The Accuracy of Retrospective Recall of Childhood ADHD: Results from a Longitudinal Study

Attention-deficit/hyperactivity disorder (ADHD) is a common condition in children and adolescents (worldwide prevalence: 2.6–4.5%, Polanczyk, Salum, Sugaya, Caye and Rohde 2015). Over the last two decades, longitudinal studies have repeatedly shown that a substantial proportion of children with ADHD (41% - 77%) continue to struggle with symptoms and functional impairments in adulthood (Faraone, Biederman and Mick 2006; Sibley, Swanson, Arnold, Hechtman, Owens, Stehli, et al. 2017; Uchida, Spencer, Faraone and Biederman 2018). Yet, despite increased awareness of ADHD as a life-long condition, adult ADHD is often underdiagnosed and undertreated (Kooij, Bijlenga, Salerno, Jaeschke, Bitter, Balázs, et al. 2019). To facilitate classification in adolescents and adults, the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association (APA) 2013) introduced some important revisions of the diagnostic criteria for ADHD (e.g., Epstein and Loren, 2013). For instance, the DSM-5 provides examples describing how ADHD symptoms may manifest across the lifespan. In addition, the symptom threshold was reduced to five symptoms (instead of the previous six) for individuals older than 17 years, and the age-of-onset criterion now requires several inattentive or hyperactive-impulsive symptoms to be present prior to age 12 years.

In the 4th edition of the DSM (DSM-IV-TR; APA 2000), the age-of-onset criterion specified that some symptoms and impairments needed to be present before age 7, but a large number of studies questioned the utility of this criterion (see Kieling, Kieling, Rohde, Frick, Moffitt, Nigg, et al. 2010). The DSM-IV field trials and other studies (e.g., Applegate, Lahey, Hart, Biederman and Hynd 1997; Barkley and Biederman, 1997; Todd, Huang and Henderson 2008) showed that many adolescents and adults do not accurately recall ADHD-related impairments that occurred before age 7. Similarly, Kessler, Berglund, Demler, Jin, Merikangas and Walters (2005) showed that only 50% of adults with clinical features of ADHD retrospectively recalled the presence of ADHD symptoms before age 7, while 95% recalled a symptom onset before age 12, suggesting that the decision to extend the age-of-onset criterion in the DSM-5 will improve the identification of adult ADHD. The recently published 11th edition of the International Classification of Diseases (ICD-11; World Health Organization [WHO] 2018) even went a step further by requiring a symptom onset “during the developmental period, typically early to mid-childhood”, instead of specifying an upper age limit.

The European Network Adult ADHD (ENAA) proposes that the diagnosis of adult ADHD should include the assessment of childhood-onset and current symptoms of ADHD using self-reports and collateral information from significant others (Kooij et al. 2019). However, problems with retrospective recall of ADHD symptoms remain. It has consistently been shown that informants disagree not only regarding the rating of current ADHD symptoms but also when asked to recall the severity and onset of ADHD symptoms in childhood (Breda, Rovaris, Vitola, Mota, Blaya-Rocha, Salgado, et al. 2016; Dias, Mattos, Coutinho, Segenreich, Saboya and Ayrão 2008; Henry, Moffitt, Caspi, Langley and Silva 1994; Kooij, Boonstra, Swinkels, Bekker, De Noord and Buitelaar 2008; Moffitt, Houts, Asherson, Belsky, Corcoran, Hammerle, et al. 2015). Moreover, longitudinal studies have revealed that retrospective reports of ADHD are often inaccurate. For instance, Barkley, Fischer, Smallish and Fletcher (2002) found that adults’ retrospective self-reports of ADHD symptoms and parent ratings collected in childhood were only moderately correlated (r = .39 to r = .44). In a study by Loney, Ledolter, Kramer and Volpe (2007), participants rated their ADHD symptoms in adolescence (average age 14 years) and again in young adulthood (average age 21 years). The results revealed that participants’ self-ratings in adolescence did not correlate significantly with their retrospective age-14 self-ratings made during the follow-up in young adulthood. In a longitudinal study by Mannuzza, Klein, Donald, Bessler and Shrout (2002), adults participated in semi-structured psychiatric interviews designed to generate a lifetime diagnosis of ADHD. About one fifth (22%) of the participants did not receive a retrospective diagnosis of childhood ADHD according to DSM-III-R criteria (APA 1987), although all participants had been diagnosed with ADHD in childhood. Similarly, a longitudinal study by Miller, Newcorn and Halperin (2010) examined the recall of childhood ADHD in a sample of adolescents and young adults with a childhood diagnosis of ADHD. The authors found that only 63% of participants and 78% of parents reported sufficient childhood symptoms to substantiate a past diagnosis of ADHD according to the DSM-IV (i.e., at least six childhood symptoms of inattention and/or hyperactivity-impulsivity prior to age 7). Together, these longitudinal studies demonstrate that retrospective reports of ADHD symptoms are, at most, moderately correlated with childhood ratings, and that even in samples with a confirmed childhood diagnosis of ADHD, a substantial proportion of adult participants (22–37%) falsely denies the presence of childhood ADHD symptoms.

