer as in the original operationalisation by Champion
(1984)). This led to several changes in the constructs themselves as well as in the items
of the constructs (see Appendix Afor the respective items). We will discuss them in the
following subsections and derive the respective hypotheses based on the HBM.
3.1. Perceived Susceptibility
Perceived susceptibility is taken to a large extent from Champion (1984), whereas
we left out three of the original six items and replaced them with two items reﬂecting the
infectious aspect of the disease covering the likelihood of getting infected on the job and,
more generally, in daily life. Following the notion of the HBM, individuals are more likely
to use disease prevention tools such as the CWA if they think that there is a high likelihood
of becoming infected with SARS-CoV-2. Thus, we hypothesise:
The perceived susceptibility (PS) of getting SARS-CoV
2 has a positive
impact on the likelihood of using the CWA.
Healthcare 2023,11, 583 4 of 14
3.2. Perceived Severity
Perceived severity was originally operationalised as a 12-item construct which is
appropriate for diseases such as cancer (Champion 1984). We decided to split it into three
distinct dimensions to cover the speciﬁcs of an infectious disease which can also harm others
and can have certain social effects on social relationships and the career. Thus, we evaluate
the perceived medical consequences on oneself (corresponding to the analysis of other non-
infectious diseases investigated in the literature with the HBM), the perceived severity of
the medical consequences when one becomes infected and endangers others, such as family
members, friends and work colleagues and the perceived social consequences (non-medical)
that come along with infection with SARS-CoV-2, including ﬁnancial and career aspects as
well as risks to social relationships [
]. The respective items for medical consequences on
oneself and related to social consequences are taken from Champion (1984), except for the
third item of perceived severity related to social consequences, which we added to account
for social relationships. The construct of perceived severity of medical consequences on
others is self-made. The different dimensions of severity represent a threat for individuals
which should, according to the HBM, cause them to engage in disease prevention, i.e.,
increase the likelihood of using the CWA, leading to the following hypotheses:
The perceived severity of medical consequences (PMC) on oneself when
getting SARS-CoV-2 has a positive impact on the likelihood of using the CWA.
The perceived severity of medical consequences on others (PSO) by
getting SARS-CoV-2 has a positive impact on the likelihood of using the CWA.
The perceived severity of social consequences (PSC) when getting SARS-
CoV-2 has a positive impact on the likelihood of using the CWA.
3.3. Perceived Beneﬁts
We changed the construct perceived beneﬁts compared to the original construct as we
needed to adapt it to disease prevention with its speciﬁc characteristics from doing self-
breast exams (Champion 1984) to using the CWA. According to the HBM, the individual’s
beliefs regarding the effectiveness and perceived beneﬁts of reducing the disease threat are
important factors inﬂuencing the use of disease preventives. Thus, we hypothesise:
The perceived beneﬁts (PB) of using the CWA have a positive impact on
the likelihood of using the CWA.
3.4. Perceived Barriers
Perceived barriers are described as the potential negative aspects of a particular health
action which impede the implementation of disease preventives. A cost-benefit analysis
occurs wherein an individual weighs the action’s expected benefits with perceived barriers.
Thus, it can be stated that “[
. . .
] combined levels of susceptibility and severity provide the
energy or force to act and the perception of benefits (minus barriers) provide a preferred
path of action” [
]. We differentiate these barriers into two types covering the technical
knowledge and understanding of the app and privacy concerns. We relied on two new
constructs operationalising these dimensions as they are very specific to the health action.
On the one hand, there were several reports of technical difficulties surrounding the
introduction of the app in Germany (e.g., not having a suitable smartphone or operation
]. Thus, we include a self-made construct covering this aspect. On the other
hand, privacy concerns were heavily discussed prior to the introduction of the CWA in
]. In addition, prior work shows the importance of privacy in the adoption of
contact tracing apps (see Section 2). Thus, we included this construct as well. The items
for this construct are adapted from prior work [
]. Thus, the resulting hypotheses are
Healthcare 2023,11, 583 5 of 14
Hypothesis (H6). The perceived lack of technical knowledge and understanding (PTB) of
the CWA has a negative impact on the likelihood of using the CWA.
The perceived privacy concerns (PC) related to the CWA have a negative
impact on the likelihood of using the CWA.
3.5. Cues to Action
An internal or external motivational drive is necessary to produce health-promoting
behaviours. The intensity of motivation needed to trigger the decision-making process
to use the CWA is another important factor in the HBM next to perceived susceptibility,
severity, beneﬁts, and barriers [
]. As cues to action were, to the best of our knowledge, not
operationalised before for the case of infectious disease, we had to develop it ourselves.
