ThesisPDF Available

Profiles of Attitudes Towards Healthcare: Psychographic Segmentation

Authors:

Abstract and Figures

Two thousand telephone interviews conducted with adult heads of household in ten different U.S. metropolitan areas were used to gather data regarding people's health care attitudes and behaviors. A review of the health care marketing and services research was used to identify psychographic, attitudinal and behavioral dimensions particularly relevant to health care marketing and communications. A comprehensive analysis scheme was used to identify distinct psychographically-defined consumer segments and validate their existence both within and across the geographic markets surveyed. The motivation for the research comes from the need among health care marketing professionals for a reliable description of consumers across dimensions that influence health care behaviors and purchasing choices. The study results indicate that psychographically defined consumer segments do exist, they share common interactive patterns of health thinking and behavior, and can be generalized across U.S. markets. Implications for health care marketing applications are discussed.
Content may be subject to copyright.
Profiles of Attitudes Toward Healthcare:
Psychographic Segmentation
Profiles of Attitudes Toward Health Care: Psychographic Segmentation
Copyright © 1990 by Frederick H. Navarro.
Second Edition
Copyright © 2008 by Frederick H. Navarro, Fontana, California. All Rights Reserved. Except as permit-
ted under the United States Copyright Act of 1976, no part of this publication may be reproduced or dis-
tributed in any form or by any means, or stored in a database or retrieval system, without prior written
permission of the publisher.
Reader: Additional information is available at www.adaptivehealthbehaviors.com
Patterns of Adapting to Health (PATH) is a trademark of Frederick H. Navarro.
Profiles of Attitudes Toward Healthcare:
Psychographic Segmentation
Frederick H. Navarro
Note:
The results of this psychographic segmentaon research led to the creaon of the Paerns of Adapng
to Health (PATH), a commercial product and assessment tool used broadly within the healthcare in-
dustry. Since this research was conducted in the mid-1980s, the PATH have been idened and studied
across a quarter million adults within the United States.
The paernshave demonstrated high reliability across me and populaons, and are validated
predictors of tness, dietary paerns, age, sex, and medical expenditures, self-reported health status,
obesity, Type II diabetes, physician and prescripon drug ulizaon, and many others.
The quesonnaire developed for this research was reduced to 20 items and became the PATH In-
ventory, an inventory of adapve behavioral responses to health-related contexts, not atudes, con-
sistent with cognive neuroscience research into the semancs of acon, in its infancy at the me this
research was conducted. A more accurate descripon of individual answers to the Inventory is behav-
ioral responses to health-related social contexts encoded symbolically in language). This then aligns
Inventory responses with interaconist personality theory, eld theory, and the concept of personolog-
ical systems. The PATH Inventory thus denes a health psychological eld, similar to a phase space,
that denes the person and the health-related psychological eld he or she interacts with based on
background, psychological disposions, and his or her social frame. Mathemacally, the size of this
psychological eld (i.e., phase space) is 9.53 x 1013.
Given that the paernsexist at the populaon level and have no detectable external driver out-
side of the system, it is proper to treat them as self-organized paerns, which logically extend them to
complex adapve systems theory.
In the current thinking, the paernsare a dimension of the behavior of the U.S. populaon sys-
tem; the paernsrepresent mulple steady states or equilibria located on aractors within the dy-
namic health psychological phase space measurable by the Inventory.
Whatever theorecal perspecve is applied to the Paerns of Adapng to Health (PATH), the
paernshave proven to be an important determiner of both individual health and U.S. populaon
health.
For more informaon about the PATH, please visit www.adapvehealthbehaviors.com. If you
would like to apply the PATH in your research, please write me at adapvehealthbehav-
iors@gmail.com.
-
Frederick H. Navarro, PhD.
Profiles of Attitudes Toward Healthcare:
Psychographic Segmentation
A thesis submitted to the faculty of
San Francisco State University
in partial fulfillment of the requirements for the degree
Master of Arts in Psychology:
Research Psychology
Frederick H. Navarro, M.A.,
January, 1990
5
Attitudes Toward Health Care
Profiles of Attitudes Toward Healthcare:
Psychographic Segmentation
Frederick H. Navarro, M.A.
San Francisco State University
Health care industry literature has recently
focused attention on the need for better segmenta-
tion of the health care marketplace (Malhotra,
1986; Modern Healthcare, 1986). Health care pro-
viders have only recently realized that health care
consumers make up a heterogeneous marketplace,
and that a attitude/behavior segmentation strategy
is one approach that may increase the profitability
of its service offerings. Segmentation of the mar-
ket allows institutions to develop better marketing
strategies and better allocation of available re-
sources (Wind, 1978). The major concern trou-
bling the health care industry, however, is the
identification and selection of the appropriate di-
mensions for segmenting people as health care
consumers.
The majority of health care segmentation
studies reported in the literature are based on a
priori segmentation designs. These are designs
where management or researchers decide on a ba-
sis for segmentation such as product purchase,
level of usage, or brand loyalty, and cross-tabulate
the data with demographic, socioeconomic, or
ABSTRACT
Two thousand telephone interviews conducted with adult heads of household in ten different U.S. metropolitan
areas were used to gather data regarding people's health care attitudes and behaviors. A review of the health care
marketing and services research was used to identify psychographic, attitudinal and behavioral dimensions particu-
larly relevant to health care marketing and communications. A comprehensive analysis scheme was used to identi-
fy distinct psychographically-defined consumer segments and validate their existence both within and across the
geographic markets surveyed. The motivation for the research comes from the need among health care marketing
professionals for a reliable description of consumers across dimensions that influence health care behaviors and
purchasing choices. The study results indicate that psychographically defined consumer segments do exist, they
share common interactive patterns of health thinking and behavior, and can be generalized across U.S. markets.
Implications for health care marketing applications are discussed.
psychographic information to profile the prede-
fined segments.
A study of health maintenance organization)
(HMO) enrollment among employees (Berki, Ash-
croft, Penchansky, & Fortus, 1977) segmented
them into two groups: Enrollees in HMO-type
plans and Blue Cross/Blue Shield subscribers.
This was done to profile them and to investigate
the roles of perceived health risk, financial vulner-
ability, and access to care in determining HMO
enrollment. Showing that these and related factors
can predict HMO enrollment would provide health
insurance suppliers with information that would
make it possible to project likely HMO enrollment
given knowledge of these factors in any given
pool of employees.
Klegon (1981) investigated HMO desirability
among consumers by segmenting them into those
with a regular physician and those without a regu-
lar physician. This research found that HMO de-
sirability, taken as the dependent variable, varied
between the two segments, and proposed that tar-
get populations should segmented by both HMO
6 Frederick H. Navarro
desirability and the presence or absence of a reg-
ular physician. Thus, this research showed that
segmenting across one a priori defined dimension
was not enough to describe the health care mar-
ketplace adequately.
In a study of return intentions among obstet-
rics patients (Anderson, 1982), women were seg-
mented into two pairs of patient categories to
examine the association of patient care attributes,
hospital feature attributes, and birthing experi-
ence attributes with intentions to return. The first
segmentation focused on dividing patients into a
choice (women played an active role) group and
a no-choice (physician-controlled) group; the
second segmentation split women into first birth
and prior birth groups. The goal of the study was
to identify those features and appeals that would
create the most satisfaction among the different
segments of women. Again, this study indicated
that segmenting across one dimension of health
care consuming behavior or experience is not
enough to provide an accurate view of multiple
factors and how they interact together to shape
the marketplace. A priori segmentation tech-
niques have also been used to investigate the dif-
ferences between users and nonusers of preven-
tive health care practices and emergency walk-in
clinics (Ortinau, 1986) and the difference be-
tween blood donor segments, with level of blood
donating behavior as the basis for segmentation
(Burnett and Leigh, 1986). Relatively few health
care consumer segmentation studies use cluster-
segmentation approaches. Cluster-based segmen-
tation designs identify segments by clustering
respondents on a set of two or more variables.
These variable can be benefits desired, behaviors,
or attitudes. A cluster-based segmentation design
was used to identify benefit segments for the am-
bulatory care pharmacy market (Carroll & Gag-
non, 1983). The authors used a profile of benefits
desired to identify five health care consumer seg-
ments with like needs and desires. Knowledge of
these homogenous segments defined by the bene-
fits they jointly desire made it possible to outline
the best combination of product or service fea-
tures to serve each segment.
Finn and Lamb (1986) applied both cluster
and discriminant analyses in an attempt to identi-
fy hospital benefit segments using 15 hospital
attributes. In this study, four distinct benefit seg-
ments were identified. Given that hospitals have
limited resources, these authors suggest that hos-
pitals should position themselves more favorably
to the attitude/behavior segments they feel they
can best serve.
