Ongoing monitoring and support

Source baseline: CMS GUIDE Request for Applications, Appendix B, Section 4 ↗, with the linked duties in Sections 1, 2, 6, 7, and 8 where a monitoring contact identifies work in those service areas. The Proxi column identifies candidate automation, not functionality already deployed. The executed GUIDE Participation Agreement and current CMS instructions must be checked before field use if they differ from the public RFA.

Detailed ongoing-monitoring procedure#

This procedure expands all 37 Ongoing Monitoring and Support tasks, from navigator assignment through portfolio workload review. Tasks 14–35 are parts of the qualifying contact or follow-through from it; they are not 22 additional navigator contacts.

Source and authority basis#

The public baseline is the CMS GUIDE Request for Applications, Appendix B §§4.1–4.4 ↗ (SRC-CMS-GUIDE-RFA-V1). It makes the care navigator the primary point of contact, requires the care navigator to provide ongoing contact with the beneficiary and/or caregiver, names the purposes of that contact, establishes the original tier-based minimum frequencies, permits in-person, telephone, and audio-visual modalities, and bars SMS from satisfying the minimum contact frequency. It does not specify contact minutes.

The current CMS GUIDE Payment Methodology Paper v3.0 ↗ (SRC-CMS-GUIDE-PMP-3.0) adds the Residential Care Community (RCC) tier beginning July 1, 2026. It describes low-complexity dyad contact as at least quarterly and RCC contact as at least bimonthly, states that those touchpoint requirements are based on calendar months and reset after each touchpoint, and points to the Participation Agreement appendices for the detailed requirements. It also says Participants may not bill the DCMP for a month in which no GUIDE care-delivery service was provided and may not bill merely for maintaining infrastructure.

The participant's executed PY2026 GUIDE Participation Agreement is not present (O-PA-001). These cards therefore distinguish public CMS text, current project operating assumptions, and open participant-specific implementation questions.

Shared-interaction and labor convention#

All minute values are provisional Proxi workload-calibration assumptions, not CMS facts, service clocks, guaranteed staffing times, or field observations.

• Task 12 owns one scheduled qualifying human navigator interaction. For member scheduling, 23_GUIDE_Low_Burden_Monthly_Service_Plan.md proposes 8 / 15 / 25 direct-interaction minutes for stable / ordinary / extended nonclinical contacts, subject to O-039 field calibration. The prior 20- or 25-minute total active-navigator envelope is not a safe cap for a 25-minute direct interaction because preparation, documentation, and real follow-through are additional work; O-039 must replace or reconcile that placeholder before staffing approval.
• Task 11 preparation and Task 13's individual-tier adaptation do not create a second contact. Human preparation/documentation is counted once with the Task 12 episode but outside the direct-conversation band when it actually consumes time; components performed during the conversation allocate the same direct band rather than adding another interaction.
• Later Ongoing Monitoring tasks performed during the same conversation must allocate portions of the same direct-interaction band rather than adding another complete contact per row. Genuine follow-through performed afterward is separate active work and must be measured rather than hidden inside talk time.
• Tasks 1–10 are assignment, introduction, choice, route, cadence, calendar, consent, and reminder work. Most clean paths are software-first. Beneficiary, caregiver, interpreter, payer, outside-provider, and other external-party time is excluded unless Proxi pays that person.
• L1 Philippines time is 0 on the clean path and never satisfies the required GUIDE navigator contact under D-005. When an actual manual administrative problem occurs, the episode may use one shared 13-minute L1 exception event from the existing workload model; it is not added once per task.
• Every clinical or medical minute belongs to the separate U.S. L3/L4 clinical workforce outside the Puerto Rico navigator operation, consistent with D-015. Clinical minutes are event-driven and remain in the separate 74-minute clinical-support and 106-minute higher-clinician annual pools. No clinical review is added merely because the beneficiary has dementia or lacks a caregiver.

