Recanting of substance use reports in a longitudinal prevention study
Introduction
Evaluations of school-based drug prevention programs often rely on responses to self-administered questionnaires on drug abuse (e.g., Clayton et al., 1996, Botvin et al., 1995, Ellickson and Bell, 1992, Pentz, 1993, Ringwalt et al., 1991, Rosenbaum et al., 1994, Rosenbaum and Hanson, 1998). Such studies often employ longitudinal data collection strategies to track the emergence of ‘lifetime’ substance use reports among those who are or are not exposed to prevention education. At the same time, a growing body of research underscores the lack of validity of substance use reporting in both household surveys (Fendrich and Vaughn, 1994, Mensch and Kandel, 1988) and school settings (Siddiqui et al., 1999, Fendrich and Mackesy-Amiti, 2000) by showing inconsistencies in drug use reporting. One particular type of inconsistency that has been demonstrated in previous research is the tendency for subjects to deny previously disclosed drug use when subsequent questions about that behavior are asked either within the same interview or on subsequent follow-up interview waves. This type of response behavior is particularly troublesome for longitudinal evaluation research studies designed to monitor emergent drug use. In previous research tracking young adults in the National Longitudinal Survey of Youth who have passed typical ages of drug initiation, it has been shown that recanting of previously disclosed lifetime drug use is highly prevalent, that it increases with each successive interview, and that it varies by substance and by subject characteristics (Fendrich and Kim, 2001).
At least two underlying cognitive processes known to influence survey reporting (Sudman et al., 1996) have been identified as influencing recanting of drug reporting in prior research: Recall and ‘response editing.’ Response editing is the intentional underreporting of drug use on surveys as a result of attempts to respond in a socially desirable manner (e.g., Fendrich and Mackesy-Amiti, 2000). Recall, or memory of past behavior, may be more salient for studies focused on retrospective reporting among young adults, where much of the behavior reported on may have occurred in the more distant past. Recall may be less salient for studies focused on children entering adolescence, however, since the behaviors they are reporting on are likely to be emergent and ongoing. Surveys of adolescents may be more likely to coincide with the period of drug use initiation. After identifying patterns of inconsistency within the context of the same cross-sectional school-based survey, Fendrich and Mackesy-Amiti (2000) concluded that while editing probably plays an important role, respondent lack of motivation and carelessness also need to be considered.
In explaining reporting inconsistency, a number of drug use and subject characteristics have been examined. Fendrich and Vaughn, 1994, Fendrich and Kim, 2001 showed that there was more denial of previously disclosed cocaine use than marijuana use. Siddiqui et al. (1999) found that marijuana reporting was less inconsistent than other more prevalent substances such as cigarettes and alcohol; these authors provide no comparisons with respect to cocaine. Several research studies (Fendrich and Kim, 2001, Mensch and Kandel, 1988, Siddiqui et al., 1999) have found significantly higher rates of inconsistency for minority informants. This parallels research showing lower levels of validity for delinquency measures among black respondents (although these findings are not consistent across studies; Thornberry and Krohn, 2000). While other demographic variables such as gender, age, and location have been significant correlates of inconsistency in previous research, the precise direction of these effects vary by study (Siddiqui et al., 1999).
One possible theoretical rationale for the variation across substances found by Fendrich and Vaughn (1994) is related to question threat and underreporting (Bradburn et al., 1979). Since subjects may be more likely to find questions about ‘hard’ drugs more threatening, they are more likely to underreport this type of drug use; recanting may be an indicator of denial in response to question threat. In contrast, in the context of a school survey, Siddiqui et al. (1999) suggest that inconsistency may have an inverse relationship with prevalence. These authors suggest that rarer behaviors (such as marijuana use) are reported on more consistently than common behaviors such as cigarette smoking and alcohol use since unusual behaviors are easier to recall.
The overall salience and demographic correlates of drug use recanting in studies of youth have varied across longitudinal studies (Single et al., 1975, Mensch and Kandel, 1988, Bailey et al., 1992, Fendrich and Vaughn, 1994, Pederson, 1990, Siddiqui et al., 1999, Fendrich and Kim, 2001). It should be underscored, however, that prior studies have shown considerable variation with respect to key design issues such as the number of follow-up interviews, the length of time between follow-ups, and the age of study subjects. For example, some analyses have focused on follow-ups spanning long periods of time such as 3 or 4 years between interviews (Bailey et al., 1992, Fendrich and Kim, 2001, Fendrich and Vaughn, 1994, Mensch and Kandel, 1988). Analyses employing the National Longitudinal Survey of Youth (NLSY) are generally based on data collected during young adulthood (Fendrich and Kim, 2001, Fendrich and Vaughn, 1994, Mensch and Kandel, 1988). Some studies focus exclusively on high school seniors (e.g., O'Malley et al., 1983). Other studies employ samples in early to mid-adolescence (Siddiqui et al., 1999). To our knowledge, only two prior studies have employed more than two time points in examining this phenomenon (Fendrich and Kim, 2001, Siddiqui et al., 1999).
Previous analyses of school-based cross-sectional studies have raised the possibility that exaggeration and not underreporting may affect responses (Fendrich and Johnson, 2001). In this scenario, recanting may underscore an underlying lack of validity to responses initially provided in youth school-based drug surveys. One might hypothesize that, as youth gain in maturity through the course of adolescence, they approach surveys more seriously and have a greater concern about the truthfulness of their answers. Accordingly, one might expect to see less recanting of drug use when initial disclosure is made later in adolescence.
