Download Determining the Number and Dimensions of Health Behaviors: A Study on Navy Personnel and more Lecture notes Health sciences in PDF only on Docsity!
DEMONSTRATION OF REPLICABLE DIMENSIONS OF HEALTH BEHAVIORS
Ross R. Vickers, Jr.'
Terry L. Conway 2
Ic L^4 nda K. Hervig'
NTIS CRA&I
DI1C TAB
U'nonio Jvi,.d[.
Applied Physiology Department' Juslicillon
and
Health Psychology^ Department^2 ByDstrlbijllOn^ ...... I
Naval Health Research Center
P.O. Box 85122 Av.albihty Codes
San Diego, CA 92138-9174 Dist AveIlo
Disi Spicil
* Report No. 88-41 was supported by the Naval Medical I search and
Development Command, Navy Medical Command, Department of the Navy, under
Research Work Unit MRO4101.00A-6004; and by the Naval Military Personnel
Comm-nd, Department of the Navy, under Work Order No. NOO02288VRWW508.
The views presented are those of the authors and do not reflect the
offi-:ial policy of the Department of the Navy, Department of, Defense, nor
the '.^ S.^ Government.
SUMMARY
Appropriate health behaviors are necessary to ensure health and well-being, thereby keeping military personnel ready to perform their jobs which may demand exceptional efforts at key times. An understanding of factors influencing health behaviors would be more readily achieved if general dimensions could be identified to delineate sets of health behaviors that consistently co-occur. 'Such dimensions may represent the effects of causal factors influencing multiple behaviors and^ may,^ thereby,^ provide^ an^ empirical basis for identifying causal factors that have widespread behavioral effects. Hodifying these causal factors may be an efficient way to improve health behavior. Well-defined health behavior dimensions are a requirement for these undertakings, but such dimensions have not been established. -Prior research has suffered from the use of only brief lists of^ health^ behaviors,^ failure^ to systematically select health behaviors to represent hypothesized health behavior dimensions, and failure to replicate findings across samples. The present study was designed to extend prior efforts by determining the number of dimensions of health behavior that could be reliably identified in two samples of Navy personnel. A set of 40 health behavior items was chosen to represent four major dimensions of health behavior that prior work suggested were present in groups representing a wide range of social and demographic backgrounds.- One sample of participants consisted of 812 men assigned to duty on U.S. Navy ships during 1984 who^ volunteered^ to^ participate^ in^ a^ survey^ study^ of general health habits conducted as part of program evaluation efforts for the Navy's Health and Physical Readiness Program. A second sample consisted of 605 recruits participating in a study of the effects of different interventions to stop smoking in Navy basic training. Data on the 40 health behaviors were collected by self-report questionnaires. Principle components analysis was conducted with 2, 3, 4, and 5 components extracted in each sample. The stability of the solutions across samples was determined by computing coefficients of congruence, by cross-validating regression weights for the factor scores, and by determining the number of items with component loadings greater than .30 in both samples. Different solutions also were compared in terms of the number of items that could be assigned to at least one component and how many of these were assigned to just a single component. S. I
INTRODUCTION
Health behaviors can be broadly defined as actions undertaken to maintain or improve health (Kasl & Cobb, 1966). One issue in health behavior research is whether such behaviors must be considered individually or can be grouped into general categories to (^) better understand them. It has been demonstrated (^) repeatedly that health behaviors tend to co-occur and that between 2 and 5 dimensions or clusters are needed to summarize the empirical patterns of association between behaviors (Williams & Wechsler, 1972, 1973; Harris & Guten, 1979; Langlie, (^) 1977, 1979; Tapp & Goldenthal, 1982; Vickers & Hervig, 1984; McCarthy & Brown, 1985; Norman, 1985; Kannas, 1981; Steele & McBroom, (^) 1972). While it is reasonable to regard the presence of multiple categories of co-occurring health behaviors as (^) well established, there presently is no consensus regarding the number or (^) precise content of the categories required to describe these behaviors. The present study was undertaken to help resolve these issues by determining the number of replicable dimensions of health behavior in two large samples of young men. The conclusion that health behavior is multidimensional has important implications for the conceptualization, measurement, and modification of health behavior. Conceptually, multidimensionality means that health behaviors are neither monolithic nor independent. Instead, theoretical models must incorporate intermediate concepts that encompass multiple behaviors, but do not attempt to treat health behavior as a (^) monolithic entity. From a measurement perspective, the implication is that multi-item measures are feasible. However, it is necessary to define the domains (^) of each concept, clearly defining the referent behaviors as a basis for defining observations that can be used for measurement. The behavior modification implications are linked to the assumption that behaviors which co-occur regularly share some common causes, while (^) the differentiation of behaviors into multiple categories implies differences (^) in causes across dimensions. (^) If so, well-defined categories will provide a basis for more effective attempts (^) to identify manipulable antecedents of health behaviors, thereby providing a better basis for choosing (^) the targets of interventions. The most critical problem preventing health researchers from realizing the benefits of multidimensional models (^) of health behavior is the inconclusive nature of the evidence regarding (^) the number of dimensions to be considered. To date, the typical study has not systematically sampled
hypothesized categories or dimensions of health behavior, has involved only
a few health behaviors, and has not verified the replicability of the factor
or category structure demonstrated. Although there are isolated instances
of studies that involved large numbers of behaviors (Williams & Wechsler,
1972, 1973; Vickers & Hervig, 1984), systematic sampling from a defined
conceptual domain (Langlie, 1977), and systematic replication across samples
(Norman, 1985), the authors are unaware of any available study combining
these attributes.
