Community Dent Oral Epidemiol 2012 Ó 2012 John Wiley & Sons A/S

What influences the use of dental Zoe Marshman1, Jenny Porritt1, Tom
Dyer1, Ceri Wyborn2, Jenny Godson3
and Sarah Baker1

services by adults in the UK? 1

School of Clinical Dentistry, University of
Sheffield Sheffield, UK, 2Yorkshire and
Humber Public Health Observatory York,
UK, 3Yorkshire and Humber Strategic
Marshman Z, Porritt J, Dyer T, Wyborn C, Godson J, Baker S. What influences Health Authority Leeds, UK
the use of dental services by adults in the UK? Community Dent Oral Epidemiol
2012. © 2012 John Wiley & Sons A/S

Abstract – Objectives: Optimizing access to and utilization of dental services

remains a major public health challenge. The aim of this study was to use
Andersen’s behavioural model to investigate the factors that influence
utilization of dental services and predict oral health outcomes, and to identify
how access could be improved. Methods: Secondary analysis was conducted of
data from a regional postal survey (n = 10 864) of adults in the UK. Items were
chosen to reflect variables of Andersen’s behavioural model including
predisposing characteristics (deprivation), enabling resources (perceived
difficulty accessing a dentist), need (perceived treatment need), health
behaviours (reason for attendance and time since the last visit to the dentist)
and oral health outcomes (oral health impacts (symptoms, functional limitation
and social) and global oral health). Structural equation modelling was used to
estimate the direct and indirect pathways between the variables within the
model. Results: When a combination of indirect and direct effects were taken
into account, perceived difficulty accessing the dentist was associated with
higher perceived treatment need (b = 0.25, P < 0.01), increased oral health Key words: oral health; dental services;
impacts (b = 0.23, P < 0.01) and worse global oral health (b = 0.21, adults
P < 0.01). Overall, the variables included within this model explained 17.4% of Zoe Marshman, Oral Health and
the variance for dental attendance, 55.4% of the variance for the length of time Development, School of Clinical Dentistry,
Sheffield University, Claremont Crescent,
since people had last visited the dentist, 21.7% of the variance for oral health
Sheffield S10 2TA, UK
impacts and 42.9% of the variance for people’s global oral health. Conclusions: Tel.: 0114 2717893
Perceived treatment need and difficulty accessing dental services were found to Fax: 0114 2717843
be key predictors of oral health outcomes. Further research is needed to e-mail: Z.Marshman@sheffield.ac.uk
develop and evaluate effective interventions to improve access to dental Submitted 15 February 2011;
services. accepted 17 January 2012

Optimizing access to and utilization of dental ser-
vices remains a major public health challenge inter-
nationally with significant impacts on individuals,
communities and costs to society (1). For example,
in the UK, access to dental services has been a
political imperative since the 1990s with many pol-
icy initiatives to improve the situation. In 2006, a
new National Health Service (NHS) primary dental
care contract was introduced in England to expand
capacity and ensure services were ‘easy to access’
for patients (2). However, a review of the impact of
the new contract concluded that the aim to
improve access had not been realized with fewer
courses of treatment carried out and fewer people
seeing an NHS dentist (3). Little research has been
conducted in the UK to investigate factors that
influence service use or to identify ways to opti-
mize utilization.
Several studies have investigated influences on
dental service utilization using models of health-
care as theoretical frameworks (4–9). The most
well-known of these are the behavioural models
developed by Andersen et al. (10). The original
model suggested that people’s use of health ser-
vices is dependent on their predisposition to use of
services, factors that enable or impede use and
their need for care. This model was later refined to
describe the interrelationships between population
characteristics (predisposing factors, enabling
resources and need), health behaviours and health