We split cues to action (named “motivation” by Champion (1984)) into two separate
constructs to be able to differentiate the potentially different impacts of intrinsic motivation
and extrinsic motivation on individual behaviour [
]. Intrinsic motivation reﬂects an
individual’s inner drive to use the CWA to combat the disease (including registering
positive test results and regularly checking for risk encounters). In contrast, extrinsic
motivations are drivers coming from external events such as infections of close relatives,
the overall number of infections in Germany and the world, as well as incentives for the
social life or work. The resulting hypotheses are as follows:
The intrinsic motivation (IM) to avoid COVID-19 infections for oneself
or others has a positive impact on the likelihood of using the CWA.
The extrinsic motivation (EM) to avoid COVID-19 infections for oneself
or others has a positive impact on the likelihood of using the CWA.
Based on the adapted HBM and our research hypotheses, we develop the research
model and describe the data collection and processing in the section.
4.1. Research Model
Based on our research hypotheses derived from the HBM, we develop the following
logistic regression model to assess the likelihood of an individual using the CWA based
on the dependent observable variable CWA
(user of CWA = 1, non-user = 0). We also
include several control variables in order to address the issue of endogeneity due to omitted
variables. Thus, we include age, gender, education, income, smartphone experience and
the health condition of participants (whether they themselves or family/close relatives
belong to a COVID-19 risk group (Equation (1)).
CWAUSE =β0+β1PS +β2PMC +β3PSO +β4PSC +β5PB +β6PTB +β7PC +β8IM +β9EM +βjControlj(1)
4.2. Data Collection
We considered the ethical issues of the user study before collecting the data. The
ethics board of the authors’ university provides an extensive checklist, and our study was
qualiﬁed as exempt from an ethics review and therefore approved. We conducted the
online study with a certiﬁed panel provider in Germany from the end of December 2020 to
January 2021 (the panel provider is certiﬁed following the ISO 20252 norm). The survey was
programmed with the survey software LimeSurvey (version 2.72.6) [
] and hosted on a
university server. We sampled the participants in a way to achieve a representative sample
for Germany. For that purpose, we set quotas to end up with approximately 50% females
and 50% males in the sample as well as a distribution of age following the EUROSTAT2018
]. Furthermore, we set a quota to end up with half of the sample using the CWA
Healthcare 2023,11, 583 6 of 14
and the other half not using it. This quota serves as the dependent variable for our later
logistic regression analysis.
First, we descriptively discuss our sample and the factors of the HBM. Following
this analysis, we discuss the results of the binomial logistic regression. All analyses were
conducted in R (version 4.0.3).
5.1. Descriptive Analysis
Our sample consists of 1752 participants. Following EUROSTAT 2018, participants
are representative of the German population with respect to age and gender. The same
diversity can be observed for income and education (Table 1). Some 896 participants use the
CWA (51.14%), and 856 do not (48.86%): 1299 use Android (74.14%), 436 use iOS (24.89%)
and 17 stated to use other mobile operating systems (0.97%). This distribution of operating
systems is representative for Germany .
Table 1. Sample Demographics († 5 GCSEs at Grade C and above).
Demographics N % Demographics N %
18–29 years 371 21.17% Female 894 51.03%
30–39 years 316 18.04% Male 853 48.69%
40–49 years 329 18.78% Diverse 4 0.23%
50–59 years 431 24.60% Prefer not to say 1 0.05%
60 years and older 305 17.41% Education
Net income No degree 8 0.46%
500€−1000€160 9.13% Secondary school 187 10.67%
1000€−2000€402 22.95% Secondary school † 574 32.76%
2000€−3000€404 23.06% A levels 430 24.54%
3000€−4000€314 17.92% Bachelor’s degree 240 13.70%
More than 4000€292 16.67% Master’s degree 285 16.27%
Prefer not to say 180 10.27% Doctorate 28 1.60%
Since we deliberately divided the sample into two approximately equal groups (CWA
users and non-users), we need to check for statistically signiﬁcant differences in the de-
mographics between these groups. This is required in order to rule out the confounding
inﬂuences of these variables on our results. We conducted a Shapiro–Wilk test for normality
for all variables, and the variables are all not normally distributed. Thus, we conducted
Wilcoxon rank-sum tests to analyse whether there are signiﬁcant differences between CWA
users and non-users. Age and gender show no differences since we deliberately sampled
our participants with equal distributions with respect to age and gender. There are statisti-
cally signiﬁcant differences between users and non-users of the CWA for the remaining
demographics. The income is statistically signiﬁcantly higher for the users compared to
the non-users. However, the median is the same, which is why we argue that the absolute
difference is not having a substantial confounding effect on our later analysis. The same
argumentation holds for education, with a median of 4 for users and 3.5 for non-users and
smartphone experience in years, with a mean 8.77 for users and 8.35 for non-users. We also
calculate descriptive statistics for the used constructs. The means and standard deviations
are shown in Table 2. As we measured the answers on a 7-point Likert scale, values equal
to 4 indicate that participants neither agree nor disagree with the statements. We can see
that the mean values for all variables from the HBM are statistically signiﬁcantly different
Healthcare 2023,11, 583 7 of 14
between users and non-users of the CWA, except for the construct perceived severity
related to social consequences. All variables of the HBM have values and differences, as
one would expect between the groups.
