Predicting Discordance Between Self-reports of Sexual Behavior and Incident Sexually Transmitted Infections with African American Female Adolescents: Results from a 4-city Study

Abstract

This study examined correlates of the discordance between sexual behavior self-reports and Incident Sexually Transmitted Infections. African American adolescent females (N = 964) from four U.S. cities were recruited for an HIV/STI prevention trial. Self-reported sexual behaviors, demographics, and hypothesized psychosocial antecedents of sexual risk behavior were collected at baseline, 6-, 12-, and 18-month follow-up assessments. Urine specimens were collected and tested for three prevalent STIs (chlamydia, gonorrhea, trichomonas) at each assessment. Seventeen percent of participants with a laboratory-confirmed STI reported either lifetime abstinence or recent abstinence from vaginal sex (discordant self-report). Lower STI knowledge, belief that fewer peers were engaging in sex, and belief that more peers will wait until marriage to have sex were associated with discordant reports. Discordance between self-reported abstinence and incident STIs was marked among African American female adolescents. Lack of STI knowledge and sexual behavior peer norms may result in underreporting of sexual behaviors.

Introduction

African American adolescents experience disproportionately high rates of sexually transmitted infections (STIs) [1]. Of particular concern are rates of STIs among young African American women; recent national estimates indicate that African American female adolescents between the ages of 15 and 19 experience the highest rates of chlamydia and gonorrhea [2]. Research that informs HIV and STI prevention efforts and public health policy involves many challenges. A key challenge is the accurate assessment of sexual behavior. Research on sexual behaviors is typically limited to an examination of self-reports of past behavior. The accuracy of self-report data may be impacted by a variety of factors including the cognitive demands of recalling past behaviors and motivational biases that can lead people to mis-report their behavior [3–5].

Test–retest reliability studies examine the consistency of self-reported behavior. A review of these studies suggests that sexual behavior can be reported consistently, but there has been considerable variability in reliability estimates [3]. Studies with adolescents show that kappa reliability estimates for lifetime sexual behaviors (i.e., presence or absence of a behavior across one’s lifetime) have varied from .34 to 1.0 depending on the behavior assessed and the study’s sample [6–10]. For studies that have measured reliability for counts of past sexual behaviors, findings have been mixed. Two studies with adolescents have found modest agreement for frequency counts of sexual intercourse occasions [ICC = .44 [9] and kappa = .48 [11]], with more robust test–retest reliability for number of sexual partners [ICC = .81 [9]].

Whereas test–retest studies provide evidence regarding self-report consistency, validity studies compare self-reported behaviors to objective indicators of sexual behavior. For example, Kahn and colleagues compared young women’s self-reports of abnormal cytology following a Pap smear to medical chart data and found that 14% of those with abnormal cytology provided inaccurate self-reports [12]. Incident STIs have also been compared to self-reported sexual behaviors as an indicator of validity [13]. For instance, one study found that 15% of men and 23.5% of women who reported always using a condom tested positive for an STI [13, 14]. Biological markers of sexual activity including prostate specific antigen (PSA) and Y-chromosome polymerase chain reaction (Yc-PCR) have been utilized to examine the validity of self-reported sexual intercourse and condom use. In such studies, results suggest that a subset of participants who report either no sexual behavior or consistent condom use (i.e., a condom used during each penile-vaginal sexual episode) have evidence of sperm in vaginal fluid, suggesting either inaccurate self-report [15–18] or condom use errors [19–21]. Although these studies point to discrepancies between self-report and objective markers of sexual risk behaviors, less is known about factors affecting the validity of self-reported sexual behaviors among adolescents.

Research examining the validity of self-reported sexual behaviors has focused largely on the role of individual characteristics (e.g., age, gender) and situational factors (e.g., survey modality) [3, 5, 22]. Adolescents may be particularly prone to inaccurate reporting of health behaviors [22]. Indeed, a minority of youth acknowledges dishonesty in survey responses such as initially reporting being sexually active and then subsequently denying sexual activity [23–25]. Studies have also compared rates of reported behaviors between groups or assessment types. Such studies assume that higher reports of sexual risk behaviors are indicative of increased validity because of individuals’ tendency to underreport risk behaviors [4, 5, 22, 26]. For example, gender differences among adolescents show that males typically overreport (and females underreport) sexual risk behaviors [23]. Overall, studies that have compared rates of self-reported risk behaviors across assessment modalities found that computerized assessment formats typically result in greater risk behavior disclosure relative to paper-and-pencil questionnaires and face-to-face interviews [3]. Thus, individual characteristics and the conditions under which the assessment is conducted may impact the validity of self-report of sexual behaviors.