doi: 10.1111/j.1600-0528.2012.00675.x
1
Marshman et al.
Population characteristics
Pre-disposing
characteristics
Deprivation
Enabling resources
Perceived difficulties
accessing a dentist
Need
Perceived treatment
need
Fig. 1. Model of dental service use and oral health outcomes based on Andersen’s behavioural model (1995).
outcomes (Fig. 1) and has been used to investigate
variables that explain service use and ways to
improve it.
Reisine applied the original model to dental ser-
vices utilization to university employees in the
USA assessing predisposing factors (including age,
sex, education, dental attitudes), enabling factors
(family income, participant’s perception of diffi-
culty visiting the dentist), clinical measures of need
and use of dental services (the number of visits
over the past 2 years). She found sex to be the most
influential variable with income and age having no
significant effects, although she acknowledged that
enabling factors may be more important in a gen-
eral population (8).
Andersen’s model has also been tested in rela-
tion to oral health in the general population with
support found for the hypotheses proposed (9).
Baker tested the revised model and the interrela-
tionships between predisposing factors (qualifica-
tions, income, social class), enabling factors (oral
health education, perceptions of treatment
expense) and need and whether their influence on
oral health behaviours (toothbrushing, dental
attendance) and outcomes (oral health-related
quality of life) were direct or mediated based on
data from a national Adult Dental Health Survey
in the UK. She found support for the model but
that the impact of the predisposing factors on need,
use of services and oral health outcomes was indi-
rect. Debates continue about the nature of the path-
ways (direct and indirect) through which
socioeconomic factors influence oral health and the
appropriateness of the different approaches to
measurement taken (area-based indicators versus
income or occupation measures) (11, 12). Baker
2
Health behaviour Outcomes
Perceived
Use of health health status
services
Length of time since Oral health
last visited dentist impacts
Perceived
Use of health health status
services
Reason for dental Global oral
attendance health
suggested further work with Andersen’s models to
explore the contribution of the socioeconomic pro-
file of the individual, the importance of age and
development of the model’s dental public health
applications.
Other studies have also used versions of Ander-
sen’s models to investigate utilization of oral health
services with older people in the USA (5, 6) and
New Zealand (5), young people in Denmark (4)
and adults in sub-Sahara Africa (7) adding further
support for the use of this model.
In summary, support has been found for the
application of Andersen’s behavioural models to
dental service utilization in employed and general
populations using a range of different indicators to
reflect the variables in the model (8, 9). Research is
required to establish the influence of factors, partic-
ularly sex (8), age and socioeconomic status, and to
develop the dental public health application of this
research (9).
The aim of this study was to use Andersen’s
model in a secondary analysis to investigate the
factors that influence access and utilization of den-
tal services and oral health outcomes, from a sur-
vey of adults in the UK, and to identify where
improvements in access could be made. The ver-
sion of Andersen’s model selected was based on
hypothesized pathways that have been supported
within the dental literature.
Materials and methods
In 2008, a postal survey of adults in the Yorkshire
and Humber region of the UK was conducted to
investigate oral health and service utilization (13).
 Influences on dental service use

The questionnaire covered adults’ experience of Table 1. Responses to items included within the Adult
using dental services, self-reported oral health sta- Dental Health Survey (N = 10864)
tus and the impact of the mouth on everyday life. Variable N (%)
The questionnaire was sent to a sample of 25 200
Predisposing factors
adults (16 years and over), representative of those Deprivation (Index of Multiple Deprivation) 10 849
in the region, using the database of patients Least deprived 1447 (13.3)
registered with a general medical practitioner. Less deprived 2503 (23.0)
Nonrespondents were sent two reminders (at 3- to Average 2219 (20.4)
More deprived 2210 (20.3)
4-week intervals). Ethical approval was provided
Most deprived 2470 (22.7)
by Bradford Research Ethics Committee, and Enabling resources – perceived difficulties accessing a
research governance approval was also obtained dentist
from the hosting health organizations. Difficulty accessing routine care 10 769
Yes 2232 (20.5)
Don’t know/can’t remember 854 (7.9)
Measures No 7547 (69.5)
Items from the questionnaire were chosen to reflect Need
seven variables of Andersen’s behavioural model Perceived treatment need 10 585
of service utilization (10). The variables included Would need treatment 2628 (24.2)
Don’t know 2621 (24.1)
population characteristics (predisposing character- Would not need treatment 5336 (49.1)
istics, enabling resources and need), oral health Use of health services
behaviours and oral health outcomes (Table 1). Reason for attending dentist 10 627
To have a regular check up 7352 (67.7)
To have an occasional check up 818 (7.5)
Population characteristics Only when trouble with teeth 2145 (19.7)
The population characteristics were predisposing Never been 312 (2.9)
characteristics (deprivation), enabling resources Length of time since last visit 10 706
(perceived difficulty accessing a dentist) and Up to 1year ago 7761 (71.4)
Between 1 and 2years ago 678 (6.2)
need (perceived treatment need). Deprivation Between 2 and 5years ago 801 (7.4)
was assessed based on the participant’s postcode More than 5years ago 1334 (12.3)
using a composite area-based measure called the Never been 132 (1.2)
Index of Multiple Deprivation 2007, which mea- Health outcomes
Oral health related impacts
sures seven domains of deprivation at the small
Painful aching in mouth (oral 10 797
area level (with populations of around 1500) symptoms)
(14). The domains include income, employment, Never 5424 (49.9)
health, education, skills and training, barriers to Hardly ever 2352 (21.6)
Occasionally 2339 (21.5)
housing and services, living environment and
Fairly often 414 (3.8)
crime. Areas are ranked according to quintiles Very often 268 (2.5)
with the ‘least deprived’ and ‘most deprived’ Discomfort eating food (functional 10 810
quintiles comprising those neighbourhoods fall- limitations)
ing among the least or most deprived 20% in Never 4746 (43.7)
Hardly ever 2455 (22.6)
England. Occasionally 2677 (24.6)
Perceived difficulty accessing a dentist was Fairly often 539 (5.0)
assessed using the question ‘Is it difficult for you to Very often 393 (3.6)
get routine (e.g. check-up and fillings) dental care?’ Self-conscious (social impacts) 10 788
Never 6108 (56.2)
Responses were scored using ‘Yes’ = 1, ‘Don’t Hardly ever 1611 (14.8)
know/can’t remember’ = 2, and ‘No’ = 3, with Occasionally 1831 (16.9)
higher scores representing increased perceived Fairly often 643 (5.9)
ease accessing dental services. Participants’ per- Very often 595 (5.5)
Global oral health 10 769
ceived treatment need was measured by responses Excellent 994 (9.1)
to the question ‘If you went to the dentist tomor- Very good 3160 (29.1)
row, do you think you would need treatment?’ Good 3856 (35.5)
Responses included ‘I would need treatment’ = 1; Fair 1975 (18.2)
Poor 588 (5.4)
‘Don’t know’ = 2; and ‘I would not need treat-
Very poor 196 (1.8)
ment’ = 3, with higher scores reflecting less per-
ceived treatment need (15).