The present study was designed to further health behavior research by
providing additional information regarding potential benchmark dimensions
for health behavior. Forty health behaviors were selected to represent four
major empirical categories of health behavior described by Vickers and
Hervig (1984). Broadly speaking, the categories represented (a) behaviors
which reduce the risk of overtaxing the body's adaptive capacity, (b)
behaviors which involve risk taking, primarily as a pedestrian or driver,
(c) behaviors which should help prevent the onset of illness, and (d)
behaviors which might improve health rather than merely prevent illness.
These categories were not necessarily expected to exhaust the important
components of health behavior, but they did provide a framework for sampling
health behaviors that was sufficiently general to encompass the majority of
behavioral groupings suggested by prior research.
METHOD
Sample
Two samples of Navy personnel completed health behavior checklists
voluntarily after receiving descriptions of research studies which included
these lists as part of more general research designs. The first sample
consisted of 812 men assigned to duty aboard Navy ships. The typical
respondent in this sample was 25.9 (S.D. = 6.0; range = 18-50) years of age.
The primary ethnic groups were Caucasians (79.1%), Blacks (8.9%),
Malayans/Filipinos (5.9%) and Hispanics (4.7%). Nearly all of the
participants had 12 years (68.4%) or more (25.4%) of formal schooling.
Enlisted personnel comprised 92.9% of the sample and officers 7.1%. The
average length of service at the time of the survey was 6.0 (S.D. = 5.5)
years.
The second sample consisted of 605 male Navy recruits who completed the
the 5-component solution, the second largest coefficient of congruence had
to be chosen for one component to provide a better overall match for the
full set of components (see Results).
Two additional methods of comparing component solutions were used to
confirm the component matching based on the coefficients of congruence. The
similarity index (Cattell, Balcar, Horn & Nesselroade, 1969) was one
addition. This index is computed by specifying an absolute value for
component loadings that determines whether or not an item is salient to that
component. In the present application, there were no components which had
both positive and negative salient items, so components defined by large
negative loadings were reflected and all salient loadings were greater than
zero. Under these conditions, the similarity index is the ratio of the
number of items salient to both components divided by the total number of
items salient to at least one of the components being compared. Thus, the
similarity index would reach a maximum value of 1.00 when exactly the same
items were salient to both components and a minimum value of .00 when there
was no overlap in the sets of salient items. In this study, the similarity
index was computed twice, once with .30 as the criterion and once with.
as the criterion, to evaluate the effect of criterion choice on estimates of
similarity (Walkey, 1986). The number of items salient for the components
being compared is presented in the results to indicate the number of
salience matches contributing to the similarity index and as a guide to the
number of items which might be considered as potential elements of scales to
represent the component.