Table 2. Descriptive Statistics for the Constructs.
Variable (Cronbach’s ∝)
CWA 0 1
Signiﬁcance of Difference
N = 856
N = 896
Perceived Susceptibility (∝= 0.79) 3.40 (1.35) 3.81 (1.31) <0.001
Perceived Medical Consequences (∝= 0.87) 3.86 (1.49) 4.49 (1.17) <0.001
Perceived Severity on Others (∝= 0.83) 4.32 (1.64) 4.92 (1.42) <0.001
Perceived Social Consequences (∝= 0.74) 3.04 (1.53) 3.12 (1.52) 0.308
Perceived Beneﬁts (∝= 0.90) 3.05 (1.35) 4.72 (1.22) <0.001
Perceived Technical Barriers (∝= 0.88) 2.74 (1.41) 1.60 (1.07) <0.001
Privacy Concerns (∝= 0.96) 4.64 (1.70) 2.57 (1.52) <0.001
Intrinsic Motivation (∝= 0.89) 3.25 (1.68) 5.97 (1.18) <0.001
Extrinsic Motivation (∝= 0.85) 2.33 (1.27) 3.41 (1.45) <0.001
Susceptibility and the medical severity dimensions (on oneself and others), and thereby,
the overall judgement on the potential threat of SARS-CoV-2, are signiﬁcantly higher for
users of the CWA. However, it must be noted that the perceived likelihood of getting SARS-
CoV-2 (susceptibility) is not perceived as pertinent to the users of the CWA as well. In
contrast, the severity related to medical consequences on oneself and on others is assessed
as more critical by users, whereas non-users are indifferent. A clearer difference between
the groups can be seen in the beneﬁts of the CWA (users tend to agree that it has beneﬁts
and non-users rather disagree), the privacy concerns and the intrinsic motivation. Both
groups tend to disagree that there are technical barriers and that they are extrinsically
motivated (although statistically signiﬁcantly different). In order to assess the reliability of
the constructs, we calculated polychoric correlations between the items of the respective
constructs and used this matrix to conduct a factor analysis. Polychloric correlations are
usually used for categorical variables [
]. The loadings for each item of the constructs are
at least 0.5, with an explained variance of at least 50.3% or higher. Values for Cronbach’s
Alpha are shown in Table 2. Thus, the reliability of the constructs is established, and we
can use the variables as they are for the regression analysis.
5.2. Regression Analysis
We conduct the logistic regression analysis in the following way. First, based on our
ﬁnding that the constructs are valid and reliable, we calculate the mean sum scores of the
constructs in order to have a single construct which is usable in the regression analyses.
Second, we calculate the base model only, including the constructs of the HBM. Third, we
calculated the full model with demographics with a sample size of 1571 as we needed to
exclude participants who preferred not to provide their gender or income. The results of
the regression analyses are shown in Table 3.
In order to evaluate and discuss the results of the logistic regression model, we ﬁrst
derive a reduced model (see Equation (2)) with all statistically signiﬁcant variables and
estimate whether the full model explains the dependent variable better than the reduced
one based on the likelihood ratio test. We have to reject this hypothesis which indicates
that the reduced model, and consequently all variables shown as signiﬁcant, explain
the use of the CWA with the same or better accuracy. In line with these results, the
Bayesian information criterion (BIC), as well as the Akaike information criterion (AIC),
Healthcare 2023,11, 583 8 of 14
are lower for the reduced model compared to the full model (BIC
= 1096.611 versus
BICfull = 1176.289/AICreduced = 1032.298 versus AICfull = 1036.943).
Table 3. Logistic Regression Model on Using the CWA (*** p< 0.001, ** p< 0.01, * p< 0.05).
(1) Base Model
(N = 1752)
(2) Full Model
(N = 1571)
DV: Are you using the Corona-Warn-App on your smartphone? (y/n)
Perceived Susceptibility 0.177 *
Perceived Medical Consequences −0.080
Perceived Severity on Others −0.163 **
Perceived Social Consequences 0.052
Perceived Beneﬁts 0.210 **
Perceived Technical Barriers −0.631 ***
Privacy Concerns −0.329 ***
Intrinsic Motivation 0.749 ***
Extrinsic Motivation 0.416 ***
Income 5 (€500 to €1000) −0.831 **
Age, gender, education, smartphone
experience, SARS-CoV-2 risk group
(participant herself or family/close relatives)
As we conduct a logistic regression analysis, we need to evaluate overdispersion,
which can happen if there is more variance in the data in comparison to the assumed
distribution. We check for overdispersion by calculating the ratio of the residual deviance
and the residual degrees of freedom of the logistic model. Literature suggests that the ratio
should not be considerably larger than 1 [
]. The ratio in our case is 0.647, indicating that
overdispersion is not an issue. As there is no such metric as R
for logistic regressions, we
use McFadden’s R
as a pseudo R
measure. McFadden’s R
ranges from 0 to 1 with values
from 0.2 to 0.4, indicating an “excellent ﬁt” [
]. In our case, the reduced model has a
equal to 0.536, indicating an excellent predictive power of the independent
variables. We additionally conducted two tests to assess the statistical signiﬁcance and
relevance of each independent variable. The Wald test indicates that all variables in the
reduced model contribute to the explanatory power of the model (p< 0.05 for all variables
of the reduced model). Furthermore, we assessed the relative importance based on the
t-values of the variables, which also conﬁrms that all variables in the reduced model are
signiﬁcant (t-values > 1.96 for all variables of the reduced model). Finally, we evaluate
multicollinearity, which occurs when independent variables are correlated with each other.