3
Marshman et al.

Health behaviours dentures?’ (functional limitations); and ‘In the last

Health behaviours relating to dental care utiliza-
tion were assessed using two separate items within
the questionnaire. Participants were asked ‘About
how long ago was your last visit to the dentist?’
Responses included ‘Up to 1 year ago’ = 1;
‘Between 1 and 2 years ago’ = 2; ‘Between 2 and
5 years ago’ = 3; ‘More than 5 years ago’ = 4; and
‘Never been’ = 5 (15). Participants were also asked
about their reasons for visiting the dentist: ‘In gen-
eral why do you go to the dentist?’ to which
responses included ‘To have a regular check
up’ = 1; ‘To have an occasional check up’ = 2;
‘Only when you have trouble with your teeth’ = 3;
and ‘Never been’ = 4.
Oral health outcomes
The oral health outcomes measured included spe-
cific oral health impacts and global oral health.
Oral health impacts were assessed by calculating
the total score from three questions taken from the
Oral Health Impact Profile (16): ‘In the last
12 months have you had a painful aching in your
mouth?’ (oral symptoms); ‘In the last 12 months,
have you found it uncomfortable to eat any foods
because of problems with your teeth, mouth or
12 months, have you been self-conscious because
of your teeth, mouth or dentures?’ (social impacts).
Responses included Never = 1; Hardly ever = 2;
Occasionally = 3; Fairly often = 4; Very often = 5.
Global oral health was assessed using the question:
‘Would you say the health of your teeth, lips, jaws
and mouth is?’ Responses included Excellent = 1;
Very good = 2; Good = 3; Fair = 4; Poor = 5; Very
poor = 6 (17).
Analysis
Structural equation modelling (SEM) with observed
variables using AMOS 18.0 (SPSS, Chicago, Illinois,
USA) was used to estimate the direction of direct
and indirect paths between the seven variables
within the model. SEM with observed variables is
a technique that measures the extent to which a
prespecified structural model fits a data set. Two
alternative models of dental care utilization were
compared using four goodness-of-fit indices and
two statistics used for the comparison of two
models (Table 2) (18–20). Model 1 included 14
direct pathways and was based on a simple linear
version of Andersen’s behavioural model (Fig. 1)
(10). However, findings from the dental literature

