The Attribution Problem in Dementia Care

A skilled nursing facility or memory-care unit must attribute clinical actions, disclosures, preferences, and access decisions to specific residents, and it must do so under a regulatory surface that assumes the resident participates in their own identification. HIPAA disclosure rules, the CMS Conditions of Participation, and OBRA-87 resident-rights provisions all anchor accountability to a resident who can be reliably identified and whose stated preferences can be attributed back to them. A resident with advanced dementia cannot reliably present a credential, cannot reliably remember a PIN, may remove a wristband, and may not recognize a badge as belonging to them. Conventional biometric enrollment fails for a second, subtler reason: it stores a template captured at admission and matches every later sample against it. Over months and years, the resident's gait slows, their voice changes with illness and medication, their facial geometry shifts with weight loss, and their behavioral interaction patterns degrade. The enrolled template drifts out of alignment with the living resident, forcing either steadily worse match quality or periodic re-enrollment that opens a gap in which the system cannot cryptographically vouch that the re-enrolling person is the same person who enrolled before.

This is precisely the failure mode the continuity-based architecture of United States Patent Application 19/647,395 is built to avoid. The platform's Biological Identity layer (Chapter 9 of that application) defines identity not as a stored template but as behavioral and biological continuity over time: an accumulated trust slope in which each new observation is validated as a plausible successor to the resident's prior trajectory. Drift is not an error to be suppressed; it is the expected signal. A system that measures deviation from a resident's own recent trajectory, rather than distance from a fixed enrollment template, accommodates aging and decline as a matter of course.

How Continuity-Based Identity Fits Elder Care

The disclosed Biological Identity pipeline acquires biological signals, extracts noise-tolerant features, produces a non-invertible stable sketch, generates a temporally bound, domain-scoped biological hash, and validates that hash for continuity against the resident's trust slope. The system never stores a raw biometric and never compares against a frozen reference image. For elder care, three properties of the disclosed architecture do the load-bearing work.

First, drift is accommodated structurally. The predictive identity module described in the application classifies observed change into stable, drifting, periodic, and volatile features, and constructs a forward acceptance envelope. The application's own deviation-classification model distinguishes environmental deviations from genuine physiological deviations, explicitly naming aging, illness, medication change, and fitness change as natural identity evolution rather than identity failure. In a dementia population, where decline is the baseline expectation, the system widens the acceptance envelope and, when the resident's signals have drifted past the point where the current sketch configuration stays stable, performs phase-based reseeding: the stable-sketch configuration is refreshed from current signals while a cryptographic cross-link preserves the identity chain across the refresh. There is no enrollment gap.

Second, resolution can be passive and consent-gated. The disclosed architecture supports contact-based, semi-contact, and non-contact acquisition tiers, and three resolution modes (one-to-one verification, one-to-many identification, and hybrid narrowing) selected under structural consent gating. A resident who cannot perform a deliberate identity assertion is not stranded: in an environment configured for passive observation, non-contact and semi-contact modalities (gait through floor or ambient sensors, voice, wrist-worn pulse and motion signals) sustain continuous background validation, with structured escalation to a higher-assurance tier only when continuity confidence falls below policy threshold. Consent gating is enforced as a structural constraint on which queries the resolution engine can run, not as an overridable software check, which matters in a setting where residents cannot themselves police how their identity is resolved.

Third, sparse and delayed validation are first-class. The application treats irregular, widely spaced observations as a supported operating mode rather than a degraded fallback, applying wider continuity thresholds calibrated to the longer inter-event interval and bounding the gap with policy-configured proof windows. A resident who is bed-bound for a week, hospitalized off-site, or simply not in range of a sensor does not lose their identity chain.

Deployment Embodiments

The disclosed technology admits several elder-care deployments, which a facility can mix according to acuity and privacy posture.

  • Ambient memory-care unit. Non-contact and semi-contact modalities provide continuous background validation across common areas and resident rooms, running in a privacy-preserving mode restricted to anomaly detection where policy requires it, escalating to one-to-one verification only at points where a clinical action or disclosure must be attributed.
  • Medication and treatment attribution. At the point of care, a contact-based or semi-contact high-assurance resolution event anchors the resident's trust slope, and the resulting validation drives the policy-governed authorization that gates the action, so that the administration record is bound to a continuously validated identity rather than to a wristband scan.
  • Wandering and elopement boundary. Continuous background validation detects a continuity anomaly when an observed signal is inconsistent with any authorized resident trust slope at a monitored egress, supporting an alert without resolving a specific bystander's identity in privacy-preserving zones.
  • Shift-handoff and visiting-clinician binding. The disclosed continuity attestation supports operational handoff verification, so that responsibility for a resident's care can transfer between staff with the handoff itself recorded against validated identities.
  • Distributed, resident-held deployment. For facilities or families prioritizing privacy, the disclosed distributed indexing embodiment keeps a resident's trust slope local to the resident or their device, with peer-to-peer resolution and no facility-held population index.

