Adult ADHD is the diagnosis I am least comfortable making. And the longer I practice, the more I think that discomfort is appropriate, because it is an accurate reading of what the diagnosis asks you to establish rather than a gap in anyone's training.
Look at what you are being asked to do. Confirm a pattern that began in childhood, in an adult who was never evaluated in childhood and whose memory of it is thirty years old. Establish impairment across multiple settings, from someone who can only give you their own account of those settings. And clear a differential that includes most of the rest of psychiatry, because inattention, restlessness, poor follow-through, and emotional dysregulation are among the least specific symptoms we deal with. There is no biomarker. There is no confirmatory test. There is a history, and how carefully you take it.
Then add the part nobody trained us for. The patient frequently arrives already certain, or already diagnosed somewhere else. That certainty is not a reason to dismiss them and it is not a reason to agree with them, but it is a pressure on the evaluation and it is worth naming as one.
The floor
Before the hard part, the simple part. To diagnose ADHD in an adult you ultimately need enough current symptoms, five or more in one domain for someone seventeen or older, persisting for at least six months. You need several of them present before age twelve. You need them showing up in more than one setting. You need clear evidence that they interfere with functioning rather than merely being present. And you need a presentation that is not better explained by something else.
That is the diagnostic floor, and written out it looks almost trivially easy to apply. The rest of this post is about why every one of those five is genuinely hard to establish in a real adult sitting in front of you, and where I have watched the reasoning go wrong, including my own.
One: childhood onset, when nobody has records
This is the most distinctly adult problem. The criterion is childhood onset, and your evidence is a thirty-year-old memory belonging to the person who wants the diagnosis.
Start with what does not settle it. The absence of a childhood diagnosis is not evidence of absence. Enormous numbers of people, girls and women especially, went through school in eras and settings where inattentive presentations were simply not identified, and "nobody ever said anything" tells you about the school, not the child. But the reverse trap is just as real. A retrospective explanation offered by an adult who has recently read about ADHD is not automatically evidence either. Both of those are true at once, and holding both is most of the skill.
What actually helps is asking for texture rather than conclusions. Not "did you have trouble focusing in school," which invites the answer the patient already has in mind, but what homework looked like at their kitchen table. Whether things got finished at midnight or not at all. Whether report cards said capable but not applying himself, which is close to a period-specific diagnostic phrase. Whether they lost things constantly, whether teachers moved their desk, whether they were the kid who read a novel under the desk during math. Specific scenes give you far more clinically useful information than a general impression, and they let you test the history for consistency across time and settings.
"I was gifted so nobody noticed" deserves a particular kind of attention, because it is both a real phenomenon and a very convenient story. It is genuinely true that high ability can mask ADHD for years, especially in structured, low-demand environments where a bright kid can coast. It is also the explanation available to anyone who wants the diagnosis and has no supporting history. What separates them is cost, so the cost behind the success is what I go looking for. Everything done the night before. Work that took three times longer than it took peers. An exhausted parent running the executive function from outside. If the grades were good and the process behind them was unremarkable, the masking story is thinner than it sounds.
When the patient remembers little, say so plainly and go looking. A parent, an older sibling, an old report card in a box. When parents deny symptoms but the patient describes elaborate compensation, you have a genuine conflict rather than an answer, and it is worth sitting with. Parents underreport for their own reasons, and adults reinterpret their childhoods through whatever framework they most recently encountered. Neither account is automatically the honest one.
Two: impairment versus trait
If I could get one thing across, it would be this section. It is where I see the most diagnostic drift, in both directions.
Procrastination, losing keys, hating boring tasks, needing a deadline to start, hyperfocusing on something interesting, forgetting why you walked into a room, drifting during meetings. Very large numbers of people recognize themselves in that list, and recognition is not diagnosis. If the criteria were satisfied by identification with ADHD-flavored experiences, a substantial share of the adult population would meet them.
The criterion is not presence, it is interference. So the useful questions are about consequence. What has this actually cost? What failed repeatedly, not once during a bad year? Were there jobs lost, or nearly lost, or a pattern of leaving before being asked to. Relationships strained by the same unfinished thing. Bills, licensure deadlines, taxes. Did they need accommodations or scaffolding that other people in the same role did not need, and what happened on the occasions that scaffolding was removed.
That last question is the one that most often turns a vague history into a clear one. High-functioning adults with ADHD are frequently running an elaborate compensatory system, and the system is invisible until it fails. Someone who has built their entire life around external structure, who color-codes everything, who cannot function without a partner managing the calendar, who has never once filed on time without help, is describing impairment even while presenting as competent. Meanwhile, if the same subjective experiences have never meaningfully impaired functioning, required disproportionate compensation, or carried a recurring functional cost, you have much less evidence that you are looking at disorder rather than trait.
Three: cross-situationality and compensation
More than one setting is the criterion, and it is the one most often applied too simply in both directions.
