Correlates of Precoital Behaviors,Intentions,and Sexual Initiation Among Thai Adolescents

Abstract

This paper examines the risk and protective factors associated with sexual behaviors among Thai youth ages 13-14 (N=420) living in Bangkok, Thailand. Cross-sectional data were collected using a random sample of households methodology. Three outcomes were assessed: (1) intention to engage in sexual intercourse, (2) pre-coital behaviors, and (3) sexual initiation. Bivariate analysis indicated that parental disapproval of sex, exposure to pornographic media, refusal self-efficacy and having a boyfriend/girlfriend had the strongest relationships with all three outcomes. Multivariate analyses found that parental disapproval of sex and exposure to pornographic media (internet or TV) were significantly associated with all three outcomes. Having a boyfriend or girlfriend was associated with pre-coital behaviors and intentions and sexual refusal self efficacy was correlated with pre-coital behaviors only. The potential competing influences of parent disapproval and exposure to pornographic media on adolescent sexual behaviors should be considered when adapting HIV prevention interventions for Thai youth.

Keywords

ANOVA family relations/processes logistic regression middle school multiple regression sexual behavior (including Pregnancy)STD prevention

Introduction

Thailand has experienced a remarkable reduction in HIV seroprevalence due to government supported interventions targeting the commercial sex industry and its patrons (Hanenberg, Rojanapithayakorn, Kunasol, & Sokal, 1994; Mason et al., 1998; Nelson et al, 2002; Rojanapithayakorn & Hanenberg, 1996). As HIV-prevention efforts have taken hold in Thailand, there has been a gradual shift from males engaging in sex with commercial sex workers to sex partnerships with unmarried females which are now viewed as more acceptable and less risky (Liu et al., 2006). Contributing to these changes are dramatic shifts in the Thai economy and culture (Vuttanont, Greenhalgh, Griffin, & Boynton, 2006) where Western views of sexuality have come in conflict with traditional Thai values.

Adolescent populations are particularly vulnerable to this “dual value system” (Vuttanont et al., p. 2072) with expectations from their traditional Thai culture that values modesty and reverence to elders clashing with a desire to experience dating and engage in sexual behaviors (Vuttanont et al., 2006). Although population based data on younger Thai populations is limited, cross-sectional studies of older vocational students have documented earlier age of sexual debut when comparing 15 to 17 year olds to 20 to 21 year olds (Liu et al., 2006), lack of condom use (Allen et al., 2003; Thato, Charron-Prochownik, Dorn, Albrecht, & Stone, 2003) and significant increases in number of sex partnerships comparing cross-sectional surveys conducted in 1991 and 2002 (Whitehead et al., 2008). These risk behaviors have heightened adolescent risk of HIV, other STIs, and pregnancy (Allen et al., 2003; Whitehead et al., 2008).

The purpose of this article is to examine the risk and protective factors associated with the sexual risk behaviors among Thai early adolescents of whom there has been little published using a theoretically driven approach to variable selection. In this article, we examine the correlates of precoital behaviors, intentions to have sex, and sexual initiation among Thai youth ages 13 to 14 who participated in a household survey of 420 adolescents in Bangkok, Thailand.

The Risk and Protective Framework (Hawkins, Catalano & Miller, 1992; Jessor, 1991) is the theoretical framework of our study and is used to guide variable selection. The framework takes into consideration the following: (a) individual, (b) peer, (c) family, (d) school, and (e) community-level factors that can contribute to or protect adolescents from high-risk behaviors including sexual risk taking (Blum & Mmari, 2005). In this exploratory analysis, we assess the impact of these risk and protective factors on precoital behaviors, intentions to have sex, and ever having sexual intercourse. Described below are the risk and protective factors under each of these five domains that are associated with risky sexual behaviors in either the Western and/or Thai literature and are therefore included in this exploratory analysis.

At the individual level, we include key psychosocial variables (Noar & Zimmerman, 2005) such as sexual refusal self-efficacy, attitudes about delaying sex, and perceived consequences of HIV infection that have been found to be negatively associated with sexual risk taking in the United States (for review see, Sheeran, Abraham, & Orbell, 1999) and in Thailand (Guruge et al., 2004; Liu et al., 2006; Rasamimari, Dancy, Talashek, & Park, 2007; Thato et al., 2003). We also include characteristics innate to the individual such as interdependence, independence, and the practice of Buddhism. A U.S. study found that women with a strong sense of familial interdependency had a lower risk of pregnancy than those with lower levels of familial interdependence (Mendez & Majumdar, 2005). Among Thai adolescents, there persists a deep obligation to the family unit (Spielmann, 1994) and also a desire to embrace more Western concepts of independence (Vuttanont et al., 2006). We therefore include interdependence as a protective factor and independence as a potential risk factor for sexual risk-taking behaviors. In Thailand, Buddhism is interwoven into many elements of Thai culture. The practice of Buddhism has been viewed as protective for sexual risk behaviors (Ubillos, Paez, & Gonzalez, 2000) and is included as a potential protective factor. Last, we include male gender (Rasamimari et al., 2007) and older age (Guruge, Isaranurug, Nanthamongkolchai, & Charupoonphol, 2004) as important risk factors.

At the peer level, we include peer norms (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) that have been found to be associated with sexual risk taking in U.S. studies (Noar & Zimmerman, 2005). We also include having a boyfriend/girlfriend which has been associated with early sexual intercourse in the United States (Blum, Buehring, & Rinehart, 2000; Halpern, Joyner, Udry, & Suchindran, 2000; Thorton, 1990).

