Pregnancy beliefs
 
 
 
 
 
AN AUDIENCE ANALYSIS OF FACTORS CONTRIBUTING TO
AFRICAN-AMERICAN AND WHITE TEENS' PREGNANCY HEALTH BELIEFS:
IMPLICATIONS FOR MASS MEDIATED HEALTH CAMPAIGNS
 
 
 
 
 
 
 
 
 
 
Rose G. Campbell
Doctoral Student
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Paper submitted to the AEJMC:  Minorities and Communication Division
 
Rose G. Campbell, M.S,, is a doctoral student, Department of Communication,
 Liberal Arts and Education Building (LAEB), Rm. 2159, Purdue University,
 
          West Lafayette, IN 47907
(317-494-3429).
 
 
 
 
 
 
 
 
AN AUDIENCE ANALYSIS OF FACTORS CONTRIBUTING TO
AFRICAN-AMERICAN AND WHITE TEENS' PREGNANCY HEALTH BELIEFS:
IMPLICATIONS FOR MASS MEDIATED HEALTH CAMPAIGNS
 
(STUDENT PAPER)
ABSTRACT
        Health professionals generally view the disparate health status of
 
     minority populations as a direct effect of socioeconomic factors.  This
 
         view does not explain, however, why the infant mortality rate for
 
   well-educated African-American women is considerably higher than for
 
      less-educated Caucasian women.  Further, this view has spurred largely
 
        unsuccessful mass media campaigns to address this problem.
        High school women were surveyed in this extensive audience analysis, thus
 
          identifying cognitive and affective factors that lead to
African-American
 
          and white teens' pregnancy health beliefs.   Important implications
for
 
         mass mediated public health campaigns that address the black infant
 
     mortality problem are provided.
         AN AUDIENCE ANALYSIS OF FACTORS CONTRIBUTING TO
AFRICAN-AMERICAN AND WHITE TEENS' PREGNANCY HEALTH BELIEFS:
IMPLICATIONS FOR MASS MEDIATED HEALTH CAMPAIGNS
Introduction
        The disparate health status of minority populations is one of our most
 
         serious problems in the United States.  In 1989, the United States
 
    Secretary of Health and Human Services Task Force on Black and Minority
 
         Health identified infant mortality as one the most important  health
 
      priority concerns based on high risk factors associated with minority
 
       populations (Office of Minority Health [OMH], 1989, p. 1).  Infant
 
    mortality is a critical health index, considered to be "the primary
 
     indicator of the overall health status of nations" because it is associated
 with access to basic needs such as "food, shelter, education, sanitation,
 
          and health care; and second, it is relatively easy to monitor with
basic
 
          vital statistics collected in all states" (Indiana  State Board of
Health
 
          Report [ISBH], 1989, p. 1).
        Major risk factors associated with infant mortality include smoking and
 
          substance abuse during pregnancy, late or no prenatal care, adolescent
 
        pregnancy, and poor dietary habits.  All of these factors contribute to
 
         having a low birthweight baby (weighing less than 5 pounds, 8 ounces at
 
         birth), which is "the most significant contributor" to infant mortality
 
         (ISBH, 1989, p. 5).  Further, pre-term babies that survive are at
greater
 
          risk of dying before their first birthday or of developing serious
 
    long-term disabilities.
        Health professionals report that seeking early and regular prenatal care
 
          (beginning in the first trimester and continuing throughout pregnancy)
 
        reduces the risk of having a low birthweight baby and subsequent infant
 
         mortality, not only because of the medical benefits, but also because
of
 
          the important prevention education provided during those visits, which
may
 
          improve the likelihood of having a healthy baby (OMH, 1990; Public
Health
 
          Centers for Disease Control [PHSCDC], 1988; St. John & Winston, 1989;
Uni
 
          ted States Department of Health and Human Services, 1989).  Of course,
 
        women should engage in healthy activities before they conceive, but
 
     changing health behavior during pregnancy also can improve pregnancy
 
      outcome significantly.  A primary goal, then, is to persuade women to
 
       obtain prenatal care early in their pregnancy.
        There is no doubt that limited access to health care services and lack of
 
          medical insurance coverage are formidable obstacles to obtaining
prenatal
 
          care services (Starr, 1982; OMH, 1989).  Yet statistics indicate that
lack
 
          of prenatal care is prevalent among many African-American women, not
just
 
          those who are in typically high risk sociodemographic populations.
This
 
          claim is supported by the Indiana State Board of Health Report (1989)
which
 showed that 81.4 percent of white women over 20 who had more than thirteen
 years of education received adequate prenatal care in 1988, as opposed to
 
          only 59.2 percent of black women in the same sociodemographic group.
 
       Further, nearly twice as many black women (19.5 percent) received
inadequate
 
            prenatal care than did white women (11.7 percent ) in poorly
educated populations that
 
           same year.  It appears that lack of adequate prenatal care is a
significant precursor of
 
            infant mortality among black women, regardless of socioeconomic
status.
Theoretical Frameworks for Examining Health Behavior
The Demographic Model and Mass Media Campaigns
        Although there is no doubt that demographic variables, (i.e., race,
 
      income, education) affect health behavior, there are many other important
 
          contributing factors not revealed by the way in which health
statistics
 
         typically are reported.  Nevertheless, many health studies continue to
 
        focus on demographic differences as key correlates of health status and
 
         health-seeking behavior.  For example, government and other health
studies
 
          illustrate pregnancy outcome differences through sociodemographic
predictor
 variables such as maternal age, marital status, income and educational
 
         level, and race (Centers for Disease Control, 1989; Lia-Hoagberg et
al.,
 
          1990; ISBH, 1989).
        This emphasis on sociodemographic factors has a significant impact on
 
        health communication research.  Many studies use samples representing
lower
 sociodemographic populations to analyze health behavior.  Additionally,
 
          the demographic model serves as the foundation for many public health
 
       campaigns in the mass media, because statistical analyses of health
status
 
          indicate that individuals with lower education and income levels are
in
 
         greater need of health education.  Consequently, the prevailing
strategy
 
          has been to disseminate accurate health messages to society's
"information
 
          poor," with the belief that increased knowledge will result in
healthier
 
          behavior.
        Unfortunately, few public health campaigns have succeeded at changing
 
        targeted individuals' health behavior and attitudes (Atkin, 1990; Rice &
 
          Atkin, 1989).  One explanation, the knowledge gap hypothesis, suggests
that
 low socioeconomic status and low education level are responsible for campaign
 
            failures because these factors create access and comprehension
barriers for some
 
     populations (Brown, Ettema & Luepker, 1981; Donohue, Tichenor & Olien,
1975; Gaziano,
 
          1983; Izard, 1985; Tichenor, Donohue & Olien, 1970).  Therefore, the
infusion of mass
 
          media information in high socioeconomic subgroups (who subsequently
share a higher
 
       education level) creates a gap in knowledge because those in higher
socioeconomic groups
 
            will acquire the information more efficiently than those in lower
socioeconomic groups
 
           with less education.
        Although important, the knowledge gap hypothesis does not provide answers
 
          for why those in lower socioeconomic levels who do receive and recall
 
       information from public health campaigns do not comply with campaign
 
      recommendations.  For example, the knowledge gap hypothesis does not
 
      explain why better educated African-American women are less likely to seek
 
          prenatal care than less educated white women:  In 1988, 72.4 percent
of
 
         white women with only a high school education received adequate
prenatal
 
          care while only 59.2 percent of black women who have completed at
least one
 year of college received adequate care (ISBH, 1989).
        Another reason public health campaigns may fail to reach targeted
 
    populations is that some individuals' perceptions of doctors and the health
 care setting may create powerful barriers to seeking health care.  This
 
          has significant implications for health messages, because discomfort
with
 
          doctors and the clinical setting may prevent an individual from
complying
 
          with public health campaign appeals.  This discomfort creates
communication
 barriers in the clinical setting, and ultimately prevents the successful
 
          delivery of health care (Arntson, 1989).  Therefore, those who are
 
    uncomfortable with doctors or other health professionals are unreceptive to
 public health campaigns that recommend seeking care in a medical setting
 
          (Pendleton & Hasler, 1983).
        Another explanation for campaign failures is offered by Anderson (1989),
 
          who suggests that public health campaigns typically focus on creating
 
       awareness rather than teaching skills that will enable individuals to
 
       change behavior.  Anderson presents Bandura's self-efficacy theory as a
 
         framework for his argument.  Self-efficacy theory asserts that by
providing
 information designed to empower people to engage in a recommended action,
 
          those who typically avoid particular behaviors will gain the
psychological
 
          and cognitive skills necessary to change.  But although self-efficacy
 
       theory is a useful concept for encouraging compliance with message
appeals,
 it does not provide a mechanism for attracting those who do not perceive
 
          they are engaging in unhealthy behavior.
         Additionally, one line of research focuses on text features of messages
 
          to explain why people do not absorb information.  For example,
reducing
 
         sentence, word, and paragraph length, or improving the visual
presentation
 
          of a message, may help simplify a text, rendering it more
understandable to
 a broader audience (for a review, see Rowan, 1988).  Though important to
 
          consider when explaining intellectually challenging information, these
 
        strategies based on readability and organization do not offer methods
for
 
          gaining the attention of an uninvolved audience.
        Clearly, the widely accepted demographic explanation for disparate health
 
          status does not provide a complete conceptual model for understanding
 
       health behavior.  And theories that advocate specific message features do
 
          not address fully the issue of audience involvement.  Indeed, there
are
 
         many intervening variables that are important to consider in the
 
  development of effective public health campaigns.
A Review of Health Beliefs and Behavior Models
        The questions that remain unexplained by the discussions of health
 
     behavior thus far are:  (a) Why do some people continue to engage in
 
      behavior that contradicts what orthodox medicine views as healthful,
 
      specifically, why do some women fail to seek prenatal care; and (b) Why do
 
          some individuals fail to attend to and assimilate relevant information
 
        received through mass media channels?  The primary reason public health
 
         campaigns often fail is that they are not informed by a comprehensive
model
 of health behavior (Kasch, 1983; Pettigrew & Logan, 1988).  Health
 
     professionals' efforts are not framed by theory that explains why people
 
          differently incorporate information gleaned through mass media and
 
    interpersonal channels, and how this information affects their
          health-seeking behavior.
        It is true that many health communication researchers today present
 
      overarching theories of attitudes, cognition, motivation and behavior to
 
          explain health outcomes (for example, see Anderson, 1989; Marin,
Narin,
 
         Perez-Stable, Otero-Sabogal, & Sabogal, 1990; Reardon, 1988).  These
 
      models, however, do not provide a conceptual framework to demonstrate
 
       relationships among mediating variables, including demographics, and
health
 behavior.  For example, some models such as attitude-behavior models
 
       described by Ajzen's theory of planned behavior (1985), and Fishbein and
 
          Ajzen's theory of reasoned action, are based on determinant attitudes
 
       toward particular health behaviors (for a review, see Fishbein, Ajzen &
 
         McArdle, 1980; and O'Keefe, 1990).  These theoretical models, based on
 
        persuasion and attitudinal change research, may offer an overly
simplistic
 
          approach because they fail to explain how situational factors
important to
 
          the development of health beliefs shape individuals' health behavior.
 