Two large birth cohort studies confirmed that inaccurate reports of childhood ADHD are common, even if the age-of-onset criterion is extended to symptom onset before the age of 12 (as suggested by the DSM-5). Findings from the Dunedin Multidisciplinary Health and Development Study (Moffitt et al. 2015) revealed that only 23% of adults with a confirmed childhood diagnosis of ADHD had parents who correctly recalled that their offspring had core ADHD symptoms or had been diagnosed with ADHD before the age of 12 (true-positive recall). The parents of the remaining participants with confirmed childhood ADHD (77%) did not recall that their offspring had core ADHD symptoms before the age of 12 (false-negative recall). Only 4% of non-ADHD comparison subjects had parents who incorrectly recalled ADHD core symptoms during childhood (false-positive recall). In the 1993 Pelotas Birth Cohort study (Breda, Rohde, Menezes, Ansemi, Caye, Rovaris, et al. 2019), individuals with a current ADHD syndrome at the age of 22 (i.e. at least five ADHD symptoms in the last 6 months; symptoms in two or more settings; ‘a lot of’ or ‘very much” impairment) were asked about the presence of symptoms before the age of 12. Only 33% of the individuals with confirmed childhood symptoms positively endorsed the presence of at least several ADHD symptoms in childhood (true-positive recall), and 67% falsely denied the presence of childhood symptoms (false-negative recall). Of the individuals without childhood ADHD, 30% positively endorsed the presence of childhood symptoms (false-positive recall), and 70% denied the presence of childhood symptoms (true-negative recall). Thus, the results of these birth cohort studies suggest that inaccurate recall of childhood ADHD is a common phenomenon and that false-negative recall was more frequently observed (67–77% of individuals with confirmed childhood ADHD) than false-positive recall (4–30% of individuals without childhood ADHD; Breda et al. 2019, Moffitt et al. 2015).

Since inaccurate symptom recall may lead to misdiagnosis and impede access to appropriate treatment options, knowledge about the factors influencing the accuracy of retrospective ADHD symptom recall is needed in order to inform clinical decision makers about potential biases. To date, only a small number of studies have focused on this issue. Breda et al. (2019) investigated factors associated with true- and false-recall of ADHD symptoms in the above-mentioned population-based birth cohort sample. False-negative endorsement of childhood ADHD was more likely in male participants, non-white participants, and participants with less years of schooling. False-positive endorsement of childhood ADHD was more likely in individuals with social phobia and current ADHD symptoms. Similarly, the longitudinal study by Miller et al. (2010) found that the severity of current ADHD symptoms, but not the severity of childhood symptoms, was positively associated with the accuracy of retrospective symptom recall in a sample of adolescents and young adults with confirmed childhood ADHD (Miller et al. 2010).

We first analyzed changes in ADHD symptom severity over time. Based on the existing literature (e.g., Faraone et al. 2006), we expected that ratings of ADHD symptom severity would show a significant reduction from childhood to young adulthood. Next, we examined cross-informant agreement. Based on previous evidence, we expected to find low to moderate correlations between participants’ and collateral informants’ retrospective ratings of childhood symptoms (Breda et al. 2016; Dias et al. 2008; Henry et al. 1994; Murphy and Barkley 1996; Zucker, Morris, Ingram, Morris and Bakeman 2002). Similarly, we expected to find low to moderate correlations between self- and proxy ratings of current ADHD symptoms in adulthood (Kooij et al. 2008; Murphy and Barkley 1996; Mörstedt, Corbisiero, Bitto and Stieglitz 2015).

