Concordance of parental and adolescent health behaviors

Pergamon

0277~9536(93)EOO52-G

CONCORDANCE

OF PARENTAL AND ADOLESCENT HEALTH BEHAVIORS INGEBORG

*Research

Sm. S-i. Med. Vol. 38, No. 9, pp. 1299-I 305, 1994 Copyright 0 1994 Elscvier Science Ltd Printed in Great Britain. All rights reserved 02179536194 $6.00 + 0.00

‘Institute of Community Center for Health Promotion,

Rossowl and JOSTEIN RISE* Dentistry University of Oslo, Norway Environment and Lifestyle, University

and of Bergen,

Norway

Abstract-This paper reports upon an empirical study of health behaviors in adolescents and their parents. The study aimed at assessing: effects of parental health behaviors on that of their adolescent child; whether mother’s and father’s health behaviors have additive effects on the respective health behaviors of their child; and whether eventual effects of parental health behaviors decrease with increasing age of the child. The data stemmed from the Norwegian national Health Survey in 1985 and comprised separate interviews with two parents and an adolescent child in 337 families. Results from logistic regression analyses showed

that the strongest association found between parental and adolescent health behaviors was for fat intake, and the probability of having a low fat intake was 5 times higher if the mother had a low fat intake than if she did not. With the exception of mother’s frequency of exercise, all other parental health behaviors were positively and statistically significantly associated with the corresponding health behavior of their adolescent child. Parental fat intake, smoking behavior and alcohol consumption appeared to have additive effects on the corresponding behaviors of their children. No statistically significant interaction between any of the parental health behaviors and age of the adolescent was found. Hence, the effect of parental health behaviors on that of their adolescent child does not seem to decrease with increasing age of the adolescent. The results are discussed with reference to the functions of modeling. Key words-health

behaviors,

parents,

adolescents,

modeling

INTRODUCTION During childhood and adolescence most individuals develop and establish a lifestyle within which a range of health behaviors are embedded. With few exceptions the family represents the primary influencing social institution and framework for the performance of health behaviors [l]. Yet, much of the research on family and health has failed to conceptualize the family qua family and thereby not considered the family as a total unit [2]. Numerous empirical studies have demonstrated significant associations between parental and adolescent health behaviors [3,4]. However, a major shortcoming of most studies in this area is that parental health behaviors are often reported by the adolescents and not by the parents themselves [S]. Thus, the correlation between ratings of own behavior and ratings of similar others’ behavior may be inflated due to motivated social projection, and in this way the raters psychologically make others similar to themselves and thereby attain a sense of selfvalidation, a phenomenon known as the false consensus effect [6]. This was aptly demonstrated by Aas er al. [7] who observed that the correlation between the drinking behavior of parents and children was significantly higher when based upon reports by the adolescents themselves, compared to when each respondent reported their own drinking behavior. In a methodological study on young adults’ reports of

parental drinking O’Malley et al. [8] found that half of the respondents either over-reported or underreported the frequency with which their parents consumed alcohol. This means that the direct parental modeling effect may have been over-estimated. A related issue is that most of the studies have focused on the association between adolescent health behavior and that of one of the parents [3]. Since parental behaviors may not necessarily be congruous, the measure of one parent’s behavior may not be a valid measure of the parental influence in terms of modelling effect. Thus, assessment of parental influence on adolescent behavior should therefore be sought by applying measures on both parents’ behaviors. Furthermore, it may be of interest to assess the relative importance of the two parents as models for their children’s health behaviors, and to assess whether the parental influential effects are of an additive or interactional nature. For example, Nolte et al. [9] observed that there was an additive effect of parents’ smoking behaviors on that of their adolescent child, so that the probability of an adolescent’s daily smoking was twice as high if both parents smoked and half as high if neither parents smoked, compared to if either parent smoked. There seems, however, to be few empirical studies on additive effects of parental health behaviors on that of their children. A third issue of great empirical, thee, -tical as well as practical concern is whether the parental influence

1299

1300

INGEBORG Rossow and JOSTEINRISE

on adolescents’ health behavior persist as they move from adolescence into adulthood, or whether other influential socializing agents take over [5]. In their comprehensive longitudinal study Lau et al. observed that the total effect of parental influence was roughly equivalent to peer influence on their health behaviors after 3 years of college [5]. This was the case for drinking, eating and exercise behaviors, while parental influence was much greater for use of seat belts