Catastrophically disorganized at home but highly effective at work does not by itself argue against ADHD. If work is intensely structured, externally paced, deadline-driven and socially monitored, and home is unstructured and unmonitored, the difference may be telling you about the environments rather than about the person. The structured setting is doing the executive work. Conversely, "I could not concentrate in college" is not by itself evidence for ADHD if college was the first time the problem appeared, because college is also the first time many people lose all external structure at once.
Which leads to what I think is the most useful reframe in this whole area. ADHD does not appear at thirty-seven. But its consequences absolutely can.
The vulnerability is lifelong. What changes is the ratio between demand and scaffolding. A person can go decades in a well-matched environment, with a structured job, a partner who runs logistics, no children, a short commute, and no crisis, and function. Then they get promoted into unstructured work. Or have a second child. Or start working from home, where the office that was silently regulating them disappears. Or the partner who ran the calendar leaves. Nothing new began. The compensation stopped covering it.
This distinction matters practically, because the history that sounds most suspicious, everything was fine until two years ago, is exactly the history that could be either a genuine late unmasking or a new mood, sleep, or substance problem wearing ADHD's clothes. You do not resolve that by deciding which sounds more likely. You resolve it by going backwards and asking what the earlier environments actually demanded, and whether the person was quietly paying a cost inside them.
Four: chronology as the differential
The mimic list, sleep and trauma and mood and substances and thyroid, is easy to recite and does very little work on its own. Almost every condition on it can produce inattention, and simply naming them does not tell you which one you are looking at.
What does work is chronology. Instead of asking whether depression is present, ask when concentration first became a problem and what it has done since. Is the attentional difficulty state-dependent or trait-like? Does it lift when mood is well, or does it persist through euthymic stretches? Did it predate the cannabis, or arrive with it? Was executive dysfunction there before the trauma? What happens to it after a genuinely good stretch of sleep? Those questions separate conditions that a symptom checklist cannot, because they are about pattern over time rather than presence at one moment.
Cannabis deserves specific handling because it comes up constantly. Chronic daily use lowers diagnostic confidence substantially, because intoxication, withdrawal, and sleep disruption each confound the same domains you are trying to assess. Full abstinence is not always feasible or necessary before diagnostic assessment, but I would not try to establish ADHD while someone is intoxicated or in acute withdrawal. Ideally the substance use is at least stabilized. The workable approach is the chronological one: look at the temporal relationship, whether symptoms predated use, and what things looked like during any substance-free periods. Practically, you often work with an imperfect picture and say so in the chart.
Trauma is the hardest of them, particularly in high-functioning women who have compensated for years. What looks like mood cycling can just as easily be shifts in activation, executive fatigue, and nervous system load. ADHD and trauma become very hard to disentangle, and the honest position is often that you cannot fully separate them in one visit. Sequencing helps more than certainty: treat what is clearly present, then look at what remains.
Sleep is the one most often skipped. Untreated apnea produces executive dysfunction, irritability, and attentional failure that will look like ADHD on any instrument you own. If someone has apnea and questionable adherence to treatment, you do not have a clean evaluation yet, whatever else you have collected.
Bipolar disorder is where the chronological question does the most work, because the symptom lists genuinely overlap. Distractibility, increased activity, impulsivity, talkativeness, and reduced sleep appear in both. The distinction is not which symptoms are present but how they are distributed in time. ADHD is trait-like and persistent across time, although its expression changes with context and demands. Bipolar symptoms are episodic, occurring in discrete periods that represent a change from the person's own baseline. So the overlapping items have to be read against the presence or absence of a distinct mood episode rather than counted toward both diagnoses at once. Someone who is distractible every day since childhood and someone who becomes distractible for a fortnight alongside reduced sleep and elevated mood are describing different things with the same word.
Head injury can reorganize the whole picture, and not only the ADHD question. TBI independently produces mood instability, irritability, impulsivity and emotional dysregulation that can resemble bipolar II, and the DSM-5 has a separate code for that situation. A normal CT at the time of injury does not rule out diffuse axonal injury, and cognitive and emotional changes can persist, evolve, or become more apparent over time, particularly alongside chronic sleep disruption. The chronology is what matters. If the attentional and executive difficulties began only after the injury, with no convincing pre-injury developmental pattern, acquired dysfunction becomes the better explanation than ADHD. If a clear ADHD pattern predates the injury, both may be relevant, because a head injury can also worsen executive dysfunction that was already there. Either way it may warrant a different medical and cognitive workup rather than an ADHD workup, and treating post-TBI symptoms is a separate clinical question in its own right.
What the instruments can and cannot tell you
I have written at length about brief screeners and about what real psychological testing actually is, so I will keep this short.
The ADHD instruments are a heterogeneous group, and it is worth being accurate about that rather than lumping them. The WFIRS is an impairment measure. The AAQoL is a quality-of-life measure. The WURS is retrospective, aimed at childhood. Conners and Brown are rating scales. The Vanderbilt is primarily a pediatric parent and teacher instrument. They do genuinely different jobs. What they have in common is the thing that matters here: not one of them independently establishes onset, cross-setting impairment, differential diagnosis, or etiology. They organize and document symptom and impairment information. They do not make the diagnosis.
