Understand Dementia and Cognitive Status
Learn through assessments about Dementia and if an Alzheimer's review would be helpful.
Understand — in about 3 minutes. Understanding your cognitive status shouldn't be hard. In two short, private check-ins you get a clearer picture of how you're really doing — on any phone or computer.
- Early Detection
- Coordinates Family Dialogue
- Enables Mental Health and Cognitive Monitoring
Two families sit in the same neurologist’s waiting room with the same diagnosis and the same new prescription. One will still be at home together in three years. The other will be in crisis by spring — not because the disease was different, but because nobody was counting the nights, the worry, and the quiet losses that were visible at the kitchen table for years before they were visible in any office.
Why this exists
Dementia gives about ten years of warning.
Almost nobody is counting.
The biology of Alzheimer’s begins changing roughly a decade before anyone is given a diagnosis. Blood tests that can measure part of that biology now exist — but they are cleared for people who are already showing signs and symptoms. So the warning decade usually passes with nothing written down, and the first real record of it begins on the day the trouble has become impossible to miss.
What is missing in those years is not a laboratory. It is a record. Sleep, worry, mood, memory, judgment, handling money, following a conversation — the things that change at the kitchen table long before they change in an office. A family can see them. A chart that was never kept cannot.
This platform keeps that record. Once a quarter, four short assessments are taken back to back and saved as one dated sitting — two answered by the person, two by someone close to them. They are set out in full further down this page.
None of them is a diagnosis, and none of them can name a disease. What they do, quarter after quarter, is show direction — whether things are steady, and if they are moving, which one moved first. That is the difference between a worry and an observation, and it is the kind of thing a doctor can actually act on.
The report appears the moment the sitting is finished, in plain language, with the dates on it. It is written to be printed and brought to an appointment. The decisions stay where they belong — with the person, the family, and the clinician.
What we do
How M3 works
Step 1
Assess
Take the short, private M3 review of your mood, anxiety, sleep, energy — and pain, on the DVPRS pain scale. About 3 minutes, on any device.
Step 2
Review
Get an easy-to-read report that shows depression, anxiety, PTSD, bipolar, and pain together — so more of what matters is seen at once.
Step 3
Plan
Take your report to a clinician and find the next step and support that fits you.
What you actually answer
Four published instruments, one dated record
Each one reproduced unaltered and scored against its own published thresholds. No boundary in this platform was invented.
Answered by someone who knows you well
Sixteen questions for a family member or friend about how memory, judgment and everyday thinking compare with ten years ago. The people closest are often the first to notice real change — sometimes years before a test in a clinic picks it up.
Jorm, short form, 1994. Named by the Alzheimer’s Association among the informant tools for the Medicare Annual Wellness Visit.
How it worksAnswered by you
One question about pain right now, with a plain description at every number rather than a bare 0 to 10 — then four about the past 24 hours: activity, sleep, mood and stress. Pain is the most treatable thing on the page and the most often missed, because dementia takes away the words a person would normally use to report it.
Defense and Veterans Pain Rating Scale v2.1, Defense & Veterans Center for Integrative Pain Management. Psychometrics: Polomano and colleagues, Pain Medicine, 2016.
How it worksAnswered by the same person as the IQCODE
Ten everyday jobs that need judgment rather than physical ability — money and paperwork, shopping, meals, appointments and medications, travel. It measures what a person can still do now, where the IQCODE measures what has changed.
Schmitter-Edgecombe and colleagues, 2014; diagnostic cut-offs Rahman and colleagues, 2025. Washington State University.
It is shown here for example. A licence has been requested and is not yet granted, and we are hopeful of concluding it.
Answered by you
Twenty-seven questions covering depression, anxiety, bipolar spectrum and post-traumatic stress in one screen. Depression in an older adult can look exactly like early dementia — and a depression-only screen reaches about a third of diagnosable mental illness, so the other three are asked here too.
Gaynes and colleagues, Annals of Family Medicine, 2010.
How it worksThree of the four are ours to use outright. The fourth, the IADL-C, is shown for example while its authors consider our request, and this platform does not open to the public until that is answered — an instrument is only fully ours when it is licensed and validated for the way we would use it, not when one of the two is true. See the evidence
Why two axes, not one
Either number alone can be read the wrong way.
Read together, they cannot.
Depression in an older adult mimics cognitive decline. Cognitive decline destabilises mood. And the two are frequently present at once. That is why a thinking score on its own can look like a healthy person, a mood score on its own can look like an anxious one, and both readings can be wrong about the same person on the same day.
So this platform never prints one without the other. Every sitting lands in one of four places, and each one means something different.
Thinking moved · mood steady
The one a mood screen would miss
Observed thinking changed while mood, sleep and worry held steady — a pattern less likely to be explained by mental health alone. This is exactly the documentation an evaluation needs, and a depression questionnaire on its own would have found nothing at all.
Both moved
The one nobody should untangle alone
Mood mimicking decline, decline destabilising mood, or both at once. That is a clinical job, not a kitchen-table one. What the record does is hand a clinician the map instead of a memory of a difficult year.
Both steady
The flat line, which is the product
Nothing is changing. This is the report working — the baseline just got one quarter deeper. A year of flat readings is not a wasted year; it is the thing that makes any future change visible the quarter it happens.
Mood moved · thinking steady
The one a memory test would miss
The mental-health axis moved; observed thinking has not. Common, treatable, and worth care in its own right — and in some people it comes before cognitive change. A cognition-only test here returns a clean result and honest reassurance, and both are incomplete.
Mood, worry and sleep → changed
Two of these four are cases a single instrument gets wrong — one in each direction. That is the whole argument for measuring both, and it is why the 2024 revised clinical criteria accept either a change in thinking or a recent change in mood, anxiety or motivation as documentation of the earliest stage. Most instruments produce one of those routes. A sitting here produces both, on one dated page.
