Evaluating the Patient Who Arrives With the Diagnosis
A patient who already knows the DSM-5 criteria can endorse every one of them honestly, which means your checklist has stopped measuring anything. The evaluation has to move from confirming symptoms to reconstructing a life.
Shafi Lodhi, MD · September 2026
The patient who arrives with the diagnosis already in hand has not made your job harder. They have made one specific part of it impossible. Criterion verification — reading a symptom aloud, offering an example, recording the agreement — no longer measures anything, because this patient has studied the criteria and can endorse each one accurately and in good faith. So the evaluation has to change shape. You stop asking whether the symptoms are present and start asking the patient to produce the granular, dated, idiosyncratic evidence of them: the kind of detail nobody memorizes from a video and nobody invents consistently across thirty years. Many of these patients do have ADHD. Some have something else that is hurting them just as much. The method has to be able to tell you which.
The knowledge problem
Most structured ADHD instruments were designed for a patient who does not know what the right answer is. That assumption no longer holds. A patient who spent four months reading before booking the appointment arrives knowing that DSM-5 requires several symptoms with onset before age 12, that inattentive presentation is underdiagnosed in women, and that the phrase impairment in two or more settings matters. They know what you are looking for. They are not trying to deceive you. They simply have the answer key.
Consider what that does to a recognition-format instrument. The clinician reads a criterion, supplies an illustrative example, asks whether it fits. The patient says yes. What has been measured is not the symptom but the patient’s recognition of the symptom as described, and a patient with prior knowledge will recognize it whether or not they have it. This is the structural problem I built the LODHI-A around, and the clearest line of difference between it and the recognition-based instruments in common use. A recall task survives prior knowledge. A recognition task does not.
It is tempting to call that dishonesty. It is not. The patient who nods at your example is telling the truth as they understand it; the instrument asked a leading question and got the answer leading questions get.
The reinterpreted autobiography
The harder phenomenon is not endorsement. It is narrative.
When a person encounters an explanatory framework that finally makes their difficulty legible, they do not simply add it to their list of beliefs. They go back through their life and read it again. The unfinished degree, the friendships that ended without a fight, the performance review that said inconsistent, the childhood sense of being slightly out of step: all of it gets reorganized around a single cause. By the time they reach your office the story is coherent, chronological, and delivered with real conviction. It has been rehearsed, not cynically, but the way any of us rehearse the story of how we came to be the way we are.
This is ordinary cognition. Autobiographical memory is reconstructive; we retrieve the gist and rebuild the detail to fit, and we trust the rebuild more than we should. ADHD makes a powerful organizing lens because its symptoms are dimensional. Everyone has lost keys, abandoned a project, failed to read the whole email. Twenty years of those moments, sorted for relevance and stripped of context, will look exactly like a lifelong pattern — because in a sense it is one, just not necessarily a disorder.
Sincerity is not evidence. It is only the absence of one specific problem.
So a coherent narrative is not confirmatory, and neither is the obvious fact that the patient is not fabricating. What matters is whether the story holds its shape when pressed for detail it did not anticipate being asked for.
Structural solutions rather than suspicion
I do not interview these patients differently. That is the whole point. A method that changes when you suspect the diagnosis was self-supplied requires you to guess correctly about the patient before you evaluate them, and you will guess wrong in both directions. The protections have to be structural: present in every interview, invisible in any single one.
The first is simple and absolute: never supply the symptom description. No examples, no menus, no would you say it’s more like A or more like B. The moment I offer language, the response is contaminated and I cannot tell whose language it is. This costs time. Patients stall, and I let them stall.
The second is to treat criteria-shaped language as a prompt rather than data. When a patient reports difficulty with task initiation, or executive dysfunction, or object permanence for their own belongings, I have learned nothing yet. That is vocabulary, and vocabulary travels. So I ask for the last time it happened. Not in general: the last time. What was the task. Where were you sitting. What did you do instead, for how long, and what was the consequence the next morning. Lived memory answers with furniture and weather and names in it. Reconstructed memory tends to restate the theme in slightly different words. Neither answer settles the diagnosis, but the difference between them is the most useful signal I get all hour.
