%20(1).webp)
"My mum died in 2019. My dad wasn't around. I've got one sister and she's four years younger than me, so she wouldn't remember."
You have thirty minutes left, a patient who meets every other criterion comfortably, and a diagnostic framework that requires you to establish something about their life before the age of twelve. This is not an edge case. In a service seeing adults in their forties and fifties it is close to routine, and how you handle it separates a report that holds up from one that gets returned.
Collateral history is the least discussed part of adult ADHD assessment and the part most likely to be picked apart. This guide covers how to run an informant interview properly, how to weigh what you are told, and what to do when there is no informant to interview.
Collateral history is everyone's job in an assessing service, but it lands differently depending on your role:
Two failure modes, pulling in opposite directions:
Adults with ADHD are poor historians of their own symptoms, in both directions. Some under-report, because a lifetime of difficulty has been normalised into personality — you cannot identify a deficit in something you have never experienced otherwise. Others over-report, having recently found a framework that explains thirty years of confusion and now reading every memory through it.
Neither is dishonesty. Both are what happens when you ask someone to rate the frequency of a behaviour from twenty-five years ago on a five-point scale.
The research on retrospective recall bears this out consistently: adults' accounts of their own childhood ADHD symptoms correlate only modestly with contemporaneous accounts. This is precisely why NICE NG87 and DSM-5 both push toward corroboration, and why the DIVA-5 structure builds an informant column into the instrument rather than treating it as optional.
The requirement is simple: someone who knew the patient well, before age twelve, across more than one context if possible.
In descending order of usefulness:
Note who is missing from that list. A partner met at twenty-eight is not a Criterion B informant. Neither is a close friend from university. They may be excellent witnesses to current impairment, and their account is worth having for Criteria C and D — but they cannot speak to onset, and a report that uses them as though they can invites exactly the query you are trying to avoid.
Record the relationship, the years known, and specifically which period of childhood the informant can speak to. An older sibling who left home when the patient was nine has a window, and the window has an edge.
Confirm consent. The patient consents to you speaking to the informant; the informant consents to the information being used and recorded. Both should be documented, and it takes one line.
Then explain the purpose, because informants routinely misunderstand it. Parents in particular often arrive braced for judgement — they assume you are assessing their parenting, or that a diagnosis implies they did something wrong. A sentence at the start defuses most of this: I'm not assessing how they were brought up. I'm trying to build a picture of what they were like as a child, because that is part of what the diagnostic criteria ask for.
Say explicitly that you want examples rather than judgements, and that "I don't remember" is a perfectly acceptable answer. An informant who feels obliged to have an opinion about all eighteen symptoms will manufacture one.
Telephone is fine. Video is fine. Written responses are the weakest option, because you cannot follow up, and following up is where most of the value is.
Structure matters here more than in the patient interview, because informants wander. Six sections, roughly thirty minutes.
Be specific on the losing-things item in particular. "Did they lose things?" produces a shrug; "how many PE kits, coats or lunchboxes did you replace in a year?" produces a number. Numbers go in the report.
Accident history is an underused prompt. Parents who cannot rate impulsivity on a scale will often tell you, unprompted, about three visits to A&E before the age of ten.
Presentation shapes what an informant noticed. Disruptive children got recorded; quiet ones often did not. Our guides to ADHD symptoms in young boys and signs in young children and girls set out the asymmetry that explains why so many adult referrals are women in their forties.
End with these, and write down the answers verbatim:
How would you describe them as a child, in a few words?
What did teachers or school reports say about them?
These two questions produce more usable material than the structured items above them. "Bright but never finished anything." "Away with the fairies." "A lovely boy who could have done so much better." Those phrases go into the report in quotation marks, and they do more evidential work than any rating scale.
The structure above is a list of symptoms, and reading a list of symptoms to a willing parent is a reliable way to generate false positives. Informants want to help. If you ask "was he easily distracted?", a parent who is hoping their adult child finally gets an answer will very often say yes.
Three habits protect against this:
Open before closed. Start each domain with an open question and only use the symptom list to fill gaps. "What was homework like in your house?" gets you a scene. "Did he have difficulty sustaining attention on homework?" gets you a yes.
Ask for the example first, the judgement second. If an informant endorses a symptom, ask what it looked like. An endorsement with no example behind it should be recorded as an endorsement with no example behind it — which is to say, weakly.
Ask about the comparison group. ADHD criteria are developmental: the behaviour has to be out of step with peers, not merely present. "Compared with his brother at the same age" or "compared with the other children in the class" turns an absolute judgement into a relative one, which is what the criteria actually ask for.
It is also worth asking at least one question in the negative — something the patient probably did not struggle with. An informant who endorses everything, including the item you expected them to reject, is telling you about their expectations rather than the patient's childhood.
A parent's answers are calibrated against the world they were parenting in, not against DSM-5.