Across the years
The same four questions mean different things at different ages
A sitting takes about the same twenty minutes whether you are 55 or 105. What changes is what it is for.
Start the record
Nothing to compare against yet — that is the point. Most of the fourteen modifiable risk factors named in 2024 are midlife levers. A first sitting is the line everything later is measured from.
Now it means something
About 1 in 9 Americans 65 and older is living with Alzheimer’s. But the useful question now is not do I have it — it is has anything moved since last time.
Direction, not a snapshot
Seventy-four percent of Americans living with Alzheimer’s are 75 or older. A dated run of sittings is worth more in the room than any single score — it shows which way things are going.
What still works, day to day
What a person can still do says more than what they can recall on demand. The everyday-function questions move to the front, and what the family answers becomes the record.
For the conversation
The instruments still apply; the reason for asking changes. Less about detecting anything, more about what help is needed and what still works. For the conversation, not a verdict.
What matters now
Detection is rarely the point by this age. The questions that keep earning their place are the ones about pain and what still brings enjoyment — the parts of the record a family and a clinician can still act on.
Figures. Fourteen modifiable factors and the share of cases potentially preventable: Lancet Commission on dementia prevention, intervention and care, 2024. One in nine, and seventy-four percent: Alzheimer’s Association, 2026 Alzheimer’s Disease Facts and Figures. No age band here is a risk score — these are descriptions of what a record is useful for, not predictions about anyone, and nothing on this page is a diagnosis or a recommendation about treatment.
Who it's for
Adults 55 and older
This platform is built for people 55 and older — the age range the FDA-cleared blood biomarker tests are written for, and the age from which risk starts to climb in earnest. It is also the age at which a baseline is still worth having, because there is something to compare against later. Pick the door that fits you.
On Original Medicare
Complete review, no cost. Confirm coverage with your Medicare number (MBI).
Open to anyone 55 or older who wants a baseline
Cost free. A dementia review to share with your family and your doctor. No code needed.
Paying on your own
$11. Pay securely by card, HSA, or FSA.
The Science
Watch how M3 works — and see the validation
A one-minute explainer with Dr. Gerry Hurowitz, psychiatrist, plus the peer-reviewed study behind the M3 Checklist™.
About this tool. The Dementia Review is part of M3MentalHealth.org and lives at m3mentalhealth.org/dementia. It is designed for adults age 55 and older. The Dementia Review is informational and is not a diagnosis. To find Medicare-approved providers visit medicare.gov/care-compare or call 1-800-MEDICARE.
If you are in crisis or thinking of harming yourself, dial or text 988 (Suicide & Crisis Lifeline) or go to your nearest emergency room.
Your well-being may be challenging, but understanding your mental health and pain shouldn't be. Two short check-ins. A clearer picture of how you're really doing. Your personal report is private, easy to read, free for Original Medicare members, and payable with most Health Savings Accounts (HSA/FSA).
Traditional mental health screenings often focus on identifying a single condition, primarily depression. This narrow focus can lead to the oversight and misdiagnosis of other prevalent disorders such as anxiety, bipolar disorder, and PTSD. Utilizing the M3 Checklist, a gold standard and user-friendly tool, allows for the comprehensive evaluation of various treatable mental health issues, including depression, anxiety, bipolar disorder, family dysfunction, and post-traumatic stress disorder.
Watch how M3 works
Mental Health
Take the M3 Checklist to assess mood, anxiety, and daily function in just a few minutes. It's a quick way to check in on your emotional health and identify potential risks beyond depression and recognize anxiety.
Assess Pain
Complete the DVPRS pain scale to understand how physical discomfort impacts your daily life. Mood and pain are often connected; treating one can often help improve the other.
Free with Original Medicare
No cost to Original Medicare members for the assessment or follow-up care.
$11 — Card, HSA or FSA
Pay with Visa, Mastercard, HSA, or FSA. We email an itemized receipt you can submit for HSA/FSA reimbursement.
Organization promo code
From your clinic, employer, or health-plan partner. No Medicare number or account needed to start.
If you are in crisis or thinking of harming yourself, dial or text 988 (Suicide & Crisis Lifeline) or go to your nearest emergency room.
To find Medicare-approved providers visit medicare.gov/care-compare or call 1-800-MEDICARE.
Returning?
No payment requiredAlready used the platform? Sign in below to view your past assessments, see your trend chart, or — if you're due — take a new check-in. You'll never be asked for Medicare or payment info to view your history.
Forgot the email? Use your patient ID instead
Welcome — let's get started
Enter the three pieces of information Original Medicare and other Insurers use to confirm your coverage — your name, email, date of birth. If you're not on Original Medicare, choose one of the alternate payment options below. You can return any time to see how you're doing — no password to remember.
A little background — optional, and none of it changes your scores
Choose your group
Pick the one that describes you. Every option below is cost free on this review.
🆕 New here? See how it works — the 6 simple steps
About this tool. The Dementia Review is designed for adults age 55 and older. The Dementia Review is informational and is not a diagnosis. To find Medicare-approved providers visit medicare.gov/care-compare or call 1-800-MEDICARE.
If you are in crisis or thinking of harming yourself, dial or text 988 (Suicide & Crisis Lifeline) or go to your nearest emergency room.
Discussion of Care Options · the working reference
The Clinician Handbook
Everything the care options panel is assembled from, laid out so you can check it rather than trust it: the four instruments and the authority behind every band boundary, the workflow from first review through alternatives to monitoring, the FDA approval sheet, and — the part that matters most — each option’s documented benefit set beside the responses in this person’s own record.
The one rule, before anything else
This handbook and the panel it explains put two lists side by side: what an option is documented to do, and what this person actually answered. They do not join them. The joining is the clinical act and it is yours.
Nothing here recommends starting, changing or stopping a treatment. There is no ranking, no formulary, no preselected option and no scoring of options against each other. Every option carries its argument against, and the code refuses to render one that does not — an option with an empty against column is a marketing claim, and it is dropped with an error rather than shown.