Although cluster-based segmentation designs
are in the minority, the wisdom of taking the ap-
proach over a priori segmentation has been point-
ed out by several researchers. Wells (1975)
points out that cluster-based segmentation de-
signs do "not assume that members of any target
group are similar." Instead of trying to discover
what blood donors, HMO joiners, hospital pa-
tients, or those without regular doctors have in
common, the technique admits the possibility that
"members of a 'buyer' segment...may buy a given
brand for different reasons. They may be very
heterogeneous in their needs, demographic char-
acteristics, and information required. Therefore, a
very specific multidimensional definition of the
basis for segmentation is required." (p. 327).
There are, unfortunately, some important
problems inherent to cluster-based segmentation
designs. First is the problem of the reliability seg-
ments identified through clustering procedures.
As stated by Wells (1975), "If segments vary
greatly in size, or disappear altogether, from one
analysis to another, it makes no sense at all to
develop products, messages, or media schedules
on the assumption that the segments are re-
al." (p.204). This was also a major limitation cit-
ed by Finn, et al. (1986) where they state that,
"Further research of a cross-sectional nature is
needed to determine...whether [cluster-based]
segments are stable across institutions, popula-
tions, and/or situations." (p.32). Second, the
7
Attitudes Toward Health Care
number of clusters to be specified is not a clear
and simple decision. Many clustering algorithms,
both hierarchical and nonhierarchical, provide as
many clusters as the researcher requests: the al-
gorithms do not rely on the actual homogeneity
or heterogeneity of a sample of respondents to
determine the number of clusters which best de-
scribe the available cases (Aaker & Day, 1983).
The number and nature of clusters or segments
across geographic markets may vary; therefore,
one study's findings in terms of attitude/behavior
segment structure may not be validly generalized
to another area.
A third problem lies in the selection of di-
mensions. The poor performance of general char-
acteristics as components for segmentation analy-
sis is well documented (Wells, 1975). Segments
based on general characteristics, such as life-
style, often relate weakly to the purchasing situa-
tion, and do not provide the insight necessary for
developing effective industry-specific marketing
strategies. Product or service-specific customer
characteristic approaches (e.g., benefit segmenta-
tion), on the other hand, are an extreme example
of this segmentation strategy. They may not per-
form well if product purchase is situation de-
pendent and driven (Young, Ott, & Feigin, 1978).
For example, selection of a hospital is one deci-
sion that may be dictated by the need for treat-
ment. One hospital might be used for maternity
care and another for cardiac care, depending on
the advice of the physician or the particular spe-
cialties of the hospitals. Therefore, any segment-
ing approach based on hospital selection criteria
that ignores the situation or context of that selec-
tion may be only marginally effective. A fourth,
but not necessarily final, problem with cluster-
based segmentation studies is one not usually
discussed in the literature, but one certainly de-
serving of attention. This is the "problem" of
having every case or respondent in a sample as-
signed to a cluster. If the direct goal of cluster-
based segmentation research is to identify those
consumers who are homogeneous across multidi-
mensional issues, the indirect goal should neces-
sarily be the identification and exclusion of re-
spondents or cases that do not share homogene-
ous traits with formed clusters. In the majority of
published research studies where cluster-based
segmentation designs have been employed, there
is rarely, if ever, discussion of the respondents or
cases that could not be classified. In my opinion,
forcing all cases into available clusters must de-
tract from the integrity of the formed clusters in
two ways: First, forced clustering weakens the
homogeneous nature of the formed cluster seg-
ments, so that actual differences between clusters
are less distinct, second, forced clustering in-
creases the probability that the size estimates of
identified cluster segments are overstated -- that
is, truly homogeneous segments are increased by
the addition of cases that share fewer homogene-
ous traits with them. Therefore, any thorough
study using cluster-based segmentation tech-
niques must make allowances for those cases that
do not belong in previously identified segments.
Study Overview
The present research addresses the above-
listed major limitations of cluster-based segmen-
tation analysis within the context of the health
care industry, by applying independent cluster
and validation analyses to each of ten different
health care marketplaces, and comparing cluster-
based segments defined on the basis of health
attitudes and behaviors across markets. The clus-
tering and validation analyses applied to collect-
ed health attitude/behavior data allow for some
cases to be unassigned to produce homogeneous
sets of health care consumers defined by similar
interactive patterns across attitude/behavior
measures and it assesses potential segment struc-
ture differences within and across markets. The
purpose of this study is to assess whether homo-
geneous sets or segments of cluster-based health
care consumers, defined by common interactive
patterns of health thinking and behavior, exist in
8 Frederick H. Navarro
today's health care marketplace and whether these
cluster-based health care consumer segments can
be generalized across geographic markets. To car-
ry out this research, an investigation of potential
multidimensional bases for segmentation was car-
ried out. As an outcome of this preliminary re-
search, a psychographic approach was selected.
Psychographics refers to the quantification of peo-
ple's described attitudes and behaviors. Psycho-
graphics can be employed in general lifestyle re-
search, or can be applied to the measurement and
quantification of people's described attitudes and
behaviors in a specific context or situation, such
as the seeking or purchasing of healthcare ser-
vices. A set of critical dimensions quantifiable
through psychographic techniques was identified
for inclusion in the project. Survey items were
devised to measure the identified dimensions, and
two thousand random telephone interviews were
conducted with consumers, divided evenly be-
tween ten different geographic areas. Respondents
interviewed in each geographic area were inde-
pendently clustered based on their measured psy-
chographics, tested for statistically significant dif-
ferentiation, and validated. Cluster solutions for
each market were then compared across all mar-
kets; similar cluster-based segments were com-
bined, tested again for statistically significant dif-
ferentiation, and profiled. Evidence demonstrating
the reliability of psychographically defined, clus-
ter-based segments, both within geographic mar-
kets and across geographic markets, is offered by
comparing obtained within and across market
cluster solutions with Monte Carlo synthetic pro-
files, which mimic results expected under the op-
eration of random chance.
Method
Identifying Bases for Segmentation
The major goal of this step in the research was
to identify health care specific psychographics;
that is, attitudinal and behavioral dimensions af-
fecting health care consuming behaviors. Due to
time restraints and funding limitations, it was not
possible to conduct the focus group research gen-
erally used to uncover critical factors or develop
hypotheses. Because of this, the search for dimen-
sions was limited to review of the available re-
search literature. The goal of this review was to
identify psychographic dimensions that are gen-
eral enough to be applicable across a wide range
of health care purchasing situations, but not so
general as to be ineffective at usefully segmenting
the health care consumer marketplace.
Five segments of the women's health care
market, defined in terms of four dimensions of
attitudes and behaviors concerning health care,
were identified by Harrel and Fors (1985) using
qualitative research methods. Health concern, fre-
quency of health service use, self confidence in
decision making, and propensity to experiment
(shop) with health care services served to define
these segments. Harrel and Fors (1985) also spec-
ulated that levels of health care information seek-
ing, receptivity to health care advertising, concern
for family health, and health enhancement empha-
sis (e.g., nutrition vs. physical fitness) may differ
across segments. All these dimensions are poten-
tially useful elements for a health care segmenta-
tion model because all have direct health care
marketing implications.
The bases for segmentation proposed by Har-
rels and Fors (1985) were applied in a pilot study
of five hundred women throughout a five-county
area for a North Carolina hospital. A cluster-based
segmentation approach was utilized for this study.
Clustering using two different hierarchical cluster-
ing algorithms was centered on twenty-two home-
made psychographic, health care specific, attitude,
interest, and opinion (AIO) measures (Wells &
Tigert, 1971) modeled closely along the lines pro-
posed by Harrels and Fors (1985). Using these
measures, seven distinct women's health care atti-
tude/behavior segments very similar to those hy-
pothesized by Harrels and Fors (1985) were iden-
tified by comparing cluster solutions using two
different clustering algorithms, combining like
segments, and confirming distinctiveness through
9
Attitudes Toward Health Care
multiple discriminant analysis. Final segment
membership information, when cross-tabulated
with assessment of respondent interest in specific
programs and services under consideration by the
hospital, showed wide variation of interest levels
across the psychographic segments. This provided
the North Carolina hospital with clear direction
regarding which services to develop for which
segments. These preliminary research results
yielded good evidence supporting the dimensions
selected and psychographic measures developed
as potentially useful for segmenting female health
care consumers, issues of psychographic measure
reliability notwithstanding.
Drawing further from the literature, additional
health care attitudinal and behavioral dimensions
applicable to both men and women were identified
for inclusion in the segmentation procedure. For
example, the level of health care information seek-
ing is one dimension very relevant to marketing. It
provides an indication of how intensely health
care communication and promotion must be im-
plemented (Wells & Tigert, 1971). Consumers
with low levels of information seeking will re-
quire wide spread efforts to reach them, and high-
ly potent appeals to attract their attention. The re-
verse is true of consumers with high levels of
health care information seeking. These latter con-
sumers are also more likely to demand more so-
phistication and greater informational content.