[51. In this study we wanted to empirically elucidate three issues which correspond to those elaborated above. Firstly, to study modeling behavior in terms of assessing the effect of parental behavior upon that of their adolescent children. Secondly, to assess whether mother’s and father’s health behaviors have additive effects on that of their adolescent child. Thirdly, the hypothesis that the parental influence becomes weaker as the adolescents grow older, will be tested in terms of an interaction between age and parental behavior on the adolescents’ health behavior; thus the effect of parental behavior should decrease with increasing age. MATERIAL

AND METHODS

The data of the present study stemmed from the Norwegian National Health Survey of 1985. The sample was nationwide and drawn in a two-stage random manner. The sampling unit was the household, and all eligible household members were interviewed separately by trained personnel [lo]. Interviews were obtained from 10,576 persons in 4333 households. The original data file was organized with each household member as subject unit. This file was reorganized so that the household became the unit of analysis [I 11. Only household members of 16 years or above were interviewed about tobacco smoking, alcohol consumption, exercise and fat intake. Thus, the study material comprised a national and presumably representative sample of 337 families with both parents present and one child in the age group 1620 years living at home. In Norway the majority of children grow up in two-parent families, and in 1984 aprox. 82% of all families with children were two parent families [ 121. Due to internal missing observations on one or several variables in 55 families, multivariate analyses were performed for 272 families with complete responses. Health behavior variables This study focused on four health behaviors; smoking, alcohol consumption, fat intake, and exercise. These are behaviors with consequences for high prevalent chronic diseases, they are frequently performed in both the adolescent and the adult population, and they are of primary concern in the health educational activities within the health services. In

Norway the legal ages for purchase of tobacco and alcohol are 15 years and 18 years, respectively. For each family member the variable on one health behavior was based on several questions related to this particular behavior. Thus smoking behavior was based on the following questions: “Do you smoke?“, “If yes, do you smoke daily or occasionally?“, and “If no, have you ever smoked before?“. Alcohol consumption was based on the questions: “Have you ever drunk alcohol during the past year?“. “If yes, how often do you drink alcohol?“. Exercise was measured by the questions: “Do you engage in any kind of exercise or sports?” (exercise and sports were exemplified), and “If yes, how often do you exercise?“. The variable on fatty dietary components was constructed from the questions: “What kind of milk do you usually drink or use in food preparation?” and “What kinds of butter or margarine do you usually

use?“.

In order to enable the variables to enter logistic regression analyses, the variables were dichotomized. Number of valid observations, category values and cut off point for dichotomizations are given in Table I. Control variables As both age and gender are reported to be of major relevance for occurrence of health behaviors in adolescence [4, 131, gender and an interval level variable on age of the adolescents were included as control variables and possible effect modifiers in the multivariate analyses. Living in an urban or rural area was included in the models of alcohol consumption as it improved the fit of these models significantly. Interaction

terms

To enable statistical test of any interaction between the effect of parental health behaviors and age of the adolescent, multiplicative interaction terms of the dichotomies of the parents’ health behaviors and the age of the adolescent were computed. Statistical

analyses

Logistic regression analysis was preferred to linear regression analysis for two reasons: firstly, the logistic regression coefficients are more directly interpretable and may hence be more useful for practical purposes than are the linear regression coefficients of noninterval scaled variables. Secondly, it was assumed that the effect of parental health behaviors may take on a more sigmoid than straight linear form of association. Logistic regression analyses were performed by use of logit model. The logistic regression coefficients were used to calculate the probability of an adolescent’s engagement in a specific health behavior based on the logit for the specified model. Level of statistical significance for each partial effect was calculated by use of Wald test. Pearson’s Goodness of Fit was estimated to evaluate fit of the logistic model [ 141.

1301

Parental and adolescent health behaviors

adolescent but also dependent on the parents’ alcohol consumption. Thus, the probability of weekly alcohol consumption was more than twice as high if the mother reported weekly alcohol consumption than if she reported less frequent alcohol consumption, and a similar effect was found with respect to father’s alcohol consumption. As can be seen from Fig. 1 the mother’s and father’s frequency of alcohol consumption was highly correlated, and the dummy variables on number of parents consuming alcohol weekly were not statistically significantly associated with the adolescent’s alcohol consumption. However, the additive index on parental alcohol consumption was statistically significantly associated with the adolescent’s alcohol consumption, so that if both parents reported infrequent alcohol consumption (less than once a week), the probability of weekly alcohol consumption was lower than if one of the parents reported weekly alcohol consumption, and the probability of weekly alcohol consumption was highest if both parents reported weekly alcohol consumption (Table 3). The adjusted odds ratios of parental effects on fat intake was higher than for any other health behaviors. Thus, it appeared that the parents’ influential effect on their children’s fat intake was stronger than on any of the other observed health