We plot the pairwise correlations and ﬁnd that intrinsic motivation and perceived beneﬁts
might pose a problem regarding collinearity (see Figure A1, Appendix B). We will consider
the potential effect of this correlation in the discussion of the results.
Healthcare 2023,11, 583 9 of 14
In order to interpret the statistically signiﬁcant coefﬁcients, we calculate the average
marginal effects for the reduced model, as the coefﬁcients themselves are not interpretable
as they are (see Table 4). We refer to the marginal effects in the following discussion as they
can be measured in percentage points and can be interpreted in isolation.
Marginal effects (average partial effects) of the independent variables of the reduced model.
DV: CWA Use (y/n) Marginal Effect
Perceived Susceptibility 0.017
Perceived Severity on Others −0.020
Perceived Beneﬁts 0.021
Perceived Technical Barriers −0.063
Privacy Concerns −0.035
Intrinsic Motivation 0.074
Extrinsic Motivation 0.047
Income 5 (€500 to €1000) −0.098
We ﬁnd that participants with a higher perceived susceptibility, higher perceptions
with respect to the beneﬁts of the CWA as well as a higher intrinsic and extrinsic motivation,
have an increased probability of using the CWA. We ﬁnd the lowest marginal effect of
0.017 for perceived susceptibility, corresponding to a 1.7 percentage point increase. The
highest marginal effect can be observed for intrinsic motivation, with a one-unit increase in
it corresponding to a 7.4 percentage points increase in the probability of using the CWA.
The variables which have a negative effect on the probability of using the CWA are the
perceived severity of others, perceived technical barriers, privacy concerns, and income.
The largest negative effects can be observed for perceived technical barriers (6.3 percentage
point decrease in the probability of using the CWA) and privacy concerns (3.5 percentage
point decrease in the probability of using the CWA).
Interestingly, the hypotheses for none of the other threat dimensions of the HBM
(severity of medical consequences on oneself, others and social consequences) can be
conﬁrmed. We even have to reject H3 (positive effect of perceived severity on others) as the
marginal effect indicates that a one-unit increase in this variable decreases the probability
of using the CWA by 2 percentage points if participants feel that infection with the virus
harms others. This is interesting as the concept of warning others in the case of an infection
is the key function of a contact tracing app.
Our results indicate that the use of the German CWA app is mainly inﬂuenced by
ﬁve variables. The likelihood of using the CWA is positively inﬂuenced by participants’
intrinsic motivation and extrinsic motivation. Interestingly, several variables representing
threat dimensions of SARS-CoV-2 are not signiﬁcant (e.g., perceived medical consequences)
or have relatively small effect sizes (perceived susceptibility). Perceived severity on others
has a signiﬁcant (small) negative effect. The latter ﬁnding is surprising since the mean
of perceived severity on others is signiﬁcantly larger for CWA-Users than for Non-CWA-
Users (cf. Table 2). Further research will be required to determine if it is an artefact of the
regression or indicates a serious effect. The general understanding of participants that
infections of others are a reason for using the CWA can be seen in the positive effect of
extrinsic motivation on use (although it has to be noted that participants tend to disagree
in both groups to items of this construct (see Table 2)).
The three other variables with the largest effects are perceived technical barriers,
privacy concerns and income. This indicates that there is a relatively high intrinsic mo-
tivation with a large effect on app use on the one hand (also visible by the large mean
of 5.77 in the group of users), but this factor is attenuated by technical barriers and a
Healthcare 2023,11, 583 10 of 14
lack of understanding (e.g., explanations missing, compatibility with a smartphone not
given) and privacy concerns. In addition, our results indicate that socioeconomic factors
might inﬂuence app use as a lower income tends to decrease the likelihood of using the
app in our sample, whereas a higher level of education increases the likelihood. These
ﬁndings partially correspond to prior work showing that the older population and those
in lower-income households might also face technical barriers to using the app, e.g., by
having limited smartphone access or an old smartphone [
]. This ﬁnding is also discussed
by other researchers pointing towards the larger theme of increasing inequality during the
pandemic as citizens who do not have the respective resources or who do not want to en-
gage in certain behaviours (e.g., because of missing information about the technology or the
disease) are basically shut out from employing all possible resources to protect themselves
and others [
]. In the context of demographic inﬂuences, it is interesting that neither age
nor gender has a signiﬁcant effect. The same holds for the variables indicating whether
participants themselves or close relatives and family members belong to a SARS-CoV-2 risk
group. For the risk group variables, the same argumentation could be made for perceived
severity on others, namely that participants do not perceive a contract tracing app as a
necessary disease prevention tool in their inner social circles.