The UK adult ADHD assessment standard published in 2024 makes the same point from the other direction. It describes the reference standard as a semi-structured interview conducted within a full psychiatric and neurodevelopmental review, using open questioning and probing to draw out real-life examples of symptoms and impairment, and it cautions specifically against closed-question, checklist-driven assessment and against leaning too heavily on rating scales. It also suggests two hours or more for an adequate assessment. That is a useful number to sit with, because it tells you what the instruments are actually competing against, and it is not a competition they can win.
Performance-based testing, the QbCheck and Creyos family, adds a different kind of information, because it measures performance rather than self-report. That distinction is real. But it is neither required nor diagnostic, and it is worth knowing what the guidance actually says. NICE's 2024 review recommends QbTest alongside standard clinical assessment for ages six to seventeen, explicitly as a supplement to professional judgement rather than a replacement for it. For adults it concluded that more research is needed, and it placed QbCheck, the remote version, in that same more-research-needed category. That matters when an adult practice is deciding how much diagnostic weight to give these products: the recommendation does not extend to the population most of us are assessing. Whether any of this is worth incorporating into your practice is a workflow and economics question, and I deal with it properly in Part 2.
One thing about instruments does belong here, because it is diagnostic reasoning rather than logistics. A score anchors. It anchors the patient, who arrives holding a number, and it anchors you, because once a summed self-report is in the chart the burden quietly shifts from evaluating whether criteria are met to explaining why the number does not mean what it appears to mean. These instruments are also transparent. An adult who has spent three months reading about ADHD can answer one in the direction they expect without any intent to deceive. The instrument cannot tell recognition from endorsement.
The patient who is already certain
When someone is convinced they have ADHD but the developmental pattern, cross-setting impairment, and functional cost do not line up, I am fairly direct that I do not have enough to make the diagnosis at this time. That is not a screening failure and it is not a refusal to help. It is appropriate restraint, and it is worth saying to yourself in those words, because the pressure in the room will not describe it that way.
I am hearing attentional distress, but I am not seeing the pattern of impairment ADHD requires. There are other explanations that fit better, and I do not want to label this inaccurately.
From there, usually one of two directions. The first is to treat what is clearly present, the mood or anxiety or sleep or substance use or burnout, and be explicit that we can revisit ADHD once those are better controlled and we can see what is left. That reframes the plan as information-gathering rather than dismissal, and it noticeably lowers defensiveness, because the patient hears a sequence instead of a no.
The second is referral, and I want to be precise about what for, because it is easy to leave the wrong impression. ADHD remains a clinical diagnosis. Neuropsychological testing is not required to make it, and referring is not what you do simply because a case is difficult. What justifies it is a differential that has broadened past what you can responsibly resolve in an outpatient visit, a question of cognitive impairment from another cause, or a high-stakes accommodation or disability determination that warrants a more specialized evaluation than you are set up to provide. In those situations I am very comfortable saying that this is beyond what I can determine here. That is not punting. It is practicing within scope.
The patient who arrives already diagnosed
A different problem, and an increasingly common one. The diagnosis was made elsewhere and you are being asked to continue it, sometimes with a testing report attached, sometimes with a prior authorization request for a specific medication.
An inherited diagnosis is information, not an obligation to agree. It also does not warrant reflexive suspicion, and it is not a reason to put a stable patient through a re-evaluation for its own sake. What it means is that continuing treatment is a decision you are signing your name to, so you need enough history to know what you are treating. That matters most when the surrounding picture is complicated: other controlled substances already on board, a sedative regimen, an open substance use question, a comorbidity that changes the risk calculus.
And an outside result is still just a result. It does not establish etiology because it came with a letterhead, any more than one you generated yourself would. The reasonable posture is to ask what it actually demonstrates, and what is still missing.
What is left
Take all of that seriously and the instruments shrink to their real size. Collateral adds a perspective you cannot manufacture. Performance testing adds data you cannot observe. Rating scales organize and document. None of them makes the diagnosis, and after all of them, clinical judgment still does most of the work.
That is not a discouraging conclusion, though it can read like one. It is the reason your evaluation is worth something. A patient can obtain a computerized test result without you. What they cannot obtain without you is someone who takes the developmental history properly, distinguishes trait from impairment, holds the differential open long enough to be useful, and says no when the criteria are not met.
Which is also, incidentally, the thing that protects you. Not a stack of printouts, but a visible line of reasoning showing what you concluded, on what basis, and what would have changed your mind. In Part 2 I will take that apart operationally: what the process actually looks like from the first request through to treatment, what to send and when, how to get collateral without chasing people for a month, whether the computerized tools are worth buying at your volume, and what belongs in the chart.
References