5.1.1 Length of Recall Period
Participants are typically asked questions about their behavior over a specific time period. To increase the accuracy of retrospective self-reports, researchers have recommended recall periods of 3 months or less. Research has shown sexual behaviors can be reliably assessed by self-report measures for intervals as long as 3 months but reliability decreases at longer intervals (Kauth et al, 1991). In a study comparing daily diaries to a SAQ 1, 2, and 3 months after diary completion, Graham et al (2003) found recall of condom use was stable across the 3-month period but participants reporting more frequent condom use had more errors. Jaccard et al (2002) compared weekly mailed self-report questionnaires with retrospective reports at 1, 3, 6, and 12 months finding no difference between type of method at the 3- and 12-month assessments, but not at the 1- or 6-month assessments. They concluded that retrospective SAQs accurately represent behavior for at least 3 months. Little evidence suggests that recall periods longer than 3 months provide accurate information (Schroder et al, 2003a; Sheeran and Abraham, 1994).
Reviews of the sexual risk behavior literature show that 28-49% of studies ask participants to recall condom use over a 3-6-month time frame whereas 15-18% provide no specific time frame (Noar et al, 2006; Sheeran and Abraham, 1994). Noar et al (2006) suggest that specific and brief recall periods should yield optimal responses but additional research is necessary to evaluate the reliability of various recall periods. Moreover, research shows high-frequency behaviors may be more difficult to recall over longer periods of time, whereas rarer behaviors may not occur over short recall periods. Since low-frequency behaviors are more salient, participants may be inclined to believe an event occurred more recently (i.e., telescoping) potentially leading participants to exaggerate sexual behaviors. Patterson and Strathdee (2005) recommended using absolute frequency counts for rare behaviors and relative frequency measures (e.g., proportion of time condom was used) for more frequent behaviors.
To enhance recall of retrospective self-reports, Weinhardt et al (1998) suggest using three strategies: (1) Provide anchor dates for recall periods, (2) Encourage participant to use calendars to aid the recall of memorable events, and (3) Prompt participant recall of extensive periods of sexual behavior (e.g., abstinence, consistent sexual activities). All three of the suggested strategies may be accomplished using the timeline follow-back (TFLB; Sobell and Sobell, 1996) procedure. The TLFB uses calendars marked with landmark events, personally meaningful dates, and other memory aides to facilitate accurate recall. Because of the interactive format used for the TLFB, memory is aided by the recalling of one event in reference to another event. Using the TLFB, behavioral patterns are recorded in greater detail and over multiple time points. Sexual risk information yielded by the TLFB include number of sexual partners, frequency of sexual events (vaginal, anal, oral), frequency of unprotected sexual events (vaginal, anal, oral), alcohol and/or drug use prior to sex, number of occasions and quantity of alcohol and/or drug use prior to sex, and STI history. Research has confirmed the stability of the 3-month retrospective self-reports using the
TLFB procedure (Carey et al, 2001; Weinhardt et al, 1998a).
Failure to specify and/or define type of partner or sexual act may result in inaccurate self-reports of sexual behavior. Researchers recommend using measures that are specific to sexual partners and specific to sexual acts, rather than general measures (Fishbein and Pequegnat, 2000; Schroder et al, 2003b; Sheeran and Abraham, 1994). In Sheeran and Abraham's (1994) review of condom use measures, they found 79% of the measures did not specify type of partner and most (65%) did not specify the type of sexual act being assessed (i.e., vaginal, anal, or oral sex). An updated review of the literature examining 56 studies found 57% of measures did not specify partner type, 16% specified primary versus non-primary partners, and 16% were tailored to partner type (Noar et al, 2006); most studies reported the type of sexual activity (67%). Because different levels of risk are associated with various sexual practices, it is important for researchers to specify both sexual partner and sexual act to increase the accuracy of the data.
Research examining aggregate (i.e., summed across all sexual partners) versus partner-specific (i.e., questions specific for each sexual partner) formats has found partner-specific SAQs produce more accurate self-reports of sexual behavior than do aggregate question formats (McAuliffe et al, 2007). Thus, partner specificity (i.e., name of actual partner rather than primary versus non-primary labels) "may help cue the recall of sexual activities by providing both a context and a focus for past experiences and events" and improve the accuracy of self-reported sexual behavior (McAuliffe et al, 2007). Moreover, the use of multiple terms, without clear definitions, for partners (primary, steady, exclusive, regular versus secondary, casual, nonexclusive) makes comparisons between studies challenging.
Was this article helpful?