RESULTS Bivariate associations between parents’ and child’s health behaviors and between the two parents’ health behaviors were calculated by Spearman rank correlation coefficients and are illustrated in Fig. 1. The covariation in health behaviors between pairs of family members appeared to be larger among spouses than among pairs of parent and child for both health compromising behaviors, although most significantly for alcohol consumption. The estimated multivariate models of adolescent tobacco smoking, alcohol consumption, fat intake, and exercise with the respective health behaviors of the mothers and fathers are given in Tables 2-5, respectively. The results showed that the probability of daily smoking among the adolescents increased with age, but it was also highly dependent on each of the parents’ smoking habits as well as on both parents’ smoking habits on an additive basis. Thus, the probability of daily smoking among the adolescents was lowest if none of the parents were daily smokers, and it was higher if both parents reported daily smoking than if one parent was a daily smoker. The probability of alcohol consumption at least once a week was dependent on age and gender of the

Table 1. Percentaae

distributions

of studv variables.

Codes for cateeories

are given in oarentheses

Daily (I) 39.3

Previous (2) 21.4

Never (3) 39.3

Father’s smoking n = 323

Daily (1) 44.9

Previous (2) 31.2

Never (3) 23.9

Adolescent’s n = 323

Daily (I) 24.3

Previous (2) 5.0

Never (3) 70.7

1 x month (2) 13.9

< I x week(3) 17.6

> I x week (4) 23.5

< 1 x month(l) 24.3

I x month (2) 15.0

< I x week(3) 19.9

> I x week (4) 40.8

< I x month (I) 53. I

I x month

(2)

19.7

< I x week (3) 15.3

>I xweek(4) II.9

No fat in milk or butter (I) 41.8

Fat in milk or butter (2) 37.5

Fat in milk and butter (3) 20.7

Father’s fat intake n = 302

No fat in milk or butter (I) 29.8

Fat in milk or butter (2) 45.7

Fat in milk and butter (3) 24.5

Adolescent’s n = 336

No fat in milk or butter (1) 38. I

Fat in milk or butter (2) 45.5

Fat in milk and butter (3) 16.4

Never (I) 46.1

< I week (2) 7.2

1-2 x week (3) 25.7

34

x week (4) II.9

5-7 x week (5) 9.1

Father’s exercise n = 297

Never (I) 42.4


l-2 x week (3) 29.5

34

x week (4) 9.1

S7 x week (5) 9.8

Adolescent’s n = 320

Never (1) 24.7


l-2 x week (3) 26.6

34

x week (4) 26.9

5-7 x week (5) 19.0

Mother’s n = 323

Mother’s n = 324 Father’s n =3Ol

smoking

alcohol cons

Mother’s n = 319

<

I x month

(I)

45.1 alcohol cons

Adolescent’s n = 320

Mother’s n = 323

smoking

alcohol cons

fat intake

fat intake

exercise

exercise

Gender n = 337

Boys(l) 53. I

Age ” = 337

I6 24.0

week(2) 9.1

Girls (2) 46.9 17 25.5

18 16.0

I9 16.6

20 7.7

INGEBORG Rossow and JOSTEIN RISE

1302

DISCUSSION

Smoking Alcohol

Mother-father

Mother-child

Father-child

Fig. 1. Spearman’s rank correlation coefficients for health behaviors of pairs of family members.

behaviors. The probability of having a low fat intake was twice as high if both of the parents had a low fat intake than if one of the parents had a low fat intake, and the probability of having a low fat intake was 3.5 times higher if one of the parents had a low fat intake than if none of them had a low fat intake. Mother’s exercise was the only parental health behavior that was not statistically significantly associated with that of her adolescent child, although the 95% confidence interval of the point estimate for the adjusted odds ratio scarcely comprised the value 1.0. It follows that the statistically significant effect of one parents’ exercise on that of the adolescent could be ascribed to the father’s frequency of exercise, and furthermore that th+e was no significant effect observed of both parents’ engagement in exercise as compared to one parents’ exercise.