The preceding tests describe the replicability of the component
solutions in terms of the location of health behavior items in component
space. A fourth replicability estimate was provided by computations based
on the location of individuals in the component space. The factor score
regression coefficients were obtained for each component analysis in each
sample. These regression weights then were applied to the standardized item
scores within each sample to provide two sets of linear composites. One set
of composites represented estimated component scores for the sample obtained
by applying the regression weights derived in that sample to the data for
that sample (e.g., shipboard weights applied to the data of the shipboard
sample); the second set of linear composites represented estimated component
scores obtained applying the regression weights derived in the other sample
(e.g., recruit weights applied to the data of the shipboard sample). The correlations between the two sets of composites then were computed within each sample to determine how similar the scores produced by the two sets of weights were. If the matched factors defined by the coefficients of congruence produced very similar regression weights for the computation of factor scores, these "cross-validation" coefficients would be close to 1. (Everett, 1983). The second analysis concern was the definition of^ behavior^ composites that could be used as marker variables to represent the replicable health behavior dimensions. This concern directed attention to the identification of specific behavioral instances which could be employed to represent those dimensions. Identification of specific behaviors as representative of a given component was based on an average weighted component loading of .45 or more with a loading of .30 or greater in both samples, provided that the item met these criteria^ for^ only^ a^ single^ component.
RESULTS Component Replication Analyses On the whole, the 2-component solution was the most replicable across the two samples (Table 1), but there was no clear failure to match componenrtz until the 5-component solution was reached. Even for the 5-component solution, it was possible to match components so that the various replication coefficients were comparable in magnitude to those obtained in the 3- and 4-component solutions. However, Table 1 does not show the close similarity of shipboard component 4 and^ the^ recruit^ component 2 in the 5-component solution. The coefficient of congruence for this pairing was^ .77^ with^ cross-validation^ correlations^ of^ .69^ and^ .61^ and similarity coefficients of .48 and .55. These values were larger than those obtained matching shipboard component 2 with recruit component 2 as shown in Table 1. However, if shipboard component 4 had been matched with recruit component 2, then shipboard component 2 would^ have^ been^ matched^ with^ recruit component 4. This match would^ have^ produced^ a^ low^ coefficient^ of^ congruence (.36), low cross-validation correlations (.10 and .20 for the shipboard and recruit samples, respectively), and low similarity indices (.08 and .00, for the low and high criteria, respectively). The combined implication of these statistics was that shipboard component 4 was the best match for t,4o recruit
Components 4A and^ 4B^ clearly^ were^ comprised^ of^ subsets^ of^ the^ behaviors
defining component 2A. Component 4C clearly was part of Component 2B.
Finally, 4 of 5 items with weighted average loadings greater than .40 on
Component 4D had their largest loading^ on^ Component^ 2B.
The replicability of the health behavior components also was estimated
by comparing the present 4-component solution to that reported by Vickers
and Hervig (1984). Coefficients of congruence were computed based on the
component loadings for the 34 items common to the two studies. Approximate
matches for the shipboard sample were: Component^ 4A^ -^ Vickers^ &^ Hervig
(V&H) Component 4 (.81); Component 4B - V&H Component 3 (-.84); Component 4C
- V&H Component 2 (.68); Component 4D^ -^ V&H^ Component^2 (-.67)^ or^ V&H
Component 1 (.52). Approximate matches for the recruit sample^ were:
Component 4A - V&H Component 3 (-.75) or^ V&H^ component^1 (.70);^ Component^ 4B
- V&H Component 3 (-.59)^ or^ V&H^ Component^4 (.66);^ Component^ 4C^ with^ V&H
Component 2 (.80);^ Component^ 4D^ with^ V&H^ component^4 (.74).
Table 2
Averaged Component Loadings^ for^ Health^ Behaviors:
2- and 4-Component Solutions
Solution: 2-Component 4-Component
Component: 2A 2B 4A 4B 4C^ 4D
Preventive Habits
(a) Wellness Behaviors
14 Exercise# .61* .05^ .16^ .53*^ .05^.
31 filth Info# .55* .15 .18 .50* -. 10.
8 Reg Check# .55^ .16^ .28^ .55^ -.^17.
22 Dent Check# .55* .04 .18 .61* -. 07 -.
30 Disc Hlth# .53* .09 .15 .50* -. 05.
23 Limit Food# .53* .11 .16 .47* -. 05.
32 Floss# .50* .03 .21 .50* -. 03.
11 Weight# .50* .11 .13 .46* -. 04.
25 Vitamins# .47* -. 11 .02 .56* .08.
1 Diet .46*^ .09^ .30^ .32^ .01^.
35 Food Suppl# .45^ -.^14 -.^07 .57^ .12^.
20 Avoid Germs .45* .30^ .24^ .32^ -.^21.
29 Avoid Poll .42* .19 .23^ .23^ -.^02 .43*
37 Inoculation .39* .12^ .28^ .36*^ -.^12 -.^08
34 Brush Teeth .38* .00 .17 .32 .05.
24 Avoid OTC Med .37*^ .12^ .18^ .30^ -.^03.
9 Religion .35 .27 .10 .31 -.^19.
Table 2 Continued Averaged Component Loadings for Health Behaviors: 2- and 4-Component^ Solutions
Solution: 2-Component 4-Component Component: 2A 2B 4A 4B 4C 4D (b) Accident^ Control 3 Emerg Phone# .34^ .25^ .61^ .01^ -.10^. 7 Destroy Med# .40^ .20^ .57^ .09^ -.07^. 6 First Aid Kit# .37 .09* .56* .10 .03^.
