The Science
A quarterly sitting uses four published screening instruments. None of them is a diagnosis, and not one of the boundaries they draw was invented here — every threshold on this page comes from the research that established it. This page is the evidence behind each one, and behind reading them together.
A sitting takes about twenty minutes and asks four sets of questions. Two are answered by the person themselves, and two by someone who knows them well. Together they look at the same person from four directions rather than one.
What comes out is not a diagnosis and not a risk score. It is a dated record you can put in front of a clinician, and a second one three months later that can be compared with the first. If you would rather read about the things you can actually change, start with What can be done and come back here for the evidence.
A 16-question informant assessment, reproduced from the published short form.
The IQCODE — the Informant Questionnaire on Cognitive Decline in the Elderly — is a short, confidential 16-question assessment. It takes about five minutes and produces a plain-language score with a band and a report you can print and bring to an appointment. It is a screening instrument, not a diagnosis.
Instead of testing the person directly, it asks someone who has known them for at least ten years to rate how their memory and everyday thinking have changed compared with ten years ago. Research shows that the people closest to someone are often the first to notice real cognitive change — sometimes years before a test in a clinic picks it up.
Because it measures change from the person’s own baseline rather than performance against a norm, it works across education levels and backgrounds — and a 2021 Cochrane review supports its use as a first-step screen in health care settings.
Sixteen items, grouped into the three areas this platform scores and prints separately on every report.
7
items
Remembering recent events and conversations, important details about family and friends, the day and date, and where things are kept. The heart of the IQCODE.
4
items
Whether they can still pick up new skills — working a new gadget or appliance, learning new information, adapting to anything unfamiliar. Learning is often affected early.
5
items
Following a story, making everyday decisions, handling money and finances, working out everyday sums, and reasoning things through. The skills independence rests on.
Each of the 16 items is rated on the published 5-point scale, from 1 (much improved) to 5 (much worse), with 3 meaning “not much change.” The 16 ratings are averaged. That average — a number between 1.00 and 5.00 — is the score, exactly as the published instrument defines it.
| Average | Band | What it means |
|---|---|---|
| 3.00 or below | No meaningful change | The answers do not show the pattern of change this screen is built to flag. That is reassuring — and the next sitting is what turns one reassuring result into a trend. |
| 3.01 – 3.29 | Watch | Below the published cutoff, but no longer flat. Worth another sitting next quarter rather than a wait-and-see. |
| 3.30 – 3.59 | Decline warranting evaluation | At or above the published community cutoff of roughly 3.3–3.4, the range researchers use to flag possible cognitive decline. A calm, unhurried appointment is the sensible next step. |
| 3.60 and above | Decline — secondary-care threshold | At or above the published threshold validated in hospital settings. Meaningful change has been observed across several areas. Please help this person see a doctor soon — and remember that many causes are treatable. |
A separate change rule also matters: an increase of 0.30 or more from the previous sitting counts as movement on the cognitive axis, even when the band has not changed. That is why a sitting is taken quarterly rather than once.
| Typical free quiz | This platform | |
|---|---|---|
| Reproduces a validated instrument | Rarely | Yes — the 16 published items, unaltered |
| Asks the person who knows them best | No | Yes — informant-style, as clinicians do |
| Compares now with ten years ago | No | Yes — change is what matters |
| Covers reversible causes clearly | No | Yes — thyroid, B12, mood, medicines, sleep, hearing |
| Clear referral to a doctor | Often vague | Yes — it is the central message |
| A report to bring to the appointment | No | Yes — printable, dated, and built for a clinician to read |
| Repeats on a schedule | No | Yes — quarterly, so a single number becomes a direction |
Comparison framing adapted from the dementia screen published at Psychology.com, which offers a separate guided version of this instrument.
You will answer these questions about someone you know well — a parent, partner, or close friend. For each one, compare how they are now with how they were ten years ago. Focus on change over that time, not on how good their memory has always been. If you are worried about your own memory instead, the M3 Checklist below is the part of the sitting you answer about yourself.