This study is based on a relatively unique data set for assessment of parental and adolescent health behaviors. The data were nationwide and stemmed from a large, representative sample of the Norwegian non-institutionalized population, they comprised variables on health behaviors of both of the parents, and they comprised variables on a variety of health behaviors. Furthermore, the responses on health behaviors are given from each family member separately. Hence, observed associations between health behaviors of parents and adolescents are less likely to be due to systematic measurement errors as can be expected if respondents report health behaviors of other family members. The present study provides empirical support for the assumption that health behaviors of parents and their adolescent children are strongly associated. Sallis and Nader [I] argue that family influences on health behaviors are multi-directional; in terms of reciprocal influences among health behaviors of family members. The cross-sectional design of this study limits the interpretation of the results with respect to causal inferences. It may however, be advocated that parents to a larger degree than their adolescent children possess the means to control both the physical and social antecedents of health behaviors within the family. This is because the parents are more likely to purchase foods and beverages and to provide equipment for sports and leisure time activities, and they are more likely to give and pursue rules and family schedules which may serve as social antecedents of the health behaviors. Thus, it may be argued that the observed associations of parents’ and children’s health behaviors mainly can be interpreted as parental influential effects. The results of the present study indicate that there is a significant association between parental and adolescent behaviors in a variety of areas with known consequences for health. With the exception of mother’s exercise all parental health behaviors were significantly associated with that of their adolescent child. Lau et al. [5] also reported no significant association between parental (i.e. mother’s) and adolescent’s engagement in physical activities. The apparently strongest association between parental and adolescent health behaviors was found with respect to intake of fat. Laskarzewski et al. [ 151 also found that parental and child intake of fat and calories were significantly correlated. Considering the varying strengths in associations between parental and adolescent behaviors, it may be anticipated that these variations reflect the relative degree to which the different kinds of behavior are attached to the family as a social arena for the performance of the behavior. Hence, intake of fat is often performed at common meals within the family, whereas exercise and sports activities often are performed in other social contexts.

Parental

Pearson’s

Goodness

and

adolescent

Table 2. Probability of fit = 0.706

Independent variables

health

behaviors

1303

of daily smoking

Regr. coeff.

SE

Wald test

Adj. odds ratio

95% CI

Mother’s smoking Age of adolescent

1.13 0.31

0.29 0.11

3.93’ 2.80.

3.10

1.73, 5.53

Interaction term mother’s smoking and age

0.19

0.22

0.86

Pearson’s Goodness

of fit = 0.654

Independent variables

Regr. coeff.

SE

Wald test

Adj. odds ratio

95% CI

Father’s smoking Age of adolescent

0.89 0.3 I

0.30 0.11

3.00’ 2.75’

2.44

1.34, 4.45

Adj. odds ratio

95% CI

2.05 1~2.73

1.01, 4.2 l:(1.25, 5.9)

Interaction term father’s smoking and aee Pearson’s

Goodness

-0.21

0.23

-0.91

of fit = 0.738

Independent variables Both parents smoke No parents smoke Ace of adolescent

Regr. coeK

SE

Wald test

0.72 - I.01 0.36

0.36 0.39 0.12

I .98’ -2.55* 2.93’

*P < 0.05

Modeling is assumed to be a powerful and complex mechanism for family influence [l]. The parent-child associations in health behaviors can be explained by the four functions of modeling [16]. The children’s behaviors can be learned, maintained or changed by observing the parents perform an action, by observ-

Table 3. Probability Pearson’s Goodness of fit = 0.51 I Independent variables Mother’s alcohol consumption Age of adolescent Gender Urban dwelling Interaction term mother’s alcohol consumption and age Pearson’s Goodness

Father’s alcohol consumption Age of adolescent Gender Urban dwelling Interaction term father’s alcohol consumption and age Goodness

Independent variables Additive index parents’ ale. cons. Age of adolescent Gender ‘P < 0.05.

of weekly alcohol consumption

Regr. coeff.

SE

Wald test

Adj. odds ratio

95% CI

0.83 0.55 -1.18 0.07

0.40 0.15 0.43 0.39

2.07. 3.64+ -2.74’ 0.19

2.29

1.06, 5.10

0.79

0.38

2.04’

Regr. coeff.

SE

Wald test

Adj. odds ratio

95% CI

0.85 0.64 - 1.29 -0.20

0.41 0.16 0.46 0.41

2.07’ 3.95’ -2.81’ -0.49

2.35

1.05, 5.21

0.44

0.33

1.34

Regr. coeff.