        Additionally, these models do not explain the actions of audiences with
low
 involvement with the health message, e.g., those who do not see themselves
 as a target for the message or generally do not think about the
 
  consequences of their behavior.
        There are some models that provide a more comprehensive view of factors
 
          that influence health behavior.  These current models do not, however,
 
        provide explanations for why some factors may be more important than
others
 when it comes to health care decision-making.  One such model, the Health
 
          Beliefs Model (HBM), has been adapted and readapted by many health
 
    researchers as a multidimensional instrument to predict health behavior
 
         (Becker, Kirscht, Haefner, & Drachman, 1979; Eisen, Zellman, &
McAlister,
 
          1985; for a review, see Janz & Becker, 1984).  The HBM focuses on four
 
        perceptual measures that affect the probability that an individual will
 
         engage in preventive health behavior:  perceived susceptibility to an
 
       illness, perceived seriousness of contracting an illness; perceived
 
     benefits of the preventive behavior; and perceived costs of the preventive
 
          behavior.  Several researchers have used the model successfully to
predict
 
          compliance with specific medical recommendations, such as Shatz's
study of
 
          diabetic patients' medical compliance patterns (1988).
        The HBM has proved helpful in predicting behavioral outcomes and it
 
      suggests some important correlates of health behavior. Often, however,
 
        studies employing the Health Beliefs Model are limited in scope because
of
 
          the assumptions inherent in the HBM.  First, the model assumes that
people
 
          make conscious decisions about their health behavior, while in real
life,
 
          people may choose to avoid the decision-making process altogether.
Second,
 most tests of this model focus on specific health behaviors as dependent
 
          measures, such as beliefs associated with avoiding fertility control
or
 
         using fertility control, rather than long-term dispositional,
attitudinal
 
          variables that underlie these behaviors.  This makes the HBM a
difficult
 
          model to use for generalizing and predicting future behavior in other
 
       health contexts.  Third, the model offers no explanations for how beliefs
 
          about susceptibility to illness, seriousness of health outcomes, and
 
      benefits and costs of desired behavior develop over time.
        For the purpose of this study, one key limitations of current health
 
       beliefs models is that they describe people's beliefs about their own
health.
 Rather, I am interested in exploring individuals' beliefs about how their
actions affect
 
            the health of another human being.  Specifically, a pregnant woman
may have different
 
          beliefs and attitudes about her own health and that of the developing
fetus.  In fact,
 
           several studies have shown that teenaged and single mothers
frequently have highly
 
       ambivalent feelings about being pregnant (Eisen et al, 1985; Tinsley &
Holtgrave, 1989).
 
            These ambivalent feelings sometimes are associated with efforts to
ignore the developing
 
            fetus or to engage in behavior that reduces that likelihood of
having a healthy baby.
The Proposed Mediated Effects Model of Pregnancy Health Behavior
        There are at least two models that can account for pregnancy behavior.
 
          Extant studies offer support for a direct effects model and a mediated
effects
 
          model. The direct effects model suggests that one's membership in a
particular
 
   socioedemographic and racial or ethnic group predicts pregnancy behavior.
The mediated
 
            effects model holds that (a) these demographic factors are
associated with certain
 
       cognitive and affective states, and that (b) it is these cognitive and
affective states
 
            that best account for health behavior.
        This thesis maintains that the mediated effects model is the more accurate
 of the two.  To test this assertion, certain factors need to be
 
  identified, factors shown to be associated both with health behavior and with
 
          demographic factors predictive of health behavior.  These factors are:
(a) health locus of
 control, and particularly fetal health locus of control (i.e., the extent to
which a
 
          woman feels personal responsibility or control over the outcome of her
pregnancy); (b)
 
           perceived social support from family and friends (i.e., the extent to
which one feels that
 significant others are available for advice, comfort, and/or physical or
material
 
       support); and (c) propositional knowledge about health, particularly
health behaviors
 
          likely to affect pregnancy outcomes.
        This mediated effects model proposed for testing is depicted in Figure 1.
 As Figure 1 shows,
 fetal health locus of control is conceptualized as the principal set of
 
          beliefs affecting pregnancy health behavior.  Fetal health locus of
control
 is in turn associated with many cognitive and affective factors.  Of
 
       these, it is expected that perceived social support and propositional
 
       health knowledge are the factors most likely to be correlated with fetal
 
          health locus of control.  Additionally, several demographic factors
such as
 race, income, and educational level are likely to be associated with these
 cognitive and affective states.
        There are many features of the proposed mediated effects model that should
 be tested.  The core relationship being proposed in the model is that of
perceived
 
           social support and propositional health knowledge with fetal health
locus of control.  If
 
            the mediated effects model is viable, then these cognitive and
affective factors should
 
            better account for variance in fetal health locus of control than
should the demographic
 
            variables of age, race, and socioeconomic status.  This study tests
this assumption.  The
 
            following sections establish the foundation for asserting that fetal
health locus of
 
         control is associated with both health behavior and with demographic
factors, and that
 
           perceived social support and propositional knowledge should be
associated with these
 
         demographic factors and fetal health locus of control.
 
HEALTH BELIEFS AND HEALTH BEHAVIOR
Health Behavior *
(Seeking Appropriate Prenatal Care)
 
Fetal Health Locus of Control
 
Characteristics of
         Propositional Health
         Social Support
                  Knowledge (Lay Theories
 
                      and Facts)
 
 
 
 
Interpersonal Network
            Interpersonal and Mass
 
                           Media Sources
 
Culture
 
 
Sociodemographic Factors
(Race, Income Level, Education)
 
 
 
 
 
 
 
 
 
*In this study, the health behavior of interest is seeking appropriate
 
        prenatal care.  "Appropriate care" is outlined by the Kessner Index:  A
 
         measure of timing and quantity of care--at least nine prenatal
check-ups
 
          with the first visit beginning in the first trimester of pregnancy.
 
Figure 1.  Mediated effects model illustrating multiple influences on
 
       health behavior. Fetal Health Locus of Control
        The concept of fetal health locus of control is adapted from Wallston and
 
          Wallston's (1978) Multidimensional Health Locus of Control (MHLC).
Locus
 
          of control is a psychological construct measuring the extent to which
 
       individuals attribute events that happen in their lives as being the
result
 of their own actions or outside forces.  The theory underlying the locus
 
          of control construct asserts that a reinforcement, reward, or
punishment of
 a particular behavior will not necessarily be instructive unless one sees
 
          a causal relationship between his or her actions and the
reinforcement,
 
         reward or punishment (Rotter, 1966).
        Wallston, Wallston, Kaplan, and Maides (1976) developed an instrument to
 
          assess the locus of control construct in a health decision-making
setting.
 The instrument includes Rotter's original two components, internal
 
     (individuals generally hold themselves responsible for events that happen
 
          in their lives) and external (individuals generally hold powerful
others,
 
          fate, luck, or chance as being responsible for life's events), but the
 
        questions were reworded to address the health context.  Because of the
 
        specialized nature of medicine, many people often seek professionals
 
      ("powerful others") for advice and consultation regarding their health.
 
          Therefore, those individuals who continue to seek help regularly from
 
       health professionals may score relatively high on the external dimension.
 
          This type of external orientation is desirable in some matters of
health.
 
          Consequently, Wallston and Wallston (1978) developed the
Multidimensional
 
          Health Locus of Control (MHLC) which included three dimensions:
Internal,
 
          Powerful Others (health care professionals) and Chance (fate or luck).
 
         Chance and Powerful Others both are measures of tendencies toward
external
 
          orientation, but scoring highly on the Chance component would better
 
      predict noncompliant behavior than would scoring highly on the Powerful
 
         Others dimension.
        Repeated testing of the MHLC scale demonstrates its efficacy in predicting
 health seeking behavior (for a review, see Coelho, 1985; Cooper & Fraboni,
 1990; Larde & Clopton, 1983; Marshall, Collins & Crooks, 1990; Stanley,
 
          Hyman & Sharp, 1984).  Those who score highly in the Internal domain
have
 
          the tendency to take responsibility for their health and to believe
that
 
          their behavior directly affects their health status.  They are more
open to
 appeals on behavior modification and health outcome.  Conversely, those
 
          who score highly on the Chance scale may not seek health care as often
as
 
          others, and may maintain unhealthful behavior because they do not
believe
 
          it can positively or negatively affect their health condition.  Those
with
 
          a strong Powerful Others orientation may rely on doctors (or family
and
 
         friends) for health advice (Amaro et al., 1987; Grady, Kegeles, Lund,
Wolk
 
          & Farber, 1983; Noland, Riggs & Hall, 1985; Wallston, Smith, King,
 
    Forsberg, Wallston & Nagy, 1983; Wallston & Wallston, 1978).  Further,
 
        studies have shown that there may be cultural differences in control
 
      orientations.  In a study by Weitzel and Waller (1990), African-Americans
 
          scored higher on the chance component of health locus of control than
did
 
          whites, particularly when education level was lower.  Although the
study
 
          employed a relatively small sample and the disparity in income between
the
 
          African Americans and whites sampled was large, the results suggest
that
 
          those with lower education levels, and African Americans in
particular, are
 less likely to seek care for illness.
        Because this study is concerned with how women perceive their behavior as
 
          affecting another human being--the developing fetus--a fetal health
locus
 
          of control scale (FHLC) was selected as an appropriate instrument for
 
       assessing the locus of control orientation of young women who one day
will
 
          be pregnant (Labs & Wurtele, 1986).  This instrument measures the same
 
        three dimensions of the MHLC, but the questions are designed to address
 
         women's perceptions of their actions in relation to their future unborn
chi
 
          ld.  For example, one item in the Internal subscale reads, "My unborn
 
       child's health can be seriously affected by my dietary intake during
 
      pregnancy"; one item in the Chance subscale reads, "No matter what I do
 
         during pregnancy the laws of nature determine whether or not my child
will
 
          be normal"; and an item in the Powerful Others subscale reads, "My
baby
 
         will be born healthy only if I do everything my doctor tells me to do
during
 
     pregnancy" (response choices range from "strongly agree" to "strongly
disagree").
        The FHLC scale has proven useful for measuring maternal health beliefs,
 
          but it has not been tested extensively.  Therefore, it will be
conceptually
 useful to examine both FHLC and MHLC in order to illuminate health beliefs
 more fully.
Factors that Predict Health Locus of Control
        The fundamental questions addressed by this study is why are
          African-American women less likely to seek prenatal care than women
from
 
          other ethnic groups, and how can mass media messages encourage healthy
 
        pregnancy behavior ?  Clearly this connection is not genetically
 
  determined.  Many researchers suggest that people develop beliefs and
 
       subsequent behavior patterns within their distinct social networks, and
 
         these networks operate within specific cultures (Jaco, 1979).
Therefore,
 
          in addition to learning about individuals' attributional control
 
  tendencies, we also need to identify culturally encouraged assumptions
 
        about health care.  Culture may be defined in many ways, as an
economically
 measured phenomenon or as a more abstractly delineated state.  For the
 
         purpose of this study, culture refers to the values, beliefs,
knowledge,
 
          and behavioral customs shared by a particular group of individuals.
Often
 
          these groups are bound by racial or ethnic ties.  A cultural
perspective of
 attributions of control offers further understanding of individuals'
 
       health behavior (Pennebaker, 1982).  Prescribing a conceptual model that
 
          examines the influence of culture (race, ethnicity) on a health
beliefs
 
         construct has important implications for health communication
researchers.
 Studies continue to show that one's health beliefs may be the most helpful
 unit of analysis for predicting health behavior.  There is evidence that
 
          in some cases health beliefs are better predictors of health service
 
      utilization and health outcomes than demographic factors (TinsleyE&
 
     Holtgrave, 1989).  For example, a report from the Census Bureau shows that
 
          Asians' and Pacific Islanders' poverty rate is higher than whites, yet
 
        their infant mortality rate is significantly lower than the white infant
 
          mortality rate (Centers for Disease Control, as reported by Eberstadt,
Wall
 
            Street Journal, January 20, 1992).  This suggests that culture and
subsequent health
 
         beliefs supersede demographic factors in terms of health outcomes.
However, there are no
 
            particular methods for assessing one's cultural background; rather,
the investigator is
 
            attempting to discover what specific elements of culture influence
the development of
 
          health beliefs and behaviors.  This thesis asserts that one's
perceived social support
 
           network, propositional health knowledge gleaned from mass media and
social networks, and s
 
            ociodemographic variables of race, and income will best reveal the
influence of culture on
 health locus of control beliefs.  There is much research that supports the
notion that
 
            these variables will explicate health beliefs in a cultural context.
        Social support.  One way culture can be transmitted is through one's social
support
 
         system.  Cultural beliefs, and subsequent health beliefs are
transmitted through one's
 
           social contacts with family and friends.  Social support may
influence health locus of
 
           control in several ways, such as
(a) extensiveness (i.e., number of people in one's network; (b) quality
 
         (the extent to which one feels socially supported; (c) type (i.e.,
 
    emotional, physical, informational, material, social participation).
 