Price sensitivity and quality of care are dimen-
sions of growing relevance to the marketing of
health care services. Blendon and Altman (1984)
discuss the growing concern with personal health
care costs. They indicate that there may be an in-
creasing number of consumers who consciously
consider price or direct "out-of-pocket" costs
when making health care purchase decisions. This
is also supported by Paul (1987) where the author
states that "more quality is preferable to less, other
things being equal -- but 'other things' count for
more now." (Pg. 4) The whole price/quality trade-
off issue comes from the fact that individuals have
often received high quality care without directly
paying for it. However, because insurers, busi-
nesses, and the government have been forced to
attempt control of the amount they spend on
health care, the burden has been shifted directly to
the consumer in the form of increased deductibles
and larger copayments (Paul, 1987). Some re-
searchers have pointed out that there are many
product areas where high price does not necessari-
ly signal better quality (Gerstner, 1985). In the
area of health care, however, there may be a diver-
gence of perception regarding the price/quality
relationship across different consumer groups.
Price consciousness, concern for quality, and the
level of information seeking are related to the
broader concept of purchasing involvement. Work
by Slama and Taschian (1985) focused on identi-
fying socioeconomic and demographic character-
istics associated with purchasing involvement.
They conclude that "involvement with purchasing
influences purchase behavior and that different
consumer types (i.e., market segments) can be
identified on the basis of purchasing involve-
ment" (p. 72). Thus, health care information seek-
ing and price sensitivity, as psychographic indica-
tors of health care purchasing involvement, are
important dimensions for consideration in a multi-
dimensional approach such as this. Two additional
dimensions are the degree of health involvement
and the propensity to be health proactive. From a
promotional standpoint, health care consumers
who do or do not have a high degree of health in-
volvement could respond differently to health care
marketing messages. Messages communicating
programs which assume high involvement, such
as concern for proper nutrition or physical fitness,
could miss the low involvement consumer, while
messages for programs targeted to low involve-
ment consumers could lose the high involvement
ones.
The second dimension, propensity to be health
proactive, which could include preventive health
behavior such as frequent check-ups, addresses
consumers' concern with the long- versus short-
term benefits of any health behavior or treatment.
10
Research has shown that this proactive behavior is
positively correlated with socioeconomic status
(Coburn & Pope, 1974).
Marketing strategies which speak to the long-term
benefits of a program or regimen may appeal to
the more affluent segment of the population, but
may not appeal to those consumers whose primary
concern is dealing with today and letting tomor-
row take care of itself. Senior citizens are one seg-
ment of the population that this can apply to.
Thus, knowing whether to promote long- as op-
posed to short-term benefits could be of vital im-
portance in trying to reach the right audience.
Attitudes toward medical professionals, and
the medical profession in general, is another im-
portant issue. Blendon and Altman (1984) dis-
cussed a growing cynicism and lack of faith in
doctors. Haug and Lavin (1981) and Powills
(1981) also cited the waning influence of physi-
cians. Lack of trust or a cynical attitude toward
health care professionals may influence a consum-
er's acceptance of promoted medical services or
products. Consumer targeted advertising and pro-
motional efforts that rely on the authority of phy-
sicians (e.g., "One hundred doctors stranded on a
deserted island more often chose Bayer Aspirin"),
may appeal to some consumers, but not others.
Therefore, a good argument can be made for
the need to include these attitudes as a basis for
segmentation.
Hulka, Zyzanski, Cassell, and Thompson
(1970) designed a 41-item scale to measure atti-
tudes toward physicians and primary medical care.
Several item "themes" were borrowed from their
scale to represent this dimension in the context of
this research. As a result of this literature review,
eleven broad dimensions were identified which
offered the greatest potential for useful and action-
able health care specific, psychographic consumer
attitudinal and behavioral variables. These are
summarized below:
1. Health emphasis and involvement (physical
fitness vs. nutritional fitness)
2. Propensity to experiment with health care
alternatives
3. Involvement in decision-making, choosing
health care services
4. Propensity to avoid health care
5. Attitudes toward medical professionals
6. Involvement in family health
7. Health care information seeking
8. Receptivity to health care advertising
9. Propensity to be health reactive or proactive
10. Quality consciousness
11. Price concern.
These dimensions were used as a framework
for generating descriptive, attitudinal/behavioral
statements describing possible consumer view-
points. The final battery of health related AIO
measures consisted of 41 statements, with two to
five statements applicable to each dimension.
Some examples of the statements can be found in
Appendix 1 (for information on the full battery,
please contact the publisher). Consumer agree-
ment with each statement was measured on a five-
point, Likert-type, strongly agree-disagree scale.
Data Collection
To obtain the funding required for this study,
five hospitals were recruited as sponsors, and 200
interviews were conducted in each of their respec-
tive service areas. The five sponsoring hospitals
were located in geographically diverse areas. This
increased the likelihood that identified health care
attitude/behavior segments will be applicable to
the majority of U.S. markets.
Probability sampling was not used to select
geographic markets because the goal was not to
develop health care attitude/behavior segments
representative of the entire U.S. population, but to
determine if similar segments exist across geo-
graphic markets. A total of 2,000 interviews were
conducted in ten geographic markets. The areas
can be found in Appendix 2.
Subjects. Nine hundred and eighty six adult
males and 1,014 adult female: were interviewed
11
Attitudes Toward Health Care
by telephone. Random digit dialing ensured that
listed and unlisted telephone households would be
represented; random telephone numbers were gen-
erated by prefix according to the proportion of
telephone households within each prefix. Only
heads of household were interviewed, and up to
three call backs were made to each household.
Questionnaire In addition to 41 psychograph-
ic items, the questionnaire also included additional
healthcare measure-assessment items for later
cross-tabulation with segment membership infor-
mation. These additional items included questions
regarding health insurance shopping behavior,
satisfaction with choices and available coverage,
use of hospitals, determiners of hospital choice
(e.g., respondent vs. doctor vs. health plan), medi-
cation shopping behavior, presence of a personal
doctor as well as interest in switching doctors, use
of preventive health check-ups, common personal
healthcare referent sources (e.g., spouse, parents,
etc.), and common media sources of healthcare
information. The questionnaire concluded with
traditional demographic (age and sex) and socio-
economic measures (education, income, and occu-
pation). The average time per completed interview
was seventeen minutes.
Data Analysis
Psychographic Inventory Reduction
Two R-type factor analyses (Kim & Mueller,
1978) performed on 200 and 100 initial question-
naires were used, prior to segment development,
to identify the number of psychographic state-
ments necessary to precisely determine distinct
dimensions from those on the psychographic in-
ventory of statements. Two factor analyses were
performed as a reliability check of the factor ana-
lytic solutions. Fifteen and fourteen factors, re-
spectively, were identified in each R-type factor
analysis using the Kaiser normalization, eigenval-
ue less than one criterion (Kim & Mueller, 1978),
and Varimax rotation These results are shown in
Appendix 3. The number of factors identified rela-
tive to the proportion of variance explained, 65
percent and 68 percent in both analyses, demon-
strated that 34 of the 41 statements were success-
fully measuring distinct, independent dimensions,
and that factor structures were similar across these
samples.
Examination of psychographic item factor
loadings in each analysis revealed that 21 of the
34 statements were retained in each solution when
the two psychographic statements with the highest
loadings on each identified factor were selected. A
review of these 21 items revealed that all but two
of the segmentation dimensions (propensity to
avoid healthcare and health apathy) were repre-
sented by the retained statements. In the North
Carolina survey, however, the statements measur-
ing propensity to avoid healthcare dimensions
were the key to identifying the most attractive
women's healthcare attitude/behavior segment,
which were judged to be attractive based on their
willingness to seek health care services. This ap-
parent contradiction led to a reexamination of item
factor loadings in both solutions. This reexamina-
tion showed that the statements assessing the pro-
pensity to avoid healthcare and health apathy nar-
rowly missed being included in both solutions.
Given the potential relationship between health
care utilization rates and these dimensions, these
items were included for use in the segmentation
analysis. A final count of 25 psychographic
measures were used in the cluster analysis.
Cluster Analysis
Q-factor analysis was used to identify homo-
geneous clusters within each of the ten geographic
locations. Recent research has demonstrated the
superiorly of Q-factor analysis over alternative
clustering routines (Funkhouser, 1983; Hagerty
1985). The author's own comparisons of both
techniques on the same data revealed that Q-factor
analysis produced more statistically separate clus-
ters than did a hierarchical approach (Ward,
1963).
Split-half samples of 100 respondents were
12 Frederick H. Navarro
used to produce two Q-correlation matrices
(correlation matrices correlating subjects rather
than variables), and each matrix was factor ana-
lyzed using orthogonal Varimax rotation. Because
individual measures may be positively or nega-
tively correlated with generated factor scores, it is
likewise possible for any given subject to be either
positively or negatively correlated with a set of
factor scores. As such, two initial subject clusters
are potentially identifiable for each factor. Both
the positive and negative loadings of all subjects
on all factors, therefore, were examined, and sub-
jects were assigned to the factor clusters on which
they had the largest positive or negative loading.
Subjects who did not have a minimum loading of
at least ±0.33 on a factor would not assigned using
the criterion developed by Stephenson (1953).
Approximately one to two subjects per survey ar-
ea could not be assigned using this rule.
