ARTICLE IN PRESS

Social Science & Medicine 61 (2005) 14391451

www.elsevier.com/locate/socscimed

How can trust be investigated? Drawing lessons

from past experience
Jane Goudgea,, Lucy Gilsona,b
b

a
Centre for Health Policy, School of Public Health, University of Witwatersrand, SA
Health Economics and Financing Programme, London School of Hygiene and Tropical Medicine, UK

Available online 28 January 2005

Abstract
Although the concept of trust has gained popularity in public debate and academic analysis over recent years,
it continues to be regarded by many as difcult to dene and so to investigate. In this paper we provide guidance
on how to conduct future work on trust in the health sector, by reviewing the methods used in earlier studies. The
paper draws on a range of the available literature which investigates trust in different settings from a variety of
disciplinary perspectives.
The review suggests that appropriate denitions of trust are highly context dependent. Where little is known about
how trust functions, qualitative research to explore how respondents view trust and trusted behaviour is important in
advance of quantitative investigation. The results of qualitative inquiry facilitate the development and renement of
hypotheses about how trust functions and can be used to generate questions for use in structured questionnaires.
Quantitative inquiry is valuable because it allows larger scale investigation and generates data that can be used, for
example, to assess the statistical signicance of different determinants to overall levels of trust. The review indicates that
trying to use existing data to answer a new research question, without ensuring that respondents answers refer to the
form of trust under investigation, may lead to the generation of use of inappropriate data. In this respect, it highlights
the need to test pre-existing research tools to ensure that they remain valid and relevant and retain their reliability in
different settings. Although there may appear to be common structural features of specic relationships that might
allow international measurement tools to be developed for use in the same type setting across geographical locations,
cultural differences might invalidate such tools. Specic investigation of such differences is, therefore, needed to explore
the potential validity of international tools for investigating trust.
r 2005 Elsevier Ltd. All rights reserved.
Keywords: Trust; Methods; Measurement

Introduction
The concept of trust has become popular in public
debate and academic analysis. In high income countries

Corresponding author. Tel.: +27 11 489 9940;

fax: +27 11 489 9900.
E-mail address: jane.goudge@nhls.ac.za (J. Goudge).

this interest is associated with concern for the decline of

trust in governments and professionals, and in developing countries has been prompted by debates around the
notion of social capital. For the health sector, trust has
long been recognised as crucial to the patient/provider
relationship (Parsons, 1951). Recent work has also
begun to highlight its potential value in understanding
the performance of health care organisations (Goold,
2001) and health systems (Gilson, 2003).

0277-9536/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2004.11.071

ARTICLE IN PRESS
1440

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Yet the notion of trust is often regarded as

ambiguousdifcult to dene and to investigate. Trust
has only recently begun to be measured and analysed
systematically in the health sector, and almost no
empirical investigation of this kind has so far been
conducted in developing countries. Gilson (2003) presented a synthesis of theoretical perspectives, looking at
the meaning, bases and outcomes of trust and its
relevance to the health system. This companion paper
reviews the methods applied in investigating trust,
drawing on literature from different contexts and
disciplinary perspectives. Seeking to encourage rigorous
empirical investigation of the role of trust in a range of
health system relationships and across countries, this
paper describes methodological approaches that have
been used in such studies, drawing out guidance about
how to conduct future work on trust.
Two search strategies were used to identify Englishlanguage papers that provided a reasonable level of
detail about their methods. First, two data bases
covering the health and broader social science elds
(Medline and the Bath Information and Data Services
Bibliography of the Social Sciences) were searched using
a wide range of terms. No period was specied but this
search primarily identied papers from 1993-2002.
Abstracts were reviewed and those papers discussing
their methods in adequate detail were selected. Second,
additional references were identied either through the
initial searches or as a result of discussions with
knowledgeable people. All the selected publications
were then reviewed and a core sample of papers
that provided sufcient detail about different types
of methodological approaches in different contexts, or
about important conceptual issues, were chosen for
review. Where papers used the same overall methodological approach, only those illustrating key differences
were included in this core sample. More health papers
were selected than those from other elds. This paper,
therefore, presents an illustrative but comprehensive
review of the major methodological approaches that
have so far been used in investigating trust across
contexts.
The remainder of the paper has three sections.
Section 2 summarises the settings in which trust has
been investigated and highlights some of the key
similarities and differences in how trust is dened.
Section 3 then examines the main methodological
approaches used in investigating trust. Finally, Section 4
summarises the key lessons drawn from this review.