Although an understanding of personal and situational factors that may affect sexual behavior self-report validity is important, few studies have examined the extent to which involvement in an HIV/STI prevention intervention or theoretically derived mediators of sexual behavior influence self-report validity. Participation in a sexual risk reduction intervention may increase social desirability bias [22] and result in underreporting of sexual risk behaviors. Within the context of a randomized controlled trial, one would anticipate that participants assigned to the HIV/STI prevention condition would be more prone to underreport risk behaviors relative to participants not exposed to sexual health promotion programming. However, to our knowledge, no studies have investigated whether participation in a sexual health intervention is associated with differences in self-report validity. An individual’s sexual health knowledge, attitudes, and beliefs may also influence their disclosure of sexual behaviors. There is a limited literature to suggest that constructs from Social-Cognitive Theory [27–29] may be associated with discordance between self-reported sexual behaviors and objective biological markers of sexual activity. For example, the perceived social norms for sexual behaviors may affect the degree to which young individuals report sexual activity [3, 4], with individuals more likely to endorse sexual behaviors consistent with peers’ perceived normative behaviors [3]. Other social-cognitive constructs including self-efficacy and knowledge relevant to behaviors being assessed may also affect the validity of adolescents’ self-report. Indeed, one study found that lower knowledge regarding HPV was associated with inaccurate self-reports regarding abnormal cytology findings following a pap smear [12]. Thus, given the importance of self-report, this study examined factors that may impact the validity of African American female adolescents’ self-reported sexual behaviors. In addition to examining individual-level (i.e., demographic characteristics) factors, this study explored whether social cognitive constructs or participation in a sexual risk reduction intervention with repeated assessments were associated with differences in the validity of self-reported sexual activity.

Methods

Participants

African American adolescent females (N = 964) were recruited in four mid-sized cities: Syracuse, NY (n = 237, 25%), Providence, RI (n = 247, 26%), Macon, GA (n = 267, 28%), and Columbia, SC (n = 213, 22%). Participants were 14–17 years old (M age = 15.1, SD = 1.0). The majority (93%) were living in their family’s home, most often with their mother (86%) and less frequently with their father (21%). Most participants (74%) qualified for a free or reduced price lunch at school. Eighty-four percent were in high school (i.e., grades 9 through 12), 15% were in junior high (i.e., grades 7, 8), and 1% were not in school at the time of study enrollment.

Procedures

Participants were part of a larger study evaluating sexual risk-reduction interventions for African American adolescents (Project iMPPACS) [30]. The design of Project iMPPACS was a 2 (sexual risk reduction or a general health promotion intervention) by 2 (media present or media absent) by 6 (time: baseline, 3, 6, 12, and 18, and 36 months post-recruitment). African American adolescents were recruited in cohorts of 25–30 youth for random assignment to one of two interventions: Focus on Youth (FOY), the small-group treatment condition [31] or Promoting Health Among Teens (PHAT), the control status condition [32]. The media program was randomly assigned to one city in each region. Youth in the media cities were exposed to the media intervention for at least 18 months after their recruitment and participation in the small-group intervention. Participants were recruited from partnering community based organizations, through peer referral, and via respondent driven sampling. Eligibility criteria included: (a) self-identifying as African American; (b) 14–17 years of age; and (c) ability to understand English. Analyses reported in this article are limited to the female participants. Written informed consent from adolescents and also their parent/legal guardian was obtained prior to study participation.

Data were collected via audio computer assisted self-interviews (ACASI) prior to randomization to study condition. In this study, ACASI allowed participants to view the question on a laptop computer screen and also hear the question via headphones. Participants were compensated $30 for completion of the baseline ACASI and $40 after completing each of the two half-day small group (13–15 participants) intervention sessions. Follow-up assessments were conducted at 3-, 6-, 12-, 18-, and 36-months post-intervention; participants were compensated $30, $40, $50, $60, $60 at each assessment, respectively. Participants also provided urine specimens that were tested for three STIs (chlamydia, gonorrhea, and trichomonas) at baseline, and 6-, 12-, and 18-month follow-ups. Analyses for the current study are limited to the baseline, 6-, 12-, and 18-month follow-ups. Adolescents with a positive STI test received CDC recommended single-dose directly observed treatment and prevention counseling at each time point. Retention rates of females at the 6-, 12-, and 18-month surveys were 92, 92, and 94%, respectively. The Institutional Review Boards at all participating institutions approved all study protocols.