At the family level, we include parental characteristics found to be protective for sexual risk taking in the United States and Thailand including parental monitoring (Harris et al., 2006; Longmore, Manning, & Giordano, 2001; Resnick et al., 1997), parental closeness (Miller, Benson, & Galbraith, 2001; Paul, Fitzjohn, Herbison, & Dickson, 2000; Resnick et al., 1997); parental disapproval about sex (Resnick et al., 1997), and parent–child communication (Allen et al., 2003; Hutchinson, 2002; Romer et al.,1999).

At the school level, we include school attachment and self-reported academic performance both of which have been found to be associated with delay in sexual debut (Billy, Brewster, & Grady, 1994; Lammers, Ireland, Resnick, & Blum, 2000; Raine et al., 1999; Resnick et al., 1997).

Last, at the community/societal level, we include exposure to pornographic media on the internet or television and perception of neighborhood disorganization, as potential risk factors. In the Thai literature, exposure to sexualized media has been correlated with sexual risk behaviors and attitudes (Guruge et al., 2004). Thai studies have also shown that adolescents’ perceptions of greater neighborhood disorganization are related to higher levels of minor and serious delinquency (Byrnes, Miller, Cupp, Zimmerman, & Chookhare, in press).

Method

Sampling Procedures

Using a cross-sectional design, families (N = 420) were randomly and proportionally selected from seven districts of Bangkok, Thailand. Sampling was based on the former Bangkok Metropolitan Administration (BMA) that divides districts into three zones (the inner, middle, and outer zones). There are 10 districts in the inner zone (14.7% of the Bangkok population), 27 districts in the middle zone (56.1% of the city’s population), and 13 districts in the outer zone (29.2% of the city’s population). The selection of districts was based on a sampling fraction of each of these zones that was proportional to the total Bangkok population using the probability proportional to size (PPS) sampling method (with case multiplication technique). Using district-level population data from 2006, one district was selected from the inner zone, four from the middle zone, and two from the outer zone (n = seven districts). Within each district, 35 blocks were selected using the PPS method of selection, resulting in a total of 245 blocks (35 blocks × 7 districts). There were slightly more than 4,000 households per district or 30,471 households across all seven districts. Surveyors made contact with 28,174 of the 30,471 households (82.6% of the total).

The Mahidol University researchers in conjunction with Thailand’s National Statistical Office (NSO; 2000) conducted a household census and identified a total of 957 households (3.4%) determined to have a 13- and/or a 14-year-old present. Of the total, 79 (8.3%) refused to be interviewed and 116 households (12.1%) were unwilling to participate in the study, leaving 762 families who agreed to participate (79.6% of all households with 13- and/or 14-year-olds). Of the 762 families, 66 families were randomly sampled in each of the seven districts, yielding 60 completed adolescent surveys and 60 completed parent interviews from each district for a total sample of 420 completed pairs of adolescent and parent interviews across all seven districts (60 × 7 = 420). For this analysis, only the adolescent survey data, not the parent survey data, were used (n = 420), yielding 85% power to detect a standardized multiple regression coefficient of .15 making standard assumptions about a Type II error (α_2-tails = .05).

Adolescent survey data were collected using audio-computer-assisted self-interviews (ACASI) on a laptop computer with one adolescent per family. Face to face interviews were completed with one mother per family. Formative interviews indicated that parents were less comfortable with computers, therefore, in-person interviews were conducted. Because only the adolescent data were used for this analysis, the different data-collection strategies did not impact our findings.

Sample Characteristics

Adolescents in our sample (N = 420) were on average 13.45 years of age (SD = .50), 50.5% were female, and represented mostly ethnic Thais (91.2%; data not shown). Few adolescents reported ever engaging in sexual intercourse (4.5% female and 3% males).

Measures

To ensure that survey measures were equivalent (i.e., measuring the same construct when comparing the U.S.-based measures to the same measures when adapted and translated in Thai), a number of steps were taken. Interview instruments were adapted in partnership with the Thai and U.S. research teams. Suggestions for changes were made to ensure they measured the same underlying domain of interest. A Thai research member translated the survey into Thai and a second Thai researcher back-translated into English to reduce the possibility of a biased assessment. Back-translated versions were reviewed by the entire research team to assess whether the original intent of the questions were communicated. Last, Thai parents and teens, not involved in the primary study, provided feedback on the instrument to assess wording clarity and interpretation.

Dependent variables

Three dependent variables were derived from the adolescent self-report cross-sectional surveys. The first dependent variable was intention to have sex. Respondents were asked, “How likely do you think it is that you will have sexual intercourse by the time you finish high school?” Responses used a 4-point scale ranging from 1 = very sure I won’t to 4 = very sure I will. The second measure, an index of precoital behaviors, was assessed with a six-item latent index, adapted from a similar measure (Martino, Elliott, Collins, Kanouse, & Berry, 2008). Items inquired about kissing, “French-kissing or tongue kissing,” “touching a girl’s breast” (for boys) or “someone having touched your chest or breasts” (for girls), touching someone else’s “private parts below the waist,” and someone “touching your private parts below the waist.” Responses used a 4-point ordinal scale where 1 = no, never, 2 = yes, once, 3 = yes, a few times, 4 = yes, many times. The scale was found to have good reliability in our Thai sample with Cronbach’s alpha of 0.87 for males and 0.86 for females. The third and final dependent variable was having experienced sexual intercourse. Respondents were asked, “Have you ever had sexual intercourse (By this we mean ‘going all the way’ or ‘doing it’)” measured as a single item dichotomous variable (yes vs. no).