No doses appear anywhere in this section. That is a deliberate boundary, not an oversight, and it is enforced by an automated test that strips prices from the page and then fails the build if any milligram figure remains.
The parts this was built from
A bill of materials. If a number appears anywhere in this module, its source is in this section.
1.1 The four instruments
The platform administers four instruments and nothing else. Each is used verbatim, and each is scored only by the method its own published literature specifies. Where this platform sets a band boundary, the boundary comes from a published source named in the table — we do not derive cut-points ourselves.
| Instrument | What it reads | Who answers | Items | Band boundaries as implemented | Scoring authority · rights status |
|---|---|---|---|---|---|
| IQCODE Short Form |
Cognitive change over ten years — not current ability | Informant | 16 | Average of answered items, 1–5. ≤3.00 no meaningful change 3.01–3.29 watch — descriptive only, fires nothing 3.30–3.59 community threshold for decline warranting evaluation ≥3.60 secondary-care threshold |
Jorm 1994, and the two 2021 Cochrane reviews. Granted for use at no charge. Professor Jorm’s grant is conditioned on the instrument never being billed, and the platform carries an automated test that fails the release if the IQCODE ever appears on a billing path. |
| IADL-C | Everyday function, and separately compensation — the workarounds a person builds | Informant | — | Mean of answered items. <1.29 below the published boundary ≥1.29 MCI boundary — sens 0.77, spec 0.69 ≥4.41 dementia boundary — sens 0.71, spec 0.89 |
Schmitter-Edgecombe 2014; boundaries from Rahman 2025 in 488 research participants. Licence requested from Washington State University, not yet granted. This is stated plainly because the compensation pair is the most novel element in the platform and it rests on an instrument we do not yet have rights to. |
| M3 Checklist | Depression, anxiety, PTSD and bipolar spectrum, plus functional interference and substance items | The person | 27 | Raw 0–4 sums. Total 0–108: Low ≤1 · Mild ≤32 · Moderate ≤50 · Severe to 108 Depression /28 · Anxiety /48 · PTSD /16 · Bipolar /16, each with its own four-tier band |
Gaynes 2010 for the published validation; the band tables from the M3 Checklist Algorithm,
which is M3 Information’s confidential and proprietary material and is
therefore implemented but never published. The public science page carries only what Gaynes
published. Owned by M3 Information, LLC. |
| DVPRS | Pain intensity now, and four supplemental items: interference with activity and sleep, effect on mood, contribution to stress | The person | 1 + 4 | Intensity 0–10 with a descriptor at every number. Low ≤0.99 · Mild ≤3 · Moderate ≤6 · Severe to 10 The four supplemental items are reported individually and are never averaged into the intensity reading. |
Polomano 2016. Free to use unaltered. The DVPRS publishes no per-item cut-points for the four supplemental items, so any coloring of those items is our display convention and is labelled as such on the report. |
What is deliberately not in this platform
The MMSE is not here, and neither is any other licensed cognitive test. Not as a fallback, not as an option, not behind a toggle. This is a standing constraint on the build, not a gap waiting to be filled.
The consequence is worth stating to yourself before you read a record: this platform carries no direct cognitive testing of the person at all. Both cognitive readings are an informant’s account. That is a real limit and it is the subject of section 1.5.
1.2 What is ours, and what is borrowed
Of four instruments, one is owned by M3 Information. The IQCODE can never be billed. The IADL-C is not yet licensed. The DVPRS is free but only unaltered. You should know this when you read a claim about the platform, and it is set out here rather than left to be discovered.
The part that is genuinely ours
Not the instruments. The two axes and what happens where they cross — a cognitive-and-function axis read from an informant, a mood-and-pain axis read from the person, the same four instruments repeated on a schedule, and the compensation reading held beside the function total so that rising workarounds and falling function can be seen as the single event they usually are. The assembly is the invention. The instruments are borrowed, properly attributed, and used unaltered.
1.3 The guidance documents used as guideposts
- The NIA’s own diagnostic sequence — which places mental health evaluation at step four, before any conclusion about neurodegeneration. This is why the platform asks about mood and pain every quarter rather than treating them as a footnote.
- The NIA’s sequencing for agitation — strategies for physical and emotional comfort before medicines. Both agitation options in the panel carry this as their “what would need to be true first.”
- AHA and ACC blood pressure guidance, which names dementia among the reasons to treat to below 130 systolic.
- The 2024 Lancet Commission on the fourteen modifiable risk factors.
- FDA approvals, taken from FDA’s own announcements or the sponsor’s approval release, never from secondary summaries. Part three is that sheet.
1.4 The price and value sources
Prices are list or cash prices, dated. They are not what anyone with coverage pays, and a stale price presented as current is the most likely way this material becomes untrue.
- ICER — the Institute for Clinical and Economic Review — for the value-based price benchmark on lecanemab and both of its votes, reported together rather than only the convenient one.
- GoodRx cash prices for the generics and brands named, compiled 18 September 2026.
Reporting that something is poor value is not a reason for anyone to refuse a treatment their clinician recommends, and the panel says so in those words.
1.5 The limits that are built in, not incidental
- Two of the four instruments need an informant who has known the person ten years and will answer honestly. For a person living alone the platform delivers two axes and no cognitive reading at all.
- The IADL-C’s informants in the validation sample were primarily White and highly educated, and their own cognitive status was never assessed — which matters most when the informant is an aging spouse.
- The IQCODE reads change, not level. A fall in the score means the informant’s account changed. It does not mean the person recovered.
- Nothing here establishes eligibility for anything. No questionnaire can confirm amyloid.
The workflow
From the review in front of you, through the alternatives, to what happens at the next sitting. The loop is the product; a single pass through this diagram is worth much less than the fourth pass.