Physical pain and Alzheimer’s disease are closely related. Chronic pain often coexists with dementia, the two may share underlying biology such as neuroinflammation, and — most consequentially for a family — dementia removes the words a person would normally use to report pain. Pain is measured here every quarter, on the same dated record as thinking and mood, and it scores neither of them.
The Defense and Veterans Pain Rating Scale was developed by the Defense & Veterans Center for Integrative Pain Management at the Uniformed Services University of the Health Sciences, and its psychometrics were published by Polomano and colleagues in Pain Medicine in 2016 in 307 service members and veterans.
Why this instrument and not another. Two reasons, and the second is the one that decided it. First, the DVPRS carries its own colour-coded severity zones — green, yellow and red — so a mild / moderate / severe label on this platform comes with the instrument and is not our invention. Second, and more important for the people this platform is built for, it puts a descriptor at every number — a plain-language word for each one rather than a bare 0 to 10. A person is not choosing “6”; they are choosing hard to ignore, avoid usual activities. On a form filled in by someone whose words are going, that is not a cosmetic difference.
0–10
Pain intensity, right now. Every number carries its own description — from no pain at 0, through sometimes distracts me at 3 and hard to ignore, avoid usual activities at 6, to as bad as it could be, nothing else matters at 10.
This is the pain reading. The four below are reported beside it.4
Then the past 24 hours. How much pain interfered with usual activity, interfered with sleep, affected mood, and contributed to stress. Each 0 to 10, each with its own wording.
Two recall windows on one form: the first item asks about now, these four ask about the past day.0
Numbers averaged together. The four supplemental items are shown individually and never combined into one interference score.
The instrument validates no composite, and its own factor analysis splits them two ways.Scoring. The pain reading is the intensity item itself. The four supplemental items are shown next to it, separately. We do not average them, and that is a deliberate refusal: the 2016 paper reports them separately, and a two-factor solution explained 79% of the variance against 66% for one factor — intensity, activity and sleep loading together, mood and stress loading together. Averaging four items across two real factors would produce a tidier number and a worse one.
Bands. Mild 0–4 (green), moderate 5–6 (yellow), severe 7–10 (red) — the DVPRS’s own colour zones. We hold 0 out as its own tier inside the mild zone, because the form’s own word for 0 is no pain and telling someone who reported no pain that they are “mild” would be a display error, not a reading. That moves no boundary. The 2016 validation derives no cut-points of its own; the bands it quotes are Serlin and colleagues’ 1995 bands, and they are the same boundaries: 1–4, 5–6, 7–10.
Change. No minimal important difference has been published for the DVPRS. So this platform reports the movement and says nothing about whether it matters, because nobody has established what matters. Any threshold we printed here would be ours, and it would look like the instrument’s.
Licence. The DVPRS is granted for free use “as is, without alteration,” and as a US federal government work it is effectively public domain. It is used here unaltered, with one substitution disclosed on the screen itself: the printed form says circle the one number and this platform uses a slider, so the instruction verb differs and nothing else does.
They overlap heavily. Pain is common in people living with dementia, and how common depends enormously on who is asked and how. A 2023 systematic review of twenty-five studies of nursing-home residents with dementia found reported prevalence ranging from 8.6% to 79.6% — varying with the assessment method, the definition of pain and the stage of the stay — reaching about 80% towards the end of life. The figure is often quoted flatly as “up to 80%”; the range is the more honest number, and the spread is itself the finding. Pain in dementia is measured badly, so it is found inconsistently.
They may share brain changes. Neuroinflammation and changes in neurotransmitters appear in both chronic pain and cognitive decline, and are the leading proposed link between them. This is an active research question rather than a settled mechanism, and it is described here as a hypothesis, not a finding.