Values shown as three numbers are low / typical / high provisional human minutes for the named event. Included component minutes allocate attention inside Task 12; they are not automatically additive and do not imply every component occurs in every contact. An extended nonclinical contact may redistribute the 25-minute direct band, continue as a documented navigator follow-up, or open separately counted clinical/service work rather than silently multiplying every row. If Task 1 navigator acceptance, Tasks 2–5 setup, Task 11 review, or Tasks 14–35 work is performed in the same episode, shared work is counted once; actual preparation, documentation, and post-contact follow-through remain visible rather than being forced into the direct-interaction band.

Cross-pillar preparation, no-repeat transfers, receiving-owner acceptance, and shared-event labor follow 25_GUIDE_Eight_Pillar_Service_Integration.md. This SOP still applies its own contact, result, and completion rules.

Member contact budget and monthly service selection#

One primary navigator coordinates person-facing work. Before a planned contact, the team must be able to state why conversation is appropriate now, which service will be furnished, what useful result the beneficiary or caregiver should receive, and why their participation is needed. Cadence determines when a required contact is due; person value determines its agenda and endpoint; billing classification occurs only after the actual service. “Monthly check-in,” “touch base,” “see how they are doing,” and an empty billing month are not service purposes.

A required stable contact is a brief person-led ongoing-monitoring and support service. It gives the beneficiary or caregiver a genuine opportunity to raise, choose, or decline topics and completes at least one person-specific result: an answered question, an advanced due action, selected support, or attributable confirmation that a relevant current goal/action still fits and no additional help is wanted. It closes with a usable next step or help route. Duration or checklist breadth does not prove value.

Do not force every detailed screen into every contact. Begin with an open invitation for a change, concern, or chosen priority, then address the person-selected or actually due focus. Ask plan-fit, medication-use/access, open-action, caregiver, and no-caregiver questions only when relevant to the selected service, current evidence, an open obligation, or the person's concern. A domain not covered remains not asked, not “no change,” and receives no service credit.

The Proxi Daily Companion may offer a configured accessible beneficiary interaction, capture source-attributed self-report, support current actions, receive a caregiver pulse, and prepare a routed human-follow-up request between scheduled contacts. It is not continuous monitoring, emergency response, clinical review, synchronous human availability, or guaranteed immediate help. Its result is input to navigator preparation or affected-pillar work, not a qualifying GUIDE human contact. A caregiver pulse remains caregiver-owned and may trigger routed review, but never becomes a beneficiary fact, burden diagnosis or score, urgency decision, completed assessment or support call, completed service, or payment evidence. Silence and stale or conflicting information remain unknown and never provide reassurance, including in Caregiver View. D-023 settles these exclusions; O-061 governs activation mechanics only.

Use a substantive inbound navigator interaction in place of a redundant planned contact when it independently meets Task 31. Extra nonclinical contacts require beneficiary/caregiver agreement; extra clinical frequency requires an authorized clinical basis. In low-dyad and RCC off-contact months, perform a real non-contact service only when patient-specific work is actually due. An empty billing month is not a reason for outreach.

The direct conversation may be approximately 8 minutes for a stable brief service, 15 minutes for an ordinary one-need service, and 25 minutes for extended nonclinical, multi-action, or accessibility-supported work. A second moderate/high individual-tier contact may be 5 / 10 / 15 minutes and should focus on one active action, barrier, skill, or goal rather than repeating the full screen. “Complex” never expands L2 authority: any clinical or safety issue routes to the separate beneficiary-location-authorized U.S. clinical or emergency pathway and is counted separately. CMS specifies no minimum minutes. These proposed ranges are scheduling assumptions, not service clocks or staffing guarantees, and remain open under O-039.

Total active episode labor around the direct bands#

The 8 / 15 / 25-minute figures are direct-conversation capacity bands. A completed contact episode also contains preparation, documentation, and—only when the episode creates promises—follow-through oversight. These are real paid Puerto Rico L2 minutes outside the conversation. They are counted once with the episode and are never compressed into the direct band.