On the other hand, as youth mature they may become more aware of the adverse consequences of disclosure. In this scenario, one might expect more recanting of drug use later in adolescence. At this point, however, little is known about the influence of maturity on the reporting process. Multi-year longitudinal studies of youth interviewed in the early stages of substance initiation and followed through late adolescence might be particularly informative about the potential role that psychosocial development and maturity plays in relation to the editing and disclosure process. Furthermore, the emergence of older adolescents in such a sample ensures sufficient variation for exploring this issue for harder drugs such as cocaine, which generally show a very low base rate in surveys of young adolescents (Fendrich and Johnson, 2001).
With these concerns in mind, we undertook an analysis of the recanting of substance use in a panel study that included a pre-test and seven follow-up waves of interviews. The study was originally designed to evaluate the Drug Abuse Resistance Education (DARE) program in the State of Illinois. This data set is particularly useful for its frequency of measurement, geographic diversity and, most importantly, for the developmental span in which youths were assessed. Youths were first interviewed in early adolescence (5th and 6th grade). Interviews continued on a yearly basis through the 11th and 12th grade. Our focus was on one particular type of inconsistency, non-disclosure of previously disclosed lifetime drug use. We refer to such behavior as ‘recanting,’ although previous studies have used the label ‘denial’ to classify the same behavior (Fendrich and Kim, 2001). In particular, we were interested in characterizing the extent of recanting in this study, examining its variability across substances, and identifying demographic correlates of this behavior. To our knowledge this is the first study to track drug use recanting through the entire course of adolescence. As such, it facilitates a potential examination of the role of developmental maturation on reporting behavior.
Section snippets
Sample
The DARE evaluation study from which this analysis is based consists of one pre-test and seven planned post-tests; details of the study design have been explained elsewhere (Rosenbaum et al., 1994, Rosenbaum and Hanson, 1998). The original evaluation design enlisted 18 pairs of schools, with one school in the pair receiving the intervention and the other serving as a control school. School pairs were matched by school type, ethnic composition, number of students with limited English proficiency
Sample comparisons
Table 1 displays the demographic distribution for the 1405 youth included in the analysis sample and compares them with the demographic distribution of the 1798 youth who were in the original study. The analysis sample is predominantly urban and suburban, with the majority of students classifying themselves as white. Most of the students in the analysis sample were in the 6th grade at baseline and reported living in households with both their mother and their father. There was a fairly even
Discussion
One important limitation of this study is that it is focused on a sample attending schools in one particular state (Illinois) in the United States during one particular time period. To some extent this limits the overall generalizability of the findings. Note that the cultural bias of this particular study—i.e., its focus on student data from the United States, is one shared by nearly all previous analyses of the recanting phenomenon.
It is not clear that these findings would be replicated in
Acknowledgements
This research was supported by National Institute On Drug Abuse Grants # R01DA09286 and R01DA12425. Zoe Ma provided valuable technical assistance on all phases of the data analysis. In addition, the authors wish to thank the anonymous reviewers for their constructive suggestions.
References (31)
- et al.
The effectiveness of drug abuse resistance education (project DARE): five year follow-up results
Prev. Med.
(1996) - et al.
Multivariate analysis of retest artifact in the National Longitudinal Survey of Youth drug use
Drug Alcohol Depend.
(2001) - et al.
Decreased drug reporting in a cross-sectional student drug use survey
J. Subst. Abuse
(2000) - et al.
Prevalence of youth substance use: the impact of methodological differences between two national surveys
Drug Alcohol Depend.
(1997) - et al.
The characterization of inconsistencies in self-reports of alcohol and marijuana use in a longitudinal study of adolescents
J. Stud. Alcohol.
(1992) - et al.
Long-term follow-up results of a randomized drug abuse prevention trial in a white middle-class population
J. Am. Med. Assoc.
(1995) - et al.
Improving Interview Method and Questionnaire Design
(1979) - et al.
Challenges to social experiments: a drug prevention example
J. Res. Crime Delinq.
(1992) - et al.
Examining prevalence differences in three national surveys of youth: impact of consent procedures, mode, and editing rules
J. Drug Issues
(2001) - et al.
Validity of drug use reporting in a high-risk community sample: a comparison of cocaine and heroin survey reports with hair tests
Am. J. Epidemiol.
(1999)
Diminished lifetime substance use over time: an inquiry into differential underreporting
Public Opin. Q.
The validity of drug use reports from juvenile arrestees
Int. J. Addict.
Applied Logistic Regression
The reliability of self-reported age of onset of tobacco, alcohol and illicit drug use
Addiction
Cited by (61)
Adolescent recanting of alcohol use: A longitudinal investigation of time-varying intra-individual predictors
2018, Drug and Alcohol DependenceCitation Excerpt :One challenge to reliability is ‘recanting’: denial of previous reports of lifetime substance use when reassessed at a subsequent time (Fendrich, 2005; Fendrich and Rosenbaum, 2003). Rates of recanting vary between 7%–85% across studies of adolescents using alcohol, nicotine, cannabis, cocaine, opiates, and ecstasy (e.g., Fendrich and Rosenbaum, 2003; Percy et al., 2005). Due to the high priority of studying determinants of substance use initiation, researchers have applied different methodological approaches to handle recanting (e.g., first endorsement carried forward).
Adolescent smoking experimentation as a predictor of daily cigarette smoking
2017, Drug and Alcohol DependenceRecanting substance use over time
2023, Journal of Substance Use