Each embodiment composes with the disclosed governance layer: resolution-authorization policy decides who may resolve a resident's identity, in which mode, under what conditions; audit records capture the request parameters, outcome, confidence, and policy justification without ever recording the raw signal, sketch, or hash; and a right-to-explanation mechanism lets a resident or their authorized representative obtain the basis for any resolution event and its downstream consequences.

Non-Diagnostic State Awareness

The application discloses biological state inference as a byproduct of continuity validation: deviation from the resident's own individualized continuity baseline can be classified into operational state categories such as elevated stress, fatigue, or impairment, and used to modulate policy-governed actions, for example flagging an interaction for additional staff review or adapting interaction modality and timing. This capability is explicitly and structurally non-diagnostic. As disclosed, the system does not diagnose medical conditions, does not measure blood alcohol content, does not assess mental health, and does not determine fitness for any activity; it reports deviation from the resident's own established norm, classified into categories defined by observable deviation patterns rather than medical conditions. In an elder-care setting this boundary is essential: the value is operational responsiveness (noticing that a resident's state has shifted from their personal baseline and adjusting care interaction accordingly), not clinical assessment, which remains the province of licensed clinicians. The article claims nothing beyond this disclosed, deviation-only, baseline-relative inference, and asserts no accuracy figures, detection rates, or clinical thresholds, none of which are part of the disclosure.

Continuity Across Crises: Quorum Recovery

Dementia residents are precisely the population most likely to suffer an abrupt break in their biological trajectory: a surgery, a stroke, a hospitalization, or an extended absence can shift their signals past the continuity tolerances and suspend the trust slope. The disclosed quorum-based identity recovery mechanism handles this without re-enrollment. Rather than discarding the accumulated identity history and starting a disconnected new template, recovery requires a policy-defined quorum of attesting peers, individuals whose own trust slopes carry a recorded association with the resident, each validating against their own slope and emitting a cryptographically signed forward continuity link. When the quorum is met, the resident's trust slope is re-established through a new root entry cryptographically linked to the prior slope, preserving the identity chain across the discontinuity. The disclosed anti-collusion safeguards (diversity requirements across relationship categories and minimum health characteristics for attesters' own slopes) translate naturally to a care setting where the attesting quorum can be drawn from family members, long-tenured caregivers, and a treating clinician.

Why This Is the Right Architectural Layer

Facilities today reconstruct attribution after the fact from charting, badge logs, and wristband scans, the same after-the-fact reconstruction that the regulatory surface was written to displace. A login session or a scanned band asserts identity at a moment and then assumes it indefinitely, which is exactly the assumption that fails when the credentialed person is a resident who cannot hold a credential and whose biology no longer matches an enrolled template. Continuity-based Biological Identity supplies the missing structural primitive: identity that is re-validated continuously rather than asserted once, that treats decline as expected drift rather than as authentication failure, that resolves passively under structural consent gating, that recovers across medical crises without breaking the chain, and that never stores a raw biometric. The regulatory chain of attribution that HIPAA, the CMS Conditions of Participation, and OBRA-87 presume is thereby preserved as a property of the system, for a population that can no longer perform the authentication those regimes assume.

Disclosure Scope

This article is a domain application of the continuity-based Biological Identity layer disclosed in United States Patent Application 19/647,395 (Chapter 9), comprising: identity as an accumulated trust slope validated by continuity rather than template matching; the signal-acquisition, feature-extraction, stable-sketch, and domain-separated biological-hash pipeline that stores no raw biometric; predictive acceptance envelopes, drift and deviation classification, and phase-based reseeding with cross-linked identity continuity; contact, semi-contact, and non-contact acquisition tiers with structured escalation; one-to-one, one-to-many, and hybrid resolution modes under structural consent gating; delayed and sparse validation with bounded proof windows; policy-governed authorization and capability binding; non-diagnostic biological state inference relative to an individualized continuity baseline; quorum-based identity recovery with anti-collusion safeguards; and the privacy, audit, revocation, and right-to-explanation governance framework. The regulatory framing, the elder-care deployment scenarios, the care-operations context, and the populations and parties described are application context and are not themselves part of the patent disclosure. This article asserts no accuracy figures, detection or false-match rates, sensor specifications, latency or interval numbers, or clinical thresholds, none of which are part of the disclosure, and it preserves the disclosure's non-diagnostic boundary and its no-raw-biometric-storage privacy posture.