Where the laboratory result sits, and why it sits there
The biomarker gate is step five, not step one. That ordering is the whole argument of this module. A questionnaire cannot establish amyloid status, and this platform does not decide who is eligible for anything. What it can do is manufacture the qualified trigger — a dated, repeated, four-axis record of a person presenting with signs or symptoms, which is the condition the blood-based tests are restricted to.
So the sequence runs: watch, work up what is treatable, and only then ask the laboratory question — with a record in hand that says why you are asking. When the answer arrives you record it, and the option set changes to match. The disease-modifying group stays hidden in three of the four states. Not greyed out, not shown with a caveat — absent.
The FDA approval sheet
Every medicine named anywhere in this module, with the date, what the label actually says it is for, the evidence the approval rested on, and the safety line that travels with it. No doses. Compiled 18 September 2026 from FDA’s own announcements or the sponsor’s approval release.
3.1 Disease-modifying — the anti-amyloid antibodies
| Agent | Approved | What the label is for | What the approval rested on | The safety line |
|---|---|---|---|---|
| Lecanemab Leqembi · Eisai/Biogen |
6 Jan 2023 accelerated 6 Jul 2023 traditional |
Alzheimer’s disease. Treatment is to be initiated in mild cognitive impairment or the mild dementia stage, the population in which it was studied, with confirmed amyloid pathology. | An 18-month randomized trial of intravenous infusion every two weeks, showing slowing of cognitive and functional decline and measurable plaque removal. | Boxed warning for ARIA — amyloid-related imaging abnormalities. Usually temporary brain swelling, sometimes with small spots of bleeding; can progress to life-threatening edema with seizures. Higher incidence in ApoE ε4 homozygotes. Genotyping and MRI monitoring are prerequisites, not options. |
| Lecanemab subcutaneous Leqembi Iqlik |
13 Jul 2026 | A starting dose given at home, by the person or a caregiver — the first at-home initiation of any Alzheimer’s treatment. Same disease indication. | The same lecanemab evidence base; this approval concerns route and setting of the initiation dose, not a new efficacy claim. | The same ARIA boxed warning applies. Moving administration into the home does not move the MRI monitoring requirement out of it. |
| Donanemab Kisunla · Lilly |
2 Jul 2024 | Alzheimer’s disease, initiated in MCI or mild dementia with confirmed amyloid. | A randomized trial showing slowing of decline. Distinctively, dosing can sometimes be stopped once plaques clear, which is a real difference from lecanemab. | ARIA, on the same terms, with the same APOE and MRI conditions. |
| Aducanumab Aduhelm · Biogen Withdrawn |
Jun 2021 accelerated discontinued Jan 2024 |
Was: Alzheimer’s disease. No longer marketed. | Accelerated approval on plaque reduction as a surrogate, over the near-unanimous objection of FDA’s own advisory committee. | Kept on this sheet deliberately as a cautionary case, not quietly dropped: an approval on a surrogate endpoint, a contested launch price, poor uptake, and withdrawal. A clinician weighing the current antibodies should have this in view. |
3.2 Symptomatic — cognition
| Agent | Approved | What the label is for | What the approval rested on | The safety line |
|---|---|---|---|---|
| Donepezil Aricept |
1996 | Dementia of the Alzheimer’s type. Now used across stages. | Randomized trials showing modest symptomatic benefit on cognition and global function. None of these agents alters the disease, and that has never been claimed for them. | GI effects — nausea, diarrhea, vomiting — bradycardia, syncope and sleep disturbance. GI intolerance is the commonest reason people stop. |
| Rivastigmine Exelon · also a patch |
2000 | Dementia of the Alzheimer’s type, and dementia associated with Parkinson’s disease. | ||
| Galantamine Razadyne |
2001 | Mild to moderate dementia of the Alzheimer’s type. | ||
| Memantine Namenda |
2003 | Moderate to severe dementia of the Alzheimer’s type. | A different mechanism — NMDA receptor antagonism. Evidence is weaker in mild disease than in moderate to severe, and the label reflects that. | Generally well tolerated. Dizziness, headache and confusion in some people. |
| Memantine + donepezil Namzaric |
23 Dec 2014 | Moderate to severe Alzheimer’s disease — a fixed-dose combination of two agents already approved separately. | The combination of two established agents; convenience and adherence rather than a new efficacy claim. | The warnings of both components. Worth checking against price: the brand costs many times its two generic components. |
| Benzgalantamine Zunveyl · Alpha Cognition |
26 Jul 2024 launched 2025 |
“Mild to moderate dementia of the Alzheimer’s type in adults.” | A prodrug of galantamine — the same active molecule, delivered so as to bypass the stomach. The claim is tolerability, not greater efficacy, and no greater efficacy has been shown. | The galantamine class effects, with the GI burden the formulation is designed to reduce. The value question is whether generic galantamine has actually been tried and failed — which is a history question, not a score question. |
A disclosure about this table. The four legacy approvals — donepezil, rivastigmine, galantamine and memantine — are given by year rather than by exact date, because the day was not confirmed at a primary source in this pass. Every other date on this sheet was. When the legacy dates are verified they will be stated to the day, and until then the year is what this handbook will claim.
