The Adolescent Sexual Activity Index (ASAI): a standardized strategy for measuring interpersonal heterosexual behaviors among youth

William B. Hansen, Electra D. Paskett, Linda J. Carter, The Adolescent Sexual Activity Index (ASAI): a standardized strategy for measuring interpersonal heterosexual behaviors among youth, Health Education Research, Volume 14, Issue 4, August 1999, Pages 485–490, https://doi.org/10.1093/her/14.4.485

Abstract

This study reports on the properties of a standardized measure that is used to index the spectrum of sexual behaviors typical of adolescents. Prior work has shown sexual behaviors to be ordered and progressive. Using pre-coitus sexual behaviors that did not refer to sex organs and self-reports of sexual intercourse, a 13-item index was developed. The index was standardized with values ranging from 0.0 to 10.0. Measures were administered to 1493 female and 1073 male black and white adolescents aged 12–19 years in community settings. Analyses reveal the index to have high internal consistency and to perform as a psychometrically sound dependent variable. Because the index uses conservative language, its use may avoid the controversy common to much of the research on adolescent sexuality. The Adolescent Sexual Activity Index will provide researchers and practitioners with an interpretable tool for examining developmental patterns that precede sexual intercourse.

Introduction

Over 1 million teen pregnancies, mostly unintended, were estimated to have occurred in the US in 1990 ( National Center for Health Statistics, 1993a, b). The annual public cost of these births has been estimated to be as high as $20 billion ( Center for Population Options, 1992). It has also been estimated that for every $1 spent on effective prevention, $4.40 would be saved on public costs associated with unintended teen pregnancy ( Forrest and Singh, 1990). Between 1980 and 1990, some reduction in the teen pregnancy rate was observed ( Morbidity and Mortality Weekly Report, 1993a). STDs, particularly the transmission of HIV, have also been of increasing concern for teens. Gonorrhea and syphilis in adults have had recent declines in incidence ( Morbidity and Mortality Weekly Report, 1993b). However, declines among teens have not been observed, and gonorrhea is increasing among black teens ( Webster et al., 1993) and chlamydia continues to be a serious problem ( Webster et al., 1993). Condom use is generally increasing and is effective at reducing transmission rates ( Catania et al., 1992; Weller, 1993). Nonetheless, HIV and AIDS cases continue to grow ( Morbidity and Mortality Weekly Report, 1993c). There is some evidence to suggest that HIV transmission to young females through unprotected sexual intercourse continues to increase ( Nwanyanwu et al., 1993).

The solution to all these problems will require an ability to measure sexual activity. Initially, measures are needed for descriptive purposes. It is necessary to be able to describe the normal course of a behavior as an initial step toward understanding its development. Measures are needed as dependent variables in etiological studies. Finally, measures are needed as dependent variables in intervention studies.

Guttman scaling research spanning four decades has confirmed that the progression of multiple related behaviors follows a predictable sequence (Levitt, 1965; Bentler, 1968, 1969; Mahoney, 1980; Smith and Udry, 1985; Hansen et al., 1992). This order represents one means of characterizing the degree of sexual activity engaged in by adolescents. The primary drawback of Guttman scales is the fact that scales are always imperfect and scaling applications have no means for adjusting for the unreliability of individual items. An index created from multiple indicators that have been demonstrated to have Guttman characteristics is expected to solve this problem.

Funding has been withdrawn in the past for research that has had as its objective the measurement of sexual activity (Sullivan, pers. commun., 1992; Udry, 1993). It is easy to speculate that many local projects are not funded because of reactions to sex questions. Sex is a highly charged issue and it is likely that many people fear that asking explicit questions will teach youth about sex. These fears thwart productive and beneficial research but must be addressed. One solution is to use measures that avoid charged words and topics. Most recently, in a community sample of young adults, Hansen et al. ( Hansen et al., 1992) identified several pre-coital behaviors that may avoid reactivity but that are nonetheless ordered and scalable using Guttman techniques. These behaviors include (in general order) hugging, spending time, holding hands, kissing, cuddling, laying down together and being undressed together, all of which precede sexual intercourse for both males and females, and offer an alternative means for creating psychometrically sound measures.

The purpose of this study is to present data about a new measure of adolescent sexual activity, the Adolescent Sexual Activity Index (ASAI). The ASAI is the first measure of sexual activity that explicitly uses Guttman scaling about sexual activity as the basis of scale creation. The goal of this research is to create a measure that reflects the progression of sexual involvement that is psychometrically sound, useful in research and practice, and one that avoids social controversy.

Methods

Subjects

Participants were 1424 female and 1037 male adolescents who ranged in age from 12 to 19 years. All were residents of Forsyth and Davidson Counties, North Carolina. Participants were recruited as part of a larger project designed to use information about risk factors for sexual activity to develop community-based teen pregnancy prevention efforts. The specific age, gender and ethnic characteristics of the sample are presented in Table I .