Measures

Demographics

Demographic characteristics assessed were age, ethnicity, current living arrangement, eligibility for a free or reduced price lunch, religious service attendance (at least once a month versus never/rarely), and grade in school.

Self-reported Sexual Behaviors

Using items from previous research [33–35], we assessed sexual activity status. To assess lifetime oral sex, all participants were asked two dichotomous items: (a) whether they had ever “put your mouth on a partner’s private parts (some people call this ‘giving oral sex’)” and (b) whether “a partner ever put their mouth on your private parts (some people call this ‘receiving oral sex’)?” Responses to these items were combined to form a single, dichotomous indicator of lifetime oral sex. To assess lifetime vaginal sexual intercourse, all participants were asked whether they had “…ever had vaginal sex (that is, when a boy put his penis in a girl’s vagina).” To assess lifetime anal sexual intercourse, all participants were asked whether “…a male ever put his penis in your bottom or anus (some people call this receiving ‘anal sex’).” For each of the four behaviors participants endorsed a lifetime history of, they then indicated the last time they engaged in each behavior. Response options were dichotomized to indicate whether or not the behavior occurred in the past 6 months (i.e., the interval between follow-up assessments).

Laboratory Confirmed STIs

Urine specimens were tested for chlamydia, gonorrhea, and trichomonas. Becton–Dickinson’s propriety DNA amplification technology, called Strand Displacement Amplification (SDA), was used to test for chlamydia and gonorrhea. A real-time polymerase chain reaction (PCR) assay tested for trichomonas. DNA urine tests for chlamydia, gonorrhea, and trichomonas are highly sensitive and specific. All urine specimens were processed at the Emory University Microbiology Laboratory. Presence or absence of any of the three STIs at each time point was recorded.

Study conditions

Intervention condition was coded to indicate intervention conditions (a) STI/HIV prevention condition (coded as 1) or general health control condition (coded as 0), and (b) media city (coded as 1) or non-media city (coded as 0). Additionally, study condition assignment was coded as 1 for only the follow-up assessments after the treatments were administered and 0 at the baseline assessment prior to randomization.

Knowledge

Knowledge of STI transmission and treatment was measured with a measure developed and validated for use with older adolescents [36]. For each of 10 statements, participants indicated whether the statement was “mostly true” or “mostly false.” If a participant did not know, she was instructed to respond with “don’t know.” Each item was scored for correctness (“don’t know” was scored as an incorrect response). A total score on the baseline measure was calculated (higher scores indicated greater STI knowledge).

Sex Refusal Self-efficacy

Five items were used to measure the extent to which participants could refuse vaginal sex under various conditions (e.g., if your partner is pressuring you to have sex) [37]. For each item, participants rated their confidence to refuse sex using a four-point scale (i.e., “definitely could not say no,” “probably could not say no,” “probably could say no,” “definitely could say no”). Item scores were summed on the baseline measure (higher scores indicated greater self-efficacy). Cronbach’s alpha at baseline was .78.

Condom Use Self-efficacy

Five items measured participants’ self-efficacy to use condoms in different situations (e.g., when using another form of birth control) [38]. For each item, participants indicated their self-efficacy to use a condom using a four-point scale (i.e., “not at all sure,” “somewhat sure,” “sure,” “extremely sure”). If participants reported that they had never had sex before, they were asked to respond by imagining that they were in each of the hypothetical scenarios. Items were summed on the baseline measure (higher scores indicated greater self-efficacy to use condoms). Cronbach’s alpha at baseline was .81.

Peer Norms for Sexual Behaviors

Two items assessed perceived peer norms for engaging in vaginal sex [31]. Using a six-point scale (1: “none” to 6: “almost all of them”), participants were asked to report the number of same age peers who: (a) have had vaginal sex; and (b) are planning to wait until marriage to have vaginal sex. Higher scores on the baseline measure for each item indicated participants believed more of their same age peers engaged in the behavior.

Honesty

At the end of the ACASI, participants reported how honest they were in responding to the survey questions. Response options were: “completely dishonest,” “somewhat dishonest,” “somewhat honest,” and “completely honest”. Higher scores were indicative of greater honesty.