       Individuals' feelings of social support are likely to be associated with
 
          health beliefs for several reasons.  First, individuals are often
reliant
 
          on friends or family members' opinions when making personal decisions
about
 health.  Second, individuals who feel supported usually experience high
 
          self-esteem (Rotter, 1966; ).  High self esteem is related to having
an
 
         internal locus of control.
        Third, there is increasing evidence that a positive social network has an
 
          ameliorating effect on health behavior.  For example, studies have
shown
 
          that African-American women who have strong social support systems are
more
 likely to seek prenatal care, even while facing concrete obstacles such as
 lack of insurance, no transportation to clinics, or no funds for
 
   babysitters (St. John & Winston, 1989; Lia-Hoagberg et al., 1990).
 
     Further, a study by the Centers for Disease Control (CDC, 1990) found that
 
          unmarried mothers who are college graduates have a higher infant
mortality
 
          rate than do married mothers who have less than a grade school
education.
 
          Although based on only an eight-state sample in a pilot study, this
 
     evidence clearly suggests that spousal support may be more important than
 
          education in terms of pregnancy outcome.  Thus, this study will expect
that
 the association between demographic factors and fetal health locus of
 
        control will be mediated by the nature and extensiveness of individuals'
 
          perceived social support.
        Propositional health knowledge.  A second mechanism through which culture is
transmitted
 
            is through the encouragement or discouragement of the seeking of
health knowledge and
 
          through the transmission of folklore about health. Thus, there should
be cultural
 
      differences on measures of propositional health knowledge.
        Knowledge of appropriate health behavior is related to education, race,
 
          and health beliefs, and it may follow that it is similarly related to
locus
 of control and subsequent behavior.  For example, knowledge that smoking
 
          is related to premature birth is significantly different between
blacks and
 whites, and blacks have a higher rate of low birthweight babies (Klesges,
 
          et al., 1988).
        Propositional knowledge may be an important predictor of health behavior.
 Intuitively, we can posit that if a woman is unaware that prenatal care
 
          should begin during her first trimester of pregnancy, and no one has
told
 
          her otherwise, it is unlikely that she will monitor her pregnancy
closely.
 Similarly, individuals who are unaware of the consequences of a particular
 behavior may continue to engage in that behavior.  For example,
 
  African-Americans have a higher rate of heart disease and infant mortality
 
          than do whites in the United States (National Center for Health
Statistics,
 1987; OMH, 1989).  In a study about smoking, 90 percent of whites and only
 69 percent of blacks were aware that heart disease was related to smoking,
 and 76 percent of whites and only 64 percent of blacks identified
 
    premature births as being related to smoking (Klesges, et al., 1988).
 
        These statistics suggest that propositional knowledge is another
important
 
          contributor to health beliefs and behavior.
        In a study of special education adolescents, Noland, Riggs, and Hall
 
       (1985) found a significant association between propositional health
 
     knowledge and health locus of control.  Using the Children's Health Locus
 
          of Control Scale (Parcel & Meyer, 1978), the researchers found that
 
     Internal and Powerful Others subscales best predicted adolescents' health
 
          knowledge.  Belief that health is chiefly a function of chance was
 
    negatively correlated to health knowledge, whereas high internality and the
 belief that health is chiefly a function of Powerful Others were
 
   positively related to health knowledge.  These results suggest that those
 
          who tend to take more responsibility for their health (internals) are
more
 
          likely to attend to relevant health information.  Similarly, those who
tend
 to believe others (e.g., a parent, teacher, and/or doctor) are responsible
 for their health are more likely to attend to information from
 
 professionals, because they generally do not trust their own intuitions
 
         regarding appropriate health behavior.  Conversely, those who have a
 
      tendency toward the "chance" orientation are not likely to assimilate
 
       health information. They generally do not believe that changing their
 
       behavior will make a difference in their lives.
        Socioeconomic status: Race, income, education.  There is a third way in which
the
 
       influence of culture on health locus of control might be assessed.
Perhaps gross measures
 of race and income level of teenagers' guardians, for example, may best account
for
 
         variance in health locus of control and fetal health locus of control
among adolescents.
 
            If this is the case, it will not be because race causes a certain
belief orientation, but
 
            rather because there may be many ways that culture shapes health
locus of control and
 
          fetal health locus of control and perhaps these sociodemographic
variables will best
 
         account for this variance.  For example, African-Americans as a group
may perceive they
 
            face complicated and uncontrollable societal obstacles such as
racial prejudice that may
 
            contribute to their development of particular control orientations.
        Individuals' distinctive ethnic and racial cultures influence their
 
      behavior and compel them to respond to different situations in a
particular
 way (Wood, 1979).  Because many studies have demonstrated cultural
 
     differences in how people respond to illness (Pennebaker, 1982), it is
 
        likely that culture affects health beliefs in segregated racial and
ethnic
 
          groups.  Few social scientists or health communication scholars,
however,
 
          have focused on African Americans' and other minorities' health
beliefs and
 corresponding behavioral patterns; most studies look at race and ethnicity
 in relation to outcomes, such as use of services (K. Farraro, personal
 
         communication, July 22, 1990).  But much can be gleaned from extant
studies
 on health beliefs as they exist in certain cultures.
        Studying individuals' health beliefs and specifically measures of health
 
          locus of control may provide the most useful conceptual framework for
 
       understanding health outcomes, because it reflects the situational,
 
     psychological, and physiological components of health behavior.  Several
 
          recent studies provide evidence for the notion that race and ethnicity
are
 
          important predictors of the MHLC construct:  African-Americans score
higher
 on the Chance scale and lower on the Internal scale than do whites (Amaro,
 Beckman, & Mays, 1987; Weitzel & Waller, 1990).  These studies, however,
 
          employed samples from lower socioeconomic groups.  It is difficult,
 
     therefore, to generalize these results to a broader income spectrum.
        Motivation within different racial groupings also is related to HLC.  For
 
          example, those with high Chance health locus of control may not be
 
    motivated to seek care or change unhealthful behavior except in an
 
    emergency.  In a study by Young et al. (1989), differences between
 
    African-Americans and whites were examined regarding reasons for delaying
 
          prenatal care.  In the category "Utilization of Prenatal Care," more
than
 
          twice as many blacks than whites reported motivation problems with
setting
 
          and maintaining appointments as a major obstacle to seeking prenatal
care.
 Of course, several other factors such as discomfort with clinical setting,
 perceived poor treatment by physicians, lack of transportation, no funding
 for insurance or babysitting, no family support system, could affect this
 
          score, as well.
Summary of Culture and Health Beliefs
        There is little doubt that culture is associated with the development of
 
          health beliefs, such as health locus of control.  But as this review
of
 
         culture and health behavior has shown, it is difficult to assess a
 
    culture's influence on health beliefs.  This study posits that the
 
    influence of culture on health locus of control will be observed in
 
     associations of perceived social support, propositional health knowledge,
 
          and demographic variables with health locus of control measures.
Research Questions and Hypothesis to be Tested
        A review of the literature on health locus of control and its relationship
 with the sociodemographic factors of race and income level, and with the
 
          cognitive and affective factors of propositional health knowledge and
 
       social support, demonstrates the efficacy of further study of the
mediated
 
          effects model.  Therefore, this study sought to answer the following
 
      research questions:
        RQ 1.  To what extent is race/ethnicity related to high school aged
 
      women's feelings of control over the outcome of pregnancy?
        RQ 2.  To what sources do female teens turn for information about health?
The study also tested one hypothesis:
        H1.  Cognitive and affective factors of health knowledge and perceived
 
         social         support will be associated with high school aged females'
feelings
 
          of control over their own health or that of an unborn fetus, and these
 
        associations will remain strong after controlling for the group
variables
 
          of race and income.
Method
        The study was designed to investigate infant mortality by assessing
 
      sociodemographic, cognitive and affective variables that may predict young
 
          women's locus of control orientations guiding their behavior during
 
     pregnancy.
Study Design and Population
        A cross-sectional survey design was employed.  The sample consisted of
 
         female students from four inner-city public high schools in
Indianapolis,
 
          located in Marion County, Indiana.  At the time of the study, Marion
County
 had one of the highest black infant mortality rates in the country (OMH
 
          1989; ISBH Report, 1989, 1993).  This sample was selected for two
reasons:
 
          (a) all four schools contain a relatively large minority population;
and
 
          (b) students from these schools represent a broad socioeconomic range
from
 
          poverty level to upper middle class.  On the assumption that factors
other
 
          than sociodemographics will be linked to African-American and white
women's
 health beliefs important during pregnancy, it was necessary to include a
 
          balance of both African-American and white subjects from different
 
    socioeconomic classes.
        To achieve a representative sample with a broad range of academic
 
    abilities, eligible subjects were those women enrolled in required courses
 
          (such as history, health, or physical education) within each high
school in
 grades nine through twelve.   Consent from both student participants and
 
          their parents was obtained prior to conducting the survey.
Survey Response Rate and Frequency Distributions for Race and Income Groups
        A total of 215 parental consent forms were distributed to female students
 
          at the four high schools in the summer and fall of 1992.  Of these,
102
 
         high school women between the ages 14 and 19 returned signed parental
 
       consent forms and participated in the survey study, for a response rate
of
 
          47 percent.  The average age for the sample was 16.  Four of the
 
  respondents were not included in the analyses because they had failed  to
 
          complete questions pertaining to the dependent measures plus one or
more
 
          independent measures (N = 98).  [1]
        Participants included 47 whites and 51 African-Americans.  Although this
 
          ratio does not reflect true population parameters, it was necessary in
this
 study to over-sample African-Americans to examine factors that affect
 
        black white-differences in health beliefs.
        The variable "income level" was coded in the following way:  0 refers to
 
          those families with incomes that fall below the government guidelines
for
 
          determining poverty based on annual income and number of family member
 
        dependents (United States Department of Commerce, 1990), ranging from
 
        $5,000 to $19,000 annually (poor); 1 refers to those families whose
incomes
 range from approximately $10,000 to $22,000 annually (low income); 2
 
       refers to those families whose incomes range from approximately $18,000
to
 
          $45,000 (middle class); 3 refers to those families whose income level
range
 from $45,000 to greater than $70,000 annually (upper-middle class). These
 
          income categories were based on the poverty guidelines, with each
 
   subsequent level spanning an income range of one-third above the previous
 
          level, based on the number of family members.  For example, a family
of
 
         three with one child under 18 can make up to $10,520 annually to be
 
     considered below the poverty level, or Level 0 in this study.  For Level
 
          One, the same size family could make from $10,521 to $13,992.
Similarly, a
 
          family of eight (six children under age 18) can make up to $21,505
annually
 to remain under the poverty level guidelines (Level 0).  Family income
 
         level reflected a normal distribution for both blacks and whites--no
 
      significant differences existed between races on income level in this
 
       sample (mean scores for income level are 1.34 for blacks, sd 1.17; and
1.13
 
           for whites, sd .1.06). Academic achievement (self-reported grade
point average based on a
 
            4.0 scale) also shows no significant difference between GPA of
whites and
            African-Americans (mean score for African-Americans was 2.40, sd
.67; mean score for
 
         whites was 2.42, sd .65).
General Procedures and Materials
        The set of surveys administered included five instruments designed to test
 the relationships of demographic, cognitive, and affective variables with
 
          subjects' enduring health beliefs.  Specifically, three instruments
were
 
          selected or developed to obtain independent measures respectively of
(a) subjects' sociodemographic groupings (including mass media channel
 
        selection for health informtion); (b) subjects' propositional knowledge
 
         about general health topics including healthful behavior and pregnancy
 
        health behavior; and (c) subjects' perceived social support.
Additionally,
 
          two instruments were combined to assess the dependent measures of
subjects'
 health beliefs:  (a) multidimensional health locus of control (MHLC) and
 