types of research questions investigated in each setting

and the outcomes found to be associated with trust.
The table shows that questions and outcomes differ
quite substantially within settings. However, trust is
often found to support co-operation, with this cooperation enabling the achievement of a range of
outcomes. In health care, trusting provider-patient
relationships may generate positive therapeutic outcomes by encouraging patient disclosure and adherence
to treatment recommendations (e.g. Thorne & Robinson,
1989). In other settings trust outcomes include low levels
of crime, greater levels of collective efcacy, more stable
personal relationships, greater levels of informal exchange and greater business productivity.
Across contexts, where dened, there is also reasonable consensus that trust is understood as a judgement in
a situation of risk that the trustee will act in the best
interests of the truster, or at least in ways that will not be
harmful to the truster. From their review, Hall, Dugan,
Zheng, and Mishra (2001, p. 615) conclude that the
majority understand trust as the optimistic acceptance
of a vulnerable situation in which the truster believes the
trustee will care for the trusters interests (italics in
original).1 In some cases this judgement is seen as a
calculated decision, and in others as an intuitive or
affective response (Gilson, 2003). Where spelt out, there
also seem to be similarities across settings in the
dimensions of inter-personal trust identied, with a
strong focus on dimensions such as competence,
honesty, benevolence and integrity. These demonstrate
the general conclusion that trusting attitudes are
directed as much to motivations and intentions as to
results (italics in original, Hall et al., 2001, p. 616).
For example, distinguishing between generalised
and inter-personal trust, Larzxelere and Huston (1980,
pp. 595596) dene the former as a persons belief
about the character of people in the aggregate and the
latter as a belief in the integrity of another individual.
The denition of inter-personal trust is, however,
routinely adapted to the behaviours relevant in a specic
context. From the organisational management literature, Cummings and Bromiley (1996) dene interpersonal trust as the good faith efforts of another
individual or group to behave in accordance with
commitments, to be honest in the negotiations preceding
these commitments and not to take excessive advantage
of another even when the opportunity is available. In
business relationships, denitions of trust are also
generally based on notions of condence in the
capability, reliability and/or integrity of an exchange
partner (e.g. Lyon, 2000; Rindeisch, 2000; Deakin &
Wilkinson, 1998). However, the particular behaviours

Does the context matter in investigating trust?

Table 1 outlines the six settings from which the papers
identied for review were drawn, giving examples of the

1
Yet, at the same time, trust may hold dangerssuch as
exploitation, the abuse of power and the formation of criminal
gangs (Gilson, 2003).

ARTICLE IN PRESS
J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

1441

Table 1
Examples of research questions and outcomes of investigation of trust in a range of different settings
Setting

Research questions or objective

Outcome

Trust in interpersonal relationships

Does a communicated, conciliatory

strategy that includes some retaliation
lead to higher levels of cooperation,
than 50% cooperative behaviour carried
out randomly, in a conictual situation
between two individuals (a simulated
game)? (Lindskold, 1979)
Does the nature of trust in intimate
personal relationships differ according
to length of relationship? (Larzxelere &
Huston, 1980)

Conciliatory strategy (with some

retaliation) led to higher levels of
cooperation

Generalised trust within the community

Trust in government and other

institutions

Trust in institutions, responsible for

providing services to the public

Does trust and shared norms lead to

openness and compromise, better
governance, and higher level of
economic development? (Widner &
Mundt, 1998)
Is there a link between social
environment and healthy citizens?
(Kawachi et al., 1999)
Is collective efcacy linked to reduced
violence, and does mutual trust and
solidarity lead to willingness to
intervene and collective efcacy?
(Sampson et al., 1997)
Does pyschological sense of community
have the properties of a construct?
(Glynn, 1981)
Does civic education lead to increased
or reduced trust? (Finkel, Sabatini, &
Bevis, 2000)
Do administrators and politicians in a
relatively new democracy (Tanzania)
have more or less trust than those
working within a traditional and stable
democracy (Norway)? (Jacobsen, 1999);
Does the importance placed on access
to health affect the publics trust in key
actors in the health policy community?
(Kehoe & Ponting, 2003)
Is there an association between trust
and self-reported adherence, patient
satisfaction, and improved health
status? (Safran et al., 1998)
How do patientdoctor relationships
change over time for patients with
chronic illnesses? (Thorne & Robinson,
1989)
What is the effect of patient-centred
beliefs of physicians on patient trust?
(Krupat, Bell, Kravitz, Thom, & Azari,
2001)
Does patients knowledge of physician
payment incentives (capitation or fee for
service) affect patient trust?(Hall et al.,
2002)

Trust varies with level of commitment,

being lowest for ex-partners and highest
for those engaged and living together,
for newly weds, and for those married
over 20 years
Correlation was shown between trust
and voluntarism, optimism and
religiosity, but not with better
governance and economic development
Mistrust was shown to be correlated
with violence and property crime
A measure of collective efcacy is
negatively associated with variations in
violence

The construct of psychological sense of

community showed statistical reliability
Civic education increases knowledge
and critical perspectives reducing trust
Norwegian administrators exhibit a
higher degree of trust than Tanzanian
administrators. National differences are
more important in explaining variations
in trust than other explanatory variables
The value placed on access to health
care by health boards was shown to be a
powerful predictor of public trust
No link was found with health status,
but trust was highly correlated with
satisfaction with the physician and
adherence to treatment
Na ve trust is replaced by
disenchantment, and then
reconstructed trust
When physicians and patients hold
similar beliefs, patients are more likely
to trust their physicians
Increasing knowledge of payment
incentives does not, in the short term,
reduce trust in physicians or insurers

ARTICLE IN PRESS
1442

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Table 1 (continued )
Setting

Research questions or objective

Outcome

How does trust in health providers

differ across different communities with
varying levels of social capital? (Ahern
& Hendryx, 2003)

Although membership of an HMO is

generally associated with lower doctor
trust, this effect is not uniform across all
communities due to mediating factors
such as social capital. Social capital
plays a role in how health care is
perceived by citizens, and how health
care is delivered by providers
Patients from racial and ethnic minority
groups have less positive perceptions of
their physicians on at least 2 important
dimensions of trust
Interpersonal competence, involving
care, concern and compassion, listening,
technical competence were central foci,
with response varying by disease
The trust necessary for informal
exchanges to take place was guaranteed
either through the use of brokers or a
common social context, particularly the
work place
Differences in trust between suppliers
(private sector consultancy, business
associations and government support
agencies) and recipients of advice were
shown
Some support for thesis that trust is
conducive to good business
performance