The low / typical / high values are provisional workload assumptions under O-039, not service clocks. The topic screens in Tasks 14–29 divide the direct conversation; their minutes are never added again. A conversation that creates no promise has zero follow-through. Failed-contact recovery is a separate event. L1, interpreter, separate U.S. clinical, same-lane lead, and workforce-exception minutes stay in their own budgets. Substantive downstream work belongs to the SOP and role that actually performs it.

Portfolio calibration must weight the current episode bands by observed case mix and include actual preparation, documentation, follow-through, failed attempts, inbound substitution, focused second contacts, and accessibility work. No portfolio average may cap an individual episode.

1. Assign the primary care navigator#

When. Complete after alignment and before the first planned ongoing contact. Reopen when the navigator leaves, becomes unavailable, loses a required credential or approved organizational relationship, exceeds approved capacity, no longer meets service-area/language needs, or is replaced at the beneficiary's request or by the program.

What and how much. Establish exactly one accountable primary care navigator with an effective date, normal contact route, backup/coverage route, current workload capacity, and current required credential. A backup may cover absence, but it must be clear who is accountable at every point; two names without responsibility are not an assignment.

Data.

• An authoritative alignment record proves current GUIDE alignment and effective timing, not navigator assignment.
• The approved care-team roster, current credential evidence, service area, language capability, availability, workload capacity, employment/contract relationship, and any required Partner status establish assignment eligibility only to the extent each source records.
• A beneficiary language or communication preference directly supports practical matching. It does not assign the worker.
• An AI match score, prior relationship, calendar availability, or name in a directory is a candidate input, not authority to assign.

Potential Proxi work. Filter the approved roster by mandatory eligibility, credential, organization, service-area, language, availability, and caseload facts; explain the candidate match; detect conflicts or expired evidence; prepare and issue the assignment under the approved rule; and maintain a backup candidate. AI may rank eligible candidates but cannot create role accountability.

Human role. The Puerto Rico navigator lead enters only when the approved matching rule cannot produce an accepted navigator or a documented relationship conflict requires reassignment. HR/Workforce Administration corrects employment, credential-file, leave, or capacity facts outside the beneficiary episode. The L2 navigator accepts responsibility. Clinical staff are not needed.

Provisional clean-path Proxi human minutes by role. Puerto Rico navigator lead: 0 minutes when software produces an accepted clean-path assignment; 1–3 minutes only for a failed match or reassignment exception. L1: 0. Assigned L2 navigator: 1 minute to accept accountability. HR/Workforce: outside beneficiary labor and measured separately. L3/L4: 0. These are one-time/event minutes, not minutes per contact.

Completion evidence. Beneficiary identity; navigator identity; current credential and approved organization/roster basis; effective date; acceptance; primary and backup routes; workload/capacity check; and any replacement linkage.

What does not prove completion. Algorithm output, staff directory listing, calendar availability, a proposed assignee, sent assignment message, prior navigator history, or a credential without acceptance and current organizational authority.

Edge cases/open decisions. Existing: X-003, X-004, X-005, X-014; O-001, O-039, O-040, and O-PA-001. Proposed OPEN decision: primary/backup navigator continuity and temporary coverage standard. Minimum tests: clean assignment; expired credential; no eligible navigator; two simultaneous assignees; capacity full; navigator leaves before first contact; Partner/organization status unresolved.

2. Tell the beneficiary and caregiver who their navigator is#

When. Perform immediately after Task 1 and after every reassignment or change in normal contact route. Deliver to the beneficiary and only to a caregiver or representative whose involvement and communication basis permit it.

What and how much. Provide one accurate introduction containing the navigator's name, care-navigator role, normal contact route and hours, backup/24/7 route, what the navigator can and cannot do, language/accessibility options, and how to request a change or ask a question. The introduction notice can be automated; the substantive human relationship is established through Task 12 or another completed human interaction.

