Care Assessments

Part of the Anaya Care Handbook — the source of truth for how the product must behave. When the product needs to change, change this document first, then make the system match it.

What this covers

This page governs the six care assessments kept on every client's record. Together they answer the questions an agency must answer before it can care for someone well: What can this person do on their own? How does their mind work? Who are they as a person? Is their home safe? Can they eat and swallow safely? The answers are the foundation for the care plan, the care provider's daily tasks, and the AI's understanding of the client.

The intake assessment of a prospective client is covered in Initial Assessments; the notes care providers record when a client's condition changes are covered in Change of Conditions. Staff can capture these assessments through a guided, conversational intake — see Assessment Mode.

Key terms

• Care assessments — the six evaluations on every client record: Simplified Routine Task Inventory, Routine Task Inventory, Montessori Profile, Allen Cognitive Level, Home Safety, and Meal Assessment. Each client has at most one of each.
• ADL (Activities of Daily Living) — the basic self-care activities of everyday life: bathing, dressing, eating, toileting, moving around, continence, communication, and managing medication.
• IADL (Instrumental Activities of Daily Living) — the practical life skills that keep a household running: preparing meals, housekeeping, managing money, transportation, shopping, and managing appointments.
• Capacity — in the Simplified Routine Task Inventory, what the client can do for a task on their own: Independent · Supervision / Cueing · Minimal Assistance · Moderate Assistance · Dependent · Not Established.
• Support now — in the Simplified Routine Task Inventory, who currently performs a task for the client: Self · Family · Paid Caregiver · Unmet (needed, but no one does it).
• Risk — in the Simplified Routine Task Inventory, a flag meaning the client can do a task but doing so is unsafe.
• Allen Cognitive Level (ACL) — a standardized 6-level scale of cognitive functioning, from Level 1 (automatic reflex actions, total care) to Level 6 (fully planned, independent actions).
• ACL mode — a fine-grained step within an Allen level (e.g. 4.4 "Complete a Goal"); the assessor selects exactly one mode for the client.
• MNA-SF (Mini Nutritional Assessment — Short Form) — a standardized nutrition screening: six scored questions producing a total that classifies the client as well-nourished, at risk, or malnourished.
• IDDSI level — the international standard scale for food texture and drink thickness, used for clients with swallowing difficulties.
• Assessment status — where each assessment stands: Not Started, Draft, or Completed. Only Completed assessments inform AI generation.
• Assessment version — the sequential number a care assessment's record carries: every save keeps the prior content in the record's own history and increments the version (CA-15). One record, many versions.
• Assessment generation — a client-level stamp cut across all six care assessments at once when reassessment suggestions are applied (CA-16): every assessment bumps a version and shares the generation number, so "generation N" names one same-moment picture of the client.

The six assessments

1. Simplified Routine Task Inventory

Purpose: a quick, structured picture of what the elder can do and who is doing it today — the agency's working summary of service scope. Shorthand Simplified RTI or SRTI (the on-screen label stays the full name; the client-record route is .../assessments/simplified-rti). It follows the GCS Initial Care Assessment model — GCS being the reference care framework Anaya adopted for this instrument — in which Capacity and Support now are recorded separately because they are different facts. It is the first assessment every client has: when a client is created from a care proposal, it arrives pre-filled from intake and already Completed.

What it covers:

How it is answered (ADL / IADL): every task gets three facts plus optional notes:

1. Capacity — Independent · Supervision / Cueing · Minimal Assistance · Moderate Assistance · Dependent · Not Established
2. Support now — Self · Family · Paid Caregiver · Unmet
3. Risk — tick when the elder can do it but doing so is unsafe

Equipment uses Has & uses / Has, not used / Not present (an equipment item may also carry a Risk flag); Home safety uses In place / Present, inadequate / Not present. The inventory also records optional mobility extras (two-person assist, bedbound, falls in the last six months) and free-text observations for meal prep, ADLs, and IADLs. Values are stored as these descriptive labels (not numeric codes).