Is a persons race or ethnicity associated

with low trust in the physician?
(Doescher et al., 2000)

Trust in business relationships between

companies or individuals in the context
of either formal contracts, or informal
exchange relationships

To gain insights into the dynamics of

trust and to identify concepts for use in
surveys with more representative
samples (Mechanic & Meyer, 2000)
Is trust necessary for informal exchange
to take place? Compared informal
exchange of favour, services, goods and
information in St. Petersburg, and
Helsinki (Lonkila, 1997)
Does business advice and type of
interaction with advisor differ between
types of suppliers of advice? (Bennett &
Robson, 1999)
Do trusting relations between rms
generate cooperation and/or benecial
business outcomes, such protability?
(Sako, 1998)
Do small-scale rural produces and
traders in less developed countries rely
on trust in their exchange relations?
(Lyon, 2000)
Does trust play a role in inter-rm
collaboration in a range of industries,
including the health sector? (Goddard &
Mannion, 1998)

Trust within organisations between

employers and employees, and between
co-workers

Does trust facilitate acceptance of an

authoritys decisions? (Tyler & Degoey,
1996)
Does trusting a supervisor improve an
employees motivation (and if so how)?
(Barton Cunningham & MacGregor,
2000)
Do social norms and political systems
inuence employee behaviour and
attitudes ? (Pearce, Branyiczki, &
Bigley, 2000)

expected in the context of health care provision are

different. They include technical competence, communication and listening abilities, maintaining condentiality, honesty and an impartial concern for the patients
well-being (e.g. Mechanic & Meyer, 2000; Straten,

Resource poor farmers and traders

draw on existing trust networks,
allowing them to enter into new markets
and increase incomes
Both theory and empirical evidence
suggest that co-operation and trust can
play a central role in the efcient
organisation of contractual
arrangements
Treatment with respect and dignity
provides a basis for trust and therefore
acceptance of authorities decision
Predictability, benevolence and fairness
in supervisors behaviour are highly
correlated with trust by employees
The practice of rewarding employees on
the basis of who they are rather than
impersonal judgements of performance
affected employees trust of one another

Friele, & Groenewegen, 2002). The vulnerability associated with being ill may specically lead trust in medical
settings to have a stronger emotional component than in
other settings, where it might be more based on a
calculated judgement (Hall et al., 2001).

ARTICLE IN PRESS
J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Moreover, even within one setting the dimensions of

trust may differ in relation to its object. Trust in health
providers is, for example, largely based on experience of
the providers behaviour and the patients own inclination to trust (Thorne & Robinson, 1989). Yet, although
partly rooted in trust in inter-personal trust between
patients and providers, trust in health care organisations
is also linked to aspects of organisational functioning.
Goold (2001), for example, suggests that trust in health
care organisations might be partly based on protection
from the catastrophic costs of illness and the health of
neighbours as well as oneself. Birungi (1998) distinguishes trust in the provider from trust in the expert
systems that provide the basis for judging that providers
will act in the patients best interests (such as the
certication associated with professional training).
Using the concept of public trust in health care, Straten
et al. (2002) combine consideration of inter-personal,
organisational and system trust.
It is important to note that trust is not always clearly
dened. Only one paper was identied which established
a denition drawn from the local language of the study
setting (Lyon, 2000). This gap may reect an assumption
that the notion is broadly understood (e.g. Sampson,
Raudenbush, & Earls, 1997), or that the investigation
aims to identify how trust is understood by respondents
(e.g. Mechanic & Meyer, 2000). The distinction between
trust and trustworthiness is also rarely considered,
although the two concepts are different. We may trust
people who are not trustworthy because we have too
little knowledge of their behaviour or are mis-led by
signals like their reputation. Butler (1991) provides an
interesting example of an inquiry that sought to measure
the conditions leading to trust in a manager rather than
the construct of trust itself. He judges this approach as
providing a more useful basis for future managerial
intervention to develop trust.
Overall, the differences in research questions, outcomes and denitions across settings suggest that
investigation of trust always requires an understanding
of the specic social, personal and political processes
surrounding trust. Trust in medical providers, for
example, has been found to be higher in communities
with established health care institutions that have
longstanding leadership, that provide a training site for
professionals within the community, have established
community leadership on health issues and have
collaborative relationships with other sectors (Ahern &
Hendryx, 2003). In relation to generalised trust,
Yamagishi, Cook, & Watabe (1998) show a clear
difference between US and Japanese social relations
that is likely to affect the nature and extent of trust in
the different countries. The closer-knit and tighter social
relations in Japan are associated with lower levels of
trust in others outside the immediate circle, whereas the
more open and less rigid relations in the US are

1443

associated with higher levels of trust in others. Therefore, [a] clear understanding of trust for a trustee
necessitates understanding how the context affects
perceptions of trustworthiness (Mayer & Davis, 1995,
p. 728).