SE

Wald test

Adj. odds ratio

95% CI

0.57 0.60 - I .42

0.24 0.16 0.42

2.38* 3.60’ -2.93’

1.77

1.10, 2.87

of fit = 0.654

Independent variables

Pearson’s

ing the consequences of a parental behavior, by responding to parental behaviors as social prompts, and by setting own standards for conduct in line with the parental behaviors. According to Sallis and Nader [l] modeling can act as an antecedent of the behavior, modeling can teach the actual behavior,

of fit = 0.504

INGEBORG

1304 Pearson’s

Goodness

Independent variables Mother’s Gender

Pearson’s

Goodness

Independent variables Father’s Gender

Table 4. Probability of fit = 0.712 Regr. c&l-.

SE

I.61 1.12

0.26 0.26

fat intake

Interaction term mother’s fat intake

-0.12

Wald test 6.13’ 4.27’

Adj. odds ratio

95% CI

5.00

2.91, 8.41

Adj. odds ratio

95% CI

4.44

2.56, 7.85

-0.64

SE

I .49 1.00

0.28 0.26

5.40. 3.86*

0.15

0.23

0.65

SE

Wald test

Adj. odds ratio

95% CI

0.36 0.31 0.27

2.22* - 3.96* 4.09’

2.21 3.48

1.10, 4.45 I : (I .87, 6.46)

Wald test

of fit = 0.695

Independent variables Both low fat intake Not low fat Intake Gender

of low fat intake

Reg. co&

Interaction term father’s fat intake and age Goodness

0.19

JOSTEINRISE

of fit = 0.249

fat intake

Pearson’s

Rossow and

0.79 - 1.25 I.13

*P < 0.05

and modeling

can influence

the consequences

of the

behavior. Antecedents are events which cue behaviors such as signs, reminders, and rules. Parental behaviors may imply provision of physical cues for adolescent behaviors such as purchase and presence of cigarettes or alcohol in the home. Parents may also provide social cues for their children’s behaviors in line with their own behaviors, for instance expression of food pref-

Pearson’s

Goodness

Table 5. Probability of fit = 0.107

erences, persuasion related to physical activity or proscriptions against tobacco use. Parents may also provide social support in line with own behaviors. Social support can be helpful as a consequence of a behavior, and parents may support their children’s behavior, for instance by complimenting on low fat intake. Thus, social support as a consequence of behavior may also serve as a social antecedent [I]. We observed significant additive effects of parental

of weekly exercise

Regr. co&f.

SE

Wald test

Adj. odds ratio

95% CI

0.53 -0.28 0.68

0.27 0.10 0.27

1.95 - 2.72’ 2.50.

I .70

0.99, 2.93

0.25

0.21

1.21

Regr. co&.

SE

Wald test

Adj. odds ratio

95% CI

Father’s exercise Age of adolescent Gender

0.83 -0.30 0.67

0.29 0.1 I 0.29

2.85* -2.79* 2.31’

2.30

1.29. 4.10

Interactmn term father’s fat intake and see

0.03

0.02

1.71

SE

Wald test

Adj. odds ratio

95% CI

0.40 0.34 0.11 0.30

0.59 - 2.65’ -2.66’ 2.51.

I .27 I :2.49

0.57, 2.82 l:(l.26, 4.91)

Independent variables Mother’s exercise Age of adolescent Gender Interactvan term mother’s fat intake and age Pearson‘s

Goodness

Independent variables

Pearson’s

Goodness

Independent variables Both exercise No parents exercise Age of adolescent Ginder ‘P < 0.05

of fit = 0.564

of fit = 0.318 Regr. coeK. 0.24 -0.91 -0.30 0.76

Parental and adolescent health behaviors health behaviors on that of their child with respect to smoking, alcohol consumption and fat intake. Assuming that parents provide social cues and social support in line with own behaviors, modeling would be more effective if the parents were consistent in their behaviors than if they were discordant. The observed additive effects can also be interpreted by focusing on modeling as setting cognitive standards for self-regulation. Thus, we will assume that by observing consistent behaviors of both parents the probability of setting the same standards for own conduct in similar situations seems much higher than if the behaviors of the parents are divergent. Nader et al. [17] stated that the influence of the parental health behaviors in childhood may be reinforced or modified by the influence of peers in adolescence. During adolescence young people take part in the process of separating from parents by developing a sense of autonomy and independence. This process might involve a reorientation from parental influences towards peer influences and a priority change in values and models. Consequently a decrease in parental influences on health behaviors with increasing age during adolescence could be expected. The results of the present study do not serve to confirm such a hypothesis. On the contrary, the interaction term between mother’s alcohol consumption and age of the adolescent was statistically significant and positive. Thus, adolescents tend to adapt their mother’s drinking habits as they get older, either as a result of reaching the legal age for purchase of alcohol or as a response to parental influence, or both. Thus, the results of this study indicate that, at least with respect to the health behaviors we have studied, parents may be strong and lasting models for their children also during adolescence, and the conditions for parental influence may not change significantly from the age of 1620. However, an additional explanation is that parents may contribute to selection of peers with a lifestyle in line with their own. In conclusion, the results of the present study indicate that parents provide models for their adolescent offspring in a range of health behaviors. In particular this was the case for fat intake but also for smoking and consumption of alcohol, and in addition this modeling influence persisted at least until the age of 20. These results corroborate those which have been presented recently on a rather selected material, but runs counter to most previous studies. The conclusion is also based upon a data set with presumably unbiased reporting of engagement in health behaviors. The results may have important implications for parental health education; thus parents should be made aware of the fact that they not only exert a profound influence as models for their adolescent offspring in the early years of socialization, but