Table 2. Goodness-of-fit indices and multi-group analysis

Goodness of fit
Comparison of two models
CFI IFI RMSEA SRMR BCC (Lowest ECVI (Lowest Criteria
Model fitting (>0.95) (>0.95) (<0.08) (<0.05) value better fit) value better fit) fitted
Model 1 (N = 10864) 0.74 0.74 0.15 –a 1680.10 0.16 0/4
Model 2b (N = 10864) 0.98 0.98 0.05 0.02 175.83 0.02 4/4
(Accepted model)
Multi-group analysisb
Gender
Males (N=4832) 0.98 0.98 0.06 0.02 – – 4/4
Females (N=6032) 0.98 0.98 0.05 0.02 – – 4/4
Age
16–44years (N=3920) 0.96 0.96 0.08 0.03 – – 4/4
45–74years (N=5647) 0.98 0.98 0.05 0.01 – – 4/4
75years and over 0.99 0.99 0.04 0.01 – – 4/4
(N=1297)
Random subgroups
Group 1 (N=5432) 0.98 0.98 0.05 0.02 4/4
Group 2 (N=5432) 0.98 0.98 0.05 0.02 4/4
Alternative sample
Following list-wise 0.98 0.98 0.06 0.02 4/4
deletion (N=10220)
CFI, Comparative Fit Index; IFI, Incremental Fit Index; RMSEA, root mean squared error of approximation; SRMR, stan-
dardized root mean square residual; BCC, Browne and Cudeck criterion; ECVI, Expected Cross-validation Index.
Figures in bold are those that meet the model-fitting criteria.
Model 1: 14 direct pathways hypothesized; Model 2: 17 direct pathways hypothesized.
a
Model not successfully fitted.
b
Multi-group analysis was based on model 2 (accepted model).