Data Analyses

Descriptive statistics were first performed to characterize self-reported sexual behavior engagement at each assessment. Second, descriptive statistics were calculated to measure STI incidence at each assessment and cumulatively across the trial. Next, we examined prevalence of self-reported abstinence among those with one or more laboratory-confirmed STIs. Contingency table analyses were performed to characterize the proportion of those with an STI at each time point who reported: (a) lifetime abstinence from vaginal sex; (b) abstinence from vaginal sex during the past 6 months (i.e., the follow-up interval prior to testing positive for an STI); (c) lifetime abstinence from oral, vaginal, and anal sex; and (d) abstinence from oral, vaginal, and anal sex during the past 6 months (i.e., the follow-up interval prior to testing positive for an STI).

An indicator of discordant self-reported abstinence was then created by identifying: (a) presence of one or more laboratory-confirmed STI; and (b) self-reported abstinence from vaginal sex (the behavior most likely to result in an STI detected via urine specimen). Discordant reports were instances where the following was reported: (a) lifetime abstinence from vaginal sex during the baseline survey or (b) abstinence from vaginal sex during the past 6 months at the 6-, 12-, or 18-month follow-up surveys. For the follow-up assessments, self-reported abstinence during the past 6 months was selected because participants had been previously tested, and if diagnosed with an STI, treated with oral antimicrobial therapy during the preceding follow-up period.

Analyses were conducted to compare discordant to non-discordant reports. First, bivariate statistics compared discordant and non-discordant reports on demographic characteristics, study condition assignment, social cognitive constructs, and reported honesty. Independent samples t-tests examined differences between groups for continuous variables and Chi-square analyses examined differences between groups for categorical outcomes.

Multivariate logistic regression analyses then examined whether reporting status differed by study condition assignment or assessment point. Next, factors significantly associated with differences in discordant reporting at the bivariate level were then entered into a logistic regression analysis with study condition assignment and assessment point. The STATA cluster command was utilized to obtain a robust estimate that adjusts for within-cluster correlation to account for repeat incident STIs among participants across assessment points [39, 40].

Results

Descriptive Findings

Table 1 presents descriptive data regarding participants’ engagement in oral, vaginal, and anal sex throughout their lifetimes and during the past 6 months at each of the assessment points. Self-reported engagement in each of the sexual behaviors increased across the study. For example, 31.6, 43.2, 49.3, and 55.1% of participants reported engaging in vaginal sex during the past 6 months at the baseline, 6-, 12-, and 18-month assessments, respectively. Table 2 provides descriptive statistics of STI incidence for each assessment point. Chlamydia was the most frequently diagnosed STI with the rate of incident infections ranging from 5.1 to 8.1% across the assessment points. There was lower incidence for gonorrhea (range: 1.2–1.6%) and trichomonas (range: 1.7–2.6%) infections. Collectively, from the baseline to 18-month assessment, between 7 and 10.6% of young women were diagnosed with at least one STI. There were 234 unique participants with at least one STI at one or more assessment points (N = 322 incident STI occasions). Across the assessments, 70.1% (n = 164) of participants had one STI incident occasion; 22.6% had two STI incident occasions (n = 53); 6.8% had three STI incident occasions (n = 16); and .4% (n = 1) had four STI incident occasions.

Table 1.

Prevalence of self-reported sexual behavior by assessment occasion

	Baseline (N = 964)	6 months (n = 886)	12 months (n = 889)	18 months (n = 906)
Lifetime
Oral sex	356 (36.9%)	412 (46.5%)	499 (56.1%)	565 (62.4%)
Vaginal sex	438 (45.4%)	472 (53.3%)	531 (59.7%)	599 (66.1%)
Anal sex	125 (13.0%)	115 (13.0%)	135 (15.2%)	151 (16.7%)
Any sexual behavior (oral, vaginal, or anal sex)	506 (52.5%)	545 (61.5%)	614 (69.1%)	677 (74.7%)
Past 6 months
Oral sex	222 (23.0%)	289 (32.6%)	366 (41.2%)	435 (48.0%)
Vaginal sex	305 (31.6%)	383 (43.2%)	438 (49.3%)	499 (55.1%)
Anal sex	68 (7.1%)	71 (8.0%)	84 (9.4%)	91 (10.0%)
Any sexual behavior (oral, vaginal, or anal sex)	357 (37.0%)	429 (48.4%)	499 (56.1%)	566 (62.5%)

Frequencies (corresponding percentage) are presented for each of the sexual behaviors

Table 2.