19 Check Hazard# .52* .27 .56* .23 -. 16.
21 Fix Broken# .50* .08 .53* .25 .00. 36 Know First Aid# .44* .00 .47^ .25^ .07^ -. 13 Health Sign^ .58^ .28^ .42^ .40^ -.21^. 4 Relax .31 .07 .41 .07^ .03^.
Risk Taking^ Habits
(a) Traffic-related Risks
28 Cross Street# -. 04 -.60* -. 20 .01^ .63*^ -.^02
38 Take Chances# .24 -.58^ .11^ .14^ .62^ -. 33 Drive Fast# -.05^ -.57^ -.01^ -.11^ .60^ -. 5 Pedest Risk# -.06 -.55* -.14^ -.07^ .62^. 12 Traffic Rule# .30 .55 .28^ .18^ -.50^. 15 Stop Light# .01^ -.51^ -.^14 .02^ .57^. 40 Risky Hobbies# .14 -.50 .18 .02 .53* -. (b) Substance^ Use^ Risk 26 Not Drink# .14 .41* .00 .05 -.23^ .57* 18 Not Chem Subs# .26 .35 .08 .14^ -.17^ .50* 39 Drink/Drive -. 08 -.53* -.08 .01 .38* -.43* 16 Avoid Crime .21^ .43^ .21^ .04^ -.30^. Miscellaneous Items 17 Do Not Smoke# .16 .19 -.^19 .15^ -.03^ .55 2 Get Sleep .29 .16^ .27^ .14^ -.08^.
27 Seat Belt .35* .39* .24 .25 -.^29.
10 Avoid^ Chills^ .40^ .32^ .29^ .27^ -.^26.
NOTE: Table entries are weighted averages of the component loadings for the two samples computed using sample^ sizes^ as^ the^ weights.^ Numbers^ at^ the^ left margin indicate item^ numbers^ as^ they^ appear^ in^ the^ complete^ checklist^ (See Appendix A). ""* indicates that^ the^ component^ loading^ was^ greater^ than^. in both samples. "#" indicates an item used in the^ proposed^ health^ behavior composites.
an additional 1% had Graduate^ Equivalence^ Diplomas.^ The^ second^ sample
consisted of male Marine Corps personnel^ going^ through^ cold^ weather^ training
(n = 95) who completed the^ health^ behavior^ questionnaire.^ The^ typical
respondent in this sample was 21.9 (S.D. =^ 3.7,^ range^ =^ 18-39)^ years^ of^ age.
The primary ethnic groups were Caucasians (68%), Blacks (16%),^ and^ Hispanics
(7%). The large majority of the respondents had 12 years (83%)^ or^ more^ (9%)
of formal education. Most of the men were enlisted^ (95%)^ with^ a^ median^ of
24 months of service (range 5 months - 17 years).
Descriptive statistics for the proposed marker^ composites^ were^ computed
for these additional samples (Table 3). The resulting internal consistency
estimates, mean^ scores,^ and^ patterns^ of^ correlation^ were^ broadly^ similar^ to
those in the development samples.
Table 3
Descriptive Statistics for Proposed Scales
Inter-scale
correlations
Mean S.D.^ Alpha^ (1)^ (2)^ (3)
Shipboard Sample (n = 812)
(1) Wellness 2.87^ .77^.
(2) Accident Control 3.41 .84 .73^.
(3) Traffic Risks 2.70 .78 .75^ -.^12 -.
(4) Substance Risks 3.03^ 1.08^ .48^ -.^31 -.^15.
Recruit Sample^1 (n^ =^ 605)
(1) Wellness 2.81 .75.
(2) Accident Control 3.33^ .84^ .64^.
(3) Traffic Risks 3.09^ .78^ .67^ -.^20 -.
(4) Substance Risks 3.41^ 1.19^ .61^ -.^21 -.^08.