Data.

• The accepted Task 1 assignment is direct evidence of whom the program assigned.
• Current communication preference, accessible format, safe-contact restrictions, and caregiver/representative involvement determine the recipient and route.
• A sent message proves transmission was attempted. Delivery evidence proves only the channel outcome; it does not prove understanding or a usable relationship.
• A caregiver contact record or portal proxy does not by itself establish permission to disclose GUIDE participation or communicate through that person.

Potential Proxi work. Generate an accessible introduction from the exact assignment; deliver it by an approved channel; adapt format/language from confirmed preferences; track delivery and bounce/failure; suppress revoked recipients; and offer a direct question or rescheduling route. Proxi cannot invent the navigator's scope or treat delivery as a qualifying contact.

Human role. No Proxi human is required merely to send a clean introduction. Software applies the approved recipient, purpose, content, channel, permission, expiry, revocation, and suppression rule. L2 answers questions and makes the relationship real during Task 12 or an earlier human interaction. L1 may address a failed administrative destination through the single shared exception. Missing caregiver relationship or permission facts return to L2; only a complete verified fact pattern exposing a genuine legal-authority question not answered by the approved rule routes to Healthcare Legal Counsel while the affected delivery remains held.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0 incremental; any human introduction inside the first Task 12 contact allocates that contact's direct-interaction band. L3/L4: 0.

Completion evidence. Exact assignment; recipient and permission basis; channel, language and accessible format; delivery result; failure/fallback disposition; and the current normal and backup contact routes.

What does not prove completion. Message creation, send event alone, no electronic bounce, a mailed letter without destination evidence, caregiver receipt without permission, or later contact with a different navigator.

Edge cases/open decisions. Existing: X-002, X-005, X-007, X-013; O-001, O-011, and O-PA-001. Minimum tests: beneficiary-only delivery; permitted caregiver delivery; wrong caregiver blocked; inaccessible format; bounce/fallback; reassignment after old introduction; shared telephone number; no electronic channel.

3. Confirm who will participate in ongoing contacts#

When. Complete during setup and recheck when the beneficiary, caregiver, representative authority, privacy basis, communication ability, or practical availability changes. Do not infer a permanent participant arrangement merely because one person answered the last contact.

What and how much. Record whether future contacts may include the beneficiary, caregiver, or both; which topics each person may discuss or receive; the beneficiary's preferred participation method; the caregiver's willingness/availability; and any separate representative authority. The RFA permits ongoing contact with the beneficiary and/or caregiver, but caregiver presence never automatically removes the beneficiary from their own care relationship.

Data.

• An authenticated beneficiary choice directly proves the beneficiary's requested participation arrangement.
• A caregiver statement proves that caregiver's willingness and availability, not beneficiary permission or representative authority.
• A representation/permission record applies only to its documented scope, dates, conditions, and decision.
• Prior attendance, household relationship, emergency-contact status, or portal proxy access identifies a possible participant but does not settle current choice or authority.
• AI-extracted conversational preference remains a candidate until authenticated.

Potential Proxi work. Present the current participant arrangement; collect an authenticated choice through accessible text/voice/portal; read it back; preserve topic/recipient limits; compare it with existing authority and privacy records; retain differing views; and block only affected communication when facts conflict.

Human role. The beneficiary or decision-specific authorized representative supplies the choice; the caregiver supplies willingness. No separate Proxi human is required on the clean path. L2 confirms ambiguity or relationship-sensitive choices within Task 12. Software applies the approved identity, representative-authority, recipient, purpose, content, permission, and restriction rule after the required facts are present. Missing relationship or permission facts return to L2; a clinical capacity question routes to the authorized clinician; only a genuine legal-authority question not answered by the approved rule routes to Healthcare Legal Counsel. L2 and AI do not decide either protected question.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0 incremental; clean confirmation inside Task 12 is included in that shared allowance. L3/L4: 0 absent a capacity or clinical trigger.