Early Q-factor studies of data from both the
North Carolina study and the present study indi-
cated that rotated factor solutions accounting for
70 percent of the variance across subjects was suf-
ficient to ensure that almost all subjects in each
split-half sample met the minimum criterion for
classification in a cluster. Nine to ten rotated fac-
tors were produced in each market using this crite-
rion and evaluation of decreasing eigenvalues in-
dicated that this cut-off level was appropriate. Q-
factor identified groups were then entered into a
multiple discriminant analysis routine to 1) pro-
duce mean and standard deviation profile vectors
for each cluster, 2) evaluate the classification con-
fusion matrix, and 3) assess the statistical sepa-
rateness of clustered subjects (Klecka, 1980).
Validating Clusters within Markets
Reliability of cluster solutions within markets
was evaluated by, first, constructing a matrix of
distance indices between each cluster solution
within the split-half samples. Each distance index
is constructed using an unweighted average of
Euclidian distances between the split-half cluster
mean vectors (Lehman, 1979). This produces an
m by n matrix of distances, where m refers to the
number of clusters produced in the split-half sam-
ples with odd cases and n refers to the number of
clusters produced in the split-half samples with
even cases. The minimum distance between each
cluster mean vector in one split-half solution and
the set of cluster mean vectors in the other solu-
tion were then identified. An average of minimum
distances was used as the index of optimum simi-
larity (Lesser & Hughes, 1986). This formula ap-
pears in Appendix 4.
Reliability of cluster solutions within a market
was then evaluated by comparing the mean of split
-half minimum distances with the mean of mini-
mum distances developed by replacing each of the
cluster mean vectors of the odd split-half sample
with Monte Carlo synthetic-random mean vectors.
A significantly smaller mean for the actual split-
half sample minimum distances provides evidence
supporting reliability of the solutions. These com-
parisons are summarized in Table 1 (see compari-
sons A:B and A:C), and support the reliability of
the cluster solutions.
Multiple Discriminant Analysis
As a final reliability check, validated cluster
solutions in each market were subjected to a mul-
tiple discriminant analysis. Multiple discriminant
analysis includes many features useful in analyz-
ing group differences. For this research, one goal
was to derive Wilk's lambda coefficient, which is
a multivariate measure of group differences over
several variables (Klecka, 1980). Wilk's lambda is
an inverse measure with values ranging between
0.0 and 1.0. As Wilk's lambda increases toward its
maximum value of 1.0, it progressively reports
less discrimination (Klecka, 1980).
Another feature of multiple discriminant
analysis is the derivation of discriminant func-
tions. These discriminant functions represent di-
mensions which can describe group differences.
Like canonical variates in canonical correlation
analysis or factors in principal components analy-
sis, each discriminant function is orthogonal to the
13
Attitudes Toward Health Care
next and explains successively less variation in the
data. Although it is mathematically possible to
derive as many discriminant functions as varia-
bles, very often the first few discriminant func-
tions possess the majority of the discriminating
power and the remainder offer only trivial solu-
tions (Klecka, 1980). Wilk's lambda is the most
common measure used to assess residual discrimi-
nation among groups. When discriminant func-
tions are examined in seeking to interpret multiple
discriminant analysis output, each function is first
tested to determine if it explains sufficient residual
discrimination. This assessment is made prior to
deriving each discriminant function. Thus, the
very first test of discrimination occurs before cal-
culating the first discriminant function. If k repre-
sents the number of discriminant functions calcu-
lated, then k=0 indicates the case where the resid-
ual discrimination is measured before deriving the
first discriminant function. Since the goal of this
analysis is mainly to validate cluster differences
and not proceed to calculate and interpret discri-
minant functions, the assessment of statistically
significant differences across the groups in every
area occur when k=0.
Wilk's lambda is an intermediate variable, which
can be used to derive tests of significance by con-
verting it to an approximation of either the Chi
Square of F distribution, making it directly com-
parable with common statistical tables (Klecka,
1980). These calculations were made for each area
surveyed and the results summarized in Table 2.
In every market the differences between clusters
were statistically significant. Pairs of clusters in
the odd and even split-half samples exhibiting
high similarity in composition of respondents and
low distances were combined within each market
and validated again using mulitple discriminant
analysis, as shown in the total columns of Table 2.
The group numbers in the total columns for each
market represent the number of actual clusters
cross-validated through the split-half methodolo-
gy. Looking at all markets, between six and eleven
clusters were validated within each, with the aver-
age number of clusters validated in this way being
eight. This completed initial attitude/behavior seg-
mentation identification and validation within a
market.
Identification and Validation of Segments Across
Markets
To evaluate the similarity of cluster solutions
across markets, a set of across-market distance
matrices was calculated. First, the average across-
market minimum distance index was compared to
the synthetic profile of random mean vectors. A
statistically significant lower minimum across-
market distance index provides evidence that the
profile vectors across markets are more similar
than expected by chance. This was the case as
summarized in comparisons D:E and D:F in Table
1. Also, comparison of the within-market mini-
mum distance index with the across-market mini-
Table 1: Comparison of Segment Profiles
Source n Mean SD Comparison t-statistic df p-value
A. Split Half Indices 10 15.0 0.97
Synthetic Indices
B. N = 20 10 19.1 1.07 A:B 8.98 19.8 0.0001
C. N = 10 10 19.4 0.78 A:C 11.19 19.1 0.0001
D. Across Market Indices 45 12.6 0.74 A:D 7.36 12.7 0.0001
Synthetic Distances
E. N = 20 10 17.7 1.09 D:E 14.09 12.1 0.0001
F. N = 10 10 18.1 0.93 D:F 17.51 12.6 0.0001
14 Frederick H. Navarro
mum distance index revealed that the across-
market index was significantly less than the with-
in-market index (see comparison A:D in Table
1). This provides evidence that segments across
markets are more similar than the within-market
clusters identified in each split-half sample.
The analysis described above provides strong
evidence regarding the similarity of healthcare
consumer segments occurring across markets and
show that psychographically based healthcare
segments, as defined by the selected dimensions,
can be generalized across markets. As a final
check on the discrimination of psychographically
based healthcare consumer segments, the confu-
sion matrix results of the eleven multiple discri-
minant analyses performed on the validated seg-
ments within the ten market areas and the final
across-market set of psychographic segments
were evaluated. These results are summarized in
Table 3. Table 3 shows that in every market area
the improvement in correct reclassifications as a
result of the discriminant functions was large and
easily exceeded those expected on the basis of
random change. This was also the case for the
correct classification among the combined psy-
chographic segments across markets.
The last phase of the analysis consisted of
combining similar segments across markets on
the basis of distance between pairs of profiles.
The across market analysis of the data revealed
that nine major healthcare consumer segments
occur across-markets, and account for between
60 percent and 90 percent of consumers in any
one market. After identifying the nine, 19 vali-
dated within-attitude/behavior segments re-
mained. These segments could not be rolled into
the nine across-market groups because they were
too similar to two or more of them when evaluat-
ed on the basis of the similarity of their distance
indices and visual inspection of their line pro-
files. This made it too difficult to determine
where they should be classified (i.e., these 19
validated groups had some traits similar to more
than one of the nine major groups).
Prior to segment profiling, respondents repre-
senting the nine major across market groups, in-
cluding the unassigned, were isolated in a separate
subsample. These cases were then entered into a
multiple discriminant analysis program. Through
an analysis of the resulting confusion matrix, the
posterior probabilities of each case, and the Ma-
halanobis distances for each case, a subsample of
"ideal" types was identified and created to act as a
"classification" data file. This classification sub-
sample was created to represent the maximum dif-
ferences between the nine major groups. A posteri-
or probability level of 0.70 was used as the cut-off
point in the case by case evaluation for determining
which cluster members would be retained or
dropped from the "classification" subsample. The
classification subsample was used to create a set of
multiple discriminant functions capable of optimal-
ly classifying additional cases into one of the nine
major groups or leave them unassigned. A multiple
discriminant analysis of this classification subsam-
ple was significant (X2 = 2,950; df = 192,p .0001),
and an evaluation of the confusion matrix results
showed excellent levels of correct reclassification
(tau = .96; Klecka, 1980). This multiple discrimi-
nant analysis provided a set of optimum discrimi-
nant functions in terms of differentiating the nine
major groups and the unassigned. These functions
were then used to reclassify the remaining cases
left in the 19 validated groups, as well as the cases
that were screened out of the classification subsam-
ple.