          (b) fetal health locus of control (FHLC).  These two scales were
integrated
 to avoid drawing students' and school administrators' attention to the
 
         project's focus on pregnancy health beliefs.
Independent Measures
        Sociodemographic characteristics survey.  This instrument was developed to
assess
 
       respondents on a number of different characteristics, including health
information sources
 used, race, academic achievement, and family income level. To determine health
 
    information sources used, respondents were asked to select from a list and
rank order all
 
            the sources (newspapers, television, radio, magazines, textbooks,
mal friend, female
 
         friend, parent, school teacher or counselor, religious leader, health
care professional,
 
            and "other") they turn to in order (a) to seek information about
health issues, or (b) to
 
            obtain health information for a sick friend.
        Race of respondent was determined by asking them to identify the race
 
        category for each parent.  For the purpose of this study, a listing of
 
        "African-American" for one parent (even if the other parent was of a
 
      different race) placed the respondent in the "black" racial category.
This
 was done for 2 of 98 subjects.
        Academic achievement was assessed by asking respondents to list the
 
      overall grade point average (4.0 scale) they had accrued in high school up
 
          to the time of the survey.  Although considered as an individual
difference
 variable in study analyses, the academic achievement assessment was
 
      included in this instrument because it fit appropriately with other
 
     personal information requested in the survey.
        To appraise the family income level of respondents, respondents were asked
 to describe the jobs held by each of their parents or guardians, indicate
 
          whether each job was full or part time, and repeat the same
information if
 
          second jobs were held by parents or guardians.  Next, respondents were
 
        asked to estimate their family income level.
        For the purpose of the study, the researcher checked the respondents' job
 
          descriptions with the job earnings listings for the year 1988 in the
United
 
            States Bureau of Labor and Statistics Report for Median Incomes of
Wage Earners (1989),
 
            and compared the median income listed in the report with those
selected by respondents,
 
            taking into account all jobs described by respondents.  Because the
complete reports are
 
            only published for five-year periods, the report for the years 1983
to 1988 was the most
 
            recent comprehensive summary available.  Eighty-six percent of the
job descriptions and
 
            annual family income levels reported by respondents matched within
ten percent above or
 
            below what the labor statistics report listed as median income
levels for the jobs
 
       respondents described.  The ten percent range on the top and bottom was
allowed as a match
 for respondents' answers because the government reports median incomes rather
than a
 
          range of incomes for each job.  Therefore, it is difficult to
calculate exact
 
  correspondence between respondents and the labor statistics report.  Given
that no
 
       detailed information was available from respondents, such as wage
earners' length of time
 
            employed at each job, or whether or not workers were union members,
the ten percent
 
        latitude for agreement was deemed reasonable.  Two coders were used to
determine family
 
            income levels of the thirteen respondents whose income estimates for
the jobs they listed
 
            did not fall within ten percent above or below what the labor
statistics report listed for
 the corresponding jobs.  The purpose of the coders was to determine if the
initial coder
 
            had erred in interpreting the job descriptions provided by
respondents, to account for the
 non-agreement between respondents' answers and the labor statistics report.
Inter-item
 
            coder reliability for these cases was 93 percent exact agreement.
In these thirteen cases
 where there was no income level agreement between the respondents and the labor
 
     statistics report, respondents' answers were privileged.  That is,
subjects' responses
 
           were deemed correct.
        Propositional health knowledge.  Propositional health knowledge refers to
individuals'
 
            knowledge about general health issues such as hygiene and other
preventive care topics
 
           that they should know given their age and educational level.  It is
an important concept;
 
            people's health knowledge has shown to guide behavior.
        To measure health knowledge, a portion of a 50-item multiple choice survey
 designed to reflect the health topical areas typically studied at the high
 school level was used (Noland, Riggs, & Hall, 1985).  Questions cover such
 topics as how venereal diseases are contracted, which foods contain the
 
          greatest amount of protein, and what diseases are related to cigarette
 
        smoking.  The reading level of the scale is directed at the fourth to
fifth
 grade, and the scale was tested on high school aged special education
 
        students.  Therefore, the instrument was appropriate for this study
sample.
 
        For the purpose of this study, the general nutrition and health questions
 
          allowed the researcher to study health knowledge important during
 
   pregnancy.  Additionally, three questions about a woman's health behavior
 
          during pregnancy were added to assess knowledge of appropriate
pregnancy
 
          health behavior.  Although the new questions had not been tested
 
  previously, the questions reflect what health experts list as crucial
 
       knowledge about behavior for ensuring a healthy newborn baby. While the
 
         inter-item reliability for this scale was deemed low but acceptable
 
     (Cronbach's alpha = .56), it was not expected that a scale assessing
 
      knowledge of several different health topic areas would generate high
 
       reliabilities.
        Perceived social support network inventory (PSNI).  The social support measure
used was
 
            the Perceived Support Network Inventory developed by Oritt, Paul,
and Behrman (1985). The
 
            instrument was selected because it provides a multidimensional
assessment of individuals'
 
            perceptions of the social support they receive.  Oritt et al.'s
measure of perceived
 
         social support allows the researcher to look at preconceptions and
attitudes people have
 
            about individuals within their social network that shape the types
of interactions that
 
            will occur during stressful times.  This comprehensive scale
assessed respondents'
 
       perceived social support network on several different levels, including
the extensiveness,
 nature, and quality of a person's social support network.
        The validity and reliability of this instrument have been well-tested:
 
          test-retest reliability total scores and subscale scores range from
.72 to
 
          .88; internal consistency for the PSNI is .77 (using Cronbach's
alpha);
 
         convergent validity ranges from -.25 to .20; and discriminant validity
 
        estimates ranged from -.11 to .19.  (Note:  In general, convergent
validity
 should show high correlations, but for the purpose of this test strong
 
         inverse associations are desired.)  Care was taken to maintain the
clarity
 
          and meaning of each subscale and instructions after editing the
 
 instrument's wording for a high school population.
Dependent Variable Measures
        Multidimensional health locus of control (MHLC).  The MHLC (Wallston and
Wallston, 1978)
 
            was administered as an additional measure of health locus of
control.  Individuals who
 
           score highly on the internal subscale have the tendency to feel
personal responsibility
 
            for events that happen in their lives.  Those who score highly on
the external subscales
 
            have the tendency to believe that other forces such as luck, chance,
or powerful others
 
            are ultimately responsible for events that happen to them in their
lives.  Health status
 
            positively correlates with internality.
        An example of an Internal subscale question is "If I take good care of
 
         myself, I can keep from getting sick";  a Chance subscale question is
"My
 
          good health is largely a matter of good fortune"; and a Powerful
Others
 
         subscale question is "When it comes to my health, I can only do what my
 
         doctor tells me to do."  (Cronbach's alpha for this sample was .58 on
the
 
          internality subscale, .67 on the chance-externality subscale, and .63
on
 
          the powerful others-externality subscale. )
        Fetal health locus of control (FHLC).  The primary dependent measure used in
this study
 
            was the fetal health locus of control scale (Labs & Wurtele, 1986).
The 18-item
 
     instrument is designed to assess the extent to which women perceive that
their actions
 
           during pregnancy will affect the health of their future unborn baby.
FHLC is an important
 construct to study, because few locus of control studies have examined control
 
    orientation in terms of how one believes his or her actions affect another
human being.
 
            Some researchers have recognized the importance of looking at
women's health beliefs
 
         because women typically are responsible for managing the care of their
pregnancies, and
 
            for managing the care of their children. The Labs and Wurtele (1986)
scale sought to
 
         measure the control orientation of women who are pregnant or of
child-bearing age.  The
 
            researchers confirmed a relationship between a woman's tendency
toward internal
 
    orientation and the likelihood that she would seek prenatal care, an
important
 
   relationship for this study, as well.
        Like Wallston & Wallston's (1978) MHLC instrument, it consists of three
 
          subscales: internal, chance, and powerful others.  Scoring highly on
the
 
          internal subscale expresses an individual's general belief that she is
 
        personally responsible for having a healthy baby. An example of an
internal
 subscale question is "My unborn child's health can be seriously affected
 
          by my dietary intake during pregnancy."  Scoring highly on the chance
 
       subscale expresses an individual's general belief that there is little
she
 
          can do to change the outcome of her pregnancy; fate, luck or chance
plays a
 major factor.  An example of a chance subscale question is "Having a
 
       miscarriage means to me that my baby was not destined to live."  Scoring
 
          highly on the powerful others subscale expresses an individual's
general
 
          belief that the health of her unborn child lies in the hands of health
 
        professionals who manage her care. An example is "Only qualified health
 
         professionals can tell me what I should and should not do when I am
 
     pregnant."  As with the Labs and Wurtele study, subjects were instructed to
 answer the questions as if they were planning a pregnancy in the near
 
        future.
        For this sample, Cronbach's alpha was .63 on the internality subscale; .85
 on the chance (externality) subscale; and .71 on the powerful others
 
       (externality) subscale.  Because the fetal health locus of control
 
    instrument had not been tested extensively, the investigator calculated
 
         correlation coefficients for the two dependent measures,
multidimensional
 
          health locus of control (MHLC) and fetal health locus of control
(FHLC),
 
          and for each of their subscales.  The total scale score (means) of the
two
 
          scales were significantly correlated (r = .56, p < .01), and
significant yet not
 
            always strong associations existed between each of the subscale
scores:  internality
 
         subscale for FHLC and MHLC (r = .27, p < .01); powerful others
dimension of externality
 
            subscale for FHLC and MHLC (r = .51, p < .01); and fate, luck or
chance dimension of
 
         externality subscale for FHLC and MHLC (r. = .48, p. < .01).  A high
correlation between
 
            the scales was not expected; the MHLC assesses one's feelings of
control over one's own
 
            health, while the FHLC assesses a woman's feeling of control over
someone else's
 
     health--the outcome of a pregnancy.
 Procedures
        The five surveys combined took respondents an average of 38 minutes to
 
         complete.  The survey was administered during regular class time (a
 
     45-minute time limit) to allow as many students as possible to participate.
  Students who participated in the study were assured of anonymity and
 
        confidentiality.  Response formats for the administered scales varied,
 
        ranging from a five-point Likert-type scale, to a multiple-choice
selection
 that reflects correct and incorrect answers, to open-ended questions
 
       designed to assess individual or family characteristics.  (See Table 1for
a
 list of variables used in the study.)
Results
        The first research question was answered through a series of correlational
 analyses between race and different dimensions of multidimensional health
 
          locus of control and fetal health locus of control.
        Race and income level and multidimensional health locus of control.  As
expected,
 
       results did not support the sociodemographic model of health behavior
disputed in this
 
           thesis.  No significant relationships were found between race and the
total mean score of
 
            the multidimensional health locus of control scale (r = .06, p =
.53, ns) or income level
 
            and the total mean score of the multidimensional health locus of
control scale (r = -.05,
 
            p = .63, ns).  Similarly no significant relationships exist between
income level or race
 
            and the subscales of the multidimensional health locus of control
instrument.
        Race and income level with fetal health locus of control.  Again, results did
not
 
       support the sociodemographic model challenged by the investigator.  No
significant
 
       relationships were found between the two group variables and the total
mean scores of the
 
            fetal health locus of control scale.  The correlations were:  race
and fetal health locus
 
            of control (r = -.02, p = .82, ns); and income level and fetal
health locus of control (r
 