Incidence of sexually transmitted infections (STIs) by assessment occasion

	Baseline (N = 964)	6 months (n = 885)	12 months (n = 884)	18 months (n = 899)
Chlamydia	77 (8.0%)	50 (5.7%)	45 (5.1%)	73 (8.1%)
Gonorrhea	15 (1.6%)	13 (1.5%)	11 (1.2%)	12 (1.3%)
Trichomonas	25 (2.6%)	15 (1.7%)	19 (2.1%)	23 (2.6%)
Any STI	99 (10.3%)	65 (7.3%)	62 (7.0%)	96 (10.6%)

Frequencies (corresponding percentage) are presented for each of the sexually transmitted infections

Discordant Self-Report: STI Incidence and Self-Reported Abstinence

Table 3 displays the proportion of participants with an STI at each assessment who reported abstinence from: (a) vaginal sex during their lifetime and in the past 6 months; and (b) oral, vaginal, and anal sex in their lifetime and in the past 6 months. Cumulatively, from baseline to the 18-month follow-up period, 13.9% of individuals with a laboratory-confirmed STI reported no lifetime history of vaginal sex. Of participants with a laboratory-confirmed STI, 9.6% reported no lifetime engagement in oral, vaginal, or anal sex. Across the 6-, 12-, and 18-month follow-up assessments, 15.7% of adolescents with an STI reported abstinence from vaginal sex during the reporting period of the past 6 months. Similarly, across follow-up assessments, among individuals with an STI, 13.0% reported abstinence from all sexual behaviors during the past 6 months. Across all assessments there were three participants who were discordant in their self-reported behaviors at more than one time point.

Table 3.

Self-reported sexual behavior among participants with one or more laboratory-confirmed STIs

	STI at Baseline (N = 99)	STI at 6 months (N = 65)	STI at 12 months (N = 62)	STI at 18 months (N = 96)
No lifetime vaginal sex	21 (21.2%)	6 (9.2%)	9 (14.5%)	9 (9.4%)
No vaginal sex during past 6 months	–	8 (12.3%)	14 (22.6%)	13 (13.5%)
No lifetime oral, vaginal, or anal sex	14 (14.1%)	5 (7.7%)	6 (9.7%)	6 (6.3%)
No oral, vaginal, or anal sex in past 6 months	–	8 (12.3%)	10 (16.1%)	11 (11.5%)

Frequencies (corresponding percentage) are presented for each of the sexual behaviors

Correlates of Discordant Self-Report

Bivariate Analysis

Table 4 presents bivariate analyses comparing discordant reports (n = 56) to non-discordant reports (n = 266).¹ Bivariate analyses indicated that discordant reports were associated with: (a) younger age, t (320) = 1.9, p = .05; (b) lower STI knowledge, t (320) = 3.3, p = .001; (c) belief that fewer peers were currently having vaginal sex, t (320) = 3.2, p = .002; and (d) belief that more peers would wait until marriage to have vaginal sex, t (320) = −3.6, p = .001. There were no differences between discordant and concordant reports based on: (a) eligibility for free/reduced price lunch; (b) regular religious service attendance; (c) study site; (d) intervention condition; (e) media condition; (f) sex refusal self-efficacy; (g) condom use self-efficacy; or (h) honesty level.

Table 4.

Reporting status as a function of participant characteristics, study condition, sexual risk behavior antecedents among those with an STI (N = 322): bivariate results

	Discordant (n = 56)	Concordant (n = 266)
Demographic characteristics
Age*	15.1 (1.0)	15.4 (.9)
Eligible for free/reduced price lunch	35 (62.5%)	193 (74.5%)
Regular religious service attendance	31 (55.4%)	126 (47.4%)
Northern site	19 (33.9%)	118 (44.4%)
Study conditions
HIV/STI prevention condition	32 (57.1%)	127 (47.7%)
Media city	34 (60.7%)	165 (62.0%)
Sexual risk behavior antecedents
Sex refusal self-efficacy	17.6 (3.7)	17.5 (3.5)
Condom use self-efficacy	16.3 (4.1)	15.4 (4.1)
STI knowledge***	4.2 (2.2)	5.2 (2.0)
Peer norm: engage in vaginal sex**	4.2 (1.5)	4.8 (1.2)
Peer norm: wait until marriage for vaginal sex***	2.9 (1.5)	2.2 (1.1)
Honesty level	3.4 (.9)	3.5 (.9)