Results
Identification and Description of Prototype Seg-
ments Across Markets
The proportion of consumers in each of the
nine most common healthcare segments after re-
classification are reported in Table 4. Given the
nature of this analysis, it remains to be seen if this
proportional mix of the segments will hold up in
future studies. Interpretation of psychographic di-
mensions provides insight into the nature of these
15
Attitudes Toward Health Care
segments. Summary statistics for the nine major
psychographic segments were used to construct
demographic and socio-economic profiles. This
analysis provides information about the age groups
represented in each segment, the distribution of
males and females, the size and structure of the
family, and relative buying power as indicated by
per capita income levels. Segment summary de-
scriptions across the attitude/behavior dimensions
are provided in Appendix 5. A second level of seg-
ment description follows from relationships be-
tween attitude/behavior descriptions and de-
mographics. These are shown in Table 5. Data indi-
cate that there are several consistent relationships
between the healthcare psychographic profiles and
demographic characteristics. Segment 3, which is
characterized as a price conscious segment, has the
second lowest per capita income. Segment 4, made
up of consumers who stress high quality, traditional
care, and a heavy reliance on physicians, tend to be
older, may be widowed, and have very few young
children. Segment 5 is made up of consumers who
have larger than average families, many young
children, tend to be female, and are relatively
young. It is not surprising that these consumers are
those most focused on family health. Segment 7,
containing those healthcare consumers who are
highly motivated to use medical services and are
not deterred by cost, have the second highest per
capita income, are mostly female, and tend to be
older than average. The segment characterized by a
focus on sports, exercise, and nutrition, referred to
here as Segment 8, is the most male dominated.
These healthcare consumers are also typically
younger and similar in age to members of Segment
5. They are also the most affluent.
The fact that there are some predictable demo-
graphic and socioeconomic congruencies with the
psychographic profiling adds credibility to the clus-
ter solutions. The finding that only some of these
psychographic segments possess such correspond-
ences also indicates that the psychographic dimen-
sions used have identified distinct healthcare con-
sumer segments that could otherwise not be identi-
fiable on the basis of demographic or socioeco-
nomic information.
Discussion
In the study conducted by Lesser and Hughes
(1986), segments of psychographic shopper types
were found to exist and could be generalized
across markets. Using a similar methodology,
this investigation demonstrated that psycho-
graphically based healthcare segments exist and
can likewise be generalized across geographic
markets. The fact that not all consumers were
classifiable into a validated psychographic seg-
ment indicates that a small portion of the con-
sumer marketplace does not share common
healthcare attitudes and behaviors. The nine clus-
ter-based psychographic segments identified can,
therefore, be thought of as existing within a
slightly larger, diverse array of healthcare attitu-
dinal and behavioral dimension combinations.
In terms of healthcare marketing and plan-
ning, the evidence presented here shows that
healthcare planners, marketers, and advertisers
can promote their services to groups of like
minded consumers on the basis of healthcare psy-
chographics. This means that campaigns de-
signed for one geographic market can be reasona-
bly transferred to other areas of the country. For
those institutions or organizations marketing
healthcare products and services on a national
basis, these findings suggest that considerable
cost savings are possible because resources need
only be spent developing one basic campaign
theme and strategy.
Finally, apart from variations in the size of
these nine segments across markets, this study
indicates that different areas of the country may
not be very different in terms of the number and
nature of the patterns of healthcare attitudes and
behaviors manifested by their respective popula-
tions.
This investigation, using a cross-sectional
methodology, has demonstrated that psycho-
graphic healthcare segments defined by common
16 Frederick H. Navarro
Table 2: Multiple Discriminant Analysis of Split-half Cluster Analysis Solutions for Every Market
Areas = Denver, CO Dallas, TX Manchester, NH
Samples* = Odd Even Total Odd Even Total Odd/ Even Total
Number of
groups = 13 15 11 13 11 8 11 12 7
Discriminant
functions
tested = 0 0 0 0 0 0 0 0 0
Wilk's lambda = .00014 .0001 .0054 .0006 .0003 .0182 .0009 .0002 .0277
X2 = 709.9 727.6 945.1 597.2 648.9 733.2 568.1 685.6 656.3
df = 300 350 250 275 300 150 250 275 150
p < .001 .001 .001 .001 .001 .0001 .001 .001 .0001
Areas = Louisville, KY Columbus, OH Southern Calif.
Samples* = Odd/ Even Total Odd/ Even Total Odd/ Even Total
Number of
groups = 13 11 8 11 15 7 13 12 6
Discriminant
functions
tested = 0 0 0 0 0 0 0 0 0
Wilk's lambda = .0002 .0008 .0184 .0008 .0002 .0425 .0002 .0004 .0335
X2 = 681.4 577.6 729.2 577.6 672.9 577.9 681.4 629.8 623.2
df = 300 250 175 250 350 150 300 275 125
p = .001 .001 .001 .001 .001 .0001 .001 .001 .0001
Areas = Pittsburgh, PN Indianapolis, IN Seattle, WA
Samples = Odd/ Even Total Odd/ Even Total Odd Even Total
Number of
groups = 15 11 10 12 11 7 18 16 10
Discriminant
functions
tested = 0 0 0 0 0 0 0 0 0
Wilk's lambda = .0005 .0003 .0067 .0005 .0016 .0355 .00001 .00001 .0065
X2 = 600.5 657.0 908.5 611.9 521.5 610.9 892.3 903.8 914.0
df = 350 250 225 275 250 150 425 375 225
p = .001 .001 .0001 .001 .001 .0001 .001 .001 .0001
Areas = Jacksonville, FL
Samples = Odd Even Total
Number of
groups = 18 16 6
Discriminant
functions
tested = 0 0 0
Wilk's lambda = .0001 .0001 .0382
X2 = 713.8 723.0 599.1
df = 425 375 125
p = .001 .001 .0001
*Sample sizes for all odd and even split-half samples is 100. Every discriminant analysis uses 25 dis-
criminating variables.
17
Attitudes Toward Health Care
patterns of health attitudes and behaviors exist, and
that they can be generalized across geographic mar-
kets. Further research might investigate these psy-
chographic healthcare segments on a longitudinal
basis. Do consumers move from one segment to
another? Is there any indication that people in Seg-
ments 5 and 8, which tend be younger, might be-
come more like Segments 4 and 7, which tend to be
older? might consumers in some of the lower in-
come levels, like Segments, 3, 5, and move into
those segments that enjoy higher incomes, for ex-
ample Segments 2, 4, 7, and 8? Research efforts
might also focus on examining actual healthcare
utilization levels across the nine psychographic
segments and seek to identify to what extent psy-
chographic type accounts for utilization differ-
ences, if any, over and above other variables,
such as age or sex?
The goal of this research was to identify and
provide the healthcare industry with a reliable,
valid, psychographic model of healthcare con-
sumer segments defined by the similarity of the
interactive patterns of their expressed health atti-
tudes and behaviors. Great care was taken in
identifying useful, sufficiently specific dimen-
sions of healthcare attitudes and behavior, and in
making sure they have actionable marketing im-
plications. The importance of well documented
attitude/behavior segmentation to the healthcare
industry is clear. The care and thoroughness with
which this study was conducted ensures that the
identified profiles of attitudes towards healthcare
provide healthcare marketers with reliable insight
into dynamics shaping trends in the healthcare
consuming public.
REFERENCES
Aaker, D. A., & Day, G. S. (1983). Market re-
Table 3: Confusion Matrix Results for Multiple Discriminant Analysis of Validated Groups
Improvement
Areas N* # Correct #Randomly in
Classification Expected Prediction
Arcadia, Glendale, 180 157 30 85%
Pasadena, CA
Columbus, OH 186 152 27 79%
Dallas, TX 181 154 26 83%
Denver, CO 174 152 19 86%
Indianapolis, IN 187 158 27 82%
Jacksonville, FL 166 143 28 83%
Louisville, FL 187 157 23 83%
Manchester, NH 180 154 26 83%
Pittsburgh, PA 184 164 18 88%
Seattle, WA 180 153 18 82%
Total Across Market 1272 657 142 42%
*Samples include only respondents who could be classified into a group.
Table 4: Distribution of Nine Segments for all
Geographic Markets
Segments Number Proportion
1 143 7.2%
2 197 9.9
3 174 8.7
4 195 9.8
5 173 8.7
6 245 12.3
7 235 11.8
8 237 11.8
9 178 8.9
Non-classified 223 11.2
Total 2,000 100.3
18 Frederick H. Navarro
search, New York: John Wiley & Sons, Inc.
Anderson, D. C. (1982). The satisfied consumer:
Service return behavior in the hospital obstet-
rics market, Journal of Health Care Market-
ing, 2(4), 25-
Berki, S. E., R. Penchansky, R. S. Fortue, and M.
L. Ashcraft (1978), Enrollment choices in dif-
ferent types of HMOs: a multivariate analysis,
Medical Care, 16(August), 682-697.
Blendon, R. J. & Altman, D. E. (1984). Public
attitudes towards health care costs, The New
England Journal of Medicine, 311(August),
613-616.
Burnett, J. J. and Leigh, J. H. (1986). Distinguish-
ing characteristics of blood donor segments
defined in terms of donation frequency, Jour-
nal of Health Care Marketing, 6(2), 38-48.
Carroll, N. V. and Gagnon, J. P. (1983). Identify-
ing consumer segments in health services
markets: an application of conjoint and cluster
analysis to the ambulatory care pharmacy
market, Journal of Health Care Marketing, 3
(Summer), 22-34.
