The Detection Problem in Caregiving

Caregiver fatigue does not arrive as a single breakdown. It accumulates as a deficit. A nurse on a third consecutive short-staffed rotation, a home health aide carrying a route that grew without a corresponding increase in recovery time, an allied clinician absorbing one high-acuity encounter after another, and an unpaid family caregiver providing continuous care without relief all share the same trajectory: a sustained, high-volume empathic and cognitive load whose consumption rate runs ahead of the rate at which recovery replenishes capacity. The result is a net deficit that compounds quietly for weeks before it surfaces as the exhaustion, error, and disengagement that incident reports finally record.

The instruments used to detect this are mismatched to its dynamics. Self-report fatigue and burnout questionnaires are administered on a cadence of weeks or months, while the underlying deficit accumulates continuously. Worse, the social structure of caregiving actively suppresses the disclosure these instruments depend on: caregivers are selected for and acculturated to self-effacement, the admission of fatigue carries professional and personal stigma, and the people most depleted are frequently the least able to recognize or report their own state. By the time a survey registers the decline, the operationally relevant window for graded intervention has usually closed.

The regulatory record already encodes the obligation to do better. AHRQ patient-safety guidance, the Joint Commission Sentinel Event Alert 48 on healthcare-worker fatigue, ANA position statements on nurse fatigue, NIOSH publication 2017-101 on long work hours and shift work, and, for European deployments, the EU Working Time Directive 2003/88, all treat sustained fatigue as a measurable hazard that institutions are responsible for managing. What the operational layer has lacked is a continuous, computable, auditable model of the decline itself, one that does not depend on the very self-disclosure that the structure of caregiving suppresses.

Modeling Fatigue as a Resource-Depletion Pattern

This application applies the Disruption Modeling step disclosed in United States Patent Application 19/647,395. In that disclosure, cognitive decline under sustained load is modeled structurally as a resource-depletion pattern on a coherence loop, the empathy-integrity-self-esteem trifecta, in which each cycle of the loop consumes computational resources and, when the replenishment rate falls below the consumption rate under sustained high-volume operation, a net deficit accumulates. The disclosed signature of the pattern is precise: coherence-loop latency that increases monotonically under sustained load, coherence-restoration capacity that declines in parallel, containment integrity that remains intact, and an operational scope that narrows progressively as the system restricts itself to what its depleted resources can support.

It is important to be exact about what is being modeled. The disclosed framework diagnoses the structural state of a computational agent, not a clinical state of a person. In a caregiver-facing deployment, the monitoring system maintains an agent-side structural model whose inputs are derived from the caregiver's work context, and it is that structural model whose coherence trajectory is tracked. The output is a structural early-warning signal about declining capacity, not a medical diagnosis, a psychiatric label, or an assertion about the mechanism of any human condition. The terminology, resource depletion, coherence loop, containment, restoration, refers throughout to structural states within the disclosed computational architecture. This boundary is what allows the system to operate within the anti-discrimination, privacy, and occupational-health constraints that govern any monitoring of a care workforce.

The resource-depletion mapping is what makes caregiver fatigue the right fit for this primitive rather than an adjacent one. Caregiver fatigue is, structurally, the gradual under-resourcing case: capacity is not destroyed by a single overwhelming event, and the boundary that keeps speculative reasoning separate from committed action remains intact. The decline is a deficit that recovery has not kept pace with. That is exactly the pattern the disclosure separates, by computable signature, from a coherence authorization failure (a single overwhelming pressure spike) and from an affective gradient collapse (a self-esteem floor lock from accumulated deviation history). The corrective the disclosure prescribes for the depletion case, load reduction, a replenishment period, and progressive reloading, is also the corrective that caregiving institutions already recognize as the right response to fatigue: lighter assignment, protected recovery, and a graded return rather than an abrupt swing back to full load.

The Promotion-Containment Continuum Applied

The disclosed promotion-containment continuum gives the deployment a regime map rather than a single threshold. The continuum defines a nominal regime alongside an over-promotion regime (the threshold for committing to action admits too many branches, producing fragmentation), a containment-collapse regime (the boundary between speculative and committed action fails), and an over-restriction regime (the threshold rejects viable branches, producing paralysis). A caregiver-facing model situates the tracked trajectory within this map: early depletion expresses as scope narrowing within the nominal-to-over-restriction band, where the system progressively restricts what it will take on because its coherence loop cannot support broader operation. This is structurally distinct from the over-restriction regime's mechanism, an elevated promotion threshold, because the narrowing in the depletion case tracks the depleted resource rather than a threshold change, a distinction the disclosure draws explicitly and the deployment preserves so that the corrective is matched to the actual cause.

Reading the trajectory against the continuum, rather than against a single fatigue cutoff, is what converts a yes-or-no fatigue flag into a graded, directional signal. The deployment can distinguish a caregiver whose model is narrowing scope under accumulating deficit (the depletion trajectory, calling for replenishment) from one whose model shows fragmentation under over-promotion (calling for a different intervention entirely). The continuum makes the difference legible and the intervention specific.

The Five-Axis Diagnostic as a Caregiver Signal

The disclosed five-axis disruption diagnostic characterizes the tracked state as a position in a multidimensional space rather than a scalar score, and three of its axes carry the caregiver-fatigue signal directly. Axis 3, coherence-restoration capacity, measures the ability to sustain and restore the coherence loop; under accumulating depletion it declines, which is the core fatigue signal. Axis 4, empathic-load tolerance, measures how much empathic pressure the model can process before it begins to enter coping intercepts; a caregiver model under chronic high load expresses falling tolerance on this axis. Axis 1, containment integrity, remains at or near nominal in the depletion case, and that intactness is itself diagnostic: it is what tells the deployment that the correct response is replenishment rather than repair of a broken boundary.