4

have revealed significant associations between
population characteristics, such as predisposing
characteristics (21, 22), enabling resources (9, 23)
and need (24, 25), and oral health outcomes in the
adult population. Therefore, model 2 included 17
hypothesized direct pathways, which included
additional pathways between population charac-
teristics (deprivation, perceived difficulties access-
ing a dentist and perceived treatment need) and
oral health outcomes (Fig. 2). Criteria that
addressed the issue of parsimony within models
and imposed penalties for model complexity were
employed (e.g. Incremental Fit Index and Browne
–Cudeck criterion).
The parameters of each model were estimated
using the asymptotic distribution-free (ADF) esti-
mator, which is a methodology specifically
designed for dealing with categorical variables,
and a total of 900 bootstraps were created from the
dataset with the aim of deriving less biased stan-
dard errors (20). The 95% confidence interval (CI)
bootstrap percentiles were used to interpret the
results as these have been shown to be more accu-
rate (20, 26, 27). However, bootstrapping proce-
dures cannot be performed on data with missing
values, so data imputation was used and missing
values were replaced by median scores for each
variable.
To test whether the demographic variables of age
and gender (8, 9) acted as moderators of the direct
pathways within the accepted model, two multi-
group analyses were conducted. This involved cre-
ating two separate models for gender (male and
female) and three separate models for age (13).
Population characteristics Health behaviour
Pre-disposing Enabling Need Use of health
characteristics resources services
Perceived
Deprivation Perceived treatment Reason for dental
difficulties need attendance
accessing a (17.4%)
dentist
Significant pathways
Non significant pathway
Fig. 2. Adapted model of dental service use and oral health outcomes (derived from Andersen’s (1995) behavioural
model and dental literature). Note: 17 direct pathways hypothesised, 16 direct pathways significant. (n%)= Percentage
of variance explained.
Influences on dental service use
Results
Sample
Of the 25 200 questionnaires, 10 864 were returned
completed (43.1% response rate). The sample con-
sisted of 6032 women (55.5%) and 4832 men
(44.5%). The age of participants ranged from 16 to
104 years (mean = 52.5, SD = 18.5). Missing data
for each of the variables can be seen in Table 1.
Based on the measure of deprivation, 1447 partici-
pants in the sample were in the least deprived
(13.3%) and 2470 were in the most deprived quin-
tiles (22.7%) in England. Table 1 describes the char-
acteristics of the study variables.
Model fit
When examining the overall fit of the two hypoth-
esized models, the analysis revealed that model 2
was a better fit of the data (Table 2). Model 2 fit all
four goodness-of-fit indices and was therefore the
accepted model of dental service utilization with
all subsequent analyses based on model 2 (Fig. 2).
When multi-group analyses were conducted using
this model, the results revealed that both the male
and female models and the three models for the
different age groups were an acceptable fit
(Table 2). The model was also rerun using only
those individuals who provided complete data
(N = 10,220). The results revealed that the model
fit was still acceptable, suggesting data imputation
did not influence the overall acceptability of the
model (Table 2). To test the robustness of the
model, the sample was also split into two equal
groups (at random) and the acceptability of the
Outcomes
Use of health Perceived Perceived
services health status health status
Length of time since Oral health Global oral
last visited dentist impacts health
(55.4%) (21.7%) (42.9%)
5
Marshman et al.
model re-analysed. Examination of fit indices
revealed the model fit to be acceptable for both
groups.
Overall, the variables included within this model
explained 17.4% of the variance for reason for den-
tal attendance, 55.4% of the variance for the length
of time since people had last visited the dentist,
21.7% of the variance for oral health impacts and
42.9% of the variance for people’s global oral
health (Fig. 2).
Significant direct pathways. When the direct effects
were taken into account, all but one of the 17 path-
ways hypothesized in model 2 were significant;
length of time since last visiting the dentist did not
significantly predict oral health impacts (Table 3).
Significant direct predictors of oral health impacts
included deprivation (b = 0.07, P < 0.01), per-
ceived difficulties with accessing a dentist
(b = 0.12, P < 0.01) and perceived treatment need
(b = 0.41, P < 0.01). Significant direct predictors
of global oral health included difficulties with
access (b = 0.04, P < 0.01), perceived treatment
need (b = 0.27, P < 0.01), length of time since the
last visit (b = 0.09, P < 0.01), and oral health
impacts (b = 0.44, P < 0.01). Therefore, the results
revealed that a higher perceived treatment need
and perceived problems accessing the dentist were
associated with increased oral health impacts and
worse global oral health. Increased deprivation
was also associated with worse oral health out-
comes, but to a lesser extent.
All of the significant predictors of oral health
outcomes existed when gender and the three age
groups were analysed separately, with just two
exceptions; for the older age group (over 75 years),
perceived difficulties accessing care was not a sig-
nificant predictor of global oral health, and depri-
vation was no longer a significant predictor of oral
health impacts. Some of the weaker relationships
between the variables within the model (Table 3)
also ceased to exist within the multi-group analy-
sis. There was no significant relationship between
perceived access and length of time since last den-
tal visit for women and those in the youngest age
group (16–44 years) or between perceived access
and reason for dental visit in the group aged
between 45 and 74 years. There was also no rela-
tionship between deprivation and perceived access
for men and between deprivation and perceived
treatment need for those over the age of 75 years.
The only additional significant pathway in the
multi-group analysis was between the length of
6
time since the last dental visit and oral health
impacts, with more recent dental visits being asso-
ciated with increased impacts for those aged
between 16 and 44 years (b = 0.06, P < 0.01).
Significant indirect pathways. When examining indi-
rect pathways separately, 12 of the indirect path-
ways that existed within the model were
significant (Table 3). Significant indirect predictors
of oral health impacts included deprivation
(b = 0.06, P < 0.01) and difficulties with accessing
the dentist (b = 0.10, P < 0.01). Significant indi-
rect predictors of global oral health included depri-
vation (b = 0.11, P < 0.01), difficulties with
accessing the dentist (b = 0.17, P < 0.01), per-
ceived treatment need (b = 0.21, P < 0.01) and
reason for visiting the dentist (b = 0.06, P < 0.01).
Whilst a number of variables indirectly predicted
oral health outcomes, perceived difficulties access-
ing the dentist and perceived treatment need were
the two variables that most influenced oral health
through a variety of indirect pathways (Fig. 2).
Discussion
Data from a large (>10 000) sample of adults in
the UK were analysed. Perceived difficulty access-
ing a dentist and perceived treatment need were
found to be key factors predictive of oral health
outcomes in adults. The model fit was acceptable
for men, women and all of the three age groups
analysed, and interestingly, perceived difficulty
accessing a dentist and perceived treatment need
remained significant predictors of oral health
impacts irrelevant of age and gender. The model
explained a large amount of variance for both
dental service utilization and oral health out-
comes, confirming the support for using the
adapted behavioural model as applied to oral
health (9). Perceived treatment need was the main
predictor of oral health behaviours and outcomes.
Individuals who had higher level of perceived
treatment need were less likely to attend regular
dental appointments. In contrast, Reisine found
sex to be the most influential variable, with
income and age having no significant effects (8).
While clinical measures of need were included in
her analyses, perceived need for treatment was
not. Baker found the pathway between perceived
treatment need and oral health outcomes to be sig-
nificant, but did not include perceived difficulty
accessing a dentist (9).
 Influences on dental service use