How are different types of research questions

investigated?
Only a small number of papers aim primarily to
explore the meaning of trust in a particular setting or for
a particular group of respondents; these primarily use
qualitative data collection and analysis approaches.
Instead, most articles measure trust or trustworthiness
using structured survey tools and then either consider
the statistical correlation between trust and an outcome variable, or examine how the level of trust or
trustworthiness differs across contexts. Some papers
place emphasis on developing a reliable tool to measure
trust in one, or multiple, settings. Others use existing
tools and/or existing data to test hypotheses. Finally, a
few studies use experimental methods in investigating
the dimensions of trust.
Using qualitative methods to explore trust and its role
The majority of studies using purely qualitative
research methods to investigate trust used semistructured interviews (e.g. Mechanic & Meyer, 2000),
although some used ethnographic techniques (Newell &
Swan, 2000) and one, a diary of social relations in which
the respondents recorded who they met each day and
what they discussed with each other (Lonkila, 1997).
Only one study used a range of data collection
approaches, in order to broaden the types of opportunities to gather data (Lyon, 2000).
Three papers looking at the relationship between
patient and health care provider are examined in greater
detail to draw out lessons for future work.
Example 1
Thom & Campbell (1997) explored 29 patients views
on the factors shaping trust, using four two-hour focus
group discussions. The participants were asked to
describe situations they had experienced that led them
to trust a physician, and situations that had caused them
to lose, or not establish, trust (p. 171). Using grounded
theory techniques to analyse the interview transcripts,
nine factors determining trust were identied. These
were: thoroughness in evaluation, understanding the
patients individual experience, caring, providing appropriate and effective treatments, communicating clearly
and completely, partnership building, honesty/respect
for the patient, structural/stafng factors.

ARTICLE IN PRESS
1444

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Example 2
Mechanic & Meyer (2000) conducted single semistructured interviews with 90 patients selected from
three different chronic illness groups. Questions, developed from an initial literature review, were open ended
to allow the subjects to describe in their own words the
most meaningful aspects of their relationships with their
doctor(s). The interviews showed that patients viewed
trust as an iterative process, testing physicians against
their own knowledge and expectations. The particular
nature of the patients disease affected their expectations; for example condentiality was more important
for patients with a stigmatised disease.
Example 3
In a 3 year, in-depth study conducted with 77
respondents with a chronic illness, Thorne & Robinson
(1989) explored the evolution of patientdoctor interactions, using grounded theory methodology. The analysis
reveals a common three-stage process. Initial na ve trust
was shattered due to unmet expectations and conicting
perspectives with health providers, leading to disenchantmenta stage of extreme anxiety, frustration and
profound mistrust. Eventually, in the resolution stage,
an alternative form of trust was reconstructed on a more
guarded basis. This reconstructed trust was contingent
on revised expectations of the roles of both patients and
provider in managing chronic illness. The relationships
of reconstructed trust tended to fall into four categories:

nature of Thorne and Robinsons study, and the

grounded approach to analysis, generated a fuller
understanding of different categories of patientdoctor
relationships and how they changed over time. Few
empirical studies have directly explored its dynamic
nature.
These examples illustrate that the data collection
approach used in qualitative work can have a signicant
impact on study ndings, yielding varying levels of
understanding about trust. Given the complexity of
trust, and the possibility of unsystematic interpretation,
it is also crucial to have condence in the rigour of the
analysis. Yet many papers provided little or no
information with which to assess the rigour, and only
a few explained the coding process and the methods used
to check consistency and reliability. Two papers provide
a level of detail useful for other researchers. Mechanic
and Meyer (2000) report analysing interview material
using codes developed from a prior literature review, as
well as additional codes derived from the material itself.
Coding reliability was assessed through using multiple
coders to consider the same transcript. Thom and
Campbell (1997) coded interview transcripts using
grounded theory techniques. Four readers (from different disciplinary backgrounds) labelled statements from
the focus group discussions. Statements were then
grouped into conceptual categories by consensus, with
coding changes being made as the material was coded,
before the nal version was reviewed by the researchers.

resignation:

Developing a measurement tool

Across setting, papers report similar approaches to

developing trust tools.

very little trust; patient reluctantly

accepts health care;
consumerism: patients transform their frustration
into insights about the professional values and
attitudes of health workers, in order to explain unmet
expectations;
team playing: patient recognises joint roles of the
patient and doctor, and reciprocal trust built;
hero worship: one particular provider is trusted and
praised, without denying the aws in the whole
system.

Using only focus group discussions, Thom and

Campbells study is limited by the much smaller body
of material and thinner personal detail than the other
two studies. The method did not allow an analysis of
how the different factors identied as inuencing trust
actually contributed to it, or how trust between doctor
and patient changed over time. In contrast, the
Mechanic and Meyer in-depth, individual interviewing
approach allowed patients time to elaborate on the
nature of their relationship with their doctor. This
generated a more detailed understanding of the iterative
process of trust building, and some sense of how the
disease type inuenced trust. However, the longitudinal

Developing the tool

Table 2 presents details of six tools developed to
measure different trust objects in the health sector. Five
of these tools were developed in the US, three by the
Wake Forest team, with the health system tool being
developed in the Netherlands.
The approach used in developing the Dutch tool can
be compared with that used by the Wake Forest team. In
both, an initial denition of trust relevant to the specic
object of inquiry was established, and then several
subsequent, but different, steps generated a questionnaire tool.
The rst phase of work in the Dutch study involved
short telephone interviews in which 125 informants were
asked: what aspects do the general public regard as part
of the concept of public trust in the health care system?
An initial questionnaire was derived from grounded
analysis of the data collected. Eight categories of topics
were identied, each of which was assumed to be a
dimension of trust. A specic set of statements was

Table 2
Characteristics of tools developed to investigate the health sector
Trust object

Trust in primary care

physician (USA)

Trust in physician scale

(USA)

Patient Trust Scale

(Physician trust) (USA)

Medical profession (USA) Health insurance

organisation (USA)

Author

Hall et al. (2002) (Wake

Forest team)

Anderson and Dedrick,

1990

Kao, Green, Zaslavsky,

Koplan, & Cleary (1998)