Recruit Sample 2 (n = 103-116)
(1) Wellness 2.92^ .70^.
(2) Accident^ Control^ 3.50^ .74^ .57^.
(3) Traffic Risks 3.35 .75 .74 -.^17 -.
(4) Substance Risks^ 3.16^ 1.10^ .46^ -.^08 -.^05.
Marine Corps Sample (n = 95)
(1) Wellness 3.12 .70.
(2) Accident Control 3.33 .78 .67.
(3) Traffic Risks 3.25^ .70^ .64^ -.^28 -.
(4) Substance Risks^ 3.00^ 1.01^ .43^ -.^13 -.^10.
DISCUSSION
This study added to the evidence that health behaviors are
multidimensional. The primary extension of previous findings has been the
demonstration that covariations of health behaviors have a replicable
pattern, at least when studied in comparable samples, with 2 to 4 dimensions
or categories needed to summarize the patterns of association. The content
of the most specific categories can be interpreted as identifying sets of
behaviors which are related to (a) maintenance and enhancement of
well-being, (b) avoiding or minimizing the effects of accidents, (c) taking
risks, primarily related to avoidable exposure to automotive or pedestrian
hazards, and (d) consumption of substances which may adversely affect health
(e.g., tobacco and alcohol). The first three behavior categories were
well-defined in the present study, and the fourth is one of the most
consistently replicated factors in prior studies of health behaviors (Harris
& Guten, 1979; Norman, 1985; Kannas, 1981; Tapp & Goldenthal, 1982).
The replicability of the proposed dimensions of health behavior might
be disputed on the basis of the weak matches to the four components
identified in a prior study (Vickers & Hervig, 1984). However, this aspect
of the findings must be evaluated in the context of differences between the
two studies. These differences included not only the sampling of specific
health behaviors (see pg. 7) but differences in response format (dichotomous
versus 5-point Likert-scale) and component rotations (oblique versus
orthogonal). Collectively, these differences could be expected to limit
convergence across studies. The comparability of the results obtained with
the proposed factor composites in two additional samples in the present
study suggests that the present results will prove replicable, although this
remains to be confirmed.
A conceptual interpretation of the replicable dimensions is provided in
Figure 1. This figure embodies the assumption that health behavior
dimensions are hierarchically organized. The broken line connecting
Substance Use Risk to Wellness Behaviors is intended to indicate that
Substance Use Risk is conceptually an element of Risk Taking, but
empirically appears to be linked to Wellness Behavior as well.
It must be emphasized that Figure 1 represents a set of hypotheses
which may be useful as a frame of reference for posing specific research
questions for subsequent studies to better explore health behaviors. For
A second reason (^) for suggesting a hierarchical organization of health behaviors is that this proposal has important implications regarding causal effects that give rise to the dimensions. From a causal perspective, behaviors covary because they share common cause(s). Thus, the two general dimensions of health behaviors presumably arise because some causal factors influence all the behaviors within, but not across, the two dimensions. Further, the general dimensions presumably contain more restricted subsets of interrelated behaviors, because additional causal factors exist which differentially affect behaviors within the two general dimensions. Verification of the prediction that differential patterns of causes are the basis for the observed dimensions is needed to demonstrate construct validity of the proposed conceptual model of health behaviors. Previous work provides (^) reason to believe the two major dimensions have differential patterns of correlation to other variables (Langlie, 1979; Feldman & Mayhew, 1984), but a detailed comparison of the four dimensions has not been made. Better definition of the behavioral scope of health behavior dimensions and delineation of antecedents of these dimensions may lead to re-evaluation of some proposed theoretical concepts in this area. The dimensions defined here are superficially consistent with some previous conceptualizations but differ in some important ways on closer examination. For example, the Wellness dimension and Traffic Risk dimensions are substantially similar to Langlie's (1977) distinction between indirect and direct risk behaviors. However, the present results suggest that both of her dimensions are specific subsets of more general dimensions which could imply very different conceptual interpretations than those proposed by Langlie (1979). Similarly, Kolbe's (1983, as cited in Green, 1984) distinction between wellness behaviors and preventive behaviors appears to be of limited empirical importance as representatives of both types of behavior appear to be elements of the Wellness Behavior dimension. In addition, his concept of "at risk" behavior might be extended to include everyday risks of accident and injury rather than referring only to illness and disease. If so, this category would require further definitional refinement to account for the presence of two empirical factors. As a general point, current conceptual models seem to emphasize the outcomes associated with health behaviors. While those outcomes are what make health behaviors important, consideration of the reasons for covariation of certain specific behaviors may provide