Table 3. Significant direct and indirect pathways within the accepted model (model 2)
Direct pathways Indirect pathways Total pathways
b Bootstrap bias b Bootstrap bias b Bootstrap bias
Significant pathways value corrected 95% CI value corrected 95% CI value corrected 95% CI
Deprivation ? difficulties with 0.03** 0.05 to 0.01 – – 0.03** 0.05 to 0.01
access
Deprivation ? perceived 0.12** 0.14 to 0.11 0.01** 0.01 to 0.00 0.13** 0.15 to 0.11
treatment need
Deprivation ? reason for 0.15** 0.13 to 0.16 0.05** 0.04 to 0.06 0.19** 0.17 to 0.21
visiting dentista
Deprivation ? length of time – – 0.15** 0.13 to 0.16 0.15** 0.13 to 0.16
since visita
Deprivation ? oral health 0.07** 0.05 to 0.09 0.06** 0.05 to 0.07 0.13** 0.11 to 0.15
impactsa
Deprivation ? global oral – – 0.11** 0.09 to 0.12 0.11** 0.09 to 0.12
healtha
Difficulties with access ? 0.25** 0.23 to 0.27 – – 0.25** 0.23 to 0.27
perceived treatment need
Difficulties with access ? 0.06** 0.03 to 0.09 0.10** 0.11 to 0.09 0.04** 0.07 to 0.01
reason for visiting dentist
Difficulties with access ? 0.04** 0.02 to 0.07 0.05** 0.07 to 0.03 0.01 0.04 to 0.02
length of time since visita
Difficulties with access ? oral 0.12** 0.15 to 0.11 0.10** 0.11 to 0.09 0.23** 0.25 to 0.21
health impactsa
Difficulties with access ? 0.04** 0.05 to 0.02 0.17** 0.18 to 0.15 0.21** 0.23 to 0.18
global oral healtha
Perceived treatment need ? 0.11** 0.13 to 0.09 0.27** 0.28 to 0.25 0.38** 0.40 to 0.36
length of time since visit
Perceived treatment need ? 0.38** 0.40 to 0.36 – – 0.38** 0.40 to 0.36
reason for visiting dentist
Perceived treatment need ? 0.41** 0.43 to 0.39 0.00 0.00 to 01 0.40** 0.42 to 0.39
oral health impactsa
Perceived treatment need ? 0.27** 0.29 to 0.25 0.21** 0.23 to 0.20 0.48** 0.50 to 0.47
global oral healtha
Reason for visiting dentist ? 0.70** 0.68 to 0.71 – – 0.70** 0.68 to 0.71
length of time since visit
Reason for visiting dentist ? – – 0.01 0.02 to 0.01 0.00 0.02 to 0.01
oral health impacts
Reason for visiting dentist ? – – 0.06** 0.05 to 0.08 0.06** 0.05 to 0.08
global oral healtha
Length of time since visit ? 0.01 0.03 to 0.01 – – 0.01 0.03 to 0.01
oral health impacts
Length of time since visit ? 0.09** 0.08 to 0.11 0.01 0.01 to 0.00 0.09** 0.07 to 0.11
global oral health
Oral health impacts ? global 0.44** 0.43 to 0.46 – – 0.44** 0.43 to 0.46
oral health
Significant predictors of oral health outcomes represented by direct pathways are highlighted in bold.
**P < 0.01.
aVarious indirect pathways possible.

Perceived difficulty accessing a dentist was a
predictor of oral health outcomes and influenced
dental service utilization indirectly through per-
ceived need. This is the first study to indicate the
importance of this factor in oral health research,
although the influence of healthcare availability on
health was discussed by Lalonde (28). While there
had been an historical lack of dental service provi-
sion in this region, at the time of data collection,
traditional oral healthcare needs assessments sug-
gested that there was sufficient volume of NHS
dental services to meet the needs of patients in
most areas. Further research is needed to explore
this concept of perceived difficulty accessing a den-
tist, which may be related to lack of availability or
other factors such as anxiety, difficulties with phys-
ical access or affordability. However, identification
of the importance of this variable also offers an