Completion evidence. Authenticated chooser; current participant set; topic and disclosure scope; caregiver willingness; representative-authority status when applicable; dates; contact preferences; and unresolved conflicts kept visible.

What does not prove completion. Caregiver listed in the chart, last-call attendance, emergency-contact status, portal access, silence, an AI summary, or caregiver availability treated as authority.

Edge cases/open decisions. Existing: X-002; CP-E04, CP-E05, and CP-E06; O-011. Minimum tests: beneficiary chooses self only; beneficiary chooses both; beneficiary requests caregiver-only contact; caregiver declines; caregiver and beneficiary disagree; authority expires; choice changes after a contact.

4. Follow the no-caregiver route when applicable#

When. Apply after caregiver status is affirmatively confirmed as no caregiver available for the relevant period. Do not apply merely because the caregiver field is blank, a caregiver did not answer, or no caregiver attended one interaction. Reevaluate when a caregiver is later identified or lost.

What and how much. Continue service directly with the beneficiary without waiting for a caregiver. Use the current CMS-assigned route: low-complexity individual at least monthly, moderate/high-complexity individual at least twice monthly, or the RCC cadence when CMS assigns the beneficiary to the RCC tier. Identify practical communication/community-living support gaps and invoke only the additional safeguards approved under O-013.

Data.

• An authenticated beneficiary statement and current caregiver assessment can directly establish reported caregiver status for the relevant date.
• Absence of a caregiver record, unanswered outreach, living alone, or facility residence does not prove “no caregiver.”
• The current CMS tier/effective date controls the minimum route. Caregiver absence must not reclassify a CMS-assigned RCC beneficiary into an individual tier.
• No-caregiver status proves neither incapacity nor elevated clinical risk.

Potential Proxi work. Detect confirmed no-caregiver status; remove caregiver-dependent assumptions; apply the current tier cadence; address the beneficiary directly; present accessible self-support options; surface missing supports; and route a specific clinical, safety, capacity, or safeguarding concern. Proxi cannot infer incapacity or invent a proxy.

Human role. L2 provides the actual direct human contact through Tasks 12/13 and works with the beneficiary on nonclinical supports. L3/L4 enters only for an actual clinical, safety, behavioral, safeguarding, or decision-specific capacity issue. L1 may pursue missing administrative facts but does not perform the navigator contact.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0 incremental for route setup; direct human time is entirely within Task 12 and Task 13 adds no second allowance. L3/L4: event-driven only.

Completion evidence. Confirmed caregiver status and source/date; current tier and effective date; direct beneficiary route; correct minimum cadence; identified support gaps; approved safeguard status; and no service delay while searching for a caregiver.

What does not prove completion. Blank caregiver field, living alone, missed caregiver call, individual-tier assumption from age or diagnosis, facility staff availability, invented proxy, or an automated check-in.

Edge cases/open decisions. Existing: X-003, X-004, X-005; CP-E07; MED-E17; O-013 and O-PA-001. Minimum tests: capable beneficiary with no caregiver; caregiver status unknown; later caregiver identified; caregiver lost mid-cycle; RCC beneficiary without caregiver; communication barrier; new safety signal.

5. Confirm communication preferences and accommodations#

When. Complete before scheduling the first contact and recheck when language, hearing, vision, literacy, cognition-related communication need, device/access, safe contact time, channel, address, phone number, or beneficiary/caregiver preference changes.

What and how much. Maintain one current communication profile for each permitted participant: preferred and fallback permitted contact modality, language, interpreter need, accessibility/accommodation, usable device/connectivity, safe times, time zone, voicemail/message restrictions, and rescheduling route. SMS is supplemental and remains separately controlled by Task 9.

Data.