            = -.15, p = .13, ns).
        Only one significant association was found between a group variable and a
 
          dimension of fetal health locus of control.  A significant but modest
 
       relationship exists between income level and fetal health locus of
 
    control-powerful others (r = -.24, p < .05), indicating that the lower the
income
 
           level, the stronger the orientation toward powerful others
(externality).  No significant
 
            relationships exist, however, between the group variable income
level and the chance
 
         subscale of fetal health locus of control (r = .00, p = .98, ns),
between  race and fetal
 
            health locus of control-chance (r = -.09, p = .37, ns), between race
and fetal health
 
          locus of control-powerful others (r = -.04, p = .70, ns), or race and
income level and
 
           fetal health locus of control-internality (r = .16; r = -.02,
respectively; p > .05 for
 
            both correlations, ns).
        Female teens' health information sources.  The study also answers the second
research
 
           question regarding teen-aged females' key health information sources.
Respondents each
 
            identified from one to six health sources they typically turn to for
information on health
 issues.  A majority (62 percent) reported they would turn to a parent for
health
 
      information, 40 percent reported television, and 30 percent reported
newspapers or
 
       magazines as one of their key health sources.  A surprising 31 percent
turn to male
 
        friends as major sources of health information.
        When asked what sources they would seek out to get pertinent health
 
      information for a sick friend, a majority of respondents (73.5 percent)
 
         reported they would turn to doctors or nurses.  Sixty-five percent also
 
         would turn to a parent, 37 percent to female friends, 22 percent to a
 
       teacher or school counselor, and 21 percent would turn to male friends as
 
          additional sources of health advice.  Mass media are less important
sources
 when specific and urgent health information is needed:  In this
 
  hypothetical circumstance, only eight percent reported newspapers and
 
       magazines, and 16 percent reported television as sources for health
advice.
Test of the Mediated Effects Model
        Statistical analyses support this study's hypothesis (the
          cognitive/affective model of health behavior):  high school aged
females'
 
          propositional health knowledge and perceptions of their social support
 
        networks are significantly associated with feelings of control over
 
     personal health and with feelings of control over the outcome of
 
   pregnancy.  Most pertinent to this study, propositional health knowledge
 
          and perceived social support maintain strong associations with control
 
        orientations for health and pregnancy outcomes even after controlling
for
 
          the sociodemographic variables of income level and race.  This finding
 
        suggests that those who feel more knowledgeable about health or more
 
      socially supported are more likely to believe that they themselves are r
 
         esponsible for their own health and that of their future babies than
are
 
          their counterparts.  Further, these relationships remain stable after
 
       controlling for the effects of race and income level.
        General health knowledge and multidimensional health locus of control.  As
hypothesized,
 
            the results indicate a strong association (see Table 2) between
general health knowledge
 
            and the overall scale mean scores of multidimensional health locus
of control (r = -.48, p
 < .001).  Similar findings occur with health knowledge and the two subscales of
 
     externality:  For chance, r = .41,
p < .001; and for powerful others, r = -.33, p < .001.  This means that general
health
 
           knowledge is associated with female teens' feelings that powerful
others such as family
 
            members or doctors control their health, and it also is associated
with their feelings
 
           that either fate, luck, or God controls their health.
        Further, this relationship is not affected when the variance accounted for
 by race and income is controlled (see Table 3).  The first-order partial
 
          correlation between health knowledge and total mean score of the
 
  multidimensional health locus of control, controlling for race, is -.47
 
         (the zero-order correlation is -.48).  In addition, controlling for the
 
         variance associated with income also does not affect the strong
association
 between health knowledge and locus of control.  Controlling for income
 
         level, the first order partial correlation of the two variables is
-.48.
 
          Whereas the zero-order correlation for health knowledge and
          multidimensional health locus of control-chance is -.41, the
first-order
 
          partial correlations for the two variables -.41, controlling for race;
and
 
          -.41, controlling for income level.  For the powerful others subscale,
the
 
          zero-order correlation is -.33, and the partial correlations are -.33,
 
        controlling for race; and -.33, controlling for income.
        General health knowledge and fetal health locus of control.  A similar
relationship
 
         exists (see Table 4) between general health knowledge and fetal health
locus of control
 
            total mean scores (r =
- .39, p < .001).  These negative correlations show that individuals who score
lower on
 
            the general health knowledge measure score higher on the externality
dimension of control
 
            (they believe that their own health status or their unborn babies'
health is the result of
 powerful others in their lives and/or fate, luck, or chance).  Subsequently,
those who
 
            have higher scores on the health knowledge assessment score higher
on internality,
 
       represented by a lower score on the internality subscale (r =
.20, p < .05).  Analogous associations were found between health knowledge and
the
 
        subscales of powerful others and chance, both measures of externality (r
= -.33, p < .001;
 r = -.24, p < .01, respectively).   These significant associations are not
affected after
 controlling for the effects of race and income (see Tables 5 and 6).  Thus, the
partial
 
            correlations reveal little variance among the hypothesized
associations after controlling
 
            for the effects of race.
        Although the direction of these relationships cannot be assessed with this
 study design, one can posit that knowing what constitutes appropriate health
 
        behavior is efficacious to feeling in control of personal health
matters.
        Perceived social support and multidimensional health locus of control.  There
were no
 
           significant associations between perceived social support and the
total mean score of the
 
            multidimensional health locus of control scale, or between perceived
social support and
 
            the three subscales of multidimensional locus of control (see Table
2).
        Perceived social support and fetal health locus of control.  One significant
association
 
            exists between the affective variable, perceived social support, and
fetal health locus of
 control measures (see Table 4).  The strong association between social support
and the
 
            internality subscale (r = -.40,
p < .001) means that the higher the score on the perceived social support
 
          scale, the lower the score on the internality subscale.  A lower score
on
 
          the internality subscale, however, indicates a more internal
orientation.
 Further, after controlling for race and income, the association between
 
          perceived social support and fetal health locus of control remains
strong
 
          (see Tables 5 and 6).  The first-order partial correlations are:
 
  controlling for race, r = -.38, controlling for income level, r = -.40.
Discussion
        The study shows that perceptions of social support and general health
 
        knowledge are significantly associated with fetal health locus of
control.
 The study also demonstrates that cognitive and affective variables are
 
         more likely to be associated with feelings of personal efficacy toward
a
 
          fetus than are the more commonly used demographic predictors such as
race
 
          and income.  Therefore, the investigation offers support for the
 
  hypothesized indirect effects model of health beliefs.        Specifically,
 
       analyses support the indirect effects model in several ways: (a) no
 
     significant relationships were found among the variables of race and fetal
 
          health locus of control and only one significant relationship was
found
 
         among the variables of income and health locus of control;
(b)  strong associations were found between the cognitive and affective
 
         variables and the dependent measures of fetal health locus of control
and
 
          multidimensional health locus of control; and
(c)  systematic variations  exist in the strength of the associations among
 the cognitive and affective predictors of fetal health locus of control.
 
          Of these three categories of findings, the most powerful evidence for
the
 
          model is found in the reports of the systematic variations in the
 
   relationships among the independent and dependent measures.  The discussion
 that follows will elaborate on each of these findings that support the
 
         indirect effects model and describe the implications for application.
 
          Support for the Direct Effects Model
        Race and income level.  The fact that no significant association was found to
exist
 
         between race and control orientation is interesting in itself.  This
finding differs from
 
            those of Amaro, Beckman, and Mays (1987) and Weitzel and Waller
(1990) that linked race
 
            and control orientations.  In these studies, African-Americans
scored higher on the chance
 scale and lower on the internal scale than did whites.  Both studies, however,
employed
 
            samples from lower socioeconomic groups, and therefore may have
confounded race with low
 
            income.  In the current study, the sample was balanced on
sociodemographic variables of
 
            race, income, and academic achievement (self-reported grade point
average), in an attempt
 
            to avoid such confounding.  Further, because only one significant
association was found to
 exist between the group factors and fetal health locus of control, it appears
that race
 
            and income have no direct effects on female teens' control
orientations in this sample.
        General health knowledge.  Statistical analyses offer support for the cognitive
link in
 
            the indirect effects model of health behavior.  These findings
suggest that general health
 knowledge and feelings of control over one's own health maintain strong
associations,
 
           even after controlling for the effects of the sociodemographic
variables of race and
 
         income level.  In summary, health knowledge is important in two ways.
First, this study
 
            has shown that general health knowledge is significantly associated
with measures of
 
         health locus of control and fetal health locus of control, even after
controlling for the
 
            effects of race and income level.  Second, possessing accurate
health knowledge may
 
        potentially lead teens to more healthful behavior.
        Social support.  The fact that the study found social support to be
significantly
 
       associated with fetal health locus of control and not with
multidimensional locus of
 
         control suggests that perceived social support may not affect
individuals' control
 
       orientations about their own health, but it may have an effect on how
women feel their
 
           actions affect the health of other beings, their future children.
This finding suggests
 
            that the more socially supported people feel, the more personally
efficacious they feel
 
            about pregnancy outcomes.
        Systematic variations:  General health knowledge, perceived social support, and
measures
 
            of fetal health locus of control as a function of race and income
level.  By looking at
 
            systematic variations in these associations the roles of race and
income can be examined
 
            further.  (See tables 7 and 8 for the listings of systematic
variations by race and
 
        income.)  Interestingly, the strength of the associations among health
knowledge, social
 
            support, and fetal health locus of control vary systematically and
significantly as a
 
          function of race and income.  This is the case both for the
association between health
 
           knowledge and fetal health locus of control and between perceived
social support and fetal
 health locus of control.  By dividing the sample into units by race and income,
the
 
         investigator can explore further the strength of the hypothesized
associations in relation
 to race and income.  Further analyses indicate that the relationships between
health
 
          knowledge and fetal health locus of control and between perceived
social support and fetal
 health locus of control do vary systematically as a function of race and
income, lending
 
            support to the idea that race and income have indirect rather than
direct effects on
 
         health beliefs.
        Specifically, the associations between health knowledge and fetal health
 
          locus of control differ for African-Americans and whites.  For the
entire
 
          sample, general health knowledge is significantly associated with
fetal
 
         health locus of control on each dimension:  With the total scale, r =
-.39,
 
            p < .001; with internality, r = -.20, p < .05; with powerful others,
r = -.33, p < .001;
 
            with chance, r = -.24, p < .01 (see Table 4).  The strength of the
associations differs,
 
            however, for whites and African-Americans (see Table 7).  For
example, if the sample is
 
            divided on race alone, the association between health knowledge and
fetal health locus of
 
            control-internality is not significant for either group.  Rather,
the strongest
 
    associations exist between health knowledge and the total scale score for
low income
 
         whites (those whites who fall in the lower two levels of income within
the sample [r =
 
           -.53, p < .01]) and for high income blacks (those African-Americans
who fall into the
 
          upper two income levels within the sample [r = -.50, p < .01]).
Interestingly, the
 
        association is not significant for high income whites (r = -.24, p =
.29), or for low
 
          income blacks (r = -.18, p = .41).  Other systematic variations in
associations between
 
            health knowledge and fetal health locus of control occur on the
chance subscale for blacks
 and whites.  It was significant for low income whites (r = -.47, p < .01) but
not for
 
           high income whites (r = -.23, p = .30, ns); significant for high
income blacks (r = -.45,
 
            p < .05) but not for low income blacks (r = .25, p =.24, ns).  One
reason for these
 
        disparate results, such as the fact that health knowledge and fetal
health locus of
 
        control are significantly associated for low income whites but not for
low income
 
      African-Americans, may be that for low income whites, greater health
knowledge is indeed
 
            empowering.  But for low income blacks, there may be many other
obstacles (racial
 
      prejudice, less support from the white majority, etc.) that block the
effects of health
 
            knowledge on beliefs of personal efficacy.  Additionally, poverty
may have a more powerful
 impact on African-Americans, since they as a minority in the United States
comprise the
 
            majority of low income families (OMH, 1989).  There may be other
more significant factors
 
            related to feelings of personal efficacy among poor
African-Americans, such as perceived
 
            social support.
        For the sample as a whole the association between perceived social support
 and fetal health locus of control internality is significant (r = -.40, p <
 
            .001).  The association is greater, however, for African-Americans
than for whites (see
 
            Table 8).  Examining the association by race shows that perceived
social support is an
 
           important factor in health beliefs for African-Americans in both low
and high income
 
         groups.  This is not the case with the white female teens; the
association was significant
 only for whites as a group and for low income whites.  Specifically, for
            African-Americans only, the correlation coefficient is -.45 (p <
.01); for whites only,
 
            the correlation coefficient is
.29 (p < .05).  This is an important finding because it suggests that a
perceived strong
 
            social support network is a key correlate of internality for
African-Americans in
 
      particular.  Further, the two factors are strongly associated both for
African-Americans
 
            and for whites in the lower two income levels:  For
African-Americans, r = -.44 (p < .05),
 and for whites, r = -.43 (p < .05).  In the upper two income levels, however,
the
 
       associations are dissimilar:  For African-American females in the upper
two income levels,
 the correlation coefficient for perceived social support and fetal health locus
of
 
        control-internality is -.43 (p < .05, sig.), but for white females it is
only -.15 (p =
 
            .51, ns).  (Note that a low score on a scale indicates a tendency
toward internality.)
        Several significant associations exist between social support and fetal
 
          health locus of control measures for African-Americans (see Table 8).
 