Means (standard deviations) are presented for continuous variables. Frequencies (corresponding percentage) are presented for dichotomous outcomes. Independent samples t-tests were used to assess differences between continuous variables. χ² analyses were used to assess differences in dichotomous outcomes

^*p < .05;

^**p < .01;

^***p < .001

Multivariate Analyses

Table 5 displays findings from the multivariate logistic regression analysis examining whether reporting status differed based on: (a) intervention condition; (b) media condition; (c) the interaction between intervention and media conditions; and (d) assessment point. There were no differences in reporting status as a function of: (a) intervention condition; (b) media condition; or (c) the interaction between intervention and media condition. Similarly, reporting status did not differ based on the assessment point.

Table 5.

Reporting status as a function of intervention condition, media condition, intervention × media city interaction, and assessment point among those with an STI (N = 322): logistic regression results

Predictor	Odds Ratio (SE)	95% CI	p value
Assigned conditiona
FOY intervention condition	1.56 (.96)	.47, 5.18	.47
Media city	.87 (.49)	.29, 2.63	.81
FOY intervention condition × Media city	.89 (.70)	.19, 4.19	.89
Assessment pointb
6 months assessment	.46 (.27)	.14, 1.47	.19
12 months assessment	1.00 (.54)	.35, 2.88	.99
18 months assessment	.52 (.32)	.16, 1.72	.29
Constant	.27 (.07)	.17, .44	0

CI confidence interval

^aFor assigned condition, the general health promotion condition and non-media city are the comparator conditions

^bThe baseline assessment is the comparator for the assessment point

As displayed in Table 6, a second multivariate logistic regression analysis then compared reporting status based on assigned condition, assessment point, and baseline social cognitive constructs significant at the bivariate level including STI knowledge and perceived peer norms. In the multivariate logistic regression analysis that controlled for age, lower STI knowledge (AOR = .82, 95% CI = .70–.96; p = .015), belief that fewer peers were engaging in vaginal sex (AOR = .76, 95% CI = .61–.96; p = .018), and belief that more peers will wait until marriage to have vaginal sex (AOR = 1.41, 95% CI = 1.12–1.76; p = .003) were independently associated with discordant reporting.

Table 6.

Reporting status as a function of assigned condition, assessment point, participant characteristics, and social cognitive constructs among those with an STI (N = 322): logistic regression results

Predictor	Odds ratio (SE)	95% CI	p value
Assigned conditiona
FOY intervention condition	1.90 (1.23)	.53, 6.79	.32
Media city	.91 (.53)	.29, 2.85	.87
FOY intervention condition × Media city	.71 (.58)	.14, 3.52	.67
Assessment pointb
6 months assessment	.42 (.26)	.12, 1.44	.17
12 months assessment	1.17 (.66)	.39, 3.51	.78
18 months assessment	.43 (.27)	.13, 1.46	.18
Demographic characteristics
Age	.83 (.14)	.59, 1.16	.29
Social Cognitive Constructsc
STI knowledge	.82 (.07)	.70, .96	.015
Peer norm: engage in vaginal sex	.76 (.09)	.61, .96	.018
Peer norm: wait until marriage for vaginal sex	1.41 (.16)	1.12, 1.76	.003
Constant	15.47 (41.62)	.08, 3018.7	.31

CI confidence interval

^aFor assigned condition, the general health promotion condition and non-media city are the comparator conditions

^bThe baseline assessment is the comparator for the assessment point

^cAll Social Cognitive constructs were assessed at the baseline assessment

Discussion

Sexual health research typically relies upon self-reports of past behavior. Results from this study highlight discrepancies between adolescents’ self-reported sexual behaviors and STI incidence. Indeed, a minority of female adolescents diagnosed with an STI reported sexual abstinence both across their lifetime and also during the past 6 months. Across assessment occasions, 5–10% indicated either no lifetime history of sexual activity or no recent sexual behavior. These findings corroborate previous research demonstrating discordance between self-report and biological markers of sexual behavior [13–18].

An examination of factors associated with discordant self-reports demonstrated that demographic characteristics, study condition, assessment point, and reported honesty were not associated with differential levels of discordant reporting. However, lower knowledge regarding STI transmission and treatment was associated with discordant reports, similar to an earlier study that found that lower knowledge for HPV transmission was associated with inaccurate self-reports of abnormal cytology results [12]. This association may reflect overall comprehension difficulties in responding to sexual health questions. However, the reasons for this association are not fully understood and warrant further investigation.