Coburn, D. & Pope, C. R. (1974). Socioeconomic
status and preventive health behavior, Journal
of Health & Social Behavior, 15(June), 67-78.
Finn, D. W., and Lamb, C. W. (1986). Hospital
benefit segmentation, Journal of Health Care
Marketing, 6(4):26-33.
Funkhouser, G. R., (1983), A note on the reliabil-
ity of certain clustering algorithms, Journal of
Marketing Research, 20(1), 99-102.
Gerstner, E. (1985), Do higher prices signal higher
quality?, Journal of Marketing Research, 22
(2), 209-215.
Hagerty, M. R. (1985), Improving the predictive
power of conjoint analysis: the use of factor
analysis and cluster analysis, Journal of Mar-
keting Research, 22(May), 168-184.
Harrel, G. D., & Fors, M. F. (1985). Marketing
ambulatory care to women: a segmentation
approach, Journal of Health Care Marketing, 5
(2), 19-28.
Haug, M. R. & Lavin, B. (1981). Practitioner or
patient - who's in charge?, Journal of Health
and Social Behavior, 22(September), 212-229.
Hulka, B. S., Zyzanski, S. J., Cassell, C. &
Thompson, S. J. (1970). Scale for the meas-
urement of attitudes towards physicians and
primary medical care, Medical Care, 8
(October), 429-36.
Kim, J. & Mueller, C. W. (1978). Introduction to
factor analysis, Beverly Hills, CA: Sage Pub-
lications, Inc., p. 49.
Klasterin, T. D. (1983). Assessing cluster analysis
results, Journal of Marketing Research, 20
(February), 92-8.
Klegon, D. (1981). The role of consumer attitudes
and preferences in the HMO planning process,
Journal of Health Care Marketing, 1
Table 5: Psychographic Segment Demographic and Socioeconomic Characteristics
Variables Psychographic Segments
1 2 3 4 5 6 7 8 9
n = 143 197 174 195 173 245 235 237 178
Female = 49.0% 42.6% 48.3% 48.7% 54.9% 48.6% 72.8% 38.4% 66.9'
Male = 51.0 57.4 51.7 50.3 45.1 51.4 27.2 61.6 33.1
Average Age = 42.6 46.3 43.1 59.5 36.6 41.9 51.6 38.1 45.3
Average Family
Size = 2.7 2.3 2.6 1.9 3.2 2.7 2.3 2.5 2.7
Children 5 or
under = 34.0% 14.9% 20.4% 5.7% 48.3% 25.0% 11.8% 21.3% 26.5%
Per capita
income
(thousands) = $13.3 $15.7 $12.5 $15.1 $11.8 $14.0 $15.8 $16.3 $12.7
19
Attitudes Toward Health Care
(Summer), 20-28
Klecka, W. R. (1980). Discriminant analysis,
Beverly Hills, CA: Sage Publications, Inc.
Lehman, D. R. (1979). Market research and anal-
ysis, Illinois: Richard D. Irwin, Inc.
Lesser, J. A. and Hughes, M. A. (1986). The gen-
eralizability of psychographic market seg-
ments across geographic locations, Journal of
Marketing, 50(January), 18-27.
Malhotra, N.K. (1986). Market segmentation and
strategic growth opportunities for hospitals,
Journal of Health Care Marketing, 6(June),
2-6.
Modern Healthcare (1986). Hospitals beginning
to use market segmentation, 16(November),
146.
Ortinau, D. J. (1986). Discriminating users and
nonusers of preventive health practices and
emergency walk-in clinics, Journal of Health
Care Marketing. 6(2), 26-37.
Paul, T. (1987), Guest editorial: emerging em-
ployee health care concerns, Journal of
Health Care Marketing, 7(3), 2-5.
Powills, S. (1987), MD influence on consumers
waning slightly, Hospitals, 61(May), 44.
Slama, M. E. & Tashchian, A. (1985). Selected
socioeconomic and demographic characteris-
tics associated with purchasing involvement,
Journal of Marketing, 49(Winter), 72-82.
Stephenson, W. (1953). The study of behavior,
Chicago: University of Chicago Press.
Ward, J. H., Jr. (1963), Hierarchical grouping to
optimize an objective function, American
Statistical Association Journal, 58, 236-244.
Wells, W. D. (1975). Psychographics: a critical
review, Journal of Marketing Research, 2.4
(August), 27-35.
- & Tigert, D.J. (1971). Activities, interests,
and opinions, Journal of A dvertising Re-
search, 11(August), 27-35.
Wind, Y. (1978). Issues and advances in segmen-
tation research, Journal of Marketing Re-
search, 15(August), 327-37.
Young, S., Ott, L. & Feigin, B. (1978). Some
practical considerations in market segmenta-
tion, Journal of Marketing Research, 15
(August), 405-12.
Notes
1. A health maintenance organization or HMO is
a prepaid health insurance plan. Members of an
HMO typically pay a monthly premium, and for
this premium all or most of their healthcare
expenses are covered. These plans are different
from typical indemnity insurance plans because
members of HMO plans are usually limited to
using only physicians and hospitals that are part
of the HMO.
Appendix 1
Psychographic Statements
Scale:
1 Strongly disagree
2 Somewhat disagree
3 Neutral
4 Somewhat agree
5 Strongly agree
(Statement numbers indicate order presented)
Propensity to experiment with health care
alternatives
11. When I get ill, I always go to the same
health care professional or doctor.
20. When I get sick, I do what my parents
used to do for me.
32. I try to use alternative methods, apart
from medications and drugs, to cure illnesses
and keep myself healthy.
40. If I had to be hospitalized, I would com-
pare hospitals near me before deciding where
to go.
Propensity to avoid health care
24. I do not seek professional health care un-
less I am very sick or injured.
27. When it comes to my health, I rarely plan
ahead and usually take things as they come.
20 Frederick H. Navarro
Attitudes toward medical professionals
14. Most health care professionals are not as
competent as we are led to believe.
30. I only seek professional health care when
I am really sick because it is too expensive.
37. Doctors often try new drugs on their pa-
tients without knowing all the effects.
Health care information seeking
18. I actively seek out health information be-
yond what health care professionals tell me.
23. I gather health information so that I can
choose from among health care alternatives.
Receptivity to health care advertising
29. I have used a health care service that I
learned about through advertising.
35. I have no reservations about using a hos-
pital service I see advertised.
36. I actively pay attention to ads about
health care.
38. I am very suspicious of medical profes-
sionals who advertise their services.
Appendix 2
Sampling Across Geographic Markets
Location Sample Size
Louisville, KY 200
Jacksonville, FL 200
Denver, CO 200
Dallas, TX 200
Columbus, OH 200
Glendale, Pasadena,
Arcadia, CA 200
Manchester, NH 200
Indianapolis, IN 200
Pittsburgh, PA 200
Seattle, WA 200
Total 2,000
Appendix 3
R Type Factor Analysis to Reduce Psycho-
graphic Inventory: Summary of Factors, Fac-
tor Loadings, and Questions For Samples of
200 and 100
First Sample n = 200
Question, Factor Loading, Question Number
and Content
1 0.86 22. I try to keep my body in top
physical shape.
0.81 28. I am very involved in keep-
ing myself healthy.
2 0.77 13. I have regular check-ups.
0.73 33. If I get ill, I know exactly
what I will do, who I will call, and where I will
go.
3 0.53 16. I mostly depend on the
health advice of others.
0.74 20. When I get sick, I do what
my parents used to do for me.
4 0.68 37. Doctors often try new drugs
on their patients without knowing all the ef-
fects.
0.72 40. If I had to be hospitalized, I
would compare hospitals near me before
deciding where to go.
5 0.59 8. I have tried to save money by
shopping around for health care.
0.61 23. I gather health information
so that I can choose from among health care
alternatives.
6 0.48 3. I really enjoy the role as the
21
Attitudes Toward Health Care
family health care decision maker.
0.76 21. I make my own health care
decisions.
7 0.70 14. Most health care profession-
als are not as competent as we are led to be-
lieve.
0.69 30. I only seek professional
health care when I am really sick because it is
too expensive.
8 0.77 25. I will pay more for high
quality health care rather than less for care
that
is just adequate.
0.55 26. If doctors in my area
charged different fees, I would never go to the
lowest priced doctor.
9 0.48 36. I actively pay attention to
ads about health care.
0.73 39. I spend a great deal of time
each day thinking about my state of health.
10 0.83 2. If my family has average
health, I am satisfied.
-0.27 19. I regularly compare health
care providers to see if I can improve the care
I receive.
11 0.83 35. I have no reservations about
using a hospital service I see advertised.
0.39 41. If doctors and hospitals ad-
vertised their prices, I would certainly shop
more
by price.
12 0.42 5. I rarely think about the food I
eat.
0.63 9. I am very certain that I can
count on good health in
the future.
-0.43 32. I try to use alternative meth-
ods, apart from drugs
and medications, to cure illnesses and
keep myself
healthy.
13 -0.79 1. Members of my family are
responsible for their own
health.