3.3 Agitation associated with dementia due to Alzheimer’s disease
Two medicines are approved for this, and neither of them is a first step. The NIA’s own sequencing puts strategies for physical and emotional comfort before medicines, and both rows below carry that as their precondition in the panel.
| Agent | Approved | What the label is for | What the approval rested on | The safety line |
|---|---|---|---|---|
| Brexpiprazole Rexulti · Otsuka/Lundbeck |
10 May 2023 | “Treatment of agitation associated with dementia due to Alzheimer’s disease.” The first medicine of any kind indicated for this. | Randomized trials in agitation associated with Alzheimer’s dementia. | Class boxed warning, quoted verbatim from the label:
INCREASED MORTALITY IN ELDERLY PATIENTS WITH DEMENTIA-RELATED PSYCHOSIS. Also weight gain, akathisia, metabolic and extrapyramidal effects, requiring monitoring at every follow-up. |
| Dextromethorphan / bupropion Auvelity · AXS-05 · Axsome |
30 Apr 2026 | Same indication wording. The first non-antipsychotic approved for it. (Auvelity was already approved for major depressive disorder in 2022.) | Stated in full, because the picture is mixed. ADVANCE-1 (n=366) met its primary endpoint: CMAI −15.4 against −11.5 on placebo, p<0.001, and against −10.0 on bupropion alone, p=0.010. ADVANCE-2 (n=408) did not meet its primary endpoint: −13.8 against −12.6. The relapse-prevention trial ACCORD-2 met its primary endpoint — time to relapse HR 0.276, p=0.001, with relapse in 8.4% against 28.6% on placebo. | Boxed warning for suicidal thoughts and behaviors (the antidepressant class warning). Also seizures, increased blood pressure, neuropsychiatric effects including activation of mania or hypomania, serotonin syndrome, hyponatremia, angle-closure glaucoma and dizziness. Check the medication list for anything else containing bupropion or dextromethorphan. |
The door this sheet has to close
Auvelity is a fixed-dose combination of two molecules that are each generic separately. A price list beside that fact is an invitation, and it has to be refused in words rather than by omission: they are not equivalent. It is the combination that was studied in the trials, and assembling it from parts has no trial behind it. Nothing in this module will suggest otherwise.
3.4 Sleep — and a label update that is often reported as more than it is
| Agent | Date | What actually happened |
|---|---|---|
| Suvorexant Belsomra · Merck |
3 Feb 2020 | This was a label update, not a new Alzheimer’s indication. FDA approved an update to the prescribing information to include findings from a study of insomnia in people with mild-to-moderate Alzheimer’s disease. The original insomnia indication was unchanged. It is on this sheet because it is frequently reported as an Alzheimer’s approval, and because sleep is one of the four DVPRS supplemental items — so it will come up. |
Benefit against response
The part that answers the risk question. Not with a disclaimer — with the structure.
Why this module is defensible, stated without hiding behind anyone
The concern about software like this is always the same: that it tells a clinician what to prescribe. The answer is not a warning label and it is not a lawyer’s paragraph. It is what the module actually puts on the screen.
It puts up a documented benefit, with the study and the number behind it, and next to it the responses this person actually gave — their own answers, their own bands, their own direction of travel over the quarters. That is the entire display. The clinician reads across and decides. The basis for every line is the line itself, which is exactly the thing a clinician has to be able to review independently.
Four structural properties do the work, and each is enforced in code rather than promised in prose: every option carries its argument against, and one without it is refused; nothing is ranked or preselected; the display changes only when the person’s answers change, never on a schedule or a nudge; and no doses appear. Take any of the four away and the argument weakens. That is why they are constraints and not preferences.
4.1 The matrix
For each option: what it is documented to do, and what in this record makes it worth raising. The right-hand column is not a trigger. Nothing in the platform fires on these combinations, and none of them opens or closes an option. They are what to look at.
Work up the treatable causes
Standard of care · several causes are reversible · runs first and runs anywayDocumented benefit
- Depression, pain, sleep, thyroid, B12, alcohol and anticholinergic burden can all look like cognitive decline
- The NIA places mental health evaluation at step four of its own diagnostic sequence
- Generic, cheap, and sometimes reversible
What in this record makes it relevant
- M3 depression subscale at Mild or above (7 or more of 28) beside an IQCODE in the orange or red band — a mimic question before it is a dementia question
- DVPRS intensity at Moderate or above (above 3 of 10), or the sleep or mood supplemental item elevated, with or without high intensity
- M3 Q26 at 3 or more (alcohol) or Q27 at 1 or more (drugs)
- Two of the four instruments in this platform ask about the first two every quarter — so this column is rarely empty
Blood pressure to target — below 130 systolic
Randomized, with dementia as an outcome — the strongest evidence on this pageDocumented benefit
- CRHCP-3: significantly less all-cause dementia at four years and a 15% reduction sustained over seven, in nearly 34,000 adults
- AHA and ACC recommend below 130 naming dementia as a reason
- Generic agents, a few dollars a month
What in this record makes it relevant
- Nothing, and that is the point. This option is independent of every score in the platform and of amyloid status
- It applies at the green band and the red one alike. No response opens it and no response closes it
- It is listed first for that reason: the highest-quality evidence here belongs to the option the assessments have no bearing on
Cholinesterase inhibitor · memantine · benzgalantamine
Approved, symptomatic, modest effect — none alters the diseaseDocumented benefit
- May temporarily improve or stabilize memory and thinking in some people
- Decades of safety record
- Generics: memantine from about $18 a fill, galantamine from about $28, against several hundred dollars for brand equivalents
What in this record makes it relevant
- IQCODE at or above 3.30 — the published community threshold for decline warranting clinical evaluation — with IADL-C at or above 1.29
- For memantine specifically: IADL-C at or above 4.41, the dementia boundary, and/or IQCODE at or above 3.60, the secondary-care threshold
- For benzgalantamine: not a score question at all. It turns on whether generic galantamine was tried and stopped on tolerability — which this record does not capture, so you have to ask
Anti-amyloid antibody
Shown only when the biomarker gate reads positive — withheld in the other three statesDocumented benefit
- Removes amyloid plaque measurably and reliably
- Slows cognitive and functional decline in early disease
- Donanemab can sometimes be stopped once plaques clear
- An at-home subcutaneous starting dose was approved 13 July 2026, which changes access rather than effect
What in this record makes it relevant
- Confirmed amyloid. Nothing else opens this option — no score, no band, no trend
- Once open, the record’s job is the stage question: the trials recruited MCI and mild dementia. An IQCODE of 3.30–3.59 with an IADL-C of 1.29–4.40 is the window that resembles the trial population