        Among all African-Americans sampled, the significant associations are:
 
         social support and fetal health locus of control total scale, r = -.36,
p <
 
            .01; and social support and fetal health locus of control
internality, r = -.45, p < .01.
 Among African-Americans in the upper two income levels sampled, the significant
 
     associations are:  Social support and fetal health locus of control total,
r = -.40, p <
 
            .05; and social support and fetal health locus of control
internality, r = -.43, p < .05.
 Finally, among African-Americans in the lower two income levels sampled, there
is one
 
           significant association:  social support and fetal health locus of
control internality, r
 
            = -.44, p < .05.  The fact that only two significant correlations
exist between social
 
           support and fetal health locus of control measures for the white
population sampled (on
 
            the internality subscale for all whites sampled and for whites in
the lower two income lev
 
            els), suggests that social support is a  more significant factor
contributing to fetal
 
           health locus of control orientations among African-American teenaged
girls than for white
 
            teenaged girls.  (Again, note that a low score on a scale indicates
a tendency toward
 
          internality.)
        This pattern of findings strengthens the claim that factors such as
 
      perceived social support better account for variance in locus of control
 
          orientations among African-Americans and whites than do
sociodemographic
 
          variables such as income level.  It also supports the claim that the
 
      effects of race and income are indirect rather than direct.  One reason
 
         this phenomenon may occur is because there are few approbations
inherent in
 American society that contribute positively to African-Americans'
 
    perceptions of social support.  Even the most prosperous African-Americans
 
          face multiple barriers such as racism, negative stereotyping, and
distrust,
 that may subvert feelings of being well supported.
        This pattern of findings strengthens the claim that factors such as
 
      perceived social support and propositional health knowledge better account
 
          for variance in locus of control orientations among African-Americans
and
 
          whites than do sociodemographic variables such as income level.  It
also
 
          supports the claim that the effects of race and income are indirect
rather
 
          than direct.  One reason this phenomenon may occur is because there
are few
 approbations inherent in American society that contribute positively to
 
          African-Americans' perceptions of social support.  Even the most
prosperous
 African-Americans face multiple barriers such as racism, negative
 
    stereotyping, and distrust, that may subvert feelings of being well
 
     supported.
        Clearly, African-Americans' perception of the strength of their social
 
         support system is a key factor in their lives.  Teenaged black women
who
 
          endure real societal barriers and retain the perception that they are
not
 
          well-supported by family, friends, and others in their lives
apparently
 
         also believe that their actions during pregnancy will not make much
 
     difference on the outcome of pregnancy.  Those who believe they can count
 
          on individuals for emotional and material support are more likely to
feel
 
          internally responsible for their unborn children.
Implications of the Mediated Effects Model on Public Health Campaigns
        This study illuminates our understanding of why some women do not and some
 women do seek prenatal care, regardless of socioeconomic status.  A focus
 
          on individual differences also has important implications for public
health
 campaigns.  First, the model offers new directions for identifying
 
     relevant target publics.  Second, the model provides insight for message
 
          content considerations.  Lastly, study results illustrate the
importance of
 advocacy media in health communication,
        Support group members as target audiences for pregnancy behavior campaigns.  It
would be
 
            interesting to examine the types of pregnancy health messages that
would be generated from
 members of the social support network, such as male peers (who female subjects
turned to
 
            most for support), parents, counselors, and teachers, if they were
consulted by female
 
           teens, and to learn about what it takes to bolster feelings of social
support.  For
 
        example, messages similar to the "Friends don't let friends drive drunk"
campaign may be
 
            beneficial for encouraging support of pregnant young women.
Targeting non-pregnant
 
        members and potential members of a pregnant woman's social network may
indirectly affect
 
            health behavior.  Although "Good friends help a pregnant friend
through pregnancy" may
 
           seem overly simplistic, it  is a relevant and appropriate approach
given the study
 
       findings.
        Additionally, males serve an important role in high school aged females'
 
          lives.  High school women frequently turn to males both for social
support
 
          and for health information: Thirty-seven percent of respondents, when
asked
 to identify up to seven key members of their social support network,
 
       listed a male first.  The findings suggest that public health campaigns
 
         directed at females only may be missing an influential audience.
Because
 
          the study found that being informed about health is connected to
having an
 
          internal control orientation, informing young males about what factors
 
        contribute to having healthy babies may benefit teenaged women.
Therefore,
 messages should be designed to encourage males--African-Americans in
 
       particular--to be supportive and helpful to pregnant friends.
          Additionally, messages also should provide accurate and pertinent
 
   information to males about appropriate pregnancy behavior for women.
        The role of mass media and health.  Respondents frequently attend to health
information
 
            they glean from television and magazines.  Because the study results
indicate a strong
 
           association between general health knowledge and control orientations
and social support
 
            and control orientations, mass media should be used to encourage
supportive behavior by
 
            illustrating, describing, and informing audiences of desired
behavior, in direct and
 
         indirect ways.  Mass media messages should highlight the actions of
supporters as well as
 
            the healthful conduct of pregnant women in various ways to
illustrate a lifestyle of
 
         healthy and supportive behavior.  While current trends of increasing
access to services is
 important to continue, the study demonstrates that access is not the principal
obstacle
 
            to health behavior.
        For example, magazines with large female teen audiences currently do
 
       little to encourage healthy pregnancy behavior (Fellure, 1993).  In a
 
       preliminary content analysis, Fellure found that top magazines read by
 
        teenagers offered little specific information about pregnancy health
 
      behavior.  There was an abundance of information about how to avoid
 
     becoming pregnant, but none pertinent to health behavior during pregnancy.
 Further, although magazines targeted to African-Americans contained
 
      articles about black families, the father's role in particular, the
 
     articles focused primarily on celebrating and encouraging family life.
 
         While a spotlight on parental involvement certainly is a valid approach
to
 
          supporting family values, it is surprising that the African-American
and
 
          mainstream magazines reviewed gave few details about specific behavior
that
 could lead to having healthy babies (e.g., when doctor visits should
 
       begin, how much weight gain to expect, how nutritional habits affect the
 
          fetus).  This may be true, of course, because editors do not want to
appear
 as though they are encouraging pregnancy among teens.  Nevertheless,
 
       editors could frame pregnancy information so that they are encouraging
 
         future responsible pregnancy behavior.
        Additionally, editors of the teen magazines reviewed by Fellure may have
 
          neglected or failed to provide more specific pregnancy health
information
 
          because the community as a whole wishes to ignore the problem of teen
 
       pregnancy among white and black teens.  Therefore, while many popular
 
       magazines seem to do a good job of modeling socially supportive behavior,
 
          clearly there is a need for gatekeepers to provide crucial health
 
   information regularly to the audiences they serve.
Directions for Future Research
        This study has interesting implications.  First, because female teens seek
 health information and support from a variety of sources, including males,
 it would be helpful to explore males' health knowledge about pregnancy.
 
          Their advice and information may influence women's decisions about
 
    pregnancy.  For example, are women worried about weight gain and
 
  subsequently diet while pregnant because of males' perceptions of how much
 
          weight a woman should gain during pregnancy?  And are males from
certain
 
          ethnic groups more inclined than others to be concerned about too much
 
        weight gain during pregnancy?  Additionally, because respondents often
seek
 health information from television and other mass media sources, content
 
          analyses, such as the study described earlier (Fellure, 1993; see also
 
        Landis, Freimuth, & Cameron, 1993), should be conducted to determine the
 
          types of  pregnancy health messages that both males and females
receive
 
         from mass media channels.
        The relationship of social support and health behavior should be explored
 
          further, particularly among African-Americans.  The strong association
 
        between social support and fetal health locus of control orientations
for
 
          all African-American respondents regardless of socioeconomic status,
and
 
          the divergent finding that the association was significantly strong
only
 
          for whites from lower income brackets, warrants further examination.
 
       Additionally, communication scholars report extensive research that
 
     demonstrates a significant relationship between social support and actual
 
          physical and mental well-being, and recovery from illness (Burleson,
1990;
 
          Kessler & Essex, 1982; Sarason, Levine, Basham, & Sarason, 1983;
Sarason,
 
          Shearin, Pierce, & Sarason, 1987; Sullivan & Reardon, 1986).
        Therefore, research should be conducted to test the associations between
 
          social support and health beliefs and health behavior that may exist
for
 
          populations other than high school women.  For example, physicians are
 
        unable to explain phenomena such as why socioeconomic status is linked
to
 
          health status (Pappas, Queen, Hadden, & Fisher, 1993).  Pappas et al.
were
 
          unable to explain findings that showed even though Medicaid and other
 
       health programs for the poor increased over a 26-year period, the gap
betwe
 
          en social classes in death rates also increased.  Although death rates
 
        decreased overall in this period, the reduction was much more dramatic
for
 
          college-educated males:  The rate fell from 9 deaths per 1,000 in
1966, to
 
          7.6 per 1,000 in 1986, among white males with less than a high school
 
       education; and from 5.7 deaths to 2.8 deaths per 1,000 for white
 
  college-educated males.  The authors found the same dissimilarity exists
 
          for low versus high income populations.  Research examining the
 
 relationship between social support, socioeconomic status, and health locus
 of control measures in other population samples may illuminate problems
 
          such as the disparity in death rates by socioeconomic status.
Conclusion
        The question motivating this study is:  Why are African American women
 
         less likely to seek prenatal care than women from other ethnic groups,
and
 
          how can mass media help change this pattern?  Public health
professionals,
 
          however, often treat disparate health status as a product of race and
 
       income level alone.  Therefore, these officials respond to health
 
   discrepancies by providing new services and informing the target audience
 
          about the services.  But extant studies demonstrate that this strategy
 
        often is ineffective.  For example, the sociodemographic effects model
 
        fails to address enduring dispositional attitudes that guide health
 
     behavior.  Additionally, the model ignores cognitive and affective factors
 
          that contribute to health beliefs.
        This study supported the need for a mediated effects, or a comprehensive
 
          model of health behavior, that will better inform health communicators
 
        about possible deterrents to health behavior.  The core relationships
this
 
         study identified are those of propositional health knowledge and
perceived social support
 
            with fetal health locus of control.  The study results indicate that
these cognitive and
 
            affective factors better account for variance in health beliefs than
race and income
 
         level, though the strength of these associations varies with race and
income.
        There are four primary advantages to viewing health status from a mediated
 effects approach.  First, the model explains anomalous findings that
 
       demographics cannot explain, such as why African-American women are less
 
          likely than whites to seek prenatal care regardless of income level.
 
       Second, the model may be generalizable to health behavior other than
 
      pregnancy behavior.  Third, the model offers a mechanism for identifying
 
          obstacles to healthful behavior, some of which could be addressed
with
 
          public health campaigns.  For example, it helps explain why roughly
half of
 all pregnant African-American women do seek prenatal care in their first
 
     trimester and why the other half does not.  These women may differ in their
knowledge of
 
            health and feelings of social support.  Finally, the model supports
other health
 
     communication researchers who have called for an integrated approach for
health campaigns,
 one that incorporates both mass communication and interpersonal communication
strategies,
 to reach targeted populations (Pettegrew & Logan, 1987; Reardon, 1988).
        Developing comprehensive models of health behavior that describe
 
   interrelations among culture and cognitive and affective factors will
 
       contribute to health communication researchers' understanding of group
 
        behavior.  Public health professionals informed by such models may find
 
         more effective and successful approaches for reaching high risk
 
 populations.    References
Ajzen, I. (1985). From intentions to actions:  A theory of planned
 
    behavior.  In J. Kuhl & J.    Beckman (Eds.), Action control:  From
cognition to
 
            behavior (pp. 11-39).   Heidelberg: Springer.
 