Long-term pain is associated with higher later risk. In the Health and Retirement Study — 10,065 older adults followed for twelve years — persistent pain predicted memory declining 9.2% faster (95% CI 2.8–15.0) and dementia probability rising 7.7% faster (95% CI 0.55–14.2), an absolute 2.2 percentage points more dementia at ten years. A 2023 French cohort of 13,596 adults corroborates the direction at 1.23 times the rate of matched controls (95% CI 1.15–1.32).
Whitlock EL et al., JAMA Internal Medicine 2017;177(8):1146–1153
That is the detail worth pausing on. In the same cohort, persistent pain carried a 15.9% higher relative risk of losing the ability to manage medications independently and 11.8% for managing finances. Those are precisely the everyday capacities the IQCODE asks an informant about at items 13 to 15, and which this platform already weights in the report narrative. Two independent literatures point at the same three questions.
The words go first. As Alzheimer’s progresses, people lose the ability to say where it hurts, or how badly. The pain does not stop. Only the reporting stops.
Behaviour becomes the pain scale. Instead of a complaint, pain shows up as agitation, aggression, grimacing, moaning, resistance to care, or restlessness that arrives at the same hour every afternoon. Read without a record, that behaviour is easy to file as the disease advancing — and the response to “the disease is advancing” is very different from the response to “something hurts.”
So it goes untreated. Because pain in dementia is hard to express and hard to spot, it is frequently overlooked and under-treated. This is the single most consequential reason pain is on this form: it is the most treatable thing on the page, and the most frequently missed.
What this does not mean. These are associations, not demonstrated causes, and nothing in this platform treats pain as a cause of Alzheimer’s. Pain is measured because it is common, because it is treatable, and because it is the most likely innocent explanation for a bad quarter — not because a pain score predicts dementia. The DVPRS contributes no points to the cognitive band, to the mental-health band, or to the quadrant.
Five questions, every quarter, on the same page as the other two axes. The value is comparative. When behaviour changes, a clinician can look at whether pain moved in the same quarter before concluding the disease did. In the reports this platform produces, an untreated knee mistaken for agitation is a recurring story, and it is a story that only a dated record can correct.
The Alzheimer’s Association’s Cognitive Assessment Toolkit names the Short IQCODE among its validated informant tools for detection at the Medicare Annual Wellness Visit, and this platform administers it. One other is recorded here rather than passed over, because this platform used it and then replaced it. The IADL-C is administered, and carries its authors’ attribution wherever it or a score from it appears. The Functional Activities Questionnaire was administered until version 6.5 and has been replaced; why is set out below.
The Instrumental Activities of Daily Living: Compensation Scale asks someone who knows the person to rate everyday activities that need planning, judgment and money sense rather than physical ability — organising a trip, sequencing errands, managing a budget, remembering whether bills were paid, keeping financial records, filling in insurance forms, finding the way back to a meeting point, recalling where things were put, balancing a statement, returning voicemail. This platform administers the 11-item short form, derived by item response theory in 2025. Each activity is rated on an eight-level scale and the score is the mean, from 1.00 to 8.00, higher meaning greater difficulty.
Why this instrument and not the one we had. Until version 6.5 this axis was the Functional Activities Questionnaire. The FAQ counts what a person can no longer do. The IADL-C counts that too — and it also records how they are still managing. Its eight anchors do something no other functional scale we reviewed does: using a list, a reminder or a routine is recorded as a level of independence, not as a failure. A person doing something as well as ever with a strategy scores 2. A person doing it slightly less well with a strategy scores 3, and slightly less well without one scores 4. Note that the aid and no-aid answers invert between those two pairs; that is deliberate, and it is the whole instrument.
And that is the reading this platform exists to take. A person who is slipping and compensating looks nothing like a person who is slipping and not compensating. The first is holding independence together and can often be helped to keep it. The second has either not noticed or can no longer organise the noticing. A scale that counts only losses cannot tell them apart, and for a monitor that runs every quarter for years, that distinction is the point.