These results are important for policymakers as we heard many discussions in Germany
which were based on the idea that there could be a kind of trade-off between a “proper” contact
tracing app and other measures such as curfews, lockdowns or social distancing. It was often
argued that a contact tracing app with less focus on privacy and more functionalities could
enable more freedom [
]. However, our results indicate that this idea is highly questionable
not only because of the simple fact that less privacy (if there is anything like this) does not
cause the app to function better [
]. The results even indicate that greater privacy concerns
could negatively impact the likelihood of the adoption of the CWA.
All these ﬁndings indicate that people are mainly driven by an inner motivational
drive and responsibility to ﬁght the disease, which could only be harmed by high barriers
in the form of a more intrusive app collecting unnecessary amounts of data or technical
barriers such as compatibility issues.
Limitations and Future Work
This study has certain limitations. First, although we work with an observed depen-
dent variable, the independent variables are self-reported, which could cause biases (e.g.,
due to the social desirability bias). Second, due to the contextual change of considering an
infectious disease in the HBM, we had to adapt existing items (e.g., perceived beneﬁts) and
develop new items (e.g., extrinsic motivation). These are not validated, and thus do not
have the same maturity as the original items in the well-established HBM and may be less
reliable. However, in order to be able to report recent results about a COVID-19-related
topic, we decided against validating the new items ﬁrst. Third, the analysis did not include
interaction effects between variables. However, we plan to conduct a second wave of the
survey in order to conduct a longitudinal analysis of the changes in user behaviour of
the CWA along with changing infection numbers. In the consequent analysis, interaction
effects between HBM factors and demographics will be investigated in the context of a
larger study. This is necessary as our study indicates a certain socio-demographic inﬂuence
on the use of contact tracing apps. An analysis of interaction effects between the factors
of the HBM and variables reﬂecting the social status of individuals (such as education or
income) is a highly interesting avenue for future work as it addresses the question: to what
extent can policymakers expect certain social groups to participate in pandemic-ﬁghting
tools? Such ﬁndings from IS research could even inform research on herd immunity in the
context of vaccinations, which requires the majority of a society to participate in order to
come back to normality. Lastly, every country had very speciﬁc types of contact tracing
apps with accompanying political and social discussions. These speciﬁc features and dis-
cussions shape the perceptions of individuals in the respective countries, causing a limited
generalisability of our results for contact tracing apps in other countries.
Healthcare 2023,11, 583 11 of 14
We conducted a representative user study with 1752 participants in Germany on the
use of the German contact tracing app (Corona-Warn-App). We sampled participants into
two groups (users of the app, N = 896 and non-users, N = 856). We base our research on
the health belief model, which explains why individuals engage in disease prevention
behaviours. We adapted the model which was originally applied in the medical context
(e.g., mammography for breast cancer prevention) to infectious disease and the prevention
behaviour of using a contact tracing app.
We ﬁnd that the probability of using the CWA is mainly positively inﬂuenced by the
intrinsic and extrinsic motivation of individuals and negatively inﬂuenced by perceived
technical barriers, privacy concerns, and a lower income. We contribute to the current
stream of literature on the adoption of contact tracing apps, disease prevention tools in
general, and the use of the HBM in the context of technologies by introducing an adapted
HBM suited to investigate infectious diseases and new information systems to cope with
the associated challenges of those diseases. Furthermore, we argue that our results from the
IS domain can contribute to a better understanding of users and, in particular, suggest that
the perceived technical barrier should be reduced, e.g., by providing further elucidations.
It has to be determined in future work whether the results can be transferred to other
behaviours or use of measures in the pandemic (e.g., the willingness to get a vaccination or
to obey rules in the pandemic for reducing contact and hygiene) and which factors would
foster an environment for generating understanding of and participation in the measures.
Author Contributions: Conceptualization, D.H. and S.P.; methodology, D.H.; formal analysis, D.H.;
investigation, D.H. and L.T.M.; writing—original draft preparation, D.H. and L.T.M.; writing—review
and editing, D.H. and S.P.; project administration, D.H. and S.P.; funding acquisition, D.H. All authors
have read and agreed to the published version of the manuscript.
This research was funded by the Goethe-Corona-Fonds of Goethe University Frankfurt
am Main. The APC was funded by the “OA-Publikationsfonds” of Goethe University Frankfurt am
Institutional Review Board Statement:
The study was conducted in accordance with the Declaration
of Helsinki and approved by the Ethics Committee of Goethe University Frankfurt am Main.
Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.
Data Availability Statement:
The data presented in this study are available on request from the
Conﬂicts of Interest:
The authors declare no conﬂict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or
in the decision to publish the results.
All items were measured on a 7-point Likert scale and adapted from Champion (1984),
if not otherwise indicated.
1. Perceived susceptibility (PS)
PS1. My chances of being infected by COVID-19 are great.
PS2. My physical health makes it more likely that an infection by COVID-19 will have
PS3. I worry a lot about being infected by COVID-19.
PS4. My job involves a high risk to be infected by COVID-19.
PS5. My daily life involves a high risk to be infected by COVID-19.
2. Perceived severity related to medical consequences on oneself (PMC)
PMC1. The thought of COVID-19 scares me.
PMC2. COVID-19 is a hopeless illness.
Healthcare 2023,11, 583 12 of 14
PMC3. Health issues I would experience from COVID-19 would last a long time.
PMC4. If I got infected by COVID-19, it would be more serious than other diseases.
PMC5. If I had COVID-19 my whole life would change.
3. Perceived severity on others (PSO) (self-made)
PSO1. The health of my friends would be at risk if I became infected with COVID-19.
PSO2. My family’s health would be at risk if I became infected with COVID-19.
PSO3. The health of my work colleagues would be at risk if I became infected with COVID-19.
PSO4. The health of other contacts would be at risk if I became infected with COVID-19.
4. Perceived severity related to social consequences (PSC)
PSC1. My ﬁnancial security would be at risk if I became infected with COVID-19.
PSC2. If I became infected with COVID-19, my career would be at risk.
PSC3. My social relationships (family, friends) would be at risk if I became infected
5. Perceived beneﬁts (PB)
PB1. Using the Corona-Warn-App makes me feel safer.
PB2. I have a lot to gain by using the Corona-Warn-App.
PB3. The Corona-Warn-App can help me to identify contacts to infected individuals.
PB4. If I use the Corona-Warn-App I am able to warn others in case I am infected with COVID-19.
PB5. I feel that the usage of the Corona-Warn-App is beneﬁcial to combat COVID-19.
6. Perceived technical barriers (PTB) (self-made)
PTB1. I am afraid to use the Corona-Warn-App because I don’t understand how it works.
PTB2. I don’t know how to go about using the Corona-Warn-App.
PTB3. Installing the Corona-Warn-App takes too much time.
PTB4. The installation of the Corona-Warn-App is associated with technical problems.
7. Perceived barriers related to privacy concerns (PC) 
PC1. I think the Corona-Warn-App gathers far too much of my personal information.
PC2. I worry that the Corona-Warn-App leaks my personal information to third-parties.
PC3. I am concerned that the Corona-Warn-App violates my privacy.
PC4. I am concerned that the Corona-Warn-App misuses my personal information.
PC5. I think that the Corona-Warn-App collects my location data.
8. Intrinsic motivation (IM) (self-made)
IM1. I feel responsible to register a positive test result into the Corona-Warn-App.
IM2. I feel responsible to use the Corona-Warn-App regularly to inform myself of potential
9. Extrinsic motivation (EM) (self-made)
EM1. Potential infections of family members or friends with COVID
19 affect my decision
to use the Corona-Warn-App.
EM2. The number of COVID-19 infections in Germany affects my decision to use the
EM3. The number of COVID-19 infections in the world affects my decision to use the
EM4. Using the Corona-Warn-App has advantages for my social life.
EM5. The use of the Corona-Warn-App is requested for my work.
Age, gender, education, income, smartphoneexperience, Oneself or close relatives/family
COVID-19 risk group (y/n)
Healthcare 2023,11, 583 13 of 14
Figure A1. Pairwise correlations between the latent independent variables.
Eurosurveillance Editorial Team Note from the Editors: World Health Organization Declares Novel Coronavirus (2019-NCoV)
Sixth Public Health Emergency of International Concern. Euro Surveill. 2020,25, 2019–2020. [CrossRef]
Reelfs, J.H.; Poese, I.; Hohlfeld, O. Corona-Warn-App: Tracing the Start of the Ofﬁcial COVID-19 Exposure Notiﬁcation App for
Germany. arXiv 2020, 19–21. [CrossRef]
Hinch, R.; Probert, W.; Nurtay, A.; Kendall, M.; Wymant, C.; Hall, M.; Lythgoe, K.; Bulas, C.; Zhao, L.; Stewart, A.; et al. Effective
Conﬁgurations of a Digital Contact Tracing App: A Report to NHSX. Available online: https://cdn.theconversation.com/static_
ﬁles/ﬁles/1009/Report_-_Effective_App_Conﬁgurations.pdf?1587531217 (accessed on 27 April 2021).
Becker, K.; Feld, C. Was Heißt Zentralisiert Oder Dezentral? Available online: https://www.tagesschau.de/inland/coronavirus-
app-101.html (accessed on 23 March 2021).