We then examined the accuracy of ADHD symptom recall. Based on previous studies, we expected low to moderate correlations between ratings of ADHD symptom severity collected in childhood and retrospective ratings of childhood symptoms (Barkley et al. 2002; Loney et al. 2007). Given the substantial proportions of false-negative recall documented in previous studies, we further expected that both adult participants and proxy respondents recall significantly fewer ADHD childhood symptoms than were originally reported by parents when participants were aged 6 to 10 years. However, since the present sample was clinically referred and treated for ADHD, we expected the proportion of participants and proxy respondents who false deny the presence of at least several ADHD childhood symptoms to be lower than in the Pelotas Birth Cohort study (67% false-recall; Breda et al. 2019) and the Dunedin Multidisciplinary Health and Development Study (77% false-recall; Moffitt et al. 2015). Subsequently, we examined whether the severity of ADHD symptoms and ADHD-associated impairments in adulthood influence the accuracy of retrospective symptom ratings. There is evidence that recall of past symptoms is fostered by emotional and physical distress (Barsky 2002). We therefore expected that participants with more severe ADHD symptoms and higher ADHD-associated impairments in adulthood would provide higher ratings of childhood ADHD.

The diagnostic criteria for ADHD according to the DSM-5 (APA 2013) require the presence of at least several symptoms prior to age 12. To evaluate this criterion in adult patients, clinicians often have to rely on retrospective recall of ADHD symptoms by patients and significant others who knew the patients during the developmental period, usually one or both parents. However, it has repeatedly been shown that a substantial proportion of adolescents and adults do not accurately recall their childhood functioning (e.g., Applegate et al. 1997; Kessler, Berglund et al. 2005). The aim of the present study was therefore to investigate (a) the accuracy of retrospective ADHD childhood ratings provided by adult participants and their parents or significant others, and b) the influence of ADHD symptom severity and ADHD-associated impairments in adulthood on retrospective ADHD symptom ratings in order to inform clinical decision makers about potential biases.

Participants (N = 55) were members of the CAMT study who had been diagnosed and treated for ADHD during childhood and were reassessed in adulthood when they were 22 to 32 years of age (18-year FU). We first analyzed changes in ADHD symptom severity over time. As we expected based on previous evidence (e.g. Faraone et al. 2006), both self- and proxy ratings of adult ADHD were substantially and significantly lower than parent ratings of ADHD collected in childhood (η_p² = .73 and η_p² = .76), suggesting that ADHD symptoms lessened from childhood to adulthood. Also consistent with previous evidence (e.g. Döpfner, Hautmann, et al. 2015), proxy ratings of hyperactivity-impulsivity showed a substantially greater decline over time compared to ratings of inattention (η_p² = .18). Our analyses further suggest that this observed decline cannot be attributed to the fact that a substantial proportion (37.5%) of proxy ratings of adult ADHD was provided by informants other than parents. We therefore conclude that our data confirm previous findings of a developmental reduction in ADHD symptomatology (e.g., Langberg, Epstein, Altaye, Molina, Arnold and Vitiello 2008), although we cannon completely rule out that other factors (e.g., changes in item formulations, altered perception of ADHD-related behaviors) influenced our results.

We then examined cross-informant agreement for ADHD symptom ratings. Previous studies found that adults report more ADHD childhood symptoms than do their parents or other informants (Magnússon, Smári, Baldursson, Kristjánsson, Sigurbjornsdóttir and Guomundsson 2006; Murphy and Scharchar 2000), but that they tend to underreport current ADHD symptoms (Kooij et al. 2008; Zucker et al. 2002). The results of the present study differ from these findings in an important aspect: Adult participants’ retrospective ratings of their childhood symptoms were substantially and significantly lower than proxy respondents’ retrospective ratings (d = 0.60), and their ratings of current symptoms of Hyperactivity-Impulsivity differed only slightly and not statistically significant from those provided by other informants (d = 0.15). Consistent with previous findings, adult participants underreported current symptoms of Inattention compared to proxy respondents, although the size of this effect can be considered small (d = 0.35). Also consistent with previous evidence (Kooij et al. 2008; Mörstedt et al. 2015; Murphy and Barkley, 1996; Zucker et al. 2002), there was a moderate correlation (r = .44) between participants’ and proxy respondents’ retrospective ratings of childhood symptoms and a somewhat (but not significantly) higher correlation (r = .63) between participants’ and other informants’ ratings of current symptoms in adulthood. Together, these analyses suggest that, in our sample, cross-informant agreement was moderate for retrospective ratings of childhood ADHD and high for ratings of current adult ADHD.