Hall et al., (2002) (Wake

Forest team)

Content (No. of
items)

Fidelity (2)

Fidelity (3)

Fidelity (3)

Fidelity (2)

Competence (3)
Honesty (1)
Global (4)

Competence (3)
Honesty (1)
Condentiality (1)
Global (2)

Competence (6)
Condentiality (1)
Global (1)

Competence (2)
Honesty (2)
Global (5)

Competence (2)
Honesty (3)
Condentiality (1)
Global (4)

Policies at the macro level (6)

Providers competence (6)
Quality of care (9)
Information supply &
communication (6)
Quality of cooperation (3)

10
3
2
1

11
0
4
1

10
0
10
1

11
3
2
1

11
6
1
1

36
0
0
6

0.93 (National sample)

0.85

0.93

0.89

0.92 (National sample)

0.88, 0.87, 0.74, 0.85, 0.87, 0.80 (with

respect to list above)

1. [your doctor] will do

0.89 (Regional sample)

1. I doubt that my doctor 1. XXX cares more about How much do you trust

whatever it takes to get

really cares for me as a
you all the care you
person
need
2. My doctor us usually
considerate of my
2. sometimes [your
doctor] cares more
needs and puts them
about what is
rst
convenient for him/her 3. I trust my doctor to put
my medical needs
than about your
above all other
medical needs
considerations when
3. [Your doctor] only
thinks about what is
treating my medical
best for you.
problems

saving money than

about getting you the
treatment that you
need

your physician(s)y

1. Doctors in general care Patient focus of providers:

about their patients

health just as much or
more as their patients
1. To put your health and
do
well-being above
2. Sometimes doctors care
keeping down the
more about what is
health plans costs
convenient for them
2. To make appropriate
than about their
medical decisions
patients medical needs
regardless of health
plan rules and
guidelines?
3. To perform necessary
medical tests and
procedures regardless
of cost?

I have absolute condence that:

1. Doctors will take their patients

seriously

2. Doctors will pay sufcient

attention to their patients

3. Doctors will listen to their patients

4. Doctors will spend enough time
on their patients

5. Doctors will always stick up for

their patients

patients problems

1445

6. Doctors will understand their

ARTICLE IN PRESS

0.92 (Regional sample)

Fidelity questions

Zheng, Hall, Dugan, Kidd, Straten et al., 2002

& Levine (2002) (Wake
Forest team)
Fidelity (1)
Patient focus of providers (6)

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

No. of items
Total
Negative
Using trust
No. of factors/
dimension
Cronbach alpha
score

Whole health system (The

Netherlands)

ARTICLE IN PRESS
1446

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

identied for each category and included in the

questionnaire. The questionnaire was then completed
by a random sample of 1050 members of the national
health care consumer panel survey. Factor analysis of
their responses was used to determine which questions
contributed to the underlying dimensions of trust,
and these were then selected as the items for the
nal tool.
The Wake Forest tools all have their foundation in a
conceptual framework derived from a literature review.
The team generated a draft questionnaire by, rst,
reviewing existing tools against the conceptual framework to identify candidate questions for inclusion or
adaptation, as well as to identify where new questions
were needed. An expert panel of academicians from
relevant elds and from community groups was then
asked to comment on whether the questions in the tool
represented their understanding of provider trust and
whether the items were clearly and easily understood.
Each tool was subsequently piloted through eight
interviews, followed by de-brieng sessions using openended, cognitive (or think aloud) interview techniques.
Information from the expert panel, focus group discussions and initial pilot tests were used to conrm and
rene the conceptual model and to create, modify or
delete candidate questions. Next, a set of 290300
interviews was conducted with individuals from a
range of community groups and socio-economic backgrounds in order to allow selection of items for inclusion
in a draft questionnaire. This revised questionnaire
was then used in national or regional telephone surveys
with between 500 and 1200 randomly selected respondents. Finally, principal components analysis was
undertaken to assess the dimensionality of the tools,
removing items with a low item-to-total correlation from
the nal tool.

The content of tools

Table 2 provides details of the tools developed
through these approaches. The Dutch tool contains
around three times the number of items of the Wake
Forest tools and includes multiple scales of trust. In
contrast, the Wake Forest tools are all uni-dimensional,
suggesting that people do not distinguish between
different dimensions of trust in making judgements
about trust in medical providers or insurers. Unlike the
Dutch tool, the Wake Forest tools include some negative
questions, to prevent automatic responses, and avoid
using the word trust as much as possible or similar
concepts such as condence, to ensure that they measure
an independent concept rather than each respondents
own denition of trust.
Table 2 also suggests that the content of the
tools differs due to differing trust objects. Although
condentiality is part of the Wake Forest insurer tool,