- Above the IADL-C dementia boundary of 4.41, the trial population no longer matches this person — which is a reason to say so out loud, not a reason to hide the row
The two agitation medicines
Both approved for this · neither is a first stepDocumented benefit
- Auvelity: ADVANCE-1 met its primary endpoint on the CMAI; ADVANCE-2 did not; ACCORD-2 showed relapse in 8.4% against 28.6%. It avoids the antipsychotic class mortality warning
- Rexulti: indicated specifically for this, and may help where other approaches have failed and distress is severe
What in this record makes it relevant — and what should stop you
- Before either: have non-drug approaches been tried? The NIA sequence puts them first and so does the panel
- M3 bipolar subscale at Mild or above (4 or more of 16) — Auvelity contains bupropion, and activation of mania or hypomania is a listed risk
- M3 Q26 at 3 or more (alcohol) — relevant to the seizure risk
- The same bipolar items bear on the antipsychotic choice, and the report carries a safety cross-link for it
4.2 The safety cross-links that run regardless of everything above
| The response | Fires at | What happens, and where |
|---|---|---|
| M3 Q5 — suicide ideation | 1 or more “rarely”+ |
The 988 pathway fires on the person’s own report, regardless of every other score and regardless of whether this clinician section is switched on at all. It is not gated by anything. |
| M3 Q24 — interferes with work or school | 3 or more | Functional interference gateway items. They shape the documentation block and the report narrative; they do not open or close a care option. |
| M3 Q25 — affects relationships | 3 or more | |
| M3 Q26 — led to using alcohol | 3 or more | Substance gateway. Bears on seizure threshold, on interaction with everything on the list, and on whether a cognitive reading means what it appears to mean. |
| M3 Q27 — led to using drugs | 1 or more “rarely”+ |
Same. This threshold was an open question in the source document — the words read “rarely or more frequently” and a parenthetical read 3. Resolved 17 September 2026 in favor of the words. The resolution is recorded in the code rather than the ambiguity being quietly deleted. |
| The compensation pair — IADL-C | direction | Compensation rising while function holds is the MCI signature. Compensation falling while function worsens is the turn, and the module reports it explicitly when it happens. Neither reading is ever displayed without the function total beside it. |
4.3 Where this argument does not hold, stated plainly
The medication tier ladder is the exception
Everything above describes the care options panel, and for that panel the argument holds: two lists, no joining, no doses, an argument against every row.
It does not describe recommendM3Tier(), a separate and older component of
the platform, which outputs a named drug at a specific dose with a titration schedule,
selected by an algorithm from a score. That is a different category of thing. It is not
information set beside a response for a clinician to join — it is the join, performed by
software.
This handbook is not going to pretend otherwise, and pointing at the panel’s discipline does not launder the ladder. It was left open, with four options: to name drug classes rather than products, to drop the doses and keep the reasoning, to move it behind this same clinician-only gate, or to leave it and accept the position knowingly. All four are defensible; doing it by default is not.
Decided, 20 September 2026: the ladder is off the dementia path
After a review of the CDS, M3 chose findings only, no ladder for this path. Here the mental-health axis is shown as findings to work up: depression beside an elevated IQCODE, pain and the sleep it costs, alcohol or drug use, and the mania and hypomania items. The mania and hypomania items are read beside the agitation discussion and the bipolar safety gate. No tier is shown, no drug is named and no dose appears, on screen, in the printed brief or in the downloaded report. The Six-Month Pathway, which counted quarterly sittings as months, is gone from this path too, and the person’s own “Six-Month Care Plan” with it.
The ladder remains on the mental-health path, and since the same review it reads the M3’s weighted risk scores, which its cut-offs were written for. Whether it should keep doses there is still open.
What would make this handbook wrong
Six live limits. They are listed because a reference that cannot say how it fails is asking to be trusted rather than checked.
- A date goes stale. Approvals and prices change. Everything here is dated 18 September 2026; treat anything older than a quarter as a starting point.
- The IADL-C licence is refused. Washington State University has not granted it. If the answer is no, the function axis and the compensation pair go with it — not one screen, the differentiator.
- There is no informant. Two of the four instruments require one. For a person living alone, this module delivers two axes and no cognitive reading at all, and nothing in this handbook applies the way it reads.
- The record is thin. Direction of travel needs at least two sittings. One sitting is a photograph, and a photograph of a person with dementia looks much like a photograph of a person having a bad month.
- Practice effects. Repeated measurement is the whole design, and cognitive training on assessment-like tasks can raise a score without raising ability. If the IQCODE improves while the IADL-C worsens, suspect the instrument before you believe the good news.
- A treatable cause was found and the search stopped. The commonest failure of step four is treating it as a branch rather than a step. Depression and neurodegeneration co-exist comfortably.
In crisis, call or text 988 (Suicide & Crisis Lifeline). The Alzheimer’s Association Helpline is 800.272.3900, 24 hours a day. A positive suicide item on the M3 fires the 988 pathway on the person’s own report regardless of every other score and regardless of whether this section is switched on.
Before you use the panel
Discussion of Care Options
What the four instruments can and cannot support, where the laboratory result sits in the sequence, and why this section offers options with their arguments against rather than a recommendation. Read this once; the panel itself assumes you have.
The working reference is the Clinician Handbook — the parts this module was built from, the workflow from review through alternatives to monitoring, the FDA approval sheet, and each option’s documented benefit set beside the person’s own responses. This page is the argument; the handbook is the evidence behind it.
Who sees this, and who does not
This section is visible only when both of these are true:
- A clinician is signed in — not the person, not a family member, not an administrator browsing a record.
- Clinical decision support is switched on for that practice. It is off for every practice by default. M3 grants approval and the practice then activates it; either one absent means this section does not exist.
The person and their family see the data, not this layer. Their report carries the four scores, the direction of travel, the crossing and the questions worth asking. It does not carry the options matrix, and the matrix never appears in the printed or downloaded report regardless of who generates it. That separation is deliberate and it is enforced in code, not by convention.