How it is used: carried from intake (proposal → client), and a primary input when AI generates the care plan and care provider tasks. Unmet support rows and risk flags are the clearest service-scope signals.

2. Routine Task Inventory

Purpose: the deep version of the same question. Where the Simplified inventory records a score, the full Routine Task Inventory records the evidence — for each ability, the assessor works through detailed behavioral criteria and identifies the highest level of independence the client genuinely demonstrates. It is the comprehensive baseline for care planning and for tracking change over time.

What it covers — four domains:

How it is scored: each ability has a ladder of rating levels (roughly 1–6), and each level lists concrete behavioral criteria (e.g. "initiates grooming tasks and follows typical procedures to completion"). The assessor checks the criteria that apply; the highest level whose criteria the client meets is their rating for that ability. This is a professional-judgment instrument — the system records the evidence, it does not compute a verdict.

How it is used: the detailed reference behind the care plan, and the baseline against which decline or improvement is recognized. Its results are also a direct input when the AI builds task instruction steps, so a task's step-by-step guidance matches the level of independence this assessment records. In generation contexts the other pages also call this instrument RTI-E (the expanded Routine Task Inventory) — it is the same assessment described here.

3. Montessori Profile

Purpose: the assessment of who the client is rather than what they can do. Following the Montessori approach to dementia and elder care, it captures the person's history, preferences, routines, and ways of communicating — so care providers can offer activities and interactions that are meaningful to this person, not generic ones.

What it covers — five areas:

• Likes & dislikes — how they prefer to be addressed; favorite time of day; favorite social, physical, recreational, and spiritual activities; music, reading, TV, games and sports preferences; household chores they enjoy (cooking, gardening, laundry…); cultural identity; meaningful roles past and present.
• Personal history — place of birth; family members; religious affiliation; education and job history; military service; awards, clubs, spouses, best friends; significant life events; proudest accomplishments; favorite memories; pets; travel.
• Daily routine — a typical day for each day of the week; wake-up and bedtime routines; what makes them happy; what upsets them and what comforts them; calming strategies; anxiety triggers; sundowning and nap patterns; sleep disturbances; social preferences.
• Cognition & language — hearing aid and glasses; languages spoken; how they converse (single words … full conversations); orientation to the rooms of their home; reading ability and preferences; preferred communication style; nonverbal communication; tips for communicating; cognitive strengths (procedural memory, long-term memory topics, decision-making, attention span).
• Observation — how they communicate wants and needs; responsive behaviors with their triggers and time of day; engagement level (high / moderate / low / variable); when they engage best; signs of engagement and disengagement; what motivates them.

How it is scored: it isn't. The Montessori Profile is narrative — text, choices, and checklists with no numeric score. Its value is in its detail.

How it is used: drives the personalization of engagement activities and meals and gives every care provider the context to connect with the client as a person. Its Daily routine area is also a source signal the AI reads when drafting the care plan's routine anchors (PLAN-37, CA-17) — the profile itself stays narrative; structure is added at the care plan.

4. Allen Cognitive Level

Purpose: a standardized, evidence-based placement of the client's cognitive functioning on the Allen scale — the single most consequential data point for how much supervision and what kind of guidance the client needs.

What it covers: the assessor selects exactly one mode out of 29, grouped into six levels:

Each mode within a level (e.g. 4.4 "Complete a Goal", 5.6 "Consider Social Standards") carries a precise behavioral description the assessor matches the client against.

How it is used: the selected mode translates directly into care guidance — how tasks must be presented, how much supervision shifts need, and what the client can safely be asked to do. It is also a direct input when the AI builds task instruction steps. For engagement activities, the Allen Cognitive Level result is used only when the client has dementia; for clients without dementia, engagement is driven by the Montessori Profile and client profile instead.