The hazard we have to design around. In the 2014 development data, “compensatory strategy use increased in the MCI group and decreased in the dementia group.” Compensation rises and then falls. So a low compensation reading means either nothing to compensate for or no longer able to compensate, and the two ends of that curve look identical on a chart. This platform therefore never shows a compensation reading on its own. It is rendered only beside the total, and the sentence is chosen from the pair: score low and compensation low is healthy; score rising and compensation rising is the state worth catching; score high and compensation falling means the strategies have collapsed, and driving, medicines and living alone become live questions for the doctor. A 2018 study using a different instrument found the fall but not the rise, so the first half of that curve is an open question and is described here as one.
The boundaries, and which of the eight we use. Rahman and colleagues published eight cut-offs in 2025 — two boundaries, two forms, two samples. This platform uses the 11-item short form on the research-sample values: 1.29 for functional limitations consistent with mild cognitive impairment (sensitivity 0.77, specificity 0.69) and 4.41 for dementia against MCI (sensitivity 0.71, specificity 0.89). We chose the research sample over the clinical one because its cut-offs favour specificity, and we screen an unselected community population where the base rate is low and a false positive is the expensive error — it sends a well person into a diagnostic pathway. The clinical cut-offs favour sensitivity and were derived in a memory-clinic referral population, 28.6% of whom had dementia. All eight are carried in the build so that changing this is configuration rather than code. The authors are explicit that the cut-offs are for use “as a screening measure or combined with other clinical and cognitive measures” — not on their own, and this platform does not use them on their own.
The number we are least comfortable with, said out loud. The two samples put that first boundary at 1.29 and 1.81 on the same eleven items. That is a wide spread on a 1–8 scale and it is the weakest published figure anywhere in this instrument. It is stated wherever a score is shown rather than left here.
Ten of the eleven come from one domain, and that is a feature. The short form draws almost entirely from money and self-management — which the 2014 paper found is the domain that shows notable difficulty in mild cognitive impairment, while home daily living becomes prominent only at dementia. So the short form is not merely kinder to a family than 27 items; it is weighted towards the domain that moves first.
The rights position, stated plainly. The IADL-C belongs to its authors at Washington State University. It has no public self-serve licence and no published permission statement; there is no download and no form to fill in. A licence has been requested and has not been granted. Until it is, this platform does not open to the public, and nothing on any screen claims a licence we do not hold. If the answer is no, the instrument comes out the same day — the same rule we applied to the FAQ.
And the limitations we will not bury. Three. The instrument was validated as a questionnaire rather than as an interview, which is how we use it — that one is in our favour and we say so. The validation sample was primarily White and highly educated, as were the informants, and the authors write that validating these cut-offs “will require a more diverse sample of older participants.” And the informants’ own cognitive status was never assessed in any of this work.
That third gap is ours to close rather than to quote. In this platform the informant is very often an ageing spouse — whose mental health we are screening on the same schedule, on the same dated record. Whether an informant instrument’s accuracy depends on the informant’s own cognition is a question no research group is positioned to answer and this platform is. It is stated here as an open question, not as a finding, and all three limitations render with every score the instrument produces rather than being filed away on this page alone.
Peer-reviewed and validated for primary care.
The M3 Checklist is a 27-item, self-rated screening tool that assesses depression, bipolar spectrum, anxiety, and post-traumatic stress disorders together, in a single one-page instrument. Most published depression screens (like the PHQ-9) capture only one of these four — which means much of diagnosable mental illness can go undetected when only depression is screened for.
The M3 Score is a single number reflecting overall symptom severity across depression, anxiety, bipolar spectrum and post-traumatic stress — plus an optional pain measure, the DVPRS scale. It is a self-administered screening and self-assessment tool that helps a person understand how they are doing and track changes over time.
Each item is rated on a simple 5-point scale — 0 Not at all, 1 Rarely, 2 Sometimes, 3 Often, 4 Most of the time. By design the questions use three different time-frames, following standard psychiatric practice: most symptom questions ask about the last two weeks or more; the bipolar-spectrum questions ask whether the person has ever experienced them, because that history is assessed across a lifetime; and a final set asks whether any of those symptoms impact work, school, relationships, or have led to alcohol or other substance use.