Discordant reports were also associated with perceived sexual norms for greater abstinence among peers, suggesting that adolescents may respond to sexual behavior questionnaires in a socially desirable fashion congruent with perceived sexual behavior norms. This finding intimates motivational biases to distort self-reports of past behavior to avoid shame or embarrassment or to appear in a more favorable light [4, 22, 26, 41]. Indeed, motivational biases may be particularly pertinent to STI/HIV prevention research, where it is commonly assumed that individuals (especially females) underreport risk behaviors because of the sensitive, personal, and often times stigmatizing nature of such behaviors [4, 5, 22, 26]. Thus, young women may have over-reported sexual abstinence, a sexual practice perceived to be more socially acceptable, as a form of impression management [3].

In addition to motivational biases, the observed inaccuracy in self-report may also reflect cognitive factors, such as comprehension or recall problems [22]. Such errors may have occurred because participants did not fully understand the sexual behavior questions. However, previous research with adolescents suggests that youth can reliably respond to sexual activity questions when specific behaviors are described anatomically rather than through use of colloquial language [42]. Additionally, the use of ACASI reduces the literacy burden created by traditional paper-and-pencil surveys [43]. Thus, discordant reporting may be more likely attributable to not fully listening (or reading the question) rather than a lack of comprehension. Participants may have also encountered challenges recalling when sexual behavior last occurred. Thus, discordant self-report could reflect an inability to remember the specific date the behavior last occurred, rather than an intentional underreporting of behavior.

Limitations

Several study limitations should be acknowledged. First, although STI incidence requires sexual behavior, we did not include additional biological markers that could have allowed for greater specificity in measurement of sexual behavior (e.g., use of Yc-PCR to assess presence of semen in vaginal fluid as an indicator of recent unprotected vaginal sex). There is also some evidence that a small number of trichomonas infections may be acquired via nonsexual transmission routes, which could have inflated the level of discordant reporting [44]. Additionally, the tests used to diagnose STIs are imperfect and may result in false positive test results. Furthermore, STIs at the follow-up assessment may in part be due to treatment failure of previous STIs [45]. This study would have also benefited from inclusion of a measure of social desirability to explore the influence of this construct on self-report. We also did not assess participants’ understanding of survey items or effort; such assessments would have provided additional insight into the degree to which these factors were associated with discordant self-reports. These limitations may partially account for discordant self-reported abstinence and presence of an STI. In addition, our sample was limited to African-American female adolescents recruited from four medium-sized cities, so results may not generalize to other populations.

Conclusions

These findings highlight discrepancies between self-reported behaviors and biological markers. Some argue that the latter be used as a more objective outcome when evaluating STI/HIV interventions [46]. However, biological markers can signal risk behavior only during a recent time period (i.e., past 14 days for Yc-PCR) and the diagnostic tests are costly. Similarly, STI transmission does not occur subsequent to all unprotected sexual intercourse and, thus, is an insufficient proxy for sexual risk behavior. Thus, future research to improve the validity of self-report is essential.

A number of recommendations to improve reliability and validity of sexual health data have been suggested [5]. Weinhardt et al. recommend that sexual health surveys be designed with the assumption that respondents have misconceptions about sexual content and are likely to experience discomfort or difficulty responding to such questions. As such, future research should investigate methodological strategies to address respondents’ knowledge gaps regarding sexual health and reduce their tendency to respond in a socially desirable fashion in line with perceived peer norms to avoid embarrassment or discomfort. For example, assessing adolescents’ understanding of sexual behavior questions and clarifying any misunderstandings may increase the validity of self-reported behaviors. The use of validity checks of self-reported sexual behaviors both within and across multiple assessments may also improve data quality. In addition, use of assessment methods that enhance participants’ anonymity (e.g., use of participant identification number, computerized survey methods) and providing confidentiality assurances may decrease socially desirable responding [3]. Strategies should also be employed to stress the importance of accurate reporting throughout the survey (e.g., in questionnaire instructions) and provide repeated information regarding how study data will be utilized (e.g., analyzing data across participants rather than examining individual responses). Ultimately, improving the validity of self-reported of sexual behavior data will improve researchers’ abilities to accurately measure the efficacy of STI/HIV prevention interventions.