5. Home Safety

Purpose: a room-by-room inspection of the client's home, because in home care the environment is part of the overall picture. It identifies hazards and accessibility barriers and ends in concrete recommendations.

What it covers — twelve areas, 100+ checkpoints:

1. Exterior entrances and exits — walkways, handrails, lighting, thresholds, locks, visible house numbers
2. Interior doors, stairs, and halls — doorway widths, floor level changes, stair rails and lighting, clutter
3. Bathroom — tub and shower access, grab bars, bath bench, toilet height, non-slip surfaces, caregiver space
4. Kitchen — task lighting, counter and shelf reach, stove controls, heat-safe surfaces
5. Living, dining, and bedroom — chair and bed heights, rug safety, reachable phone and remote, clear pathways
6. Laundry — washer and dryer access, sorting surfaces, dryer vent condition
7. Basement — stair condition, handrails, combustibles, moisture
8. Telephone and door — phone usability, visitor identification, emergency response device
9. Storage — closet reach, dresser operation, closet lighting
10. Windows — opening mechanisms, locks, sill safety
11. Electrical outlets and controls — outlet placement, ground-fault protection, switch reach, cord hazards
12. Heat, air, light, smoke and CO alarms, water temperature — alarms present and working, fire extinguisher, thermostat reach, water heater temperature

How it is scored: every checkpoint is marked off with the assessor's notes; there is no numeric score. The assessment closes with a written recommendations summary — which the AI can draft from the findings, for staff to review and edit.

How it is used: drives home adaptations, informs the care plan's safety section, and tells care providers what to watch for in the home.

6. Meal Assessment

Purpose: the complete picture of how the client eats — covering safety (swallowing), health (nutrition), ability (feeding), and identity (preferences). Eating is where health risk and daily quality of life meet, so this assessment feeds both safety rules and meal planning.

What it covers — five areas:

• Swallowing & dysphagia — recommended IDDSI levels for food texture and drink thickness; aspiration risk; swallowing ability and oral-motor function; a checklist of dysphagia warning signs (coughing at meals, wet voice, pocketing food, prolonged chewing, drooling, food refusal, weight loss, pneumonia history); required positioning during meals.
• Nutritional status — the six MNA-SF screening scores (declining intake, weight loss, mobility, recent stress or illness, neuropsychological problems, body-mass or calf measurement), the computed total, and the resulting classification (normal / at risk / malnourished); hydration status and estimated daily fluid intake; appetite level and patterns; oral health concerns (missing teeth, ill-fitting dentures, mouth sores, dry mouth, chewing difficulty, pain).
• Feeding abilities — self-feeding level; a skills checklist (bringing food to mouth, using utensils, drinking from cup or straw, opening containers, keeping attention on the meal, pacing); adaptive equipment in use or needed (built-up utensils, plate guard, nosey cup, weighted utensils, non-slip mat, two-handled cup…); whether setup, assistance, supervision, or cueing is required; preferred eating position and environment.
• Dietary needs & preferences — food allergies with severity and reactions; dietary restrictions; intolerances; cultural and religious food requirements; medication–food interactions; preferred and disliked foods; preferred cuisines; meal timing, fluid and snack preferences.
• Recommendations — the assessor's written summary.

How it is scored: mixed — the MNA-SF total is computed by the system from its six scores and classified against the standard bands; IDDSI levels follow the international scale; everything else is recorded as findings and notes.

How it is used: governs which meals may be offered to the client, informs meal-preparation tasks and AI meal generation, and alerts care providers to swallowing risks. Its meal timing preferences are also a source signal the AI reads when drafting the care plan's routine anchors (PLAN-37, CA-17); the preferences stay narrative — structure is added at the care plan.

Lifecycle

Each of the six assessments has its own status on the client's record:

Completed is a deliberate sign-off, not a save button. Saving a form always lands it in Draft; a person must explicitly mark it Completed. The status matters because only Completed assessments are part of what the AI knows about the client when generating care plans, tasks, and meals — a Draft assessment is invisible to generation.