As an index of recall accuracy, we first compared retrospective symptom ratings with ratings of ADHD symptom severity collected in childhood (parent rating). Participants retrospectively described their childhood symptoms of Inattention and Hyperactivity-Impulsivity as substantially less severe than their parents had done when the participants were children (d = 0.66 and d = 0.91). In contrast, retrospective ratings of Inattention provided by parents and other proxy respondents differed only slightly and not statistically significant from parents’ childhood ratings (d = 0.07). Their retrospective ratings of Hyperactivity-Impulsivity also differed only slightly and not statistically from parents’ childhood ratings (d = 0.31), but equivalence testing suggested that our design may lack sufficient statistical power to detect a meaningful effect. In line with previous findings (Barkley et al. 2002; Miller et al. 2010; Loney et al. 2007), participants’ retrospective symptom ratings did not correlate statistically significantly with parent ratings provided in childhood (r = .24, ns), while proxy respondents’ retrospective symptom ratings correlated moderately with parents’ previous ADHD ratings (r = .39). Interestingly, as noted above, adult participants’ retrospective symptom ratings correlated moderately with proxy respondents’ retrospective ratings (p = .44), suggesting that adult patients’ symptom recall is more closely associated with how their parents or significant others recall and communicate about childhood symptoms than with how parents judged the severity of ADHD symptoms in childhood.

As another index of recall accuracy, we determined how many informants falsely denied the presence of at least several ADHD symptoms in childhood. Consistent with previous studies (e.g., Mannuzza et al. 2002; Miller et al. 2010), adult participants and proxy respondents recalled significantly fewer childhood symptoms than the parents had reported at study entry. This effect can be considered large for participants (d = 1.15) and moderate for proxy respondents (d = 0.51). However, the majority of the adult participants and proxy respondents (68% and 76%, respectively) reported sufficient childhood symptoms to substantiate a past diagnosis of ADHD (≥ 6 ADHD childhood symptoms). Even more participants and proxy respondents (83% and 90%, respectively) accurately recalled the presence of at least several ADHD childhood symptoms (defined as the presence of at least three ADHD symptoms between the ages of 6 and 12 years). Yet, 17% of participants and 10% of proxy respondents falsely denied the presence of at least three childhood symptoms, even though all participants had received multimodal treatment for their childhood ADHD including six sessions of psychoeducation in all treatment arms (Döpfner et al. 2004). As we expected, recall accuracy in nonclinical samples has been found to be even lower. For example, results from the Dunedin Multidisciplinary Health and Development Study revealed that only 23% of parents of adults with a confirmed childhood diagnosis of ADHD recalled that their offspring had core symptoms or had been diagnosed with ADHD before the age of 12 (Moffitt et al. 2015). Similarly, results from the 1993 Pelotas Birth Cohort study showed that only 33% of adult participants with childhood ADHD recalled the presence of childhood symptoms (Breda et al. 2019). We therefore conclude that having received clinical diagnosis and treatment during childhood improves the accuracy of retrospective symptom recall.

In addition, we determined the number of participants who over- or underreported ADHD childhood symptoms. Underreporting of childhood symptoms was a common phenomenon. The majority of participants (79%) and proxy respondents (65%) recalled less childhood symptoms than parents had reported at study entry. Our data do not confirm previous reports of symptom exaggeration in 25–48% of college students self-referred for ADHD evaluation (Sullivan, May and Galbally 2007), possible because the present sample was not faced with secondary gain potentials such as academic accommodation or other forms of assistance.

To sum up, we found that proxy respondents recalled significantly fewer ADHD childhood symptoms than parents had reported at study entry, but when severity ratings were considered, proxy respondents’ retrospective ratings correlated moderately with, and did not differ significantly from, parents’ previous ratings. In contrast, adult participants’ retrospective self-ratings were substantially lower than, and did not correlate with, their parents’ ratings at study entry. Adult participants were also more likely to underreport ADHD childhood symptoms (79%) and to falsely deny the presence of at least several ADHD childhood symptoms (17%) compared to proxy respondents (65% underreporting; 10% false-negative recall). These results might suggest that participants have limited retrospective recall of childhood symptoms. However, we did not collect self-ratings in childhood, and hence compared participants’ retrospective ratings to parent ratings collected in childhood. Since cross-informant agreement on retrospective ADHD ratings has repeatedly been shown to be at most modest (see above), a comparably lower level of accuracy for retrospective self-ratings in our data was to be expected.