7
Marshman et al.
opportunity for intervention to improve service
utilization and outcomes. Such interventions
would need to be effective at changing the public
perceptions of the accessibility of dental services.
Information would need to be designed to make
people aware of how to access dental services from
sources and in a format appropriate to the public’s
needs. Social marketing techniques (29), which
identify what will motivate the public to change
oral health behaviours, may be useful to provide
insights into suitable media and methods to dispel
myths about access difficulties. In dental service
management terms, the dental contract introduced
in England in 2006 removed the ability of a patient
to register with a dentist (2). There is some evi-
dence to suggest that being registered with a den-
tist has a positive impact on perceptions of access
(22). A new dental contract will be piloted in Eng-
land from April 2011 with the concept of registra-
tion restored (30).
Andersen highlighted the high degree of muta-
bility of enabling resources such as perceived
access to health services and contrasted this with
immutable variables such as deprivation (10). This
study used an area-based composite measure of
deprivation, rather than an individual indicator of
socioeconomic status, and found that deprivation
was not as strong a predictor of oral health impacts
as perceived access or perceived treatment need.
A similar study in the UK, which used social
class, income and education, also found enabling
resources (oral health education, treatment
expense, dental anxiety) and need to be stronger
predictors of health behaviour and outcome than
predisposing factors (9). Reisine in the US study
also found income to have no significant effect on
dental service use (8). Further research is needed to
identify other factors that might mediate the rela-
tionships between predisposing factors such as
deprivation and other variables within the model.
Indeed, psychosocial resources (such as optimism,
coping style or personal control) and social support
have been shown to influence the relationship
between socioeconomic status and health (31).
These findings contribute to debates about the
degree to which socioeconomic status and depriva-
tion explain inequalities in access and utilization of
oral healthcare (11,12).
Limitations
While the use of structural equation modelling
with observed variables provides a sophisticated
analytical procedure to test the causal processes
8
hypothesized by theoretical models (32), it should
be recognized that there are some limitations of its
use in this study. First, a median imputation
method was used to handle missing data, which
will have had the potential to reduce the variance
of the variables, possibly resulting in underestima-
tion of correlation between these variables and
other variables within the model (33). However, it
is not possible to use ADF with missing data. Sec-
ond, the use of categorical indicators (e.g. ordinal
data) within the model is not ideal; however, the
categorical variable methodology of ADF was
employed to deal with this issue (20).
While the variables tested in this study
explained a large amount of variance for both den-
tal service utilization and oral health outcomes,
half of the variance remains unexplained. There-
fore, other factors need to be considered in future
including dental anxiety (8) and costs of treatment
(4). Owing to the constraints of carrying out a sec-
ondary analysis of survey data, it was not possible
to examine these and other factors such as educa-
tion and occupation. It was also not appropriate to
include ethnicity as 96.0% of the sample was White
British. Further research is needed to identify and
investigate the contribution of such factors. In
addition, as the survey data were collected in the
UK, the contribution of factors relating to the orga-
nization and delivery of dental services (10) also
need to be considered.
Finally, although the response rate of 43.1% com-
pares favourably with the other UK postal surveys
(34), the possibility of nonresponse bias is an addi-
tional limitation of this study, which may have
influenced the significant factors identified. It may
be that as the response rate was lower from those
in more deprived areas (13) that the patterns of
access seen in this study reflect the access of less
deprived people who were more likely to respond.
Conclusions and implications
This was the first study to identify the public’s per-
ceptions of difficulties accessing dental services as
a key predictor of oral health outcomes. Despite
the limitations of the study, this finding offers an
opportunity for interventions to address public
perceptions through the use of social marketing
techniques or changes to the ways dental contracts
are developed. An area-based composite measure
of deprivation was used and found deprivation to
be a weak predictor of oral health impacts