The one exception to "starts empty": when a client is created from a care proposal, the Simplified Routine Task Inventory arrives pre-filled from intake and already Completed, so the client's record begins with the intake knowledge in place.

Versions and generations (🚧 Spec only)

Statuses say where an assessment stands; versions say what it said. Every save of one of the six assessments keeps the prior content in the record's own history and increments its version number — the record stays singular (CA-1), so nothing about "one of each" changes; the history simply lives inside it (CA-15).

On top of versions sits the assessment generation: a client-level counter stamped across all six assessments when a care manager applies accepted reassessment suggestions (RAC-17). Every assessment — changed or not — bumps a version and takes the shared generation stamp, so a generation is one coherent, same-moment picture of the client (CA-16). Care plans record the generation that grounded them (PLAN-36), which is what makes "which picture of the client produced this plan?" answerable.

Where this fits in the care lifecycle

The assessments are the start of the loop, not the end of it. Completed assessments feed the AI that drafts the care plan; the plan drives the Care Task List; and each task's instruction steps are written from these same assessment scores. When the client changes, that loop runs in reverse — a reassessment proposes per-field changes to these assessments for the care manager to apply (RAC-16, RAC-17), and changed scores ripple back out to the plan and instructions. The whole loop is mapped on the Care Lifecycle page.

Because the Routine Task Inventory and Allen Cognitive Level are the two assessments that feed instruction generation directly, completing or changing either one is a recognized trigger to refresh the affected instructions — see CA-14.

Rules

Implementation status — audited against apps/backend/src/clients, packages/shared/src/types/simplified-routine-task-inventory.ts, and apps/backend/src/common/utils/is-simplified-routine-task-inventory-completed.ts on 2026-07-18. ✅ In code · ⚠️ Partial (built, but doesn't fully match the rule) · 🚧 Spec only (not yet built). The GCS Capacity / Support now / Risk model and the simplified-rti routes are shipped: CA-1 through CA-13 are ✅ In code; CA-14 through CA-17 remain 🚧 Spec only (tagged inline below).

Rules CA-1 through CA-6 were originally published as AS-12 through AS-17 on the Assessments page and moved here unchanged.