Independent variables

As seen in Table 1, we describe the multi-item constructs that fall into one of the five categories of the Risk and Protective Framework. All multi-item constructs has response categories ranging from 1 to 4 with higher scores tending to represent greater prosocial behaviors, attitudes, or norms (Table 1). These constructs include the following at the level of the individual: Sexual refusal self-efficacy (Donohew et al., 2000), sexual attitudes (Floyd, 2009), and perceived HIV severity (Witte, Girma, & Girgre, 2002), level of interdependence and independence (Singelis, 1994), and the practice of Buddhism. At the peer level, we included peer norms about sexual behavior (Floyd, 2009) and having a boyfriend or girlfriend (yes vs. no). Variables at the family level include parental monitoring (Capaldi & Patterson, 1989), parental closeness (Harris et al., 2003), parental disapproval of sex (Jaccard, Dittus, & Gordon, 1996) and parental communication (Diloria, Kelley, Hockenberry-Eaton, 1999). At the school level, we included school connectedness, and at the community level, exposure to pornographic media and neighborhood disorganization (Elliott et al, 1996). In Table 1, we provide brief examples of the survey measures, range and response categories, number of items and their associated reliabilities from our cross-sectional data.

Table 1.

Multi-Item Constructs Assessed as Components of the Risk and Protective Framework

Category	Construct	Scale source	No. of items, α	Survey question example
Individual	Sexual refusal self-efficacy	Donohew et al., 2000	7 items, α = .91	“Would be able to say no to having sex with someone you want to date again?”^a,f
	Peer norms	Floyd, 2009	6 items, α = .67	“Most of my close friends have had sex.”^b,f
	Attitudes about sex	Floyd, 2009	6 items, α = .75	“People my age should get as much sexual experience as they can.”^b,f
	HIV severity	Witte et al., 2002	5 items, α = .87	“HIV is a serious disease.”^b,e,g
	Independence	Singelis, 1994	5 items, α = .67	“My personal identity, independent of others is very important to me.”^c,f
	Interdependence	Singelis, 1994	5 items α = .0	“I have respect for authority figures.”^c,f
	Buddhism	Created for Study	4 items, α = .70	“How often do you practice the 5 precepts?” ^c,f
Family	Parental monitoring	Capaldi & Patterson, 1989	8 items, α = .73	“How often do your parents know what you are doing after school”^c,f
	Parental closeness	Harris et al., 2003	4 items, α = .66	“How close do you feel to your mother?” ^c,f
	Parental disapproval of sex	Jaccard et al., 1996	5 items, α = .84	“How would your mother feel about you having sex at this time in your life?”^d,e,f
	Parental communication (sex)	Diloria et al., 1999	9 items, α = .91	“How many times have you talked to your parent about dating?”^c,f
School	School connectedness	Strengthening Families Program, 2000	20 items, α = .84	“I don’t really belong at school.”^b,f
Community	Exposure to pornographic media	Created for this study	2 items, α = .71	Have you ever visited a pornographic website?”^c,g
	Neighborhood disorganization	Elliott et al, 1996	20 items, α = .97	“Is there a problem in your neighborhood with people using drugs?”^c,g

Response ranges:

1 = I definitely can’t do this to 4 = I definitely can do this.

1 = Agree a lot to 4 = Disagree a lot.

1 = Not at all to 4 = Very much/all the time.

1 = Disapprove a lot to 4 = Approve a lot.

Reverse coded.

Higher scores represent prosocial behaviors or attitudes.

Higher scores represent greater engagement in behavior or greater risk perception.

Analysis

For our initial exploratory analysis, we conducted bivariate analyses using the Pearson correlation coefficient to assess the correlation between each of the three dependent variables and each of the variables that fell into one of the five categories of the Risk and Protective Framework (individual, peer, family, school, and community, significant effects presented in Table 1) establishing statistical significance at p < .05. We then examined the correlation between significant predictor variables conducting a correlational matrix using the Kendal rank order correlation. Variables from each domain (individual, peer, family, school, and community) that were significantly associated with any of the three outcomes in bivariate analyses (at p = .05) and not highly correlated with each other (τ < .40, p < .0001), were included in a series of simultaneous regression equations, specific to each risk and protective level, to assess which variables were more strongly correlated with each of the three outcomes (data not shown). We selected a p value of .05 (instead of a smaller p value) because of the exploratory nature of these analyses to detect potentially important correlates of sexual behaviors. That is, we felt that a Type II error (missing potentially significant finding) was more important that a Type I error (incorrectly concluding a relationship was significant). We used linear regression for engagement in presexual behaviors and intention to have sex and logistic regression for ever having sexual intercourse (significance at p < .05). We examined violations to homoscedasticity and multicollinearity among our predictor variables by examining residual plots for each predictor variable and by assessing tolerance values. The most significant variable (as measured by the coefficient and p value) from each of these multivariate analyses at the individual, peer, family, school, and community level, were then included in a series of simultaneous regression models for each of the three outcome variables (data not shown). Final reduced models were then estimated for each outcome variable, comparing the full to the reduced models, using change in R² for linear regression and the likelihood ratio test for logistic regression.