The result is an overall score out of 108 with per-area severity ranges, and a set of gateway questions — thoughts of suicide, and functional impairment and substance use — that are reviewed individually. A positive answer on any gateway question warrants clinical follow-up regardless of the total score.
This is the one axis M3 Information owns. The platform is as explicit about it as it is about the three instruments it licenses from other people.
There are three scores, and they answer different questions. The total runs 0 to 108 on the answers as given. A set of symptom severity subscores runs on the same answers and is the right thing to watch for change over time. And a set of disorder scores runs on a different weighting — never and rarely count 0, sometimes counts 1, often and all of the time count 2 — which is the weighting the validation used and the one the published cut-offs sit on.
The gateway comes first. A person who answers the suicide question negatively and whose four lifestyle questions show no single answer at “often” or “most of the time” and no more than one at “sometimes” has a negative gateway. In the validation study that first step removed 349 of 647 participants from further scoring — which is what made the instrument quick to score by hand.
This platform does not stop scoring there, and that is deliberate. The severity subscores are computed for everyone, and the gateway conditions how the disorder scores are read rather than whether they exist. The reason is in the validation itself: 10.9% of the people the gateway screened out nevertheless met diagnostic criteria on the structured interview. Diagnostic criteria require functional impairment and those people report none, so the report says exactly that — symptoms are present, impairment is not, clinical correlation is suggested and other causes should be considered. Zeroing them would erase the group most worth a second look.
Gaynes and colleagues compared the M3 against the Mini International Neuropsychiatric Interview in 647 consecutive primary-care patients. Table 2 of that paper gives one screening cut-off per disorder, on the weighted scale:
The paper notes that collapsing the five response categories to three cost nothing: a sensitivity analysis showed no loss in either sensitivity or specificity.
Anxiety is questions 8 to 19, and it contains the four PTSD questions. That is the instrument’s design, not an overlap to be tidied away. The paper describes PTSD as “a specific type of anxiety disorder”: anxiety is the spectrum — generalised anxiety, panic, agoraphobia, social phobia, obsessive-compulsive and post-traumatic symptoms together — and PTSD is reported additionally, on its own, because it is the subtype worth naming. A person with post-traumatic symptoms should register on both readings, and does.
What the validation does not establish. That paper reports no total-score cut-point and no AUC values at all. It derived thresholds by ROC analysis and published none of the areas under those curves, so any AUC quoted for the M3 did not come from it. The 0–108 total, its tiers, and the threshold at 33 are M3 Information’s, from the instrument’s own scoring algorithm rather than from the validation. Both are shown on the report and each is labelled with where it comes from.
A correction, recorded rather than quietly made. Version 6.6 of this platform narrowed the anxiety subscale to eight questions, in the belief that its overlap with PTSD was a defect, and applied cut-offs taken from secondary summaries rather than from Table 2. Both were wrong: the overlap is the design, and the cut-offs are 5, 3, 2 and 2. Version 6.7 reverted the first and corrected the second, against the primary sources. It is noted here because a platform that says it scores to published literature should show its corrections as readily as its citations.
Every figure below is from the Alzheimer’s Association’s 2026 Alzheimer’s Disease Facts and Figures, cited by page.
2026 Alzheimer’s Disease Facts and Figures, Overview, page 4
That single sentence is the reason this platform exists. The report puts it again in its Special Report: “physiological changes marking the earliest stages of Alzheimer’s disease can begin up to 20 years before measurable cognitive impairment” (page 98). The warning is long. What is usually missing from it is a record.
8%
of older Americans living with mild cognitive impairment receive a diagnosis — the stage at which the report says “treatment and planning may be most effective.”
Facts and Figures 2026, page 28~half
of Medicare beneficiaries who have a dementia diagnosis in their own billing records report being told of it.