• An authenticated beneficiary/caregiver preference directly establishes that person's stated preference and reported ability.
• A qualified accessibility/language assessment can establish the documented accommodation need.
• Prior successful use corroborates practical usability but does not override a new preference or prove continued access.
• Portal enrollment, device ownership, stored language, demographic field, or a completed prior call does not prove the current modality is accessible or preferred.

Potential Proxi work. Collect and read back preferences; offer only currently supported modalities; match interpreters and accessibility resources; test technical connection without counting the test as contact; apply time zone and safe-contact restrictions; and flag unavailable accommodations. AI may translate approved logistics but cannot resolve clinically material misunderstanding.

Human role. No Proxi human is required for clean self-service collection. L2 resolves preference ambiguity or relationship-sensitive restrictions during Task 12. A qualified interpreter/accessibility worker supplies the actual accommodation when needed. Clinical/capacity concerns route separately.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0 incremental. Interpreter/accessibility and L3/L4 minutes: 0 unless triggered and Proxi-paid.

Completion evidence. Current profile for each participant; source/date; preferred and fallback modalities; language and actual accommodation arrangement; time zone/safe times; restrictions; technical barriers; and unresolved accommodation gaps.

What does not prove completion. Old preference, successful prior contact, portal account, device record, payer demographics, an offered interpreter not booked, automated translation output, or lack of complaint.

Edge cases/open decisions. Existing: X-002, X-005, X-007; CC-E39; O-011. Minimum tests: interpreter required; hearing/vision need; no broadband; shared phone; unsafe voicemail; preference changes after booking; repeated failed modality; beneficiary and caregiver prefer different channels.

6. Identify the required minimum contact frequency#

When. Determine before building the initial calendar, after every CMS tier update or corrected caregiver/residence fact, and before deciding that the current calendar remains sufficient. Check the current CMS Beneficiary Alignment Report when refreshed and preserve the effective date of a change.

What and how much. Map the current authoritative tier to one minimum contact cadence:

The current Payment Methodology Paper says the low-dyad and RCC touchpoint requirements are based on calendar months and reset after each touchpoint. Exact partial-month, transition, and Participation-Agreement implementation remains open; Proxi must not invent a less frequent interpretation.

Data.

• The current CMS Beneficiary Alignment Report and effective date directly establish the CMS-reported current tier; the Payment Methodology Paper says the report is updated twice monthly.
• A comprehensive assessment, CDR/FAST/ZBI score, residence report, or caregiver status can identify a candidate change or conflict. It does not by itself replace the CMS-assigned tier used for current payment/contact administration.
• RCC residence and caregiver status must remain distinct; beginning July 1, 2026, the current public methodology assigns eligible RCC residents to the RCC tier regardless of caregiver status or dementia stage.
• An AI tier prediction, old report, billing code, or prior cadence is not authoritative.

Potential Proxi work. Retrieve the current tier report; compare effective dates; detect stale/conflicting facts; map the tier to the public minimum; calculate upcoming due windows under the approved clock convention; and alert when a changed tier makes the current calendar insufficient. Proxi does not infer or change the tier.

Human role. No human is required when the authoritative tier/effective date is current and uncontested. Program/operations resolves a missing or disputed report. L2 confirms caregiver/lived circumstances only when those facts are genuinely unresolved; L2 does not assign the CMS tier.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0. Program/operations: 0. L3/L4: 0.

Completion evidence. Current tier, source/report date, effective date, applicable public cadence, approved clock convention reference, last qualifying touchpoint, next due window, and any unresolved conflict.

What does not prove completion. Assessment severity, caregiver field alone, residence alone without current tier disposition, old Beneficiary Alignment Report, current billing code alone, model prediction, or prior calendar.

Edge cases/open decisions. Existing: X-003, X-004, X-005, X-012; O-013, O-024, and O-PA-001. Proposed OPEN decision: contact-clock implementation covering calendar months, reset behavior, partial months, tier effective dates, RCC transitions, and late/corrected reports. Minimum tests: all six tiers; RCC without caregiver; report changes mid-cycle; stale report; assessment/report conflict; missing report; tier correction arrives after a booked contact.