In our final reduced regression models, significant numbers of respondents were excluded from the analyses due to missing data. A missing data analysis was run to assess whether those who were missing on any of the independent variables in the final multivariate models reported significantly different responses on variables of interest than those included in the final models (see Results section).

Results

Bivariate Results

The three sets of bivariate results (one set in each column for each of the three dependent variables—precoital behaviors, intention to have sex by the end of secondary school, and ever had sex) are shown in Table 2. These analyses are organized by each risk and protective level (individual, peer, family, school, and community). Independent variables not significantly correlated with any of the three dependent variables (general parental communication, and independence) were dropped from Table 2 and further analyses. Bivariate results were generally similar across the three outcome variables. At the individual level, refusal self-efficacy, and attitudes about delaying sex had the strongest bivariate relationships with all three outcomes with absolute values of correlations between r = .29 and r = .14. Level of interdependence and engaging in Buddhist practices were weakly correlated with all three dependent variables, whereas perceived consequences of HIV infection was weakly, negatively correlated with precoital behaviors and intentions to have sex. At the peer level, having a boyfriend/girlfriend was highly correlated with precoital behaviors (r = .40, p < .0001) and weakly correlated with intentions to have sex and ever having sexual intercourse (r = .22 and r = .15, p < .0001, respectively). Perceived peer norms were not correlated with any of the three outcomes. At the family level, parental disapproval about sex had the strongest association with precoital behaviors, intentions, and ever had sex (r = –.31, r = –.38, r = –.24, p < .0001, respectively) followed by parental closeness (r = –.21, p < .001; r = –.15, p < .01; r = –.14, p < .01, respectively) and parental monitoring (r = –.18, p < .001; r = –.19, p < .001, r = –.11, p < .05, respectively). At the school level, school connectedness and self-reported grades were both moderately correlated with the three risk behaviors with absolute values ranging from r = .17 to r = .29. At the community level, exposure to pornographic media was strongly correlated with precoital behaviors (r = .37, p < .001) and moderately correlated with sexual intention and ever having sex (r = .20, r = .22, p < .0001, respectively).

Table 2.

Significant Correlations Between Putative Correlates and dependent variables

		Precoital behaviors	Intentions to have Sex	Ever had sex
Risk and protective level	%	R	r	r
Individual level
Female	50%	−.03	−.19***	−.07
	M (SD)
Age	13.4 (0.5)	.18***	.07	.03
Attitudes about delaying sex	3.38 (0.67)	−.20***	−.22***	−.14**
Perceived consequences of HIV	3.78 (0.50)	−.17***	−.15**	−.01
Refusal self-efficacy	3.60 (0.65)	−.29***	−.26***	−.21***
Negotiation self- efficacy	3.10 (1.06)	−.06	−.01	−.13**
Interdependence	3.60 (0.40)	−.17**	−.15**	−.17***
Buddhist practice	2.77 (0.58)	−.20***	−.12*	−.14**
Peer level
Peer norms		.02	.001	−.04
	%	r	r	r
Having boyfriend/ girlfriend	17.5	.40***	.22***	.15**
	M (SD)
Family level
Parental communication (sex)	2.00 (0.78)	.04	−0.04	−0.03
Parental closeness	3.81 (0.38)	−.21***	−.15**	−.14**
Parental disapproval of sex	3.85 (0.38)	−.31***	−.38***	−.24***
Parental monitoring	2.92 (0.53)	−.18***	−.19***	−.11*
School level
School connectedness	3.2 (0.38)	−.20***	−.27***	−.18***
Grades in school	3.2 (1.41)	−.17***	−.29***	−.18***
Community level
Exposure to sexualized media (online and televised porn)	0.34 (0.81)	.37***	.20***	.22***
Neighborhood characteristics	2.38 (1.01)	.08	−.07	.04

p < .05. **p < .01. ***p < .001 (two tailed).

Correlational Matrices Results

We next assessed the correlation between risk and protective factors from each level that were identified as significant in the bivariate analysis. As seen in Table 3, among the significant individual-level characteristics, the strongest positive correlations were found between sexual refusal self-efficacy and (a) attitudes about delaying sex (τ = .30, p < .0001), and (b) perceived consequences of HIV infection (τ = .20, p < .0001), whereas sexual refusal was negatively correlated with older age (τ = –.19, p < .0001). Practicing Buddhism had weak but significant associations with most risk and protective factors, with the strongest correlations with interdependency (τ = .16, p < .0001), sexual refusal (τ = .16, p < .0001), and perceived HIV consequences (τ = .15, p < .0001).

Table 3.

Kendal’s Rank Order Correlation Among Significant Variables at Each Risk and Protective Level

	Individual level	1	2	3	4	5	6
1.	Age	1.0
2.	Attitudes about delaying sex	-0.11**	1.0
3.	HIV consequences	-0.02	0.12**	1.0
4.	Self refusal efficacy	-0.19***	0.30***	0.20***	1.0
5.	Interdependency	-0.01	0.10**	0.11**	0.12***	1.0
6.	Buddhism	0.00	0.09*	0.15***	0.16***	0.16***	1.0
	School level	1	2
1.	School attachment	1.0
2.	Grades	.38***	1.0

Note. Peer and community levels not included in correlational matrices because these levels had only one significant predictor in the bivariate analyses.

At the family level, few parental variables were significantly correlated with each other (data not shown). At the school level, school attachment and student grades were moderately correlated with each other (τ = .38, p < .0001) suggesting the need to include only one of these school-level variables in multivariate models to reduce problems with multicollinearity (Table 3).