it was dropped from the same teams primary care

and medical profession tools, suggesting condentiality
is a more important component of insurer trust.
At the same time, some of the content is common
across tools, although under different headings and
using different questions. Table 2 outlines the
questions used across these six tools in investigating
delity.
Testing the validity
A key step in developing any measurement tool is to
assess its quality and validity. A range of criteria and
tests developed in the psychology and education elds
are generally used: face, content and criterion validity,
internal consistency and stability (Brooks, 1995). In
general, individual questions are excluded from ndings/
tools if they do not pass the tests, and the remaining
results are then used to draw conclusions about, or
investigate, trust. Failure to conduct the tests not only
compromises the validity of a tool, but prevents other
researchers from assessing the potential value of
adapting the tool to another setting.
Several studies reported measuring trust without
doing any qualitative work or psychometric tests to
determine the validity and reliability of the tool. Only a
few papers (e.g. Straten et al., 2002) used qualitative
work to test the face and content validity of the tool.
The majority of articles reported testing the internal
consistency of the respondent answers, through for
example Cronbach Alpha tests and factor analysis.
(Papers summarised in Table 2 reported Cronbach
Alpha scores ranging from 0.74 to 0.93, the high value
indicating high levels of consistency across answers.)
Few (e.g. Baer et al., 1999) tested the stability of the tool
with variation in the time of the interview and
interviewer. The Wake Forest teams detailed discussion
of test ndings, and of which questions did not produce
reliable results, and why, is unusual and provides an
example that other studies could follow. However, it
should be noted that lower levels of consistency between
a question and the rest of the tool may not indicate that
the item offers less of a contribution to the respondents
understanding of trust. Removal of the item might result
in loss of data relating to an important and unexplored
facet of trust. Qualitative inquiry could allow consideration of how important the item is to respondents and, if
relevant, allow its retention in the tool despite low
statistical correlation.
Assessing factors that influence trust
A further aspect of validity testing is whether a tool
is sensitive to variation in the characteristics of
respondents or communities in which it is applied
(discriminant validity). Sensitivity to variation is affected by sample size. For example, in studies with

ARTICLE IN PRESS
J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

relatively small sample sizes (under 1000 respondents)

race is not identied as a predictor (e.g. Hall, Camacho,
Dugan, & Balkrishnan, 2002), but in larger studies (with
thousands of respondents) race and other demographic
variables were all found to be signicant predictors (e.g.
Ahern & Hendryx, 2003). Indeed, analysis of the study
with the largest sample size (n 32; 929) suggests that
race is signicant even after controlling for education
and income (Doescher, Saver, Franks, & Fiscella, 2000).
Such comparisons can inform sampling decisions of
future studies.
Using pre-existing tools and/or data
The use of pre-existing tools requires that original
questions are tested to ensure that they remain valid and
relevant, and that the tool retains its reliability, in the
new setting. An example of the problems created by
failing to conduct such assessment is provided by
Widner and Mundt (1998). They used the US General
Social Survey question on generalised trust within the
community to explore the relationship between trust,
governance, and economic development in Uganda and
Botswana. (Some say that most people can be trusted.
Others say you cannot be too careful in your dealings
with people. How do you feel about it?) However, the
question tends to lead interviewees to respond in terms
of trust in strangers, rather than on relationships that
had the potential to contribute to economic development, as required by the focus of the study. In contrast,

1447

using the same question in a study of trust, safety, crime

and tolerance of diversity in the US, Kawachi, Kennedy,
and Wilkinson (1999) found that mistrust was signicantly correlated with violent crime, and with unemployment and poverty.

Experimental studies
A very few studies apply experimental methods in
their investigation of trust as shown in the examples of
Box 1.
These methods gather information on inter-personal
trust from participants around an organised event,
game, or experiment. After the intervention, further
data are collected to assess whether it changed the level
of inter-personal trust. The value of a laboratory style
approach is that contextual inuences over trust are
reduced, as the experiment removes individuals from
their organisational or institutional context. The approach throws light on the inherent qualities of interpersonal trust by examining the effect of the intervention
on trust, while controlling for contextual inuences.
However, such an approach appears to be of limited use
where the inuence of the setting is an important
determinant of trust. The approach of Hall, Dugan,
Balkrishnan, and Bradley (2002) offers an interesting
alternative because the intervention took place in a real
setting, comparing responses between an intervention
and control group.

Box 1
Experimental approaches to investigating trust

Trust (or co-operation) has been investigated through a task or game approach. Dirks (1999)
employed a tower building task to investigate trust. Each group of participations is asked to build a
tower from wooden blocks, with each person in the group using blocks of a unique colour. The
objective is to place as many blocks on the tower as possible (a) as an individual and (b) as a group.
Trust was manipulated by providing participants with differing perceptions about the other team
members. The authors counted the number of blocks on each tower, reported the number of times the
tower fell and computed an efficiency measure as a ratio of the groups actual performance to its
expected performance. Each group was scored for a range of variables such as co-ordination, helping,
intensity of effort, direction of effort and commitment to a plan. Data was used in regression analysis to
examine the effect of the game and the manipulations on trust.
Hall et al. (2002) investigate how the knowledge of physicians financial incentives affects patient
trust. They distributed information about physicians payment incentives (capitation and fee-for service
payments) to two groups of members of healthcare management organisations. The two intervention
groups were compared with a control group who were given no information. Greater knowledge about
incentives did not have a negative impact on patients trust of their physician, suggesting that patients
had already taken into account the effect of financial incentives in the assessment of physicians
trustworthiness.

ARTICLE IN PRESS
1448

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Fig. 1. Best practice research strategies for investigating trust.