Facts and Figures 2026, page 2854%
of Americans 45+ who reported worsening memory or thinking had not consulted a health care professional about it. 16.6% reported it at all.
Facts and Figures 2026, page 29The report is explicit about where the shortfall sits: “Underdiagnosis is most pronounced at the earliest stages of dementia when symptoms are mild” (page 28). And it lists what a missed or delayed diagnosis costs — “delayed access to treatment and supportive services, less time for care planning, higher costs of care, and negative impact on the individual’s physical and mental health or even the mental health of their family members and potential caregivers” (page 28).
Facts and Figures 2026, Workforce, page 63
The report also finds that among primary care physicians surveyed by the Alzheimer’s Association, 39% were “never” or “only sometimes” comfortable making a diagnosis of Alzheimer’s or other dementias, and 50% said the medical profession is “not very prepared” or “not at all prepared” for the growing number of people living with them. 32% make a dementia specialist referral at least monthly — but 55% report a shortage of specialists in their area (page 64).
None of that is a criticism of clinicians; it is a description of a time problem. An instrument a family completes at home, on their own schedule, and brings in already scored, is a direct answer to it.
86%
of adults 40+ want brain health information during routine care. Two in three prefer to hear it from their own provider.
Facts and Figures 2026, page 10214%
have actually had a conversation about maintaining brain health with their physician. Only 11% have discussed reducing their risk of Alzheimer’s.
Facts and Figures 2026, page 1029%
know “a lot” about how to maintain brain health — though 88% say it is very important and 99% rate it equal to or above physical health.
Facts and Figures 2026, page 101A dated report, written in plain language and built to be handed across a desk, is a way of starting the conversation that 86% want and 14% have had.
The Association reports the timing of each brain change separately, measured in people carrying rare mutations that make the age of onset predictable (page 8). Read together they are a cascade — damage markers first, structure last — and the whole of it happens before the first symptom.
The clinical stages tell you where a family fits. The Association’s 2024 staging criteria (Table 3a, page 11) describe Stage 1 as biomarker-positive with no symptoms — and note a person “may remain in stage 1 for years, even for the remainder of their lives.” Two things in the stages after it are worth reading closely, because they describe this platform almost exactly:
| Stage | What the criteria say — and why it matters here |
|---|---|
| Stage 2 transitional decline | Decline representing “a change from individual baseline within the past one to three years,” persistent for at least six months, with no functional impairment. The criteria say it may be documented through subjective report of cognitive decline — or “with recent-onset change in mood, anxiety and/or motivation not explained by life events.” The mental-health axis is not a side measure at this stage. It is one of the two ways the stage is recognised at all. |
| Stage 3 cognitive impairment with early functional impact | Decline “documented by the individual’s report or by an observer’s (e.g., study partner) report or by change on longitudinal cognitive testing.” An informant report is not a substitute for the criteria. It is written into them. |
Which is the argument for a quarterly sitting in one line: the observable window is Stage 2 and Stage 3, the two stages defined by change from a person’s own baseline — and change from baseline is a thing you can only see if someone was writing it down.
A 2025 meta-analysis of 13 studies covering 30,257 people found it takes an average of 3.5 years from the first symptom to a dementia diagnosis — and 4.1 years when onset is young. Frontotemporal dementia and younger age at onset were both linked to longer waits.
Facts and Figures 2026, page 8, reporting an autopsy study
The report’s conclusion is blunt: “Studies suggest that mixed dementia is the norm” — and not only for people diagnosed with Alzheimer’s, but for those diagnosed with other types too (page 8). Mixed pathology is especially common past 85, where people are more likely than not to have two or more causes at once.
This is why the platform never names a disease. It cannot, and neither can a single clinic visit; the question “which one is it?” is frequently the wrong question, because the honest answer is often “more than one.” What a dated quarterly record can show is which axis moved, and when — and that is useful regardless of how many pathologies turn out to be underneath it.