Multultivariate Analysis

Based on our final reduced models, significant risk factors for precoital behaviors included having a boyfriend/girlfriend (β = .35, p < .001) and exposure to pornographic media (β = .32, p < .001). Protective factors included parental disapproval of sex and sexual refusal self-efficacy (β = –.17, p < .001; β = –.11, p < .05, respectively; Table 4). Our R² coefficient was relatively high for our final, reduced model, explaining 38% of the variance in precoital behaviors. Comparing our full and reduced models, school attachment did not remain significant and was dropped in our final model for precoital behavior having no significant impact on the R² coefficient, full model, R² = .38, F(5, 329) = 42.14, p < .001, reduced model R² = .38, F(1, 329) = 52.81, p < .001, ΔF = .06, p = .81 (Table 4).

Table 4.

Full Versus Reduced Final Multiple Regression Model Predicting Precoital Behavior

	Full model		Reduced model
Predictor	β	Partial β	β	Partial β
Boyfriend/girlfriend	.35***	.35	.35**	.34m
Pornographic media	.32***	.31	.32***	.31
Parental disapproval	−.17***	−.15	−.17***	−.16
Sex refusal	−.11*	−.10	−.11*	−.10
School attachment	−.01	−.01
R ²	.38		.38
ΔR²	.00
F	42.14***		52.81***
ΔF	.06
p(ΔF)	.81
N	335		345

p < .05. **p < .01. ***p <.001

For intention to have sex, similar results were found (Table 5). Based on our final reduced model, parental disapproval of sex was the most strongly correlated protective factor (β = –.29, p < .001), followed by school attachment (β = –.18, p < .001). Risk factors included having a boyfriend/girlfriend (β = .17, p < .001) and exposure to pornographic media (β = .12, p < .001) (Table 5). Sexual refusal self-efficacy did not remain significant in our full model and was dropped in the reduced model, having a small but nonsignificant impact on the R² coefficient, full model, R² = .23, F(5, 311) = 19.34, p < .001, reduced model R² = .22, F(1, 311) = 24.97, p < .001, ΔF = 2.53, p = .112.

Table 5.

Full Compared to Reduced Final Multiple Regression Model Predicting Intentions to Have Sex

	Full Model 1		Reduced model
Predictor	β	Partial β	β	Partial β
Boyfriend/girlfriend	.15**	.15	.17**	.16
Pornographic media	.10*	.11	.12***	.12
Parental disapproval	−.27***	−.25	−.29***	−.28
School attachment	−.16**	−.15	−.18***	−.18
Sex refusal	−.09	−.08
R ²	.23		.22
ΔR²	.03
F	19.34***		24.97***
ΔF	2.53
p(ΔF)	.112
N	317		345

p < .05. **p < .01. ***p <.001.

No risk or protective factors were significantly correlated with ever had sex in our full model (Table 6). However exposure to pornographic media, AOR = 2.41, 95% CI [.98, 5.85], p = .053, parental disapproval, AOR = .31, 95% CI [.08, 1.13], p = .07, and grades, AOR = .67, 95% CI [.38, 1.17], p = .10 approached significance (Table 6). When only the marginally significant variables are included in our reduced model, exposure to pornographic media, AOR = 2.73, 95% CI [1.25, 5.96], p < .01, and parental disapproval of sex, AOR = .28, 95% CI[.09, .81], p = .02 were significantly associated with ever having sex whereas self-reported grades, AOR = .62, 95% CI [.37, 1.01], p = .06 approached significance. The associated R² coefficient was .28 for our reduced model. When comparing the full to our reduced model using the log likelihood ratio test to assess goodness of fit, the reduced, nested model significantly improved the fit of the regression model (G = 12.78, p < .001).

Table 6.

Full Versus Reduced Final Multiple Regression Model Predicting Ever Had Sex

	Full model		Reduced model
Predictor	AOR	CI	AOR	CI
Pornographic media	2.41	[0.98, 5.85]	2.73	[1.25, 5.96]
Parental disapproval	0.31	[0.08, 1.13]	0.28	[0.09, 0.81]
Grades	0.67	[0.38, 1.17]	0.62	[0.37, 1.01]
Boyfriend/girlfriend	1.93	[0.38, 9.80]
Sex refusal	0.75	[0.26, 2.14]
R ^2a	0.32		.28
ΔR²	0.04
−2 log likelihood	58.77		71.37
LRT statistic G(df)	12.6(2)***
N	337		377

Nagelkerke R².

p < .05. **p < .01. ***p < .001.

Tests for multicollinearity were done in all regression models and multicollinearity was clearly not present. The tolerance value (1 – R²) was greater than .7 for all predictors in all three equations and the Variance Inflation Factor (VIF) was less than 2.0. We also tested for heteroscedasticity in our linear regression models (predicting precoital behaviors and intentions to have sex) by plotting residuals against values of predicted scores. After viewing the plots, we concluded there was very little heteroscedasticity for our variables in both analyses, with much less than the threshold of three times greater variance of residuals for any predicted value when compared to another predicted value (Tabachnik & Fidell, 2001).