1305

that they continue to influence their children longer than has been hitherto expected.

far

REFERENCES 1.

Sallis J. F. and Nader P. R. Family determinants of health behaviors. In Health Behavior. Emerging Research Perspectives (Edited by Gochman D. S.), pp. 107-124. Plenum Press, New York, 1988. 3 Gochman D. S. Family determinants. Health Behavior. Emerging Research P&specfives (Edited by Gochman D. S.), pp. 103-106. Plenum Press, New York, 1988. 3. Baranowski T. and Nader P. R. Family health behavior. I.

4. 5.

6. 7.

8. 9. 10.

11.

12.

13.

14. IS.

16.

17.

In Health, Illness and Families. A Life-span Perspective [Edited bv Turk D. C. and Kerns R. D.).I on. . . 51-80. Wiley, New York, 1985. Norman R. M. G. The Nature and Correlates of Health Behavior. Health Promotion Studies Series No. 2. Health Promotion Directorate, Ottawa, 1986. Lau R. R., Quadrel M. J. and Hartman K. A. Development and change of young adults’ preventive health beliefs and behavior: Influence from parents and peers. .I. Hlth Sot. Behav. 31. 240, 1990. Marks G. and Miller N. Ten years of research on the false-consensus effect: An empirical and theoretical review. Psychol. Bull. 102, 72, 1987. Aas H., Jakobsen R. and Andersen N. Parents’ selfreported alcohol use compared to 13-year-olds’ perceptions: the Norwegian longitudinal health behavior study. Submitted. O’Malley S. S., Carey K. B. and Maisto S. A. Validity of young adults’ reports of parental drinking practices. J. Stud. Alcohol 47, 433, 1986. Nolte A. E., Smith B. J. and O’Rourke T. The relative importance of parental attitudes and behavior upon youth smoking behavior. J. School Hlfh 53, 234, 1983. Central Bureau of Statistics. Health Survey 1985. pp. 15-32. Norges offisielle statistikk, Kongsvinger, 1987. Rossow I. Reorganization of System File with Individual as Unit of Analysis to Sysrem File with Household as Unit of Analysis. Institute of Community Dentistry, University of Oslo, Report Series No. 2, 1990. Central Bureau of Statistics. Statistical Yearbook 1986, p. 49. Norges offisielle statistikk, Kongsvinger, 1988. Aarij L. E., Wold B., Kannas L. and Rimpela M. Health behavior in Schoolchildren. A WHO cross-national survey. A presentation of philosophy, methods and selected results of the first survey. Hlth Promotion 1, 17, 1986. Hosmer D. W. and Lemeshow S. Applied Logistic Regression. Wiley, New York, 1989. Laskarzewski P., Morrison J. A., Khoury P., Kelly K., Glatfelter L., Larsen R. and Glueck C. J. Parentchild nutrient intake interrelationships in school children ages 6 to 19: the Princeton School District Study. Am. J. clin. Nufril. 33, 2350, 1980. Rosenthal T. L. and Bandura A. Psychological modeling: Theory and practice. In Handbook of Psychotherapy and Behavior Change, 2nd edn, pp. 621658 (Edited by Garfield S. L. and Bergin A. E.). Wiley, New York, 1979. Nader P. R., Perry C., Maccoby N., Solomon D., Killen J., Telch M. and- Alexander J. K. Adolescent perceptions of family health behavior: A tenth grade educational activity to increase family awareness of a community cardiovascular risk reduction program. J. School Hlth 52, 372, 1982.