compared with other variables within the model.
The study builds on previous theoretical work
using Andersen’s behavioural model to discuss the
application of this finding by public health practi-
tioners and policy makers. Further work is needed
to develop effective interventions to change public
perceptions and so improve service use and oral
health outcomes.
References
1. Petersen PE. World Health Organization global pol-
icy for improvement of oral health – World Health
Assembly 2007. Int Dent J 2008;58:115–21.
2. Department of Health. A guide to NHS primary den-
tal services from April 2006. London: Department of
Health; 2005.
3. Department of Health. Government response to
the health select committee report on dental ser-
vices. London: Office of Public Sector Information;
2008.
4. Scheutz F, Heidmann J. Determinants of utilization
of dental services among 20- to 34-year-old Danes.
Acta Odontol Scand 2001;59:201–8.
5. Atchison KA, Andersen RM. Demonstrating success-
ful ageing using the International Collaborative
Study for Oral Health Outcomes. J Public Health
Dent 2000;60:282–8.
6. Dobalian A, Andersen RM, Stein JA, Hays RD, Cunn-
ingham WE, Marcus M. The impact of HIV on oral
health and subsequent use of dental services. J Public
Health Dent 2003;63:78–85.
7. Varenne B, Petersen PE, Fournet F, Msellati P, Gary J,
Ouattara S et al. Illness-related behaviour and utili-
zation of oral health services among adult city-dwell-
ers in Burkina Faso: evidence from a household
survey. BMC Health Serv Res 2006;6:1–11.
8. Reisine S. A path analysis of the utilization of dental
services. Commun Dent Oral Epidemiol 1987;15:119–
24.
9. Baker SR. Applying Andersen’s behavioural model
to oral health: what are the contextual factors shap-
ing perceived oral health outcomes? Commun Dent
Oral Epidemiol 2009;37:485–94.
10. Andersen RM. Revisiting the behavioural model and
access to medical care: does it matter? J Health Soc
Behav 1995;36:1–10.
11. Sanders AE, Spencer AJ, Slade GD. Evaluating the
role of dental behaviour in oral health inequalities.
Commun Dent Oral Epidemiol 2006;34:71–9.
12. Locker D. Deprivation and oral health: a review.
Community Dent Oral Epidemiol 2000;28:161–9.
13. Marshman Z, Dyer TA, Wyborn CG, Beal J, Godson
JH. The oral health of adults in Yorkshire and Hum-
ber 2008. Br Dent J 2010;209:E9.
14. Department for Communities and Local Govern-
ment. Index of multiple deprivation 2007. London:
Communities and Local Government Publications;
2008.
15. Kelly M, Steele JG, Nuttall N, Bradnock G, Morris J,
Pine C et al. Adult dental health survey: oral health
Influences on dental service use
in the United Kingdom 1998. London: The Stationary
Office; 2000, pp. 297–309.
16. Slade GD. Derivation and validation of a short-form
oral health impact profile. Commun Dent Oral Epi-
demiol 1997;25:284–90.
17. Atchison KA, Gift HC. Perceived oral health in a
diverse sample. Adv Dent Res 1997;11:272–80.
18. Byrne BM. A primer of LISREL: basic applications
and programming for confirmatory factor analytic
models. New York: Springer-Verlag; 1989.
19. Hu LT, Bentler PM. Cutoff criteria for fit indexes in
covariance structure analysis: conventional criteria
versus new alternatives. Struct Equ Modeling
1999;6:1–55.
20. Byrne BM. Structural equation modeling with
AMOS: basic concepts, applications, and program-
ming. New Jersey: Lawrence Erlbaum Associates;
2001.
21. Baker SR. Socio-economic position and oral health:
comparing proximal and distal indicators. Soc Sci
Dent 2010;1:5–10.
22. Sanders AE, Spencer AJ. Social Inequality: social
inequality in perceived oral health among adults in
Australia. Aust N Z J Public Health 2004;28:159–66.
23. Sanders AE. A Latino advantage in oral health-
related quality of life is modified by nativity status.
Soc Sci Med 2010;71:205–11.
24. Jensen PM, Saunders RL, Thierer T, Friedman B. Fac-
tors associated with oral health–related quality of life
in community-dwelling elderly persons with disabil-
ities. J Am Geriatr Soc 2008;56:711–7.
25. de Oliveira BH, Nadanovsky P. Psychometric prop-
erties of the Brazilian version of the oral health
impact profile–short form. Commun Dent Oral Epi-
demiol 2005;33:307–14.
26. Efron B, Tibshirani R. An introduction to the boot-
strap. New York: Chapman & Hall; 1993.
27. MacKinnon DP, Lockwood CM, Hoffman JM, West
SG, Sheets V. A comparison of methods to test medi-
ation and other intervening variable effects. Psychol
Methods 2002;7:83–104.
28. Lalonde M. A new perspective on the health of
Canadians. A working document. Ottawa: Govern-
ment of Canada; 1974.
29. Edmunds M, Fulwood C. Strategic communica-
tions in oral health: influencing public and profes-
sional opinions and actions. Ambul Pediatr 2002;2:
180–4.
30. Primary Care Commissioning. Supporting informa-
tion for dental contract pilots in England. London:
Department of Health; 2010.
31. Taylor SE, Seeman TE. Psychosocial resources and
the ses-health relationship. Ann NY Acad Sci
1999;896:210–25.
32. Wallander JL, Varni JW. Effects of paediatric chronic
physical disorders on child and family adjustment.
Child Psychol Psychiatry 1998;39:29–46.
33. Brown RL. Efficacy of the indirect approach for esti-
mating structural equation models with missing
data: a comparison of five methods. Struct Equ Mod-
eling 1994;1:287–316.
34. Owen-Smith V, Burgess-Allen J, Lavelle K, Wilding
E. Can lifestyle surveys survive a low response rate?
Public Health 2008;12:1382–3.