Amaro, H., Beckman, L. J., & Mays, V. M. (1987).  A comparison of black and
 white     entering alcoholism treatment.  Journal of Studies on Alcohol, 48
(3),
 
            220-228.
 
Anderson, R. B. (1989).  Reassessing the odds against finding meaningful
 
          behavioral    change in mass media health promotion campaigns.  In C.
Botan
 & V. Hazelton (Eds.), Public Relations Theory, (pp. 309-321).  Hillsdale, NJ:
 
        Erlbaum.
 
Arnston, P. (1989).  Improving citizens' health competencies.  Health
 
      Communication, 1, 29-34.
 
Atkin, C. K. (1990) Mass communication and public health. Newbury Park:  Sage.
 
Becker, Maiman, L . A., Kirscht, J. P., Haefner, D. P., Drachman, R. H., &
 
          Taylor,
D. W.  (1979).  Patient perceptions and compliance:  Recent studies of the
 
          Health Belief Model.  In  R. D. Haynes, Tayler, D. W., &  D. L.
Sackett
 
         (Eds.), Compliance in Health Care.  Baltimore:  Johns Hopkins
University Press.
 
Boone, M. S. (1989). Capital crime: Black infant mortality in America.  Newbury
Park:
 Sage.
 
Brown, J., Ettema, J., Luepker, B. (1981).  Knowledge gap effects in a
 
        cardiovascular information campaign.  Public Opinion Quarterly, 47,
516-527.
 
Burleson, B. R. (1990).  Comforting as social support:  Relational
 
    consequences of supportive behaviors.  In S. Duck (Ed.), Personal
 
 Relationships and Social Support, pp 66-82.  London:  Sage.
 
Coelho, R. J. (1985).  A psychometric investigation of the multidimensional
 health locus of control scales with cigarette smokers.  Journal of Clinical
 
            Psychology, 41(3), 372-376.
 
Cooper, D., & Fraboni, M (1990).  Psychometric studies of forms A and B of
 
          the multidimensional health locus of control scale. Psychological
Reports, 66,
 
            859-864.
 
Donohue, G. A., Tichenor, P. J., & Olien, C. N. (1975).  Mass media and the
 knowledge gap:  A hypothesis reconsidered.  Communication Research, 4, 179-191.
 
Eberstadt, N. (1992, January 20).  America's infant mortality problem:
 
         Parents.  The Wall Street Journal, p. A12.
 
Eisen, Zellman, & McAlister (1985).  A health belief model approach to
 
        adolescents' fertility control:  Some pilot program findings.  Health
 
      Education Quarterly, 12(2), 185-210.
 
Fellure, S. R. (1993).   Analysis of Information on Prenatal Care as found in
Popular
 Magazines.  Unpublished paper, Purdue University, Department of Communication,
West
 
         LAfayette, IN.
 
Fishbein, M., Ajzen, I. & McArdle, J. (1980). Changing the Behavior of
 
        Alcoholics:  Effects of Persuasive Communication.  In I. Ajzen & M.
 
     Fishbein (Eds.), Understanding attititudes and predicting social behavior
(pp.
 
      217-242). Englewood Cliffs, NJ:  Prentice-Hall.
 
Gaziano, C. (1984).  The knowledge gap:  An analytical review of media
 
        effects.  Communication Research, 10, 447-486. Grady, K., Kegeles, S.,
Lund, A., Wolk, C.
 & Farber, N. (1983).  Who volunteers for a breast self-examination program?
Evaluating
 
            the bases for self-selection.  Health Education Quarterly, 10 (2),
79-94.
 
Indiana State Board of Health (1989).  Indiana infant mortality report.
 
Indiana State Board of Health (1993).  Indiana infant mortality report.
 
Jaco, E. G. (1979).  Patients, physicians, and illness. New York: Free Press.
 
Kasch, C. (1984). Interpersonal competence and communication in the
 
     delivery of nursing care.  Advances in Nursing Science, 6, 71-78.
 
Kessler, R. C., & Essex, M. (1982).  Marital status and depression:  The
 
          importance of coping resources.  Social Forces, 61(2), 484-507).
 
Klesges, R. C., Somes, G., Pascale, R. W., Klesges, L. M., Murphy, M.,
 
        Brown, K., Williams, E. (1988).  Knowledge and beliefs regarding the
 
      consequences of cigarette smoking and their relationship to smoking status
 
          in a biracial sample.  Health Psychology, 7(5), 387-401.
 
Labs, S. M., & Wurtele, S. K. (1986).  Fetal health locus of control scale:
  Development and validation.  Journal of Consulting and Clinical Psychology, 54
 
          (6), 814-819.
 
Landis, R. J., Freimuth, V., Cameron, D. (1993).  A Decade of AIDS Coverage in
 
            Three African American Magazines.  Paper presented at the Annual
Meeting of the
 
    International Communication Association, Washington, D. C.
 
Larde, J. & Clopton, J. R. (1983).  Generalized locus of control and health
 locus of control of surgical patients.  Psychological Reports, 52, 599-602.
 
Lia-Hoagberg, B., Rode, P., Skovholy, C. J., Oberg, C. N., Berg, C.,
 
      Mullett, S., & Choi, T. (1990).  Barriers and motivators to prenatal care
 
          among low income women.  Social Scientific Medicine, 30(4), 487-495.
 
Marin, B. V., Marin, G., Perez-Stable, E. J., Otero-Sabogal, R., & Sabogal,
 F. (1990).  Cultural differences in attitudes toward smoking:  Developing
 
          messages using the theory of reasoned action.  Journal of Applied
Psychology, 20
 
            (6), 478-493.
 
Marshall, G. N., Collins, B. E., & Crooks, V. C. (1990).  A comparison of
 
          two multidimensional health locus of control instruments.  Journal of
 
     Personality Assessment, 54(1 & 2), 181-190.
 
National Center for Health Statistics (1987).  Reports on Health and Illness in
 
            American Society:  1977-87 and 1990 goal.  Division of Vital
Statistics, National Vital
 
            Statistics  System.
 
Noland, M. P., Riggs, R. S., & Hall, J. W. (1985).  Relationships among
 
         knowledge, health locus of control, and health status in secondary
special
 
          education students. The  Journal of Special Education, 19 (2),
177-187.
 
Office of Minority Health Resource Center (1989).  United States Department
 of Health and Human Services Report, revised 11/20/89. Room 118-F, HHH
 
         Bldg., 200 Independence Ave. S.W., Washington, DC. 20201
 
O'Keefe, D. (1990). Persuasion: Theory and research (pp. 166-169). Newbury Park,
CA:
 
            Sage.
 
 Oritt, E. J., Paul, & Behrman, J. A. (1985).  The perceived support network
 inventory.  American Journal of Community Psychology, 13(5), 565-582.
 
Pappas, G., Queen, S., Hadden, M, & Fisher, G. (1993).  The increasing
 
        disparity in mortality between socioeconomic groups in the United
States,
 
          1960 and 1986.  New England Journal of Medicine, 329(2), 103-109.
 
Parcel, G. S., & Meyer, M. P. (1978).  Development of an instrument to
 
        measure children's health locus of control.  Health Education
Monographs, 6(2),
 
            149-159).
 
Pendleton, D. and Hasler, J. (Eds.) (1983). Doctor-patient communication.  New
 
           York: Academic Press.
 
Pennebaker, J. W. (1982).  The psychology of physical symtoms. New York:
 
  Springer-Verlag.
 
Pettegrew, L. & Logan, R. (1987).  The health care context.  In C.R. Berger
 & S. N. Chafee (Eds.), Handbook of communication science.  675-699.
 
Public Health Service Centers for Disease Control (1988, 1989, 1990).
 
       Reports on Health Statistics.
 
Reardon, K. K. (1988).  The role of persuasion in health promotion and
 
        disease prevention:  Review and commentary.  Communication Yearbook, 11,
277-297.
 
Rice, R. E., & Atkin, C. K. (1989).  Public Health Campaigns.  Newbury Park:
 
         Sage.
 
Rotter, J. B. (1966).  Generalized expectancies for internal versus
 
     external control of reinforcement.  Psychological Monographs:  General and
Applied
 
            , 80(1), 1-28.
 
Rowan, K. E. (1988).  A contemporary theory of explanatory writing.  Written
 
            Communication, 5, 23-56.
 
St. John, C. & Winston, T. (1989).  The effect of social support on
 
     prenatal care.  The Journal of Applied Behavioral Science, 25(1), 79-98.
 
Sarason, I. G., Levine, H., Basham, R., & Sarason, B. (1983).  Assessing
 
          social support:  The social support questionnaire.  Journal of
Personality and
 
            Social Psychology, 44, 127-139.
 
Sarason, B. R., Shearin, E., Pierce, G. R., & Sarason, I. G. (1987).
 
       Interrelations of social support measures:  Theoretical and practical
 
       implications.  Journal of Personality and Social Psychology, 52, 813-832.
 
Sarason, I. G., Sarason, B. & Shearin, E. (1986).  Social support as an
 
         individual difference variable:  Its stability, origins, and relational
 
         aspects.  Journal of Personality and Social Psychology, 50(4), 845-855.
 
Schoch, E. B.  (1993, January).  Infant death rate for blacks on the rise
 
          again.  Indianapolis Star, p. 1.
 
Schatz, P. E. (1988).  An evaluation of the components of compliance in
 
         patients with diabetes.  Journal of the American Dietetic Association,
88(6),
 
       708-712.
 
Stanley, R. O., Hyman, G. C., & Sharp, C. A. (1984).  Health locus of
 
       control:  Support for recent multidimensional developments.
Psychological
 
           Reports, 54, 329-330.
 
 Starr, P. (1982).  The social transformation of American medicine.  New York:
Basic
 
           Books, Inc.
 
Sullivan, C. F., & Reardon, K. K. (1986).  Social support satisfaction and
 
          health locus of control:  Discriminators of breast cancer patients'
styles
 
          of coping.  In M.L. McLaughlin (Ed.), Communication Yearbook, 9 (pp
707-722).
 
          Beverly Hills, CA:  Sage.
 
Tichner, P. J., Donohue, G. A., & Olien, C. N. (1970).  Mass media flow and
 differential growth in knowledge.  Public Opinion Quarterly, 34, 159-170.
 
Tinsley, B. J., & Holtgrave, D. R. (1989).  Maternal health locus of
 
      control beliefs, utilization of childhood preventive health services, and
 
          infant health.  Developmental and Behavioral Pediatrics, 10(5),
236-241.
 
United States Bureau of Labor Statistics (1989). Handbook of labor statistics
 
           (Bulletin 2340).  Washington DC:  U.S. Government Printing Office.
 
United States Department of Commerce.  (1990).  Poverty in the United States:
 
           1990 (Current Population Reports;  Consumer Income Series P-60, No.
175).  Washington
D. C.:  U. S. Government Printing Office.
 
Wallston, K. A., & Wallston, B. S. (1978).  Development of the
 
 multidimensional health locus of control (MHLC) scales.   Health Education
 
          Monographs, 6(2), 160-170.
 
Wallston, K. A., Smith, R. A., King, J. E., Forsberg, P. R., Wallston, B.
 
          S., & Nagy, V. T. (1983).  Expectancies about control over health:
 
     Relationship to desire for control of health care.  Personality and Social
 
         Psychology Bulletin, 9(3),  377-385.
 
Wallston, B., Wallston, K., Kaplan, G. & Maides, S. (1976). Development and
 validation of the health locus of control (HLC) scale.  Journal of Consulting
 
            and Clinical Psychology, 44, 580-585.
 