Participants were recruited through a variety of innovative strategies and were surveyed in 35 different sites. Most were recruited from non-public school education environments (N = 1291) which included local private secondary schools, and state and private colleges. The next largest group was recruited through a variety of recreation facilities (N = 554). An additional 223 were recruited from a public party held specifically to recruit participants. One-hundred and seventy-six were recruited from attenders at local health care facilities. Most of these were in attendance for pregnancy-related care. One-hundred and two subjects were approached in a variety of public places including parks. Ninety-eight were solicited to participate in local housing projects. Finally, 17 were recruited from youth worksites.

The sample was relatively experienced sexually. Among males, 80.3% reported having had sex at least once and 45.7% reported having sexual intercourse in the past 30 days. Among females, 66.6% reported having ever had sex and 39.2% reported sex in the past month.

Measures

Participants answered 13 questions about their current and past sexual practices. Items to be included in the survey were derived in part from a previously employed survey ( Hansen et al., 1992). The previous survey had shown 16 pre-coitus behaviors to be highly scalable using Guttman scaling techniques. From this list of items, those which referred to specific body parts (the penis, vagina or breast) were removed based on key informants' recommendations. Items retained in the survey asked about participation during the 30 previous days in the following heterosexual activities: (1) hugging, (2) holding hands, (3) spending time alone, (4) kissing, (5) cuddling, (6) laying down together, (7) having someone put his or her hands under one's clothing, (8) putting one's hands under someone else's clothing, (9) being undressed with sex organs showing and (10) engaging in intercourse. In addition to these items, four additional items were asked: (11) the frequency of sexual intercourse during the previous 30 days, (12) the number of different sexual partners during the previous 30 days and (13) the number of sexual partners during the previous 12 months.

Results

Scale construction

A single scale was constructed using the 13 items. Items were standardized using the SAS PROBIT transformation procedure. Values used for this transformation were response percentile mid-point values that corresponded to the percent of subjects who answered each response. 1

For example, if an item had two responses, 40% of subjects answering response 1 and 60% answering response 2, the transformed values would correspond to probit transformations of 0.20 (0.40/2) and 0.70 [(0.60/2) + 0.40], respectively. Transformed values (probits) would equal –0.84162 and 0.52440. Transformation values for the standardized ASAI items are available for research purposes from the authors.

For dichotomous data, these transformations are roughly equal usual standardizing procedures. For items with more than two responses (items 10, 11 and 12), the probit transformation creates values that fit a normal-curve but, unlike the standard transformation, are not equidistant. Two additional transformations were performed to create a scale that had a uniform character. The average of minimum values was adjusted to be zero by adding the absolute value of the minimum to all scores. All values were then multiplied by 10 times the reciprocal of the average maximum value to create an index with a maximum value of 10.0. Thus, an 11-point index was created that had a minimum value of 0 and a maximum value of 10.

Once data had been transformed, the index was created by averaging values across items. This allowed subjects with some missing data to have index values. The use of averaging responses allows for the possibility that there will be out-of-range index values. Only two scores were out of range, one each at –0.05 and 10.5. Thus, there was little threat that averaging significantly reduced the overall precision of the index or its interpretability.

Internal consistency of items

Items in the ASAI were analyzed to establish the internal consistency of items using Cronbach's α. The overall α for the index when variables were standardized was 0.93. The index did not improve with the deletion of any items. The internal consistency was high for females (α = 0.94) and males (α = 0.93), and for blacks (α = 0.94) and whites (α = 0.93). Cronbach's α values were slightly smaller for very young subjects and were lowest for 12 year olds (α = 0.88). All other age groups had α coefficients of 0.91 or higher.

Age and race

One method for assessing construct validity is to examine the performance of the scale under common conditions that can be expected a priori to have predictable differences such as when tracking the development of sexual activity developmentally. Table II presents results of the ASAI broken down by age, gender and ethnicity. As expected, the average score generally increases with age. However, even by 12 years, subjects in all groups had average scores above 2.0. Clearly patterns of intimate behavior start early in adolescence. Males, both black and white, were developmentally accelerated compared to females. Black males generally had higher ASAI scores than white males at younger ages. The relative stability of the standard deviations suggests that these scores were not due to the presence of outliers who inflated scores. Rather, for each group, the extent of sexual involvement has predictably similar variability. Black and white females had remarkably similar developmental patterns according to the ASAI.

The ASAI as a correlational outcome measure

One strategy for assessing usefulness to research is to see whether the index can facilitate the identification of correlates of behavior. Table III presents the results of analyses in which the most common measure of sexual activity, reported intercourse during the past 30 days, and ASAI scores were independently correlated with 13 scales that had been developed for use with the project. Of interest in these analyses is a comparison of the relative predictive power of the single item versus the index. Including multiple measures can be expected a priori to increase correlations simply because redundant information tends to enhance reliability.