alternative bases for conceptualization that will enrich our understanding of these behaviors. The foregoing considerations have been suggested to illustrate that the
proposed hierarchical model for health^ behaviors^ provides^ a^ potentially useful framework for additional research. Although appropriate caution must be taken when generalizing from the samples studied to populations with different socio-demographic attributes, the hierarchical model represents a set of related^ hypotheses^ which^ can^ be^ explicated^ and^ clearly^ tested^ in future research. One key problem for future research is to improve the delineation of the subcategories of health behaviors comprising^ the^ two general categories outlined here. The second major research problem posed by the proposed hierarchical model of health^ behavior^ is^ to^ identify plausible explanations for^ the^ covariances^ of^ behaviors^ that^ give^ rise^ to the proposed dimensions of health behaviors. The hierarchical model of health behavior presented here is one possible organizing framework^ for reviewing what is^ known^ about^ health^ behaviors^ and^ their^ antecedents^ and^ for conceptualizing and measuring health behaviors when addressing these^ two general research problems. It cannot be stated too strongly that the proposed hierarchical structure^ and^ the^ labelling^ of^ health^ behavior dimensions must be taken as tentative hypotheses to be tested in^ such studies. The proposed dimensions should not be taken at this time as well-defined, empirically validated^ theoretical^ constructs.^ However,^ the payoff from additional research designed to test the hierarchical model should be a better understanding of health behavior dimensions which will provide a stronger basis for programs to improve health and well-being -- even if the model ultimately proves inappropriate.
Vickers, R.R., Jr. & Hervig, L.K. (1984). Health behaviors: Empirical
consistency and theoretical significance of subdomains. San Diego, CA:
Naval Health^ Research^ Center,^ Technical^ Report^ 84-18.
Walkey, F.H. (1986).^ Replication^ of^ factors^ in^ variations^ on^ a^ synthetic
correlation matrix. Educational^ and^ Psychological^ Measurement,^ 46,
Williams, A.F. & Wechsler, H. (1972). Interrelationship of preventive
actions in health and other areas. Health Service Reports, 10,^ 969-976.
Williams, A.F. & Wechsler, H.^ (1973).^ Dimensions^ of^ preventive^ behavior.
Journal of Consulting and Clinical Psychology, 40,^ 420-425.
Appendix A Health Behavior Checklist
1. I eat a balanced diet.
- 3. (^) II getkeep enough emergency sleep. numbers near the phone.
- I choose my spare time activities to help me relax. 5. I take chances when crossing the street, etc. 6. I have a first (^) aid kit in my home. 7. I destroy old or unused medicines. 8. I see a doctor for regular checkups. 9. I pray or live by principles of religion. 10. I avoid getting chilled. 11. I watch my weight.
- I carefully obey traffic rules so I won't have accidents. 13. I watch for possible signs of major health problems (e.g., cancer,
hypertension, heart disease).
- I exercise to stay healthy. 15. I^ cross^ the^ street^ against^ the^ stop^ light. 16. I avoid high crime areas. 17. I don't smoke. 18. I don't take chemical substances which might injure my health (e.g.
food additives, drugs, stimulants).
19. I check the condition of electrical appliances, the car, etc. to avoid
accidents.
- I stay away from places where I might be exposed to germs.
- I fix broken things around^ my^ home^ right^ away.
- I see a dentist for regular checkups. 23. I limit my intake of foods like coffee, sugar, fats, etc.
- I avoid over-the-counter medicines. 25. I take vitamins. 26. I do not drink alcohol. 27. I wear a seat belt when in a car. 28. I cross busy streets in the middle of the block.
- I avoid areas with high pollution. 30. I discuss health with friends, neighbors, and relatives. 31. I gather information on things that affect my health by watching
television and reading books, newspapers, or magazine articles.
32. I use dental floss regularly. 33. I speed while driving.
- I brush my teeth regularly. 35. I take health food supplements (e.g. protein additives, wheat germ,
bran, lecithin).
36. I learn first aid techniques.
4 37. I get shots to prevent illness.
38. I take more chances doing things than the average person.
- I drink after driving.
- I engage in activities or hobbies where accidents are possible (e.g.
motorcycle riding, skiing, using power tools, sky or skin diving, hang-gliding, etc.).