Neuerer, D.; Olk, J.; Hoppe, T. Politiker Stellen Strengen Datenschutz Der Corona-Warn-App Infrage. Available online:
der-corona-warn-app-infrage/26570478.html (accessed on 10 March 2021).
Schar, P. Corona: Datenschutz Vor Menschenleben? Available online: https://www.eaid-berlin.de/corona-datenschutz-vor-
menschenleben/ (accessed on 11 March 2021).
Harborth, D.; Pape, S. A privacy calculus model for contact tracing apps: Analyzing the German Corona-Warn-App. In
ICT Systems Security and Privacy Protection. SEC 2022. IFIP Advances in Information and Communication Technology; Meng, W.,
Fischer-Hübner, S., Jensen, C.D., Eds.; Springer International Publishing: Cham, Switzerland, 2022; Volume 648, pp. 3–19, ISBN
Pape, S.; Harborth, D.; Kröger, J.L. Privacy concerns go hand in hand with lack of knowledge: The case of the German
Corona-Warn-App. In Proceedings of the ICT Systems Security and Privacy Protection. SEC 2021, Oslo, Norway, 22–24 June
2021; pp. 1–14.
9. Rosenstock, I.M. Historical Origins of the Health Belief Model. Health Educ. Behav. 1974,2, 328–335. [CrossRef]
Walrave, M.; Waeterloos, C.; Ponnet, K. Adoption of a Contact Tracing App for Containing COVID-19: A Health Belief Model
Approach. JMIR Public Health Surveill. 2020,6, 1–10. [CrossRef]
Guillon, M.; Kergall, P. Attitudes and Opinions on Quarantine and Support for a Contact-Tracing Application in France during
the COVID-19 Outbreak. Public Health 2020,188, 21–31. [CrossRef]
Blom, A.G.; Wenz, A.; Cornesse, C.; Rettig, T.; Fikel, M.; Friedel, S.; Möhring, K.; Naumann, E.; Reifenscheid, M.; Krieger, U.
Barriers to the Large-Scale Adoption of the COVID-19 Contact-Tracing App in Germany: Survey Study. J. Med. Internet Res.
23, e23362. [CrossRef] [PubMed]
Green, E.C.; Murpy, E.M.; Grybosky, K. The Health Belief Model. The Wiley Encyclopedia of Health Psychology. Cambridge Handb.
Psychol. Health Med. Second Ed. 2020,2, 97–102.
Glanz, K.; Rimer, B.K.; Viswanath, K. Health Behavior and Health Education: Theory, Research, and Practice; APA: Washington, DC,
USA, 2002; ISBN 9780787996147.
Healthcare 2023,11, 583 14 of 14
15. Champion, V.L. Instrument Development for Health Belief Model Constructs. Adv. Nurs. Sci. 1984,6, 73–85. [CrossRef]
16. Champion, V.L. Instrument Reﬁnement for Breast Cancer Screening Behaviors. Nurs. Res. 1993,42, 139–143. [CrossRef]
Champion, V.L. Revised Susceptibility, Beneﬁts, and Barriers Scale for Mammography Screening. Res. Nurs. Health
22, 341–348. [CrossRef]
Lee Champion, V. Use of the Health Belief Model in Determining Frequency of Breast Self-examination. Res. Nurs. Health
8, 373–379. [CrossRef] [PubMed]
Smith, P.J.; Humiston, S.G.; Marcuse, E.K.; Zhao, Z.; Dorell, C.G.; Howes, C.; Hibbs, B. Parental Delay or Refusal of Vaccine
Doses, Childhood Vaccination Coverage at 24 Months of Age, and the Health Belief Model. Public Health Rep.
Gerend, M.A.; Shepherd, J.E. Predicting Human Papillomavirus Vaccine Uptake in Young Adult Women: Comparing the Health
Belief Model and Theory of Planned Behavior. Ann. Behav. Med. 2012,44, 171–180. [CrossRef] [PubMed]