ARTICLE IN PRESS
J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Conclusion
This paper provides an overview of the range of
approaches that have been used in investigating trust in
different settings. Few detailed investigations of trust
have so far been undertaken in the health sector, and
most in the USA. Further investigation in other country
contexts will be important in allowing wider understanding of the relevance and role of trust in health
settings. A gap of current inquiry is the limited
investigation of the dynamics of trust in health sector
relationships, its evolution over time and the inuences
over that evolution. Inquiry into health management
issues might draw insights from the literature on
generalised trust within communities and government
(in considering how trust inuences the performance of
health organisations and providers), and on trust in and
within organisations (in considering trust as an inuence
over health provider performance).
Fig. 1 summarises the conclusions drawn from this
review about best practices for investigating trust. In any
inquiry, the rst steps must involve clarifying the
rationale for investigating trust in a particular setting
and the main research questions. It is critical to establish
a denition of trust relevant to the specic setting and
trust object. Existing experience and conceptual literature offer valuable insights at this stage.
Qualitative investigation is important in developing a
detailed understanding of how respondents view trust
and trustworthy behaviour in the setting of focus. Such
inquiry is, as demonstrated in Section 3.1, useful in its
own right. It generates insights into the complexity of
trust and its links with other factors in specic settings.
A qualitative inquiry can also be used to generate
structured questions for inclusion in a questionnaire
tool. Its sensitivity to context may generate tools that
capture a greater degree of the complexity around
the notion, than if only conceptual literature is used.
Such inquiry can allow hypotheses about how trust
functions within the relationship of focus to be
developed or rened, helping to avoid inappropriate
assumptions being made on the basis of experience from
other settings.
Quantitative inquiry is valuable because it allows
larger scale investigation. It generates data that can be
used to assess the statistical signicance of different
determinants to levels of trust, and the association of
trust with different outcomes. Such inquiry can be
conducted using a specically designed measurement
tool, pre-existing questions or experimental studies. Best
practice is likely to involve the rst of these approaches,
building on earlier phases of conceptual and qualitative
investigation. Careful attention should be paid to the
sampling approach used in developing the tool to ensure
that it does not bias conclusions around trust. Rigorous
validity testing of such tools should be conducted and

1449

reported, to generate knowledge about the suitability of

particular tools in specic settings.
Comparative, cross-country work might also be useful
in deepening understanding of the role of trust in health
settings. Since the publication of the 2000 World Health
Organisation World Health Report a wide range of
cross-country data collection activities have been implemented, intended to inform national and international debates about health system development. The
similarity of issues raised across studies might provide a
foundation for an international measurement tool. For
patient-provider trust the common issues include: the
patient focus of the provider, provider competence and
quality of care, communication and co-operation, and
supportive structures and resources. However, it is
critical to consider whether differences between country
contexts might invalidate such an approach. Patient
doctor relationships may, for example, differ with
cultural setting because of differing levels of patient
education, and the role of non-western medical traditions. Further investigation of such differences between
countries would be an essential rst step in exploring
whether an international trust tool can be developed for
use in health settings.

Acknowledgements
Thanks to Annabel Bowden for her initial work on
this review, to Gavin Mooney and three reviewers for
comments. Lucy Gilson is a part-time member of the
Health Economics and Financing Programme of the
London School of Hygiene and Tropical Medicine, UK,
which is funded by the UK Department for International Development (DFID). The views and opinions
expressed here are those of the authors alone.

References
Ahern, M. M., & Hendryx, M. S. (2003). Social capital and
trust in providers. Social Science and Medicine, 57,
11951203.
Anderson, L. A., & Dedrick, R. F. (1990). Development of the
trust in physician scale: a measure to assess interpersonal
trust in patient-physician relationships. Psychological Reports, 67, 10911100.
Baer, L. C., Baldwin, K. A., Sisk, R. J., Watts, P., Grinslade,
M. S., Brockschmidt, B., Dinger, M. K., Marion, L. N., &
McCubbin, J. (1999). Development of an instrument too
measure community acceptance of nurse practitioners and
physician assistants. Journal of Nursing Measurement, 7(1),
6377.
Barton Cunningham, J., & MacGregor, J. (2000). Trust and the
design of work: complementary constructs in satisfaction
and performance. Human Relations, 53(12), 15751591.

ARTICLE IN PRESS
1450

J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451

Bennett, R. J., & Robson, P. J. A. (1999). Intensity of

interaction in supply of business advice and client impact:
a comparison of consultancy, business associations and
government support initiatives for SMEs. British Journal of
Management, 10, 351369.
Birungi, H. (1998). Injections and self-help: risk and trust in
Ugandan health care. Social Science and Medicine, 47(10),
14551462.
Brooks, R. G. (1995). Health status measurement: a perspective
on change. London: Macmillan Press.
Butler, J. (1991). Towards understanding and measuring
conditions of trust: evolution of a conditions of trust
inventory. Journal of Management, 17(3), 643663.
Cummings, L. L., & Bromiley, P. (1996). The organizational
trust inventory (OTI): development and validation. In R. M.
Kramer, & T. R. Tyler (Eds.), Trust in organizations:
frontiers of theory and research. London: Sage Publications.
Deakin, S., & Wilkinson, F. (1998). Contract law and the
economics of inter-organizational trust. In Lane, &
Bachman (Eds.), Trust within and between organizations.
New York: Oxford University Press.
Dirks, K. T. (1999). The effects of interpersonal trust on work
group performance. Journal of Applied Psychology, 84(3),
445455.
Doescher, M. P., Saver, B. G., Franks, P., & Fiscella, K. (2000).
Racial and ethnic disparities in perceptions of physician
style and trust. Archives of Family Medicine, 9, 11561163.
Finkel, S. E., Sabatini, C. A., & Bevis, G. G. (2000). Civil
education, civil society, and political mistrust in a developing democracy: the case of the Dominican Republic. World
Development, 28(11), 18511874.
Gilson, L. (2003). Trust and the development of health care as a
social institution. Social Science and Medicine, 56,
14531468.
Glynn, T. J. (1981). Psychological sense of community:
measurement and applications. Human Relations, 34(7),
789818.
Goddard, M., & Mannion, R. (1998). From competition to cooperation: new economic relationships in the national
health service. Health Economics, 7, 105119.
Goold, S. D. (2001). Trust and the ethics of health care
institutions. Hastings Center Report, 31(6), 2633.
Hall, M., Camacho, F., Dugan, E., & Balkrishnan, R. (2002).
Trust in the medical profession: conceptual and measurement issues. Health Services Research, 37(5), 14191439.
Hall, M., Dugan, E., Zheng, B., & Mishra, A. (2001). Trust in
physicians and medical institutions: what is it, can it be
measured, and does it matter? Milbank Quarterly, 79(4),
613639.
Hall, M. A., Dugan, E., Balkrishnan, R., & Bradley, D. (2002).
How disclosing HMO physician incentives affects trust.
Health Affairs, 21(2), 197205.
Hall, M. A., Zheng, B., Dugan, E., Camacho, F., Kidd, K. E.,
Mishra, A., & Balkrishnan, R. (2002). Measuring patients
trust in their primary care providers. Medical Care Research
Review, 59(3), 293318.
Jacobsen, D. I. (1999). Trust in political-administrative
relations: the case of local authorities in Norway and
Tanzania. World Development, 27(5), 839853.
Kao, A. C., Green, D. C., Zaslavsky, A. M., Koplan, J. P., &
Cleary, P. D. (1998). The relationship between method of