Missing Data Analysis

In the final reduced multivariate models, missing data (n) ranged from 45 to 75 across the three outcomes. To assess whether our findings from our final reduced models were affected by missing cases, we compared those participants who remained in our analysis to those participants who had missing data for precoital behaviors, intentions to have sex, and ever had sex. Variables assessed as potentially differing between the two groups included the following: age, gender, parental communication about sex, parental closeness, independence, interdependence, perceived HIV risk, attitudes about waiting to have sex, Buddhist practices, neighborhood problems, and school attachment. For “ever had sex” there were no significant differences between the 377 included in the regression model and the 43 who were missing on any of these variables. For intentions to have sex, there were five variables on which the two groups (those included in the model, n = 345 and missing cases, n = 75) differed significantly (p < .05). Those with missing data were slightly younger, perceived themselves to be at higher HIV risk, had more negative attitudes about waiting to have sex, practiced Buddhist practices less, and were less connected to school. Thus our results from our multivariate model are probably weaker than if these potentially higher risk individuals had participated and generalizing to the population should be done cautiously. For the precoital behavior analysis, three variables were significantly different between the two groups (n = 345 and n = 75, p <.05). Those with missing data communicated less with their parents, practiced Buddhism more frequently, and perceived more problems in their neighborhood. These variables were less consistent (in terms of level of risk) and therefore we can probably generalize with less caution for this analysis to the entire sample population.

Discussion

The Risk and Protective Framework serves as a useful organizational framework to examine variables operating at the individual, school, peer, family, and community levels that may be associated with sexual risk taking. This is particularly true in a culture like Thailand that is undergoing dramatic social change. Our multivariate findings indicate that parental disapproval of sex (at the family level) was negatively correlated with all three outcomes, serving as a robust protective factor, whereas exposure to pornographic media (at the community level) was associated with all three outcomes and may act as an important risk factor. These cross-sectional findings underscore the tension documented in the Thai literature between parents’ strong influence on their children’s behavior on the one hand (Spielmann, 1994) and the ubiquitous influence of Westernization and adolescent exposure to sexually explicit information through the internet and television on the other (Vuttanont et al., 2006).

At the peer level, having a boyfriend/girlfriend was correlated with engaging in precoital behaviors and intention to have sex, serving as a potential risk factor for these earlier risk behaviors. These findings, consistent with the Western literature (Blum et al., 2000; Thorton, 1990), point to the opportunities that having a boyfriend or girlfriend provide for experimenting with precoital behaviors and formulating intentions to have sex. Having a boyfriend or girlfriend did not remain significant for engaging in sex, contrary to other research findings, and may be due in part to the lower statistical power to detect significant effects given the few adolescents who engaged in sexual intercourse in this sample.

At the individual level, sexual refusal self-efficacy was negatively correlated with engaging in precoital behaviors in our final reduced multivariate model but it did not remain a significant protective factor for intentions to have sex or ever had sex. Similarly, our other psychosocial variables such as sexual attitudes, refusal self efficacy, and perceived consequences of HIV, found in the Western literature to be associated with sexual risk behaviors (Noar & Zimmerman, 2005), were not associated with any of our outcome variables in our final, reduced, multivariate models. It is important to note that sexual refusal self-efficacy was moderately correlated with sexual attitudes and weakly correlated with all other pyschosocial variables at the individual level. Further, the means of these multiscale measures were high, suggesting that the Thai youth in our sample generally ascribe to prosocial attitudes and tend to regard HIV infection as a serious disease. The sample also has very low levels of sexual involvement. It may be that these psychosocial processing variables tend to have greater impact for more sexually active populations, who experience greater variation in sexual attitudes and behavior.

The practice of Buddhism was found in bivariate analysis to be associated with all three of our outcomes including intentions, precoital behaviors, and ever having sex and was weakly correlated with all other predictor variables at the individual level. However when included in a multivariate model with only the significant variables from the individual level, Buddhism was significantly associated with only precoital behaviors. Because its beta coefficient and significance value was not as high as refusal self-efficacy, it was not included as the individual level characteristic in the full or reduced multivariate models predicting precoital behaviors. Finally, at the school level, school attachment was a significant protective factor for intentions to have sex but was not significantly associated with precoital behaviors or ever having sex.

Limitations

A significant limitation of this analysis was a lack of power to assess correlates of ever having sex, given the small sample of respondents who were sexually active. The data were also cross-sectional; therefore we can only assert correlations between the independent and dependent variables not causality. Finally, because of the strong taboos that exist in Thai culture for discussing sexual behaviors, there may have been some level of social desirable responding among adolescents when completing surveys. To the extent possible, the research team tried to minimize the effects of social desirable responding by having youth self-administer the survey using laptop computers with ACASI software.

In summary, our data suggests that parent’s disapproval about sex and exposure to sexually explicit media are robust competing risk and protective factors for sexual behaviors and intentions among Thai adolescents. Our findings of the potential protective affect of parental disapproval on engaging in precoital behaviors, formulating sexual intentions, and initiating sex, are consistent with earlier research conducted in the United States among inner-city families indicating that parental attitudes towards sex can serve to inhibit engagement in early sexual behaviors (Dilorio, Dudley, Soet, & McCarty, 2004; Paikoff, 1995). Both Western and international studies have documented the association of exposure to sexually explicit media and permissive sexual attitudes (Peter & Valkenburg, 2008) and early sexual initiation (Braun-Courville & Rojas, 2009). Future longitudinal studies should examine the complex relationships between these competing risk and protective factors and HIV-prevention interventions should consider their potential influence when adapting HIV-prevention intervention for Thai youth.

Footnotes

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research for and preparation of this manuscript were supported by NIAAA 1R01AA015672-01A1 “Youth Alcohol Use and Risky Sexual Behavior in Bangkok.”

References

Ajzen

Fishbein

(1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice Hall.