21. Champion, V.L.; Skinner, C.S. The Health Belief Model. Health Behav. Health Educ. Theory Res. Pract. 2008,4, 45–65.
Yehualashet, S.S.; Asefa, K.K.; Mekonnen, A.G.; Gemeda, B.N.; Shiferaw, W.S.; Aynalem, Y.A.; Bilchut, A.H.; Derseh, B.T.; Mekuria,
A.D.; Mekonnen, W.N.; et al. Predictors of Adherence to COVID-19 Prevention Measure among Communities in North Shoa
Zone, Ethiopia Based on Health Belief Model: A Cross-Sectional Study. PLoS ONE 2021,16, 1–16. [CrossRef]
Tong, K.K.; Chen, J.H.; Yu, E.W.Y.; Wu, A.M.S. Adherence to COVID-19 Precautionary Measures: Applying the Health Belief
Model and Generalised Social Beliefs to a Probability Community Sample. Appl. Psychol. Health Well-Being
Wong, L.P.; Alias, H.; Wong, P.F.; Lee, H.Y.; AbuBakar, S. The Use of the Health Belief Model to Assess Predictors of Intent to
Receive the COVID-19 Vaccine and Willingness to Pay. Hum. Vaccines Immunother. 2020,16, 2204–2214. [CrossRef] [PubMed]
Jonker, M.; de Bekker-Grob, E.; Veldwijk, J.; Goossens, L.; Bour, S.; Mölken, M.R. Van COVID-19 Contact Tracing Apps: Predicted
Uptake in the Netherlands Based on a Discrete Choice Experiment. JMIR mHealth uHealth 2020,8, 1–15. [CrossRef]
Horstmann, K.T.; Buecker, S.; Krasko, J.; Kritzler, S.; Terwiel, S. Who Does or Does Not Use the “Corona-Warn-App” and Why?
Eur. J. Public Health 2021,31, 49–51. [CrossRef] [PubMed]
Munzert, S.; Selb, P.; Gohdes, A.; Stoetzer, L.F.; Lowe, W. Tracking and Promoting the Usage of a COVID-19 Contact Tracing App.
Nat. Hum. Behav. 2021,5, 247–255. [CrossRef]
Crossler, R.E.; Johnston, A.C.; Lowry, P.B.; Hu, Q.; Warkentin, M.; Baskerville, R. Future Directions for Behavioral Information
Security Research. Comput. Secur. 2013,32, 90–101. [CrossRef]
Adjerid, I.; Peer, E.; Acquisti, A. BEYOND THE PRIVACY PARADOX: OBJECTIVE VERSUS RELATIVE RISK IN PRIVACY
DECISION MAKING. MIS Q. 2018,42, 465–488. [CrossRef]
Clair, R.; Gordon, M.; Kroon, M.; Reilly, C. The Effects of Social Isolation on Well-Being and Life Satisfaction during Pandemic.
Humanit. Soc. Sci. Commun. 2021,8, 1–6. [CrossRef]
von Wyl, V.; Hoeglinger, M.; Sieber, C.; Kaufmann, M.; Moser, A.; Serra-Burriel, M.; Ballouz, T.; Menges, D.; Frei, A.; Puhan,
M. Are COVID-19 Proximity Tracing Apps Working under Real-World Conditions? Indicator Development and Assessment of
Drivers for App (Non-) Use. medRxiv 2020. [CrossRef]
Fraunhofer AISEC Pandemic Contact Tracing Apps: DP-3T, PEPP-PT NTK, and ROBERT from a Privacy Perspective.
Cryptol. ePrint Arch. Rep. 2020. Available online: https//eprint.iacr.org/2020/489 (accessed on 10 July 2022).
Gu, J.; Xu, Y.; Xu, H.; Zhang, C.; Ling, H. Privacy Concerns for Mobile App Download: An Elaboration Likelihood Model
Perspective. Decis. Support Syst. 2017,94, 19–28. [CrossRef]
Harborth, D.; Pape, S. Investigating Privacy Concerns Related to Mobile Augmented Reality Applications – A Vignette Based
Online Experiment. Comput. Human Behav. 2021,122, 106833. [CrossRef]
Vallerand, R.J. Toward A Hierarchical Model of Intrinsic and Extrinsic Motivation. Adv. Exp. Soc. Psychol.
36. Schmitz, C. LimeSurvey Project Team. Available online: http://www.limesurvey.org (accessed on 10 July 2022).
37. EUROSTAT. EUROSTAT. 2018. Available online: https://ec.europa.eu/eurostat/de/home (accessed on 22 January 2021).
statista Marktanteile Der Führenden Mobilen Betriebssysteme an Der Internetnutzung Mit Mobiltelefonen in Deutschland von
Januar 2009 Bis September 2020. Available online: https://de.statista.com/statistik/daten/studie/184332/umfrage/marktanteil-
der-mobilen-betriebssysteme-in-deutschland-seit-2009/ (accessed on 3 February 2021).
Holgado–Tello, F.P.; Chacón–Moscoso, S.; Barbero–García, I.; Vila–Abad, E. Polychoric versus Pearson Correlations in Exploratory
and Conﬁrmatory Factor Analysis of Ordinal Variables. Qual. Quant. 2010,44, 153–166. [CrossRef]
40. Hinde, J.; Demétrio, C.G.B. Overdispersion: Models and Estimation. Comput. Stat. Data Anal. 1998,27, 151–170. [CrossRef]
McFadden, D. Quantitative Methods for Analyzing Travel Behavior of Individuals: Some Recent Developments; EconPapers: Örebro,
Storeng, K.T.; de Bengy Puyvallée, A. The Smartphone Pandemic: How Big Tech and Public Health Authorities Partner in the
Digital Response to COVID-19. Glob. Public Health 2021,16, 1482–1498. [CrossRef]
The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.