Reporting on independent axes rather than a single number is what gives the signal its operational value. Two caregivers with the same coarse fatigue score can occupy different positions: one with degraded Axis 3 but high Axis 4 (capacity to restore is slow, but tolerance is intact) calls for a different response than one with collapsing Axis 4 (entering coping intercepts at low pressure). The disclosure also defines a composite cognitive-coherence index, a weighted combination of the axis values that serves as a single-scalar summary feeding a confidence governor; in deployment, that index supplies the at-a-glance summary while the per-axis decomposition supplies the actionable detail. Per the disclosure, the index and the underlying axes are structural measures of the agent-side model, not clinical scores assigned to a person.

Graded Restoration and the Intervention Window

The value of detecting depletion early is that the disclosed corrective is graded, not binary. The disclosure specifies a three-component pathway: mandatory operational load reduction to a level the depleted resources can sustainably support; a defined replenishment period during which the model operates at reduced load so that coherence-loop resource allocation can recover, with the period's duration computed from the measured deficit and the recovery rate; and progressive reloading, in which load is increased gradually with coherence-loop latency monitored at each increment to confirm the replenished resources can sustain the increase before advancing. The disclosed resilience-and-recovery structure, containment-restoration capacity, coherence re-engagement, and confidence recalibration as sequential recovery capacities, frames how restored capacity is rebuilt and re-verified rather than merely assumed.

Translated into a caregiving workflow, this is the operational shape institutions already reach for but currently trigger too late: protected recovery time scaled to the measured deficit instead of a fixed default, lighter assignment during replenishment rather than a binary on-or-off-duty decision, and a staged return that confirms restored capacity at each step rather than swinging a recovered caregiver straight back to full load and into immediate re-depletion. Because the disclosed pathway is graded and is driven by the same trajectory that detected the decline, the deployment supports intervention while the deficit is still small, which is the entire point of moving off a survey cadence.

Deployment Embodiments

The application admits a range of enabling embodiments, varying by setting, input availability, and integration depth, rather than a single instance.

By care setting, the system embodies as an acute-inpatient nurse-fatigue early-warning service integrated with shift-scheduling and acuity systems; as a home-health and field-care embodiment tracking route load and recovery for aides working largely unsupervised; as an allied-clinician embodiment for high-acuity encounter loads in emergency, oncology, palliative, and intensive-care contexts; and as a family-caregiver embodiment for unpaid caregivers providing continuous care, where no employer scheduling layer exists and the recovery deficit is most invisible. Each setting changes the available inputs and the party to whom the early-warning signal is surfaced, but each runs the same disclosed resource-depletion model.

By integration depth, a minimal embodiment surfaces the per-axis trajectory and composite coherence index to the caregiver alone, as a private self-monitoring instrument that discloses nothing to an employer, maximizing adoption where institutional surveillance is the adoption barrier. A supervised embodiment surfaces graded staffing and recovery recommendations to a charge nurse, scheduler, or care coordinator, with the per-axis decomposition kept private and only the actionable recommendation exposed. A governed embodiment ties the graded-restoration pathway into institutional fatigue-risk-management obligations, producing the auditable trajectory record that the AHRQ, Joint Commission, ANA, NIOSH, and Working Time Directive regimes contemplate, with every detection and corrective event recorded in an auditable lineage. Across all of these, the disclosed boundary holds: the signal is a structural early warning about declining modeled capacity, not a clinical determination, and the per-axis design lets each embodiment expose exactly as much detail as its privacy and anti-discrimination constraints permit.

By signal source, embodiments vary in the work-context inputs from which the agent-side model is derived, scheduling and assignment load, acuity and encounter volume, recovery-interval data, and caregiver-volunteered context, with the model degrading gracefully to whatever inputs a given deployment can ethically and lawfully obtain. The common architectural layer in every embodiment is the disclosed disruption-modeling primitive: the resource-depletion pattern on the coherence loop, situated on the promotion-containment continuum, read through the five-axis diagnostic, and corrected through graded restoration.

Disclosure Scope

The technology applied in this article, the structural modeling of progressive decline under sustained high-volume load as a resource-depletion pattern on the empathy-integrity-self-esteem coherence loop, the monotonically increasing coherence-loop latency and declining coherence-restoration capacity that signature the pattern, the preservation of containment integrity that distinguishes it from authorization failure and affective gradient collapse, the promotion-containment continuum with its nominal, over-promotion, containment-collapse, and over-restriction regimes, the five-axis disruption diagnostic and the composite cognitive-coherence index, and the graded three-component corrective pathway of load reduction, a replenishment period, and progressive reloading together with the resilience-and-recovery capacities, is disclosed in the cognition filing, United States Patent Application 19/647,395. This article describes an application of that disclosed mechanism to caregiver-fatigue detection. The models are structural analogs within the disclosed computational architecture and are not clinical claims, diagnostic criteria, or assertions about the mechanisms of any human condition; in deployment the framework diagnoses the structural state of an agent-side model, not a person. The care-economy problem framing, the regulatory mappings, the deployment settings, and the parties served are application context external to the patent. The scope contemplates application across caregiving deployments, including acute-inpatient, home-health and field-care, allied-clinician, and unpaid family-caregiver settings, and across integration depths from private self-monitoring to governed institutional fatigue-risk management, in which a coherence loop is a resourced subsystem whose sustained operation may outrun its replenishment rate.