One of the criteria required for a diagnosis of attention-deficit hyperactivity disorder (ADHD) is symptom onset before age 12 years.Reference Thapar, Cooper and Rutter1 When individuals first present to clinicians as adults, this requires retrospective recall of symptoms and likely limits accuracyReference Thapar, Cooper and Rutter1,Reference Asherson and Agnew-Blais2 owing to both false positives and false negatives.Reference Henry, Moffitt, Caspi, Langley and Silva3 Identifying the optimal method of assessing ADHD age-at-onset is an important question for adult psychiatrists. We compared the accuracy of four different ways of assessing age-at-onset in a prospective population cohort. We focus on those who met the other DSM-5 criteria for ADHD at age 25 years: at least five inattentive or five hyperactive/impulsive symptoms, plus impairment.
Method
We analysed data from the Avon Longitudinal Study of Parents and Children (ALSPAC),Reference Northstone, Lewcock, Groom, Boyd, Macleod and Timpson4 which includes repeated assessments since pregnancy (see supplementary material available at https://doi.org/10.1192/bjp.2021.122). Ethical approval was obtained from the ALSPAC Ethics and Law Committee and Local Research Ethics Committees (Research Ethics Committee approval references can be found at http://www.bristol.ac.uk/alspac/researchers/research-ethics/). Informed consent was obtained from participants following the recommendations of the ALSPAC Ethics and Law Committee at the time.
In total, 138 (42% male) individuals met DSM-5 symptom and impairment criteria at age 25 with complete data on age-at-onset and ADHD symptoms assessed in childhood (see below).
Age 25 assessment
DSM-5 symptom and impairment criteria were assessed using self-reports of the Barkley Adult ADHD Rating Scale (BAARS-IV).Reference Riglin, Agha, Eyre, Bevan Jones, Wooton and Thapar5,Reference Barkley6 Parents also completed the BAARS-IV: these data were used for sensitivity analyses (see below).
The BAARS-IV uses two sets of questions for age-at-onset: (a) specify age – individuals were asked to recall as precisely as possible at what age (in years) ADHD symptoms began to occur; and (b) rate behaviour between 7 and 12 years – individuals were asked to rate the frequency of 18 DSM-5 ADHD symptoms on a 4-point scale.
We generated four retrospective definitions of ADHD age-at-onset before age 12 years:
(a) the specified age that ADHD symptoms began to occur was before 12 years old;
(b) at least one ADHD symptom was endorsed as having been clinically significant (occurring ‘often’ or ‘very often’Reference Barkley6) between 7 and 12 years;
(c) several symptoms (at least three) were endorsed as having been clinically significant between 7 and 12 years (DSM-5 requires that ‘several’ inattentive or hyperactive/impulsive symptoms present prior to age 12 years7);
(d) at least six inattentive or six hyperactive/impulsive symptoms were endorsed as having been clinically significant between 7 and 12 years (DSM-5 symptom requirement for childhood ADHD7).
ADHD symptoms assessed during childhood
Symptoms had been assessed when these adults were aged 7, 8, 9 and 12 years using the five-item ADHD subscale of the Strengths and Difficulties Questionnaire (SDQ)Reference Goodman8 rated by parents, as children's self-reports are unreliable.Reference Goodman, Ford, Richards, Gatward and Meltzer9 The SDQ is a screening questionnaire with symptoms in the past 6 months categorised as low (0–5), slightly raised (6–7) or high (8–10).Reference Goodman8 Individuals with slightly raised or high symptoms (≥6) at any of these ages were defined as having ADHD symptoms when assessed in childhood: this was used to test the accuracy of adult retrospective reports of age-at-onset. This broad definition was used given the DSM-5 requirement that ‘several’ symptoms present prior to age 12 years.7
Measures for sensitivity analyses
There were two measures used for sensitivity anlaysis: (a) ADHD assessed during childhood, defined based on full ADHD diagnosis at age 7 or 10 years of age, measured using the parent-rated Development and Well-Being AssessmentReference Goodman, Ford, Richards, Gatward and Meltzer9 (see supplementary material); and (b) age 25 assessments of age-at-onset using the parent-rated BAARS-IV.
Analyses
Receiver operating characteristic (ROC) curve analyses using Stata's roccomp function were used to examine the validity of the four retrospective assessments of ADHD age-at-onset in distinguishing those with versus those without ADHD symptoms when assessed in childhood.
Results
Of those who met DSM-5 criteria for adult ADHD symptoms and impairment at age 25 (n = 138), when asked to specify the age at which symptoms started 51% (n = 71) reported onset before age 12 years. When asked to rate behaviour between 7 and 12 years, 86% (n = 119) retrospectively reported at least one ADHD symptom, 72% (n = 100) reported at least three symptoms and 44% (N = 61) retrospectively reported six inattentive and/or six hyperactive/impulsive symptoms.
Results for the four ADHD age-at-onset assessments are shown in Table 1. All approaches showed low discrimination in identifying ADHD symptoms assessed in childhood (AUC = 0.57–0.68), although there was evidence that this varied across the four approaches (χ(3) = 8.99, P = 0.03).