This page sits behind the same line. It and the Clinician Handbook open only inside the platform, under both of the conditions above: a clinician signed in through the clinician sign-in — which is separate from the person’s own sign-in — and CDS live for that practice. A clinician whose practice does not have CDS live sees a notice saying so. A person or family member who follows an old link to either page reaches a notice asking for clinician sign-in, and nothing else.
The reason is not secrecy. It is that a list of medicines with arguments for and against is useful to someone who can prescribe and weigh them, and frightening to someone who cannot. A family reading a row about an antipsychotic's mortality warning, out of the context of the person's whole picture, learns nothing they can act on and quite a lot they will lie awake about.
What this section is not
It is not a recommendation, a ranking, a protocol or a pathway. There is no “suggested option”, no highlighted row, no default. Every option carries its argument against, in the same size type, in a column of equal width.
It carries no doses. That is a deliberate design decision, not an omission. Agents are named, with indication, benefit, harm and monitoring burden; dosing belongs to the prescribing information and to your judgement about this person. Software that selects a product and a dose from a score is doing something different from software that lays out what exists.
It does not establish eligibility for anything. Most importantly it cannot establish amyloid status, which is the single gate that decides whether half the options below are even applicable. No questionnaire can.
It is not a substitute for the prescribing information, and where the two differ, the PI governs.
The sequence the panel follows
Five steps, in this order, because the order is the argument.
Step one
Detection — what the four instruments actually say
Two instruments answered by an informant, two by the person: IQCODE (cognitive change against ten years ago), IADL-C (everyday function, and whether aids are being used), M3 (mood, anxiety, PTSD, bipolar spectrum, and the impairment gateway) and DVPRS (pain now and its interference over 24 hours).
What they support. A dated statement about direction: whether something has changed, in which domain, in whose account, and by how much against this person's own earlier answers. From the second sitting onward that is a trajectory rather than a snapshot.
What they do not support. A diagnosis, an aetiology, or a distinction between Alzheimer's disease and any other cause of dementia. A high IQCODE with a high M3 depression score is as consistent with treatable depression as with early neurodegeneration, and the panel says so rather than resolving it for you.
The asymmetry worth holding onto: a rising score is worse on all four instruments. A falling score is not the mirror image. Mood and pain genuinely improve. A falling IQCODE means the informant's account changed — not that the person recovered — and the panel labels it that way every time.
Step two
Exercise and lifestyle — and the honest difference between two things that get called the same
These are placed before any medicine because the evidence for some of them is stronger than the evidence for most of the medicines, and because they are available now, to this person, whatever the laboratory eventually says.
They are two different things and the panel separates them.
| Intervention | Evidence | What it rests on |
|---|---|---|
| Blood pressure to target systolic below 130 |
Randomised, dementia as an outcome | CRHCP-3 randomised nearly 34,000 adults: significantly less all-cause dementia at four years and a 15% reduction sustained over seven. The AHA and ACC now recommend below 130 naming dementia as a reason. The strongest evidence-to-cost ratio available. |
| Multidomain lifestyle programmes FINGER, US POINTER, LatAm-FINGERS |
Randomised, benefits modest | Changing several things at once — activity, nutrition, cognitive and social engagement, vascular monitoring — has been tested and worked. The benefits are modest and their clinical meaning is unsettled, and the platform says that wherever it says the rest. |
| Physical activity | Consistent association, trial evidence mixed | Associated with staying well while ageing and plausibly with reduced decline. Recommended on general health grounds regardless of the cognitive question. |
| Hearing, vision, sleep, isolation, alcohol | Modifiable risk factors | Among the fourteen named in 2024. Several are correctable in a single visit, and uncorrected hearing loss in particular can look like cognitive decline in a consultation room. |
| Cognitive training and brain-training programmes | Weak and contested for transfer | People reliably get better at the trained task. Whether that transfers to everyday function is not established, and the commercial claims in this category have outrun the evidence for two decades. The panel lists it because people ask, and states the limit rather than omitting the row. |
One caution the panel prints beside cognitive training: practice effects. A person who trains on tasks resembling an assessment can score better without being better, which is exactly the failure mode a platform built on repeated measurement has to declare.
Step three
The Alzheimer's test — why the panel has a starting point and then changes
This is the hinge of the whole section. Amyloid status decides which options are applicable, and it arrives later than the questionnaires do. So the panel takes a position before the laboratory, states that it is provisional, and revises when the result lands.
What is now available: the FDA cleared the first single-biomarker plasma test, Elecsys pTau217, on 24 August 2026, for primary and specialty care — with a fourth blood test cleared in September. This is a genuine change: amyloid status used to require a PET scan or a spinal tap.
Read the intended use. Cleared “for use in diagnosing Alzheimer's disease in patients presenting with signs or symptoms of cognitive impairment.” Not asymptomatic screening. Which is precisely why the four instruments come first in this sequence: something has to establish that signs are present before the test applies at all.
The panel therefore holds one of four states, and shows what each one changes:
State one
Not ordered
The panel shows the whole prodrome column and no anti-amyloid options. It states what would need to be true to order the test, and what the result would and would not settle.
State two
Ordered, pending
The provisional position is held and marked provisional. Lifestyle and symptom options stand on their own merits and do not wait for the result.
State three
Amyloid positive
The anti-amyloid column becomes applicable, with eligibility, APOE ε4 status, ARIA monitoring and the cost conversation attached to it. Applicable is not indicated; the panel does not cross that line.
State four
Amyloid negative
The most informative result, and the one most often under-used. Anti-amyloid options come off the table and the question becomes what else is causing this — which sends the panel back to step one and to the treatable causes.
A negative result is not a clean bill of health. It excludes amyloid, not decline. The other dementias, LATE, and every treatable mimic remain live, and the panel says so on that state rather than letting a negative read as reassurance.