Weitzel, M. H., & Waller, P. R. (1990).  Predictive factors for
 
 health-promotive behaviors in white, Hispanic, and black blue-collar
 
      workers.  Family and Community Health, 13(1), 23-34.
Wood, C. S. 1979). Human sickness and health: A biocultural view.  Palo Alto,
CA:
 
          Mayfield Publishing Co.
 
Young, C., McMahon, J. E., Bowman, V., & Thompson, D. (1989).  Maternal
 
         reasons for delayed prenatal care.  Nursing Research, 38(4), 242-243.
 
 
 Table 1.  List of variables included in study
 
 
Independent Variables
 
1.      Newspapers or magazines
2.      Male friend
3.      Female friend
4.      Television programs and news
5.      Parent(s)
6.      School teacher or counselor
7.      Religious leader
8.      Class materials, textbooks
9.      Radio programs or news
10.     Doctor or nurse
11.     Other relative
 
12.     Race
13.     Income level
 
 
 
 
 
 
 
 
 
14.     Academic achievement
        ( self-reported grade point average )
15.     Perceived social
          support-------Affective Variable
16.     General health
          knowledge-----Cognitive Variable
 
 
                                        Dependent Variables
 
17.     Multidimensional health locus of
 
          control
        total scale
18.     Multidimensional health locus of
 
         control: internal
19.     Multidimensional health locus of
 
          control: powerful others
 
 
 
 
         (external)
20.     Multidimensional health locus of
 
         control:       chance/fate/luck (external)
21.     Fetal health locus of control total
 
          scale
22.     Fetal health locus of control:
 
        internal
23.     Fetal health locus of control:
 
        powerful others         (external )
24.     Fetal health locus of control:
 
        chance/fate/luck        (external)
 
 
 
 
Health Information Source Variables
 
 
 
 
 
 
Group Variables
 
 
Individual-Difference Variables
 
 
 
 
 
 
 
 
 
 
Health Beliefs Variables
 
 
                Table 2.  Zero-order correlations
 
        between independent
measures and multidimensional health locus
 
          of control
 
 
 
Variables
 
MHLC Total
 
MHLC Int
 
MHLC P.O.
 
MHLC Ch
 
Race
 
.06
 
.18
 
.06
 
.01
 
Income
 
.05
 
 .02
 
.10
 
.02
 
GPA
 
   -.30**
 
.11
 
.16
 
    -.29**
 
Health Kn.
 
     -.48***
 
.16
 
      -.33***
 
      -.41***
 
Soc. Sup
 
.02
 
.02
 
.12
 
.03
 
 
n = 96
 
Significance
levels:
*** p < .001
**   p < .01
*     p < .05
 
 
MHLC Total      = Total scale mean score of
 
          multidimensional health                       locus of
 
        control
MHLC Int        = Internality subscale of
 
          multidimensional health locus of control
MHLC P.O.       = Powerful others subscale
 
          of multidimensional health                    locus of
 
          control
MHLC Ch         = Chance subscale of
 
  multidimensional health locus of
 
        control Table 3.  List of first-order
 
         partial correlations between health
knowledge and multidimensional health locus
 
          of control controlling for race,
then controlling for income
 
 
 
 
MHLC Total
 
MHLC P.O.
 
MHLC Chance
 
Variables
 
Zero Order
          Correlations
 First Order Partial
 
         Correlation
 
Zero Order
          Correlations
 First Order Partial
 
         Correlations
 
Zero Order
          Correlations
 First Order Partial
 
         Correlations
 Health Knowledge
 
.48
 
.47*
 
.33
 
.34*
 
.41
 
.41*
 Health Knowledge
 
--
 
.48**
 
.33
 
.33**
 
.41
 
.41**
 
 
n = 96
 
Note:  Only the significant associations
 are examined in this table.
 
*  Controlling for race
**Controlling for income
 
MHLC Total = Total scale mean score of
 
          multidimensional health locus of control
MHLC P.O.   = Powerful others subscale
 
          of multidimensional health locus of
 
        control
MHLC Ch      = Chance subscale of
 
      multidimensional health locus of control
 Table 4.  List of zero-order
 
 correlations between
independent variables and fetal health locus
 
          of control
 
 
 
 
FHLC Total
 
FHLC Int
 
FHLC P.O.
 
FHLC Ch
 
Race
 
.02
 
.16
 
.04
 
.09
 
Income level
 
.15
 
.02
 
  -.24*
 
.002
 
Grade Pt. Avg.
 
    -.28**
 
.05
 
.19
 
   -.28**
 
Health Knowl.
 
      -.39***
 
  -.20*
 
      -.33***
 
  -.24**
 
Social Support
 
.17
 
    -.40***
 
.04
 
.03
 
n = 96
 
Significance
levels:
*** p < .001
**   p < .01
*     p < .05
 
 
 
 
 
 Table 5.  List of
 
      first-order partial
 
        correlations between
 
         independent
variables and fetal
 
        health locus of control
 
          controlling for race
 
 
 
 
FHLC Total
 
FHLC Int
 
FHLC P.O.
 
FHLC Ch
 
Variables
 
Zero-Order
Correl-ati on
 First
Order Partial
 
        Correlation
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 
Income
 
--
 
--
 
--
 
--
 
.24
 
.24
 
--
 
--
 
Social Support
 
--
 
--
 
.40
 
.38
 
--
 
--
 
--
 
--
 
Health Knowledge
 
.39
 
.39
 
.20
 
.19
 
.33
 
.34
 
.24
 
.25
 
Note:  Only the
 
         significant
 
     associations are
 
          examined in this
 
          table.
 
FHLC total      = Total
 scale mean score
 
          of fetal health
 
         locus of control
FHLC Int        =
 
        Internality
 
     subscale of fetal
 
          health locus of
 
         control
FHLC P.O.       =
 
       Powerful others
 
         subscale of fetal
 
          health locus of
 
         control
FHLC Ch         = Chance
 
          subscale of fetal
 
          health locus of
 
         control
FHLC total      = Total
 
             mean score of fetal
 
             health locus of control
FHLC Int        =
 
        Internality subscale of
 
               fetal health locus of
 
               control
FHLC P.O.       = Powerful
 
               others subscale of
 
            fetal health locus of
 
                                                   control
FHLC Ch         = Chance
 
           subscale of fetal
 
           health locus of control
 
 Table 6.  List of
 
      first-order partial
 
        correlations between
 
         independent variables
and fetal health locus
 
          of control controlling
 
          for income
 
 
 
FHLC Total
 
FHLC Int
 
FHLC P.O.
 
FHLC Ch
 
Variables
 
Zero-Order
 
    Correlation
 First Order
 
      Partial Correlation
 
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 
Zero-Order
 
    Correlation
 First Order Partial
 Correlation
 Health Knowledge
 
.39
 
.39
 
.20
 
.20
 
.33
 
.33
 
.24
 
.24
 Social Support
 
--
 
--
 
.40
 
.40
 
--
 
--
 
--
 
--
 
Note:  Only the
 
         significant
 
     associations are
 
          examined in this
 
          table.
 
FHLC total      = Total
 scale mean score
 
          of fetal health
 
         locus of control
FHLC Int        =
 
        Internality
 
     subscale of fetal
 
          health locus of
 
         control
FHLC P.O.       =
 
       Powerful others
 
         subscale of fetal
 
          health locus of
 
         control
FHLC Ch         = Chance
 
          subscale of fetal
 
          health locus of
 
         control Table 7.
 
          List of zero-order
 
          correlations
 
      between general
 
         health knowledge
 
          and fetal health
 
          locus of control
 
          showing systematic
 
          variations as a
 
         function of race
 
          and income level.
 
 
 
Group
 
Health Knowledge
 
          and FHLC Total
 
Health Knowledge
 
          and FHLC Int
 
Health Knowledge
 
          and FHLC P.O.
 
Health Knowledge
 
          and FHLC Ch
 
All Blacks
 
    -.38**
 
.24
 
    -.44**
 
.14
 
All Whites
 
    -.41**
 
.14
 
.20
 
   -.36**
 High Income Blacks
 
          ( Level 2 and 3 )
 
    -.50**
 
.35
 
.45*
 
 -.45*
 High Income Whites
 
          (Level 2 and 3 )
 
.24
 
.17
 
.06
 
.23
 Low Income Blacks (
 Level 0 and Level
 
          1 )
 
.18
 
.09
 
.45*
 
.25
 Low Income Whites (
 Level 0 and Level
 
          1 )
 
    -.53**
 
 
 
.12
 
.35
 
    -.47**
 
 
Significance
levels:
**  p < .01
*    p < .05
 FHLC total     = Total scale mean score of
 
          fetal health locus    of control
FHLC Int        = Internality subscale of
 
          fetal health locus of control
FHLC P.O.       = Powerful others subscale
 
          of fetal health locus of control
FHLC Ch         = Chance subscale of fetal
 
        health locus of control
Note:   0  refers to respondents whose
 
     annual family incomes fall below the
 
         government guidelines for determining
 
          poverty (approximately $5,000-$19,000
 
          annual income).
 
                1 refers to respondents whose annual
 
     family incomes fall above the government
 guidelines for determining poverty, but
 are still considered in low income
 
        brackets (approximately $10,000 -
 
      $22,000 annual income).
 
                2 refers to respondents whose annual
 
     family incomes are considered middle
 
         class (approximately $18,000 - $
 
     45,000).
 
                3 refers to respondents whose annual
 
     family incomes are considered above
 
        middle class (approximately $45,000 -
 
          $70,000).
 Table 8.  List of systematic variations
 
     of zero-order correlations between
 
       perceived social support and fetal
 
       health locus as a function of race and
 
          income level
 
 
 
        Group
 
Social Support and
 
          FHLC Total
 
Social Support and
 
          FHLC Int
 
Social Support and
 
          FHLC P.O.
 
Social Support and
 
          FHLC Ch
 
All Blacks
 
   -.36**
 
   -.45**
 
.16
 
.25
 
All Whites
 
.08
 
.29*
 
.12
 
.18
 High Income Blacks
 
          ( Level 2 and 3 )
 
.40*
 
.43*
 
.30
 
.28
 High Income Whites
 
          (Level 2 and 3 )
 
.03
 
.15
 
.04
 
.08
 Low Income Blacks (
 Level 0 and Level
 
          1 )
 
.25
 
 -.44*
 
.05
 
.16
 Low Income Whites (
 Level 0 and Level
 
          1 )
 
.17
 
 -.43*
 
.36
 
 .26
 
 
 
Significance levels:
* * p < .01
*    p < .05
 FHLC total     = Total scale mean score of
 
          fetal health locus    of control
FHLC Int        = Internality subscale of
 
          fetal health locus of control
FHLC P.O.       = Powerful others subscale
 
          of fetal health locus of                      control
FHLC Ch         = Chance subscale of fetal
 
        health locus of control
 Note:  0  refers to respondents whose
 
               annual family incomes fall below the
 
              government guidelines for determining
 
               poverty (approximately $5,000-$19,000
 
               annual income).
 
        1 refers to respondents whose annual
 
               family incomes fall above the government
 guidelines for determining poverty, but
 are still considered in low income
 
             brackets (approximately $10,000 -
 
           $22,000 annual income).
 
        2 refers to respondents whose annual
 
               family incomes are considered middle
 
              class (approximately $18,000 - $
 
          45,000).
 
3 refers to respondents whose annual
 
                   family incomes are considered above
 
                  middle class (approximately $45,000 -
 
                    $70,000).
 [1] . Of the 98 participants,
 two did not complete at least one quest
ion on
 the dependent measures and ther
efore were not included in analyses
 
 
     involving those variables.  Ad
ditionally, one respondent did not answe
r
 
         questions in the family
income survey and consequently was not i
ncluded in
 
          analyses involv
ing economic indicators, and one partici
pant did not answer
 
          the ac
ademic achievement question and was not
included in analyses
 
    perta
ining to that variable.