Allen

D. R.

Carey

Manopaiboon

Jenkins

R. A.

Uthaivoravit

Kilmarx

P. H.

van Griensven

(2003). Sexual health risks among young Thai women: Implications for HIV/STD prevention and contraception. AIDS and Behavior, 7, 9-21.

Billy

J. G.

Brewster

K. L.

Grady

W. R.

(1994). Contextual effects on the sexual behavior of adolescent women. Journal of Marriage and the Family, 56, 387-404.

Blum

R. W.

Beuhring

Rinehart

P. M.

(2000). Protecting teens: Beyond race, income, and family structure. Minneapolis: Center for Adolescent Health, University of Minnesota.

Blum

R. W.

Mmarim

K. N.

(2005). Risk and protective factors affecting adolescent reproductive health in developing countries. Baltimore, MD: Department of Population and Family Health Services, John Hopkins Bloomberg School of Public Health. Retrieved from http://whqlibdoc.who.int/publications/2005/9241593652_eng.pdf

Braun-Courville

D. K.

Rojas

(2009). Exposure to sexually explicit web sites and adolescent sexual attitudes and behaviors. Journal of Adolescent Health, 45, 156-162.

Byrnes

H. F.

Miller

B. A.

Cupp

P. K.

Zimmerman

R. S.

Chookhare

(in press). The roles of neighborhood disorganization and social capital in Thai adolescents’ substance use and delinquency Manuscript submitted for publication.

Capaldi

Patterson

G. R.

(1989). Psychometric properties of fourteen latent constructs from the Oregon Youth Study. New York, NY: Springer-Verlag.

Dilorio

Dudley

W. N.

Soet

J. E.

McCarty

(2004). Sexual possibility situations and sexual behaviors among young adolescents: the moderating role of protective factors. Journal of Adolescent Health, 35, 528-540.

10.

Dilorio

Kelley

Hockenberry-Eaton

(1999). Communication about sexual issues: mothers, fathers and friends. Journal Adolescent Health, 24, 181-189.

11.

Donohew

Zimmerman

R. S.

Cupp

P. S.

Novak

Colon

Abell

(2000). Sensation seeking, impulsive decision-making, and risky sex: Implications for risk-taking and design of interventions. Personality and Individual Differences, 28, 1079-1091.

12.

Elliott

D. S.

Wilson

W. J.

Huizinga

Sampson

R. J.

Elliott

Rankin

(1996). The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency, 33, 389-426.

13.

Fishbein

Ajzen

(1975). Belief, attitude, intention and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley.

14.

Floyd

(2009). Racial differences in sexual debut: Implications for designing HIV/STD and pregnancy prevention messages (Doctoral dissertation). University of Kentucky, Lexington. Retrieved from http://www.uky.edu/ETD

15.

Guruge

Isaranurug

Nanthamongkolchai

Charupoonphol

(2004). Sexual behavior among adolescents in Wattanakorn District, Sakaeo Province. Journal of Public Health and Development, 2(1), 53-68.

16.

Halpern

C. T.

Joyner

Udry

Suchindran

(2000). Smart teens don’t have sex (or kiss much either). Journal of Adolescent Health, 26, 225-231.

17.

Hanenberg

R. S.

Rojanapithayakorn

Kunasol

Sokal

D. C.

(1994). Impact of Thailand’s HIV-control programme as indicated by the decline of sexually transmitted diseases. Lancet, 344, 243-245.

18.

Harris

Mullan

Florey

Tabor

P. S.

Bearman

Udry

J. R.

(2003). The National Longitudinal Study of Adolescent Health. Retrieved from http://www.cpc.unc.edu/projects/addhealth/design

19.

Harris

Oman

R. F.

Vesely

S. K.

Tolma

E. L.

Aspy

C. B.

Rodine

. . . Fluhr

(2006). Associations between youth assets and sexual activity: Does adult supervision play a role? Child care, Health and Development, 33, 448-454.

20.

Hawkins

J. D.

Catalano

R. F.

Miller

J. Y.

(1992). Risk and protective for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin, 112(1), 64-105.

21.

Hutchinson

M. K.

(2002). The influence of sexual risk communication between parents and daughters on sexual risk behaviors. Family Relations, 51, 238-247.

22.

Jaccard

Dittus

P. J.

Gordon

V. V.

(1996). Maternal correlates of adolescent sexual and contraceptive behavior. Family Planning Perspectives, 28, 159-165.

23.

Jessor

(1991). Risk behavior in adolescence: A psychosocial framework for understanding and action. Journal of Adolescent Health, 15, 597-605.

24.

Lammers

Ireland

Resnick

Blum

(2000). Influences on adolescents’ decisions to postpone onset of sexual intercourse: A survival analysis of virginity among youths aged 13 to 18 years. Journal of Adolescent Health, 26(1), 42-48.

25.

Liu

Kilmarx

Jenkins

R. A.

Manopaiboon

Mock

P. A.

Jeeyapunt

. . . van Giensven

(2006). Sexual Initiation, substance use, and sexual behaviors and knowledge among vocational students in Northern Thailand. International Family Planning Perspectives, 32, 126-135.

26.

Longmore

M. A.

Manning

W. D.

Giordano

P. C.

(2001). Preadolescent parenting strategies and teens’ dating and sexual initiation: A longitudinal analysis. Journal of Marriage and the Family, 63, 322-335.

27.