7. Build the contact calendar#

When. Perform after Task 6 and maintain enough future coverage to avoid a preventable cadence gap. Rebuild when the current schedule does not cover the next required window, a contact is missed, or an approved higher-than-minimum cadence is added.

What and how much. Create the minimum number of booked qualifying-contact opportunities needed under the approved clock convention, with a permitted modality, correct time zone, current communication preferences/accommodations, navigator coverage, and enough recovery time for a failed contact. Schedule more frequent contact only after L2 and the beneficiary agree to added nonclinical relationship support or an authorized clinician supplies the clinical need.

Data.

• Task 6 supplies the minimum cadence; it does not prove any appointment is booked or completed.
• Authenticated availability/preferences establish scheduling constraints, not contact completion.
• A current staff calendar establishes available slots; it does not prove the assigned navigator will remain eligible or that the beneficiary will attend.
• Clinical risk or need-based higher cadence requires an authorized human determination; AI pattern detection is only a review candidate.

Potential Proxi work. Generate compliant candidate dates; apply time zone, modality, interpreter and staff constraints; present choices; book the selected times; detect coverage gaps; reserve fallback opportunities; and alert before the due window is endangered. Deterministic scheduling may execute an already-authorized plan; AI cannot decide higher clinical intensity.

Human role. No Proxi human is required for a clean connected-calendar path. The beneficiary/caregiver selects among usable times. L2 resolves relationship-sensitive conflicts or agrees to additional nonclinical contact. An authorized clinician decides clinically driven added frequency. Program staffing resolves navigator capacity gaps.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0. Program/staffing: 0. L3/L4: 0. A manual scheduling failure may draw once from the shared 13-minute L1 exception.

Completion evidence. Booked contact opportunities covering the current due windows; assigned navigator; modality; time zone; accommodations; beneficiary/caregiver choice; fallback/recovery route; and separation between minimum and approved additional contact.

What does not prove completion. A recurrence rule, tentative hold, invitation not accepted, reminder, scheduling call, automated check-in, staff availability alone, or a calendar that uses a stale tier.

Edge cases/open decisions. Existing: X-003 through X-007, X-013, and X-014; O-004 and the proposed contact-clock decision. Minimum tests: monthly/quarterly/twice-monthly/RCC windows; month-end contact; DST/time-zone change; interpreter availability; no navigator slots; beneficiary rejects proposed times; added clinical cadence with authorized source; AI-generated extra cadence blocked.

8. Rework the calendar when circumstances change#

When. Trigger when tier, effective date, caregiver route, RCC residence, modality, language/accommodation, time zone, contact success, navigator assignment, clinical/nonclinical support cadence, or beneficiary availability changes.

What and how much. Recalculate only future obligations affected by the change, preserve completed-contact evidence, cancel or replace obsolete bookings, and create enough new opportunities to meet the current route. Do not erase a missed contact, retroactively count an automated interaction, or move a completed contact into a different period.

Data.

• An authoritative tier/effective-date update can drive cadence recalculation directly.
• An authenticated beneficiary/caregiver preference or availability change can drive scheduling changes directly.
• A reported residence/caregiver change may be a candidate for route/tier review and must not silently change the CMS tier.
• A completed qualifying-contact record remains historical evidence; a calendar edit cannot change what occurred.

Potential Proxi work. Detect changes; identify affected future bookings; compare old/new due windows; preserve history; propose replacements; apply current preferences; notify permitted participants; and alert on any residual gap. Software can act on authoritative uncontested facts; AI cannot decide tier, clinical frequency, or whether a failed interaction qualifies.