Step four
Medicines — laid out as options with their arguments against
Every row in the panel has the same shape: what supports it for this person, from their own scores; what argues against it, including harms, monitoring burden, boxed warnings and interactions with what they already take; and what would need to be true first. No row is marked as the answer.
FDA approvals are the guideposts, and the panel is explicit about what an approval does and does not mean. Approved means a regulator judged benefit to exceed risk for a defined population on defined endpoints. It does not mean effective for this person, nor good value, nor better than an alternative. Those are three separate questions and the panel keeps them separate.
The two that carry the largest asymmetry between hope and evidence:
What the anti-amyloid antibodies do
- Remove amyloid plaque, measurably and reliably
- Slow cognitive and functional decline in early disease
- Leqembi and Kisunla are both fully approved for MCI or mild dementia with confirmed amyloid
- Kisunla can sometimes be stopped once plaques clear
- A subcutaneous at-home starting dose for Leqembi was approved 13 July 2026, which changes access
What argues against them
- They slow; they do not stop or reverse
- ARIA — brain swelling and microhaemorrhage. Usually asymptomatic, sometimes serious, rarely fatal. Higher in APOE ε4 homozygotes
- Genotyping and an MRI monitoring schedule are prerequisites, not options
- Anticoagulant use raises haemorrhage risk
- ICER put lecanemab's value-based price at $8,900–$21,500 a year against a $26,500 list price, voted 15–0 that it is low long-term value at that price and 12–3 that evidence was not adequate to show a net health benefit versus supportive care
Both columns are shown, always. A panel that showed only the left one would be marketing; one that showed only the right would be obstruction. The clinician's judgement needs both, and the person's family deserves a clinician who has seen both.
On agitation specifically, the panel prints the NIA's own sequencing before either approved agent: these medicines come after other strategies for physical and emotional comfort have been tried. Auvelity, approved for this indication on 30 April 2026, is the first non-antipsychotic option and carries a boxed warning for suicidal thoughts and behaviours in younger adults, seizure risk and blood-pressure effects. Rexulti is an antipsychotic and carries the class warning of increased mortality in older adults with dementia-related psychosis. Neither is a small decision.
And the cheapest finding on the page. Where a person is taking a brand cholinesterase inhibitor or memantine, the same molecule is frequently available generically at a fraction of the price — memantine from about $18 a fill and galantamine from about $28, against brand prices several hundred dollars higher. The panel surfaces that as a question, because it is one pharmacists answer in a minute and nobody asks.
Step five
Through care — what the panel does at the next sitting, and the one after
This is the part a single-visit tool cannot do and the reason the panel exists inside this platform rather than beside it.
At each administration the panel shows what moved, in whose account, and what that does to the options already on the table. Concretely:
- A domain that moved while a medicine was running is shown against that medicine's start date. The platform does not claim causation and says so; it dates the two events beside each other and leaves the inference to you.
- The compensation pair. A rising IADL-C with falling aid use is the signature the platform was built to catch: compensation climbs as someone starts to struggle and collapses as the illness deepens, so a low compensation reading is ambiguous on its own. The panel never shows it without the functional total beside it.
- A falling score that is not improvement. Handled explicitly, every time, rather than by hoping the reader remembers.
- Stopping. Worth agreeing in advance — particularly where plaque clearance can end treatment, and where a symptomatic drug is producing no measurable change across three sittings.
- The mood axis as the confound. If depression is high and rising, the panel puts that question in front of the cognitive interpretation rather than after it, because it is both more treatable and more often missed.
The panel accumulates; it does not reset. A clinician joining at the fourth sitting sees three quarters of history they did not have to take.
How to use it in a visit, and what would make it wrong
In the room
- Read the direction before the numbers. The scores matter less than which way they are going and in whose account.
- Settle which track the person is on before reading the medicine rows — prodrome with nothing confirmed, or confirmed amyloid with MCI or mild dementia. Half the panel is inapplicable in the first case.
- Rule out the treatable thing. Depression, pain, sleep, thyroid, B12, alcohol, and an anticholinergic-heavy medicine list. The platform asks about the first two every quarter for exactly this reason.
- Check the blood pressure. It is the best-evidenced intervention on the page and it does not depend on any test result.
- Use the options list as an agenda, not an answer. It is at its most useful read aloud with the family, both columns, in their order.
What would make this panel wrong, stated plainly
Every one of these is live today. None is hypothetical, and a clinician relying on this section is entitled to know them before they do.
- No informant, no cognitive axis. Two of the four instruments require someone who has known the person ten years. For a person living alone the panel is running on half its inputs, and it does not currently say so loudly enough.
- The IADL-C's cut-offs were validated in a primarily White, highly educated sample, and the informants' own cognitive status was never assessed — which matters most when the informant is an ageing spouse. The two published samples also disagree about the first boundary, 1.29 against 1.81 on the same eleven items. This platform uses 1.29, and it is the weakest number in the system.
- One IQCODE band boundary is under review. The bands sit at 3.00 / 3.30 / 3.60 and there is reason to think the middle value belongs to the 26-item long form rather than the 16-item form in use here.
- The IADL-C is used under a licence that has not been granted. Washington State University has been asked. Until they answer it is present for evaluation only.
- Prices and approvals move. Every figure in the panel carries the date it was compiled, because a stale price presented as current is the most likely way this becomes untrue.
- Nothing here detects amyloid or tau. If the panel ever reads as though it does, that is a defect in the panel and not a capability of the platform.
Compiled 18 September 2026. Nothing here is an endorsement by any agency or company named, and none is claimed.
In crisis, call or text 988 (Suicide & Crisis Lifeline). The Alzheimer’s Association Helpline is 800.272.3900, 24 hours a day. A positive suicide item on the M3 fires the 988 pathway on the person's own report regardless of every other score and regardless of whether this section is switched on.