physician payment and patient trust. Journal of the

American Medical Association, 280(19), 17081714.
Kawachi, I., Kennedy, B. P., & Wilkinson, R. G. (1999). Crime,
social disorganization and relative deprivation. Social
Science and Medicine, 48(6), 719731.
Kehoe, S. M., & Ponting, J. R. (2003). Value importance and
value congruence as determinants of trust in health policy
actors. Social Science and Medicine, 57, 10651075.
Krupat, E., Bell, R. A., Kravitz, R. L., Thom, D., & Azari, R.
(2001). When physicians and patients think alike:
patient-centered beliefs and their impact on satisfaction
and trust. The Journal of Family Practice, 50(12),
10571062.
Larzxelere, R. E., & Huston T. L. (1980). The dyadic trust
scale: toward understanding interpersonal trust in close
relationships. Journal of Marriage and the Family, 595604.
Lindskold, S. (1979). Conciliation with simultaneous or
sequential interaction. Variations in trustworthiness and
vulnerability in the prisoners dilemma. Journal of Conflict
Resolution, 23(4), 704714.
Lonkila, M. (1997). Informal exchange relations in post-soviet
Russia: a comparative perspective. Sociological Review
Online 2(2), www.socresonline.org.uk/socresonlime/2/2/
9.html.
Lyon, F. (2000). Trust, networks and norms: the creation of
social capital in agricultural economies in Ghana. World
Development, 28(4), 663681.
Mayer, R. C., & Davis, J. H. (1995). An integrative model of
organizational trust. Academy of Management Review,
20(3), 709735.
Mechanic, D., & Meyer, S. (2000). Concepts of trust among
patients with serious illness. Social Science and Medicine,
51(5), 657668.
Newell, S., & Swan, J. (2000). Trust and inter-organizational
networking. Human Relations, 53(10), 12871328.
Parsons, T. (1951). The social system. London: Routledge and
Kegan Paul.
Pearce, J. L., Branyiczki, I., & Bigley, G. A. (2000). Insufcient
bureaucracy: trust and commitment in particularist organizations. Organization Science, 11(2), 148162.
Rindeisch, A. (2000). Organizational trust and inter-rm
cooperation: an examination of horizontal versus vertical
alliances. Marketing Letters, 11(1), 8195.
Safran, D. B., Kosinski, M., Tarlov, A. R., Rogers, W. H.,
Taira, D. A., Liberman, N., & Ware, J. E. (1998).
The primary care assessment survey: tests of data quality
and measurement performance. Medical Care, 36(5),
728739.
Sako, M. (1998). Does trust improve business performance? In
Lane, & Bachman (Eds.), Trust within and between
organizations: conceptual issues and empirical applications.
New York: Oxford University Press.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997).
Neighborhoods and violent crime: a multilevel study of
collective efcacy. Science, 277, 918924.
Straten, G. F. M., Friele, R. D., & Groenewegen, P. P. (2002).
Public trust in Dutch health care. Social Science and
Medicine, 55, 227234.
Thom, D. H., & Campbell, B. (1997). Patient-physician trust:
an exploratory study. The Journal of Family Practice, 44(2),
169176.

ARTICLE IN PRESS
J. Goudge, L. Gilson / Social Science & Medicine 61 (2005) 14391451
Thorne, S. E., & Robinson, C. A. (1989). Guarded alliance:
health care relationships in chronic illness. Journal of
Nursing Scholarship, 21(3), 153157.
Tyler, T. R., & Degoey, P. (1996). Trust in organizational
authorities: the inuence of motive attributions on
willingness to accept decisions. In R. M. Kramer, & T. R.
Tyler (Eds.), Trust in organizations: frontiers of theory and
research. London: Sage Publications.

1451

Widner, J., & Mundt, A. (1998). Researching social capital in

Africa. Africa, 68(1), 124.
Yamagishi, T., Cook, K. S., & Watabe, J. (1998). Uncertainty,
trust, and commitment formation in the United States and
Japan. American Journal of Sociology, 104(1), 165194.
Zheng, B., Hall, M. A., Dugan, E., Kidd, K. E., & Levine, D.
(2002). Development of a scale to measure patients trust in
health insurers. Health Services Research, 37(1), 187202.