0.41 36. I actively pay attention to
ads about health care.
14 0.67 10. I actively seek information
about nutrition.
-0.67 27. When it comes to my health,
I rarely plan ahead and
usually take things as they come.
15 0.86 7. When I get ill, I rely on my
spouse to tell me where
I should go and who I should see.
0.72 17. My spouse makes health
care decisions for the family.
Second Sample n = 100
Question, Factor Loading, Question Number
and Content
1 -0.71 36. I actively pay attention to
ads about health care.
-0.78 40. If I had to be hospitalized,
I would compare hospitals near me before
deciding where to go.
2 -0.72 10. I actively seek information
about nutrition.
0.63 27. When it comes to my
health, I rarely plan ahead and
usually take things as they come.
3 -0.75 6. When it comes to decisions
about my health, the doctor is the boss.
-0.66 31. It is possible that I could
get sick at any time.
4 0.75 14. Most health care profes-
sionals are not as competent as we are led to
believe.
0.69 37. Doctors often try new
drugs on their patients without knowing all
the effects.
5 -0.79 7. When I get ill, I rely on my
spouse to tell me where I should go and who
I should see.
-0.72 17. My spouse makes health
care decisions for the family.
6 0.61 20. When I get sick, I do what
my parents used to do for me
0.73 32. I try to use alternative
methods, apart from medications and drugs,
to cure illnesses and keep myself
healthy.
7 -0.86 22. I try to keep my body in
top physical shape.
-0.77 28. I am very involved in
keeping myself healthy.
8 0.74 4. I regularly participate in
competitive sports.
0.60 35. I have no reservations
about using a hospital service I see adver-
22 Frederick H. Navarro
tised.
9 0.74 12. I have saved coupons for discounts on health care services.
-0.61 25. I will pay more for high quality health care rather than less for care that is
just adequate.
10 0.43 26. If doctors in my area charged different fees, I would never go to the lowest
priced doctor.
-0.80 41. If doctors and hospitals advertised their prices, I
would certainly shop more by price.
11. -0.74 29. I have used a health care service that I learned about through advertising.
0.53 38. I am very suspicious of medical professionals who advertise their services.
12 0.78 1. Members of my family are responsible for their own health.
0.33 2. If my family has average health I am satisfied.
13 -0.35 3. I really enjoy the role of the family health care decision maker.
-0.78 21. I make my own health care decisions.
14 0.69 5. I rarely think about what food I eat.
0.66 8. I have tried to save money by shopping around for health care.
Appendix 4
Distance Index
g1 g2
Index = Σmin(dij.j) + Σmin(dji.i) / (gi + g2)
i=1 j=1
Where:
g1 = number of solutions in odd split-half sample g2 = number of solutions in even split- half
sample
d = average distance between solutions
i, j = solutions in the various sample areas
Appendix 5
Summary Description of the Nine Major Across-Market Segments With 95 Percent Confidence
Intervals
Segment 1 -- 7.1 to 7.2 percent of healthcare consumers. These healthcare consumers are gener-
ally distrustful and skeptical of the medical profession. They believe that medical professionals
generally do not live up to their reputation as competent experts. Of all the groups, these con-
sumers are the most skeptical of healthcare ads and their promises. Despite their skepticism,
this group is the most likely to use the same healthcare provider and to follow predefined pat-
terns of healthcare practices.
Segment 2 -- 8.6 to 11.2 percent of healthcare consumers. Members of this segment refrain from
seeking or using healthcare services as much as possible. The cost of healthcare is definitely an
issue for them. They are a difficult market to communicate with; they have little interest in fit-
23
Attitudes Toward Health Care
ness, nutrition, or in gathering healthcare information.
Segment 3 -- 7.5 to 9.9 percent of healthcare consumers. Members of this psychographic con-
sumer group balance price and quality concern when seeking healthcare. They are most likely
to have tried to save money by shopping around for low-priced healthcare and to have experi-
mented with alternative delivery outlets.
Segment 4 -- 8.5 to 11.1 percent of healthcare consumers. These healthcare consumers want
high quality and traditional medical care. They are very much physician driven; that is, they are
heavily influenced by doctors when making healthcare decisions. If an organization is well
known for specific medical procedures or product lines and is generally perceived as prestig-
ious, it will probably attract a large share of this segment in its service area.
Segment 5 -- 7.5 to 9.9 percent of healthcare consumers. These healthcare consumers put their
family's health above all other healthcare matters. They feel responsible for and constantly seek
to enhance the condition of their family's health. They enjoy their role as family healthcare de-
cision makers, and are the least likely group to seek the assistance of their spouse when
making healthcare decision.
Segment 6 -- 10.9 to 13.7 percent of healthcare consumers. These consumers are similar to the
Segment 7 members in many respects. They are willing to use healthcare services even for mi-
nor ailments, and are not deterred by the expense. To ensure that their health is consistently
good, they are second only to the Segment 7 members in making sure that they receive regular
check-ups. These consumers tend to feel comfortable returning to the same hospital for all their
needs. This loyalty makes them less inclined to gather healthcare information or to compare
hospitals.
Segment 7 -- 10.4 to 13.2 percent of healthcare consumers. This psychographic consumer group
is highly motivated to use medical services and is not deterred by the expense involved. They
readily utilize medical services even for minor ailments. In general, they are not fitness orient-
ed, but they do focus on maintaining good nutrition habits.
Segment 8 -- 10.4 to 13.2 percent of healthcare consumers. Exercise, sports, and good nutrition
are important to this psychographic type. Being less physician driven than both Segment 4 and
Segment 7 members, healthcare consumers matching the Segment 8 profile are more likely to
try different providers and experiment with healthcare alternatives. They are looking for the
long-term benefits of healthcare and are fairly involved in seeking healthcare information.
Segment 9 -- 7.7 to 10.1 percent of healthcare consumers. Very much like Segment 8, Segment
9 members are highly health and fitness oriented. They have a strong interest in nutrition, and
are motivated to gather healthcare information. They are the most likely group to experiment
with alternatives to traditional medical care.
24 Frederick H. Navarro
25
Attitudes Toward Health Care
... The Patterns of Adapting to Health (PATH) is a psychographically-based model that identifies groups of adults whose adaptive response to local healthrelated contexts has settled around nine welldefined and reliable patterns of health behavior. The nine PATH were identified in a national study exploring the reliability of patterns of self-reported health-related behavior across geographic regions of the U.S. (Navarro, 1990). The value of the PATH has been demonstrated by its over 30 years of application within the health care and health insurance industry. ...
... A proprietary set of classification functions (Navarro, 1990) was used to identify the dominant PATH of participants in both studies based on responses to the AHBI. The classification functions also identified a No Pattern group of respondents whose health behavior showed no evidence of settling around a defined pattern (Navarro, 1990). ...
... A proprietary set of classification functions (Navarro, 1990) was used to identify the dominant PATH of participants in both studies based on responses to the AHBI. The classification functions also identified a No Pattern group of respondents whose health behavior showed no evidence of settling around a defined pattern (Navarro, 1990). The predictive direct effects of the PATH on 1) selfreported health status and 2) the risk odds of a Type 2 diabetes diagnosis after controlling for health status were tested using the analytic model in Figure 1. ...
Preprint
Full-text available
Problem. The incidence of Type 2 diabetes is expected to more than double over the next three decades. Several behavioral drivers of Type 2 diabetes have been identified. Cluster-based studies support the self-organization of health-related behavior in specific patterns with their own effects on population health. The Patterns of Adapting to Health (PATH) system of health-related behavior patterns were hypothesized to influence Type 2 diabetes incidence both directly and indirectly. Participants: 4,380 Hawaii-based health plan members including 1,412 Type 2 diabetics responded to a mail survey. Methods: Self-reported health status was measured with a subset of items from the SF-36. The Adaptive Health Behavior Inventory (AHBI) assessed multiple health-related behaviors used as inputs to identify the nine PATH. Multiple regression and logistic regression evaluated the PATH effects on self-reported health status and the odds of Type 2 diabetes after controlling for self-reported health status, respectively. Results: Five PATH had significant lowering or raising effects on self-reported health status, β =-0.05 [99%CI:-0.42 to -0.04] to β = 0.17 [99%CI: 0.35 to 0.64]. Two of the nine PATH were associated with increased odds of Type 2 diabetes, OR = 2.86 [99%CI: 1.31 to 6.24] to OR = 3.84 [99%CI: 1.41 to 6.84]. Conclusions: Two population-level patterns of health behavior were associated with increased risk odds of Type 2 diabetes after controlling for health status. Implications: Addressing individual behavioral drivers of Type 2 diabetes may be insufficient to counter the effects of population-level patterns of behavior on Type 2 diabetes incidence.
... A persona is defined as a single, specific hypothetical/fictitious person who represents a segment of the population [70] with a realistic name, a face, and a description of their character (needs, goals, hopes, dreams, and attitudes). The Blueprint persona descriptions, developed by the Blueprint Partners, include also behavioral characteristics, which could affect both short-term and long-term success with interventions directed toward managing a disease or adopting wellness [71,72], for example, a persona's trust or lack of trust in care professionals, their self-management capabilities, and specific details about their character (e.g., being prone to aggressive behavior or having the tendency to reject outside support). Twelve different personas have been developed by the Blueprint Partners, which represent different "population segments". ...