1. CA-1 — A client has at most one of each of the six care assessments. Saving never creates duplicates — it updates the single record, and each save is kept as a new version inside that record (CA-15).
2. CA-2 — Saving a care assessment form always sets its status to Draft, even if it was previously Completed. Marking it Completed is a separate, deliberate action. The one exception: when accepted reassessment suggestions are applied (RAC-17), the assessment keeps its status — the Apply action is itself the care manager's sign-off (CA-16).
3. CA-3 — Only Completed care assessments are included in the context used to generate care plans, care provider tasks, and meals.
4. CA-4 — When a client is created from a care proposal, the Simplified Routine Task Inventory is pre-filled from the proposal and starts as Completed.
5. CA-5 — Deleting a client deletes all six of their assessments; duplicating a client copies all six.
6. CA-6 — Care assessments can only be viewed and changed by the client's agency staff (owner, admin, care manager) — never by care providers, family, or users from another business.
7. CA-7 — Every care assessment shows its current status (Not Started, Draft, Completed) on the client's dashboard, so staff can see at a glance what has and hasn't been evaluated.
8. CA-8 — In the Simplified Routine Task Inventory, a functional (ADL / IADL) item counts as complete only when it has both a Capacity and a Support now recorded; equipment and home-safety items need a status, and the memory-care dementia trigger needs a Yes/No. The Risk flag, notes, and free-text observations are optional and never gate completion. Items still missing a required value must be visibly flagged, never silently skipped. (✅ In code)
9. CA-9 — In the Routine Task Inventory, a rating is always backed by the behavioral criteria the assessor checked — the system records the evidence and never computes or overrides the professional judgment.
10. CA-10 — An Allen Cognitive Level assessment records exactly one mode from the standardized scale, and the product must present each mode's level, meaning, and typical assistance needs to the assessor at selection time.
11. CA-11 — A Home Safety assessment must address every checkpoint in all twelve areas and end in a written recommendations summary. AI may draft the recommendations from the findings, but staff review and own the final text.
12. CA-12 — In the Meal Assessment, the MNA-SF total is computed by the system from its six scores and classified against the standard screening bands — never entered by hand. Food texture and drink thickness use IDDSI levels.
13. CA-13 — The Montessori Profile is narrative: it carries no scores, and no part of the product may reduce it to one. Its content personalizes engagement activities and meals.
14. CA-14 — Completing or changing the Routine Task Inventory or the Allen Cognitive Level — the two assessments that feed instruction generation directly — must flag the affected task instruction steps for regeneration (the receiving rule is STEP-5). The regeneration is routed to the care manager for review rather than applied automatically. (🚧 Spec only)
15. CA-15 — Care assessments are versioned in place: every save records the prior content in the record's own history and increments its version number. The record itself stays singular — versioning never creates a second assessment record (CA-1 still holds) — and the history is viewable by the same staff who may view the assessment (CA-6). (🚧 Spec only)
16. CA-16 — The client record carries an assessment generation counter. When a care manager applies accepted reassessment suggestions (the emitting rule is RAC-17), the platform writes the accepted changes and cuts a synchronized generation across all six assessments: every assessment — changed or unchanged — bumps its version and is stamped with the new generation number, so "generation N" always names one coherent, same-moment picture of the client. Assessments keep their status when this happens — a Completed assessment stays Completed, because Apply is the care manager's sign-off. This is the one documented exception to CA-2. (🚧 Spec only)
17. CA-17 — The Montessori Profile's Daily routine area (wake-up and bedtime routines, nap patterns) and the Meal Assessment's meal timing preferences are the source signals the AI reads when drafting the care plan's routine anchors (PLAN-37); the assessments themselves stay narrative — structure is added at the care plan, never by scoring the profile (CA-13 unchanged). (🚧 Spec only)

Who can do what

Decisions needed

1. (Resolved by CA-15/CA-16/RAC-17.) Should editing a Completed assessment revert it to Draft? Manual saves keep the revert — CA-2 stands, and re-completing remains the deliberate re-review. What the revert used to cost — silently losing what was on record — is now covered by the in-record version history (CA-15). The reassessment Apply path is the deliberate exception: it keeps the status because the care manager is signing off in that very action (CA-16).
2. Should completing a care assessment require the form to be fully filled in, the way completing an initial assessment does? Options: require completeness; or allow partial assessments to be marked Completed at staff discretion (today's behavior).
3. Three placeholder assessments appear on the client dashboard as "coming soon": Caring Touch (compassionate-touch needs), Always Fresh (dignity-focused continence support), and Care Bliss (end-of-life companionship and comfort). Options: define and build each as a real assessment in this catalog, or remove the placeholders. A fourth placeholder, Healing Ally, was removed from this catalog: wound-care collaboration is delivered by the wound care feature and reached from the client's health-monitoring tracking, not as a care assessment.
4. (Addressed by RAC-16/RAC-17.) Should assessments expire? Assessments still do not expire on a timer, but they no longer go stale silently: every reassessment re-reviews them field by field — the agent proposes per-field updates (RAC-16) and the care manager applies the accepted ones, cutting a fresh synchronized generation (CA-16). A per-assessment review interval with reminders remains a possible refinement.
5. How aggressive should the RTI / ACL regeneration trigger be? CA-14 now establishes that changing these two assessments flags the affected task instruction steps for regeneration, routed to the care manager. The remaining open question is the degree of automation: re-generate the draft automatically and queue it for review (current default), or only prompt the care manager to start it. Full auto-apply without review is excluded by the platform principle "AI drafts, humans decide."