Table 1 Discrimination of retrospective assessments of ADHD age-at-onset criterion in distinguishing those with and without ADHD symptoms when assessed in childhood, in young adults with ADHD symptoms and impairment at age 25 years
ADHD, attention-deficit hyperactivity disorder; ROC, receiver operating characteristic; AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value.
Reporting at least one symptom showed the highest sensitivity (the proportion of those with symptoms when assessed in childhood correctly identified by retrospective reports) and negative predictive validity (NPV: the proportion of those retrospectively reported not to have childhood-onset correctly identified) and the lowest specificity (the proportion of those without symptoms when assessed in childhood correctly identified by retrospective reports) and positive predictive validity (PPV: the proportion of those retrospectively identified who did have symptoms when assessed in childhood). Conversely, retrospectively endorsing at least six inattentive or six hyperactive/impulsive childhood symptoms showed the highest specificity and PPV, whereas specifying age showed the lowest sensitivity and NPV.
Sensitivity analyses
Sensitivity analysis where ADHD assessed in childhood was defined based on full diagnostic criteria (n = 122) showed a similar pattern of results although with somewhat higher discrimination (AUC = 0.60–0.81: χ(3) = 96.00, P = 1 × 10−20) (supplementary Table 1). Parent retrospective reports of age-at-onset at age 25 (n = 47) showed fairly low discrimination (AUC = 0.63–0.70) with little evidence of variation across the four approaches (χ(3) = 1.19, P = 0.76) (supplementary Table 2).
Discussion
We found variation in the discrimination of four approaches to retrospectively assess ADHD age-at-onset at age 25 years, although all showed limited validity. This is consistent with Brazilian birth-cohort findings.Reference Breda, Rohde, Menezes, Anselmi, Caye and Rovaris10 Of the four approaches, the highest proportion of participants met age-at-onset criteria when this was defined based on asking participants to retrospectively rate their behaviour between ages 7 and 12 years, and requiring the endorsement of at least one of the 18 DSM ADHD symptoms: this definition (which does not fit with the DSM-5 requirement that ‘several’ symptoms present prior to age 12 years) resulted in the highest proportion of true positives (highest sensitivity) but also the fewest true negatives. Conversely, the highest specificity (and lowest proportion of people identified) was found using the most stringent definition: the retrospective endorsement of at least six inattentive and/or six hyperactive/impulsive childhood symptoms.
The approach of asking participants to specify the age at which endorsed symptoms started resulted in the fewest true positives, i.e. this missed the most people who had ADHD symptoms when assessed in childhood. This provides tentative evidence that asking people to recall specific symptoms during a specific age period is preferable to recalling the age at which symptoms started. However, none of the four approaches showed high accuracy, which is consistent with previous work highlighting the limitations of retrospective recall.Reference Henry, Moffitt, Caspi, Langley and Silva3 Sensitivity analyses defining ADHD assessed in childhood based on full DSM-5 diagnostic criteria (and requiring the retrospective endorsement of six inattentive and/or six hyperactive/impulsive childhood symptoms) showed moderate discrimination. This suggests that recall of more severe and impairing symptoms may be better than recall of just a few specific symptoms. In practice there is likely benefit in asking about specific ADHD symptoms in childhood and acquiring additional information from other sources, such as school reports.
Although the age-at-onset criterion for ADHD is important from a developmental perspective,Reference Thapar, Cooper and Rutter1 our results, alongside increasing evidence of ‘late-onset’ ADHD,Reference Asherson and Agnew-Blais2 raise queries about its validity. Further research is needed to address the limitations of the current work, including limited sample size and non-random attrition. Defining age-at-onset is important for informing adult psychiatrists and diagnostic criteria.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2021.122.
Data availability
The ALSPAC data management plan (www.bristol.ac.uk/alspac/researchers/data-access/documents/alspac-data-management-plan.pdf) describes in detail the policy regarding data sharing, which is through a system of managed open access.
Acknowledgements
We thank all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.
Author contributions
L.R. and A.T. conceived and designed the study. L.R. analysed the data and wrote the first draft. All authors contributed to the interpretation of data for the work and provided critical revisions. All authors read and approved the submitted manuscript.
Funding
The UK Medical Research Council and the Wellcome Trust (grant ref.: 217065/Z/19/Z) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and L.R. and A.T. will serve as guarantors for the contents of this paper. A comprehensive list of grant funding is available on the ALSPAC website (www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf). The primary outcome measures for age 25 data used in the paper were specifically funded by the Wellcome Trust (204895/Z/16/Z). R.B., G.D.S. and E.S. work in a unit that receives funding from the University of Bristol and the UK Medical Research Council (MC_UU_00011/1 and MC_UU_00011/3). This research was funded by the Wellcome Trust (204895/Z/16/Z). For the purpose of open access, the authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.
Declaration of interest
None.
eLetters
No eLetters have been published for this article.