The other dementias have prodromes of their own, and they run long too. In the largest multicentre study of isolated REM sleep behaviour disorder — acting out dreams, often decades before anything else — 6.25% of people converted to parkinsonism or dementia each year, 73.5% within 12 years, with a median of 8 years to conversion (Postuma et al., Brain 2019; 1,280 patients across 24 centres). A sleep symptom the family notices first, years ahead of the clinic.
Why women’s risk is higher is mostly not what people assume. The Nature Medicine authors attribute it to survival — women live longer, and age is the dominant risk factor. The Association is careful on the same point: “it is not clear that the risk of developing Alzheimer’s or other dementias differs between men and women of the same age” (page 34). Most incidence studies find no meaningful difference at a given age. The consequence is real anyway: almost two-thirds of Americans living with Alzheimer’s are women — 4.5 million of the 7.4 million aged 65 and older (page 34).
Genetics move the number further. Lifetime dementia risk was 39% with no copies of APOE-ε4, 48% with one copy, and almost 60% with two.
In 2025, 12.7 million family and other unpaid caregivers provided an estimated 19.6 billion hours of unpaid help — valued at $446.3 billion, which the report notes is more than 17 times the total revenue of McDonald’s in 2024 (page 52). 59% rate the emotional stress of that caregiving as high or very high (page 53).
And a diagnosis is what unlocks the support: caregivers of people with a formal dementia diagnosis were “more engaged in communication during doctors’ visits and more likely to receive caregiver training” than those helping someone without one — which the report reads as showing “the importance of linking support to dementia diagnostic procedures” (page 54).
A note on “ten years” versus “twenty.” The Association dates the biology to 20 years or more before symptoms. This platform, and the book behind it, talk about roughly a decade — because that is the span over which the changes a family can actually observe at the kitchen table tend to appear. The two are not in conflict: the biology starts earlier than anyone can see it, and the observing starts when there is something to observe.
A short explainer with Dr. Gerry Hurowitz, psychiatrist — about one minute.
Explainer video · Watch on YouTube
Observed thinking and mental health move for overlapping reasons. Depression in an older adult can look like decline; decline can destabilise mood; and both can be present at once. Reading one axis alone invites the wrong conclusion in either direction. Reading both, on the same dated record, quarter after quarter, shows which one moved and when — which is the question a clinician can actually act on. Neither instrument diagnoses anything, and nothing on the report recommends starting, changing or stopping any treatment. Those decisions belong to the prescriber.
Each axis is one published instrument, reproduced unaltered and scored against its own published thresholds. Nothing on this platform is a blend, and no boundary here was invented. This section is the single place where all four sources sit together, with their rights position stated as plainly as their evidence.
16 questions, answered by someone who has known the person ten years. Free to use. Mean 1.00–5.00, higher worse, with 3.00 the fixed no-change midpoint.
11 questions, answered by the same informant. Licence requested, not yet granted. Mean 1.00–8.00, higher worse, boundaries 1.29 and 4.41.
27 questions, answered by the person. Ours. Total 0–108 on the answers as given; disorder scores on a different weighting, with published cut-offs.
5 questions, answered by the person. Free to use, unaltered. Intensity 0–10 with a description at every number, plus four supplemental questions reported individually.
Three of the four give no threshold for calling a change real. Only the M3 does, and only in its own algorithm: a change of 2–3 points may be clinically significant. The IQCODE, the IADL-C and the DVPRS publish no minimal detectable change at all. So where this platform shows movement on those three it reports the direction and the size and stops. It does not call any of them meaningful, because nobody has established what meaningful is.
One boundary under review. The IQCODE bands on this platform sit at 3.00, 3.30 and 3.60. There is reason to think 3.30 is natively a 26-item long-form value that the Cochrane reviews pool across both forms, and that the community cut-point for the 16-item short form this platform administers is 3.38 (Jorm 1994), with 3.44 as Jorm’s own clinical recommendation. That has not been confirmed against the primary source and so has not been changed. It is recorded here rather than left in a developer’s note, because a page that claims to score to published literature should say where it is not yet certain that it does.