Someone raising a child in a class of thirty-six in 1985 has a different threshold for "restless" than a parent in a class of twenty-two today. Corporal punishment, streaming, and much higher tolerance for boys being disruptive all shaped what got noticed and what got recorded. A parent may genuinely report that nothing was unusual, while the school reports describe a child who was repeatedly disciplined for the same thing.
Cultural expectations matter in the same way. In some families and communities, academic underperformance is attributed to effort as a matter of course, and behavioural difficulty is a discipline question rather than a health one. A parent operating from that frame is not withholding — they are answering a different question from the one you think you asked.
The way round it is to ask about consequences rather than perceptions. Not "was his behaviour a problem?" but "was he ever kept behind, moved seats, sent out, or discussed at a parents' evening?" Events are less filtered than judgements.
It happens often, and it is information rather than a problem.
The most common pattern is a parent who minimises. Sometimes this is generational — the behaviours described were unremarkable in a 1980s classroom of thirty-four children. Sometimes it is protective, of the child or of themselves. Occasionally a parent has their own undiagnosed ADHD and is comparing their child to a baseline that was never typical.
The reverse also occurs: a patient who recalls a placid childhood while a parent describes chaos. Adults who were consistently criticised as children often remember the criticism rather than the behaviour.
Do not resolve the discrepancy by picking a side. Record both accounts, note the discrepancy explicitly, and say how you weighed it. A report that says "the patient's account and his mother's differed on the severity of childhood restlessness; her account was consistent with the school reports provided, and was given greater weight" is a stronger document than one where the disagreement quietly disappeared.
ADHD is substantially heritable, so a meaningful proportion of the parents you interview will have it themselves, usually undiagnosed. This affects the interview in two ways, and both are worth anticipating.
The first is calibration. A parent comparing their child against their own childhood is comparing against a baseline that was never typical. "He was no worse than I was" is offered as reassurance and is often the single most diagnostically interesting sentence in the call. Follow it, gently.
The second is recall. A parent with untreated inattentive symptoms may struggle to retrieve specific dated examples, not because the events did not happen but because they were never well encoded. Where you suspect this, lean harder on documentary sources and on structured prompts rather than free recall, and be careful not to read a thin account as a negative one.
Occasionally the interview prompts a parent to recognise themselves. Have a sentence ready about how they might pursue that through their own GP — it is not your assessment to run, but leaving it unacknowledged is unkind.
Back to the patient at the start of this article. No parents, a sibling too young to remember, no contact with extended family. This does not end the assessment, and it does not oblige you to diagnose or to decline. It obliges you to work down the evidence hierarchy and to write down where you got to.
School records. Many people can obtain these, and many have never tried. Schools and local authorities hold records for varying periods, and a subject access request is free. It is worth pausing the assessment for six weeks to get them.
Other childhood documentation. GP records from childhood. Educational psychology involvement. Social care records. Any of these may contain a contemporaneous observation, even where ADHD was never mentioned.
Adult records with a long tail. Employment records, university records, or a psychiatric history stretching back to late adolescence can establish a pattern earlier than the current presentation.
Detailed, probed self-report. Weaker, but not worthless — and considerably stronger when it is anchored to verifiable landmarks. Rather than "were you inattentive at school?", ask which sets they were in and whether that matched their ability, whether they were ever kept behind, what their reports said, whether they moved schools and why. Specific, checkable memories are more reliable than global self-assessments.
Then write the limitation into the report. Something like: No childhood informant was available; both parents are deceased and no sibling could speak to the relevant period. School records were requested but no longer held. Criterion B is therefore supported by the patient's detailed account of streaming and school reporting, which was internally consistent and specific, and this limitation should be borne in mind.
That paragraph is the difference between a report that survives scrutiny and one that does not. It does not weaken the diagnosis. It demonstrates that the clinician knew exactly what the evidence would and would not carry.
Three habits, all cheap:
For GPs and trainees new to this: the corroboration summary is not a formality. It is the paragraph an ICB reviewer reads first when deciding whether to accept a diagnosis for shared care.
The corroboration summary is the last step: is childhood onset supported, is the account consistent with self-report, and have any discrepancies been explored? Three questions, three lines, and Criterion B is documented rather than assumed.
No. DSM-5 requires evidence of onset before age 12, not a specific source for it. An informant is the strongest way to obtain that evidence, but school records, childhood documentation or a detailed and clearly caveated self-report can support the criterion where no informant exists.
Anyone who knew the patient well before age 12 — most often a parent, but an older sibling, grandparent, aunt, uncle or long-standing family friend can serve. A partner or friend met in adulthood cannot speak to onset, though their account is useful for current impairment.
Record both, note the discrepancy, and state how you weighed it. Discrepancies are common and often informative. Resolving them silently is what creates problems later.
Around thirty minutes when structured. Unstructured, it can take twice as long and cover less.
Yes. Telephone and video are both appropriate and are often the only practical option. Written questionnaires are weaker because you cannot probe for examples.