Article
Full-text available
According to the World Health Organization (WHO), the worldwide obesity rate has tripled since 1975. In Europe, more than half of the population is overweight and obese. Around 2.8 million people die each year worldwide as a result of conditions linked to being overweight or obese. This study aimed to analyze the policies, approaches, and solutions that address the social and health unmet needs of obese patients, at different levels, in order to simulate the definition of an integrated approach, and to provide and share examples of innovative solutions supporting health promotion, disease prevention, and integration of services to improve the collaboration between the different health and care stakeholders involved across the country and in the lives of obese patients. A collaborative approach involving various levels of government and regional experts from different European countries was applied to identify, explore, and evaluate different aspects of the topic, from the innovation perspective and focusing on a European and a regional vision. Currently, people prefer more foods rich in fats, sugars, and salt/sodium than fruits, vegetables, and fiber. This behavior leads to a significant negative impact on their health-related quality of life. Changes in healthcare systems, healthy policy, and approaches to patient care and better implementation of the different prevention strategies between all the stakeholders are needed, taking advantage of the digital transformation of health and care. Such changes can support obese patients in their fight against an unhealthy lifestyle and at the same time reduce healthcare costs.
Preprint
Full-text available
Objective: Describe research demonstrating that the Adaptive Health Behavior Inventory (AHBI) can detect and predict a range of adaptive health behavior differences between men and women and as adults age. Methodology: A cross-sectional analysis of AHBI response data collected from four surveys of adults at different times and geographies (Sample 1: 2001 national telephone-mail survey, n = 20,685) and three online surveys conducted in 2017 and 2018 (Sample 2: Baltimore–Washington DC, n = 2,002; Sample 3: Atlanta, Georgia, n = 2,000); Sample 4: Cincinnati, Ohio, n = 2,000). Analysis: Logistic regression assessed AHBI predictive effects identifying male versus female status controlling for age. Multiple regression assessed the AHBI predictive effects in identifying adaptive health behavior differences associated with aging controlling for sex. Results: The overall R2, χ2, and odds ratios for all four logistic regression analyses were statistically significant predicting gender: Sample 1, χ2(1, 21) = 183.7, p < .001, R2 = .16; Sample 2, χ2(1, 21) = 42.4, p < .001, R2 = .31; Sample 3, χ2(1, 21) = 28.7, p < .001, R2 = .23; Sample 4, χ2(1, 21) = 28.7, p < .001, R2 = .23. The mean R2 for gender across all four samples represents a mean effect size of F2 = .30. Sixty percent (60%) of AHBI measures demonstrated some predictive reliability identifying gender differences. The overall R2, F, and β coefficients for all four multiple regression analyses were statistically significant predicting age: Sample 1, F(1, 21) = 183.7, p < .001, R2 = .16; Sample 2, F(1, 21) = 42.4, p < .001, R2 = .31; Sample 3, F(1, 21) = 28.7, p < .001, R2 = .23; Sample 4, F(1, 21) = 38.5, p < .001, R2 = .29. The mean R2 of Samples 2 through 4 was .26, represents a mean effect size of F2 = .35. Like with gender, sixty percent (60%) of AHBI measures demonstrated predictive reliability identifying lower or higher age. Discussion & Conclusions: Multiple measures within the AHBI demonstrate good and excellent predictive reliability in discriminating between women and men as well as younger versus older adults based on differences in adaptive health behavior.
Preprint
Full-text available
The purpose of this paper is to examine the psychographic method against current research and theory in social psychology, cognitive psychology, and cognitive neuroscience. The paper considers the psychographic method against research on language processing and language processing outside of conscious awareness. The paper also examines the lack of precision in the assessment of psychological attributes. It also considers issues around self-referential processing, the meaning of actions, linguistic context, mental frames, personality, and the application of cluster analysis in psychographic segmentation. In the end, this review supports a framework for redefining the psychographic method and justifying its broader application beyond its original discipline. The paper closes by offering a new definition of the psychographic method along with recommendations for improving its use.
Article
Full-text available
Мета: надати рекомендації для здійснення прицільної комунікації з ключовими групами споживачів послуг безперервного професійного розвитку (БПР) на основі попередньої сегментації. Матеріали та методи. Відповіді 1768 респондентів про важливість чинників вибору форми БПР піддані факторному аналізу, за результатами якого отримані чотири фактори, значення яких використані для проведення кластерного аналізу (метод к-середніх). Оцінка результатів здійснювалася на основі даних дескриптивної статистики, статистична вірогідність різниці показників визначалася методами попарного порівняння та дисперсійного аналізу. Результати. Виявлені чотири ключові фактори вибору форми БПР: професійність, доступність навчального курсу, рівень соціальної активності, формалізм слухача, - на підставі яких утворені сім кластерів слухачів, що відрізняються за соціодемографічними та психографічними характеристиками. Висновки. Критичними завданнями є оптимізація освітнього процесу, мінімізація документообігу, часу відриву від роботи фахівців, налагодження роботи з онлайн-медіа, облік, аналіз та комунікація з наявними споживачам освітніх послуг.
Article
Full-text available
Aim. To describe the conceptual framework of the joint research by NMAPE and UzhNU focused on the application of psychographic segmentation of the target audience on the factors of non-communicable diseases (NCDs). Object and Methods. Bibliosemantic, biostatistical, sociological methods, and the method of expert assessments were used. Reports and questionnaires of three international cross-sectional studies on the attitude of Ukrainian citizens to certain aspects of health, available WHO guidelines for the last 15 years on communication, prevention of NCDs, and promotion of a healthy lifestyle were analyzed. A Questionnaire has been developed to study the factors of occurrence and measures for preventing of non-communicable diseases (to be filled in by the public and medical doctors). Results. The developed Questionnaire for the Study of Factors and Prevention of Noncommunicable Diseases provides answers to questions in 14 areas, such as understanding the problem, a person's willingness to change behavior, take specific measures for maintaining health, assessing levels of personal engagement or barriers. Conclusion. The implementation of the survey and the analysis of respondents' answers were designed to determine the structure of the population's beliefs about NCD prevention measures, the main factors, population segments, and determinants of readiness to choose healthy behavior.
Article
Full-text available
Understanding marketing techniques is one of the skills needed by successful entre renews. Target marketing is an effective method for occupational therapy entrepreneurs to use in determining when and where to enter the marketplace. The two components of target marketing, market segmentation and the development of marketing mix strategies for each identified market segment, are described. The Profife of Attitudes Toward Health Care (PATH) method of psychographic market segmentation of health care consumers is presented. Occupational therapy marketing mix strategies for each PATH consumer group are delineated and compatible groupings of market segments are suggested.
Conference Paper
Full-text available
Consumer health technologies (CHT) are considered important catalysts for what some are calling the "dawn of a new era - the age of the empowered health consumer". To develop effective systems, we must reach the level of personal connection to capture the essence of user groups and their issues. User profiles and personas can be used to help the development team to mentally visualize and communicate how intended users of CHT's will interact with the proposed design in a meaningful way. We use an action research approach to call attention to important health related characteristics for CHT user profiles and personas, provide persona and profile artifacts for chronic diabetics, and illustrate the process and benefit of these tools in requirements generation, design, and implementation planning.
Article
Despite the widespread availability of segmentation techniques and much applied experience, many commercial users appear to be disenchanted with segmentation studies. The authors suggest that one cause is the failure of the marketing researcher to consider the competitive environment sufficiently in designing segmentation studies. They examine the problem in detail, setting forth some instances in which a segmentation approach is not useful and situations in which conventional approaches are not successful through examples from actual case studies.
Article
This research develops a Likert-type scale to measure final consumers' involvement with the purchasing activity. The scale is employed to test hypotheses concerning relationships between market characteristics and purchasing involvement. Purchasing involvement is found to be related to sex, education, income, and stage of the family life cycle.
Article
Psychographic segmentation solutions developed for 17 different cities were compared. The findings indicate that psychographic segments which are developed for markets in one geographic location are generalizable to markets in other geographic locations.
Article
This research develops a Likert-type scale to measure final consumers' involvement with the purchasing activity. The scale is employed to test hypotheses concerning relationships between market characteristics and purchasing involvement. Purchasing involvement is found to be related to sex, education, income, and stage of the family life cycle.
Article
This article presents case histories of five somewhat different uses of psychographic research, and it critically reviews the status of research in this field.
Article
Psychographic segmentation solutions developed for 17 different cities were compared. The findings indicate that psychographic segments which are developed for markets in one geographic location are generalizable to markets in other geographic locations.
Article
The author reviews the current status and recent advances in segmentation research, covering segmentation problem definition, research design considerations, data collection approaches, data analysis procedures, and data interpretation and implementation. Areas for future research are identified.