The ASAI is a continuous measure and the Pearson r was an appropriate statistic for assessing these correlations. Spearman ρ correlations were calculated for bivariate comparisons (predictor variables correlated with the measure of 30-day intercourse). To assess the relative ability of the ASAI and the single bivariate measure to be predicted, z-score differences between the correlations for each scale were calculated. 2

The z score differences were based on Pearson r to z transformations. The z scores represent the normal distribution; differences greater than 1.96 are significant at α = 0.05. The z score differences were calculated using the formula:

\ =\ 0.5\ \ Log\ [(1\ +\ r_)\ +\ (1\ \ r_)].\] Pairs of correlations were then compared using the formula: \[z\ diff\ =\ \frac(\frac)\ +\ (\frac)>.\]Several predictors, particularly for females, increased in their predictive power. For females, normative beliefs (two scales that measured the perception of engaging in sexual activity by same-age peers and the perceived acceptability of having sexual intercourse by peers) both increased, one dramatically so, in their predictive strength. Similarly, drug use and one of the values scales (a scale that measured the degree to which participants felt that premarital sex was acceptable) increased in strength when predicting the index versus the individual item. For males, only drug use was seen to have a similar increase in correlations when using the ASAI instead of 30-day intercourse as a dependent variable.

Interestingly, there were several measures that became less important when the index was used instead of the single measure. For both males and females, family history of pregnancy as predictors of sexual activity had slightly reduced predictive capability for the index compared to the scale. For females, a fixation with sexual intercourse increased in importance when recent intercourse replaced the ASAI as the dependent variable. Self-esteem (two scales that primarily addressed perceived value of self) became a negative predictor when the index was used for females. Neither outcome had much of a correlational relationship with self-esteem and limited emphasis can be placed on this result.

The overall R 2 (based on general linear model analysis including all predictors) was considerably larger for the ASAI than the single item for females (0.496 versus 0.330, respectively) and males (0.490 versus 0.251, respectively). Because the ASAI is continuous, these analyses are biased toward the ASAI. Kendall τ b correlation coefficients were calculated based on predicted values for general linear regression (for the ASAI) and for logistic regression (for the dichotomous recent intercourse item). For females, this demonstrated a slight improvement in correlations for the ASAI versus the single item (0.528 versus 0.481, respectively). A greater discrepancy was observed for males (0.537 versus 0.413) with the ASAI markedly more predictive.

Discussion

The results of this analysis indicate that the ASAI has the potential for reliably describing a broad spectrum of adolescent sexual behaviors. The index showed high internal consistency and reliably described a spectrum of sexual behaviors. We believe that the ASAI will prove to be a conceptually and heuristically useful tool. The adoption of the ASAI as a measure would allow researchers and practitioners alike to have access to a readily interpretable index. This will allow the behavioral status of individuals and groups to be better understood.

By making the ASAI an 11-point index (from 0 to 10), the inherent potential to use scores on the index add interpretability to individual and group scores that may result. For example, practitioners could readily determine the proximity of an individual to engaging in sexual intercourse by knowing the subject's score relative to the point at which sexual intercourse begins to occur, about 7.0 on the ASAI. A subject with an ASAI score of 6.0 would be expected to be at imminent risk of engaging in sexual intercourse, whereas a subject with a score of 3.0 would be expected to have less risk. An individual with a score beyond 8.0 would be characterized as sexually active.

Older youths consistently had higher index values. Because this study does not include longitudinal assessments, projecting rates at which the index marks increased sexual involvement for individuals is entirely speculative. Individuals' ASAI scores are likely to move up and down over time. Future research that would examine changes in index scores for individuals and groups over time would yield significant information about typical behavioral patterns and rates of progression.

The ASAI allows rate-of-change values to be calculated. Difference scores calculated using longitudinal data from the index will provide valuable information. Linking the index to temporal markers for individuals will allow researchers to understand patterns that typify the progression of sexual involvement from the earliest behaviors (hugging and spending time) through intercourse. Such data are not otherwise currently available in any form. The utility of this possibility needs to be examined empirically.

Much attention may be given to understanding factors that influence the temporal sequencing of behavior and the progression of sexual involvement. ASAI index change scores can be readily applied to studying psychosocial predictors. Finding psychosocial indicators that predict the general slope of increase over time for groups, and that predict increases and decreases in index values for individuals over time will significantly strengthen our understanding of sexual behavior.

We conclude the ASAI to be a promising tool that deserves further examination and application. The ASAI has high internal consistency. Future research can use the ASAI in longitudinal descriptive and predictive studies. Future research may wish to address issues of validity more directly.

Table I.

Demographic characteristics of participants (N = 2461)