The third protection is the longitudinal spine of the interview, and it is the one that does the heaviest work. Symptom criteria are public. A developmental trajectory is not. Moving through childhood, I am not asking whether the patient was inattentive at nine; I am asking what fourth grade was actually like. Which subjects, which teacher, what the report card comments said, whether a parent was ever called in. Then what happened at fourteen when the volume of work rose. Then the first year after the scaffolding came off: the first semester living alone, the first job without a supervisor two desks away. A pattern with developmental substance has a shape there — demands rise, compensation strains, something visibly gives. A pattern assembled retrospectively tends to sit evenly across the decades, because it was sorted by theme rather than laid down by time.
One caveat, because clinicians get this wrong in the direction of overconfidence. Thin autobiographical recall is not a negative finding. Adults with genuine ADHD often have patchy, undated memory for their own childhood, and some of the most convincing cases I have seen came from patients who could barely name their elementary school. Sparse detail sends me to collateral history and documents: report cards, old evaluations, a parent on speakerphone. It does not send me toward a no.
A composite case
A composite, drawn from features common to many evaluations and not from any one patient: a woman in her early thirties, a pharmacy technician who had gone back to school part-time and was failing a course for the first time in her life. She brought a printed list of her symptoms organized by DSM domain, and a working knowledge of the literature better than some residents I have taught. She had been certain for eight months.
Every criterion was endorsed. When I stopped asking about criteria and asked what was actually happening with the failing course, the account turned granular and vivid — and all of it was from the last two years. Before that the narrative was thematic, I always struggled, I was always the one who left things to the last minute, with no texture underneath. Her mother remembered a quiet, organized child and no calls from school. When I built the timeline out properly, the two years lined up with a second job on overnight shifts, five hours of sleep on a good night, and a mother newly dependent on her.
She did not have ADHD. She had a real, disabling attention problem, and she was right that something was wrong. She was wrong about what. The point is not that self-referred patients are mistaken. It is that the structure produced the answer, and suspicion would not have. Suspicion only makes you a worse interviewer with the same information.
Evaluating rigorously without making the patient feel disbelieved
This is the part most clinicians ask me about, and the solution sits mostly in the framing at the front of the hour. I say some version of this out loud: you have done a lot of reading, and that reading is often accurate, so I am going to ask for your own examples rather than offer you mine — not because I doubt you, but because your examples are the only ones that carry diagnostic weight. Then I say that I will spend an unusual amount of time on childhood, and that this is how the interview is built for everyone.
That preface turns what could feel like cross-examination into a shared method, and gives the patient something to do besides defend a position. Many arrive braced for dismissal, having been dismissed once already, and they relax on hearing that the rigor is procedural rather than personal.
Two habits help. I never argue with the self-diagnosis during the interview; I gather, and I say so. And I ask about the timing directly — what made this the month you looked it up — because the precipitant is informative in its own right, and the taxonomy of those precipitants deserves its own discussion.
Saying no, and saying yes
A non-ADHD answer delivered as a negation is a failed consultation even when the reasoning is sound. The patient came in with an explanation for years of pain. Taking it away without replacing it leaves them worse off than when they arrived, usually convinced that you did not listen.
So I do not lead with the absence. I lead with what I found, in detail, and I make clear that the impairment they described is real and that I believe it. Then the reasoning, plainly: why the pattern does not fit a neurodevelopmental course, when it actually began, what I think is driving it. Severe chronic sleep restriction. An anxiety disorder that has been running the show since adolescence. The aftermath of trauma, autism without the ADHD, a depressive illness presenting as cognitive slowing, or simply a load no one could carry. Each deserves treatment, and several respond to treatment better than ADHD does. I also say what would change my mind, and I mean it: if the sleep gets fixed and the attention problem persists unchanged, I want to see them again.
When the evaluation confirms what the patient already believed, I say so without hedging. This happens often. Many self-identified patients are correct, frequently because they have been paying close attention to their own functioning for years. I do not deliver a confirmation apologetically, and I do not water it down to protect myself against having been led. I tell them the diagnosis is supported, and then I tell them why: the specific examples they generated, the specific childhood evidence, the point in the timeline where compensation stopped working. That reasoning is worth more than the label. It is what the patient carries into the next clinician’s office, and what makes the diagnosis durable rather than merely granted.
The LODHI-A is free for licensed clinicians, and it is built for exactly this interview: patient-generated evidence, a lifespan timeline, no supplied examples. The methodology questions explain how it works, and you can request access to read the instrument itself. It will not tell you whether your patient is right. It will make sure that whatever you conclude, you got there on your own evidence.
Neuropsychiatrist · Bay Area Neuropsychiatry · Developer of the LODHI-A