Several limitations of the present study need to be mentioned. First, our sample size was restricted, which limited the power to detect small effects and thus limited the interpretation of our results. Second, since our sample was referred and treated for ADHD in childhood, our data can provide information on the percentage of participants that falsely denied the presence of childhood symptoms, but it is not suitable for detecting false positive or true negative reports of childhood ADHD. In addition, our results cannot be generalized to samples not treated for ADHD. Symptom recall might be less accurate in adults who did not undergo diagnostic assessment and treatment during childhood, since the severity of childhood symptoms has been found to be associated with recall accuracy (Kessler, Adler, Barkley, Conners, Faraone, Greenhill, et al. 2005). Moreover, adults who were being told about their ADHD symptoms by a professional during childhood might be more likely to remember having had these symptoms. The generalizability of results is further limited by the fact that 51 of 55 participants (93%) were male. Future studies should recruit both clinically referred and population-based samples of children with and without ADHD that include both male and female participants to study the effects of gender and prior diagnosis and treatment on the accuracy of retrospective ADHD symptom recall. Another limitation worth mentioning is the difference between the FBB-ADHS and the FEA-FFB / FEA-FSB. The first version of the FBB-ADHS, which was used at pre-intervention, contained 23 items assessing the occurrence of ADHD symptoms according to the diagnostic criteria of DSM-III-R and the preliminary research diagnostic criteria of ICD-10. The FEA-FSB and the FEA-FFB each contain 20 items assessing the occurrence of ADHD symptoms according to the DSM-5 and ICD-10. We were, therefore, not able to perform a direct comparison of the specific symptoms that tended to be under- or overreported at the 18-year FU.

Despite these limitations, our study adds valuable information about the accuracy of ADHD symptom recall. Our results question the validity of retrospective recall of childhood ADHD symptoms, since retrospective symptom ratings showed only low to moderate correlations (p = .24 – p = .39) with ratings of childhood symptoms assessed at ages 6 to 10 years. Moreover, 17% of adults with a childhood diagnosis of ADHD and 10% of proxy respondents falsely denied the presence of at least three childhood symptoms. We further demonstrated that participants’ retrospective symptom recall was associated with the severity of current ADHD symptoms. This finding fits well with the concept of a state-dependent recall, which postulates that recall of past symptoms and episodes is affected by a patient’s emotional and physical state at the time of recall (Barsky 2002). The relationship between current distress and recall of past emotions or symptoms has previously been supported by evidence of a negative recall bias in patients with depression (Ben-Zeev and Young, 2010), panic disorders (Margraf, Taylor, Ehlers, Roth and Agras 1987), borderline personality disorder (Ebner-Priemer, Kuo, Welch, Thielgen, Witte, Bohus and Linehan 2006), or acute pain (Eich, Reeves, Jaeger and Graff-Radford 1985).

A possible clinical implication of our findings is that adult patients who report low levels of current ADHD symptomatology might underestimate their childhood symptoms. Clinicians should be vigilant for underreporting of childhood symptoms among patients with ADHD symptoms who do not regard their behavior as ADHD-related, and may try to increase recall accuracy by asking the patient to describe specific situations and / or by providing examples of specific symptom behaviors. These examples may serve as retrieval cues that remind the patient about childhood behaviors he or she might otherwise have overlooked. In addition, clinicians may consider accepting a comparably low number of retrospectively recalled childhood symptoms as sufficient for fulfillment of the age-of-onset criterion of the DSM-5. Our data suggest that the diagnosis of ADHD should not be dismissed in adults who have a full ADHD syndrome but recall less than three ADHD childhood symptoms. Finally, the present findings demonstrate that cross-informant agreement for retrospective ratings of childhood ADHD is at most moderate, and therefore support the recommendation of the European Network Adult ADHD (ENAA) that information from different sources should be used when diagnosing ADHD in adults (Kooij et al. 2019).