Martino

S. C.

Elliott

M. N.

Collins

R. L.

Kanouse

D. E.

Berry

S. H.

(2008). Virginity pledges among the willing: Delays in first intercourse and consistency of condom use. Journal of Adolescent Health, 43, 341-348.

28.

Mason

C. J.

Kripornshai

Markowitz

L. E.

Chanbancherd

Supapongse

Jugsudee

. . . Nitaya-phan

(1998). Nationwide surveillance of HIV-1 prevalence and subtype in young Thai men. Journal of Acquired Immune Deficiency Syndrome and Human Retrovirology, 19, 165-173.

29.

Mendez

R. V.

Majumdar

(2005). Familial interdependence, pregnancy and condom use in Latino and Anglo women. Paper presented at the 2005 Annual Meeting of the American Sociological Association, Philadelphia, PA.

30.

Miller

B. C.

Benson

Galbraith

K. A.

(2001). Family relationships and adolescent pregnancy risk: A research synthesis. Developmental Review, 39, 1-38.

31.

Nelson

K. E.

Eiumtrakul

Celentano

D. D.

Beyrer

Galai

Kawichai

Khamboonruang

(2002). HIV infection in young men in Northern Thailand, 1991-1998: Increasing role of injection drug use. Journal of Acquired Immune Deficiency Syndrome, 1(29), 62-68.

32.

Noar

S. M.

Zimmerman

R. S.

(2005). Health-behavior theory and cumulative knowledge regarding health behaviors: Are we moving in the right direction? Health Education Research, 19, 165-174.

33.

Paikoff

R. L.

(1995). Early heterosexual debut: situations of sexual possibility during the transition to adolescence. American Journal of Orthopsychiatry, 65, 389-401.

34.

Paul

Fitzjohn

Herbison

Dickson

(2000). The determinants of sexual intercourse before age 16. Journal of Adolescent Health, 27, 136-147.

35.

Peter

Valkenburg

P. M.

(2008). Adolescents’ exposure to sexually explicit internet material, sexual uncertainty, and attitudes towards uncommitted sexual exploration. Communication Research, 35, 579-601.

36.

Raine

T. R.

Jenkins

Aarons

S. J.

Woodward

Fairfax

J. L.

El-Khorazaty

M. N.

Herman

(1999). Sociodemographic correlates of virginity in seventh-grade Black and Latino students. Journal of Adolescent Health, 24, 304-312.

37.

Rasamimari

Dancy

Talashek

M. M.

Park

C. G.

(2007). Predictors of sexual behaviors among Thai young adults. The Journal of the Association of Nurses in AIDS Care, 18(6), 13-21.

38.

Resnick

M. D.

Bearman

P. S.

Blum

R. W.

Bauman

K. E.

Harris

K. M.

Jones

. . . Uldry

J. R.

(1997). Protecting adolescents from harm: Findings from the national longitudinal study on adolescent health. Journal of American Medical Association, 278, 823-832.

39.

Rojanapithayakorn

Hanenberg

(1996). The 100% condom program Thailand. AIDS, 10(1), 1-7.

40.

Romer

Stanton

Galbraith

Feigelman

Black

M. S.

(1999). Parental influence on adolescent sexual behavior in high-poverty settings. Archives of Pediatrics and Adolescent Medicine, 53, 1055-1062.

41.

Sheeran

Abraham

Orbell

(1999). Psychological correlated of heterosexual condom use: A meta-analysis. Psychological Bulletin, 125, 90-132.

42.

Singelis

T. M.

(1994). The measurement of independent and interdependent self-constructs. Personality and Social Psychology Bulletin, 20, 580-591.

43.

Spielmann

(1994). The family in Thailand and drug demand reduction: problems of urban Thai society in transition. Bulletin on Narcotics, 46(1), 45-66.

44.

Strengthening Families Program. (2000). The Capable Families and Youth Study. Institute for Social and Behavioral Research, Iowa State University, Ames, Iowa.

45.

Tabachnik

B. G.

Fidell

L. S.

(2001). Using multivariate statistics (4th ed.) Boston, MA: Allyn and Bacon.

46.

Thailand National Statistics Office. (2000). The 2000 Population and Housing Census. Retrieved from http://web.nso.go.th/en/census/poph/indiregion/indi_whole .htm

47.

Thato

Charron-Prochownik

Dorn

L. D.

Albrecht

S. A.

Stone

C. A.

(2003). Predictors of condom use among adolescent Thai vocational students. Journal of Nursing Scholarship, 35, 157-163.

48.

Thorton

(1990.) The courtship process and adolescent sex. Journal of Family Issues, 11, 239-275.

49.

Ubillos

Paez

Gonzalez

J. L.

(2000). Culture and sexual behavior. Psicothema, 12(S), 70-82.

50.

Vuttanont

Greenhalgh

Griffin

Boynton

(2006). “Smart boys” and “sweet girls”—sex education needs in Thai teenagers: A mixed method study. Lancet, 368, 2068-2080.

51.

Whitehead

S. J.

Leelawiwat

Jeeyapant

Chaikummao

Papp

Kilmarx

P. H.

. . . van Griensven

(2008). Increase in sexual risk behavior and prevalence of Chlamydia trachomatis among adolescents in Northern Thailand. Sexually Transmitted Diseases, 35, 883-888.

52.

Witte

Girma

Girgre

(2002-2003). Addressing underlying mechanism to HIV/AIDS preventive behaviors in Ethiopia. International Quarterly of Community Health Education, 21, 163-176.