Human role. No Proxi human is required for a clean authoritative update and accepted replacement. L2 resolves relationship-sensitive scheduling or a request for added nonclinical support. Program staff handles tier/report or staffing disputes. Clinical staff act only when the change contains a clinical decision.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0. Program: 0. L3/L4: 0. Manual outreach/scheduling is one shared L1 exception event, not one event per changed booking.

Completion evidence. Change source/effective date; old and new route; preserved completed contacts; cancelled/replaced bookings; current due-window coverage; recipient notification outcome; and any unresolved gap/escalation.

What does not prove completion. Detected change alone, recalculated draft, cancelled appointment without replacement, beneficiary/caregiver report treated as CMS tier assignment, or notification send without calendar coverage.

Edge cases/open decisions. Existing: X-004, X-005, X-007, X-011, X-012, X-013; O-013, O-024, and the proposed contact-clock decision. Minimum tests: tier increases/decreases; caregiver lost/added; move to RCC; modality becomes inaccessible; navigator changes; contact missed after schedule update; conflicting simultaneous updates; late effective date.

9. Obtain consent before using SMS for supplemental communication#

When. Obtain before the first SMS to each intended recipient and re-evaluate after number ownership, recipient, purpose, scope, or consent changes. Stop dependent SMS immediately on withdrawal or a credible wrong/shared-number signal. SMS remains optional and supplemental.

What and how much. Record affirmative, recipient-specific SMS consent with the exact number, permitted recipient, purposes/content class, date, capture method, applicable beneficiary/caregiver participation basis, and withdrawal method. Never schedule or classify SMS as the contact that satisfies the minimum frequency.

Data.

• An authenticated affirmative response directly proves consent only for the person, number, purpose, and scope captured.
• A stored mobile number, prior SMS, portal setting, message reply, or lack of opt-out does not independently prove current consent.
• Beneficiary consent does not automatically authorize caregiver SMS, and caregiver consent does not establish authority to receive beneficiary information.
• Delivery to a telephone number proves channel delivery at most; it does not prove the intended person received or understood it.

Potential Proxi work. Present an accessible consent explanation; capture affirmative response; separate recipients/scopes; prevent SMS without consent; log withdrawal; suppress shared/wrong numbers; retain non-SMS contact routes; and label every SMS interaction as nonqualifying for cadence. AI may explain approved logistics but cannot infer consent.

Human role. No human staff is required for a clean, authenticated consent capture. The beneficiary or caregiver supplies the choice within that person's own participation and authority. L2 answers relationship-sensitive or comprehension questions. L1 may correct a number administratively but cannot infer consent. Software applies the approved identity, recipient, purpose, scope, permission, withdrawal, and suppression rule; only a complete verified fact pattern exposing a genuine legal-authority question not answered by that rule routes to Healthcare Legal Counsel while affected SMS remains blocked.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0. L3/L4: 0 unless a clinical exception is triggered. Healthcare Legal Counsel time is event-driven only for a genuine uncovered legal-authority question, not for routine consent or disclosure-rule application.

Completion evidence. Recipient, identity/authentication, exact number, scope/purpose, affirmative response, date/method, disclosure/participation basis where applicable, withdrawal route, current status, and enforced SMS suppression when not consented/revoked.

What does not prove completion. Phone number on file, prior texts, successful delivery, reply to an unconsented message, silence, caregiver relationship, generic terms acceptance, or consent that does not identify recipient and scope.

Edge cases/open decisions. Existing: X-002, X-007; CP-E14; O-001, O-011, and O-PA-001. Proposed OPEN decision: recipient-specific SMS consent, permitted content, authentication, renewal, shared-number and revocation evidence. Minimum tests: beneficiary-only consent; caregiver-only consent; both consent separately; wrong/shared number; revoked consent; changed number; inaccessible consent flow; SMS attempts never counted as contact.

10. Remind people about the upcoming contact#

When. Send according to the approved reminder schedule before a booked contact and after a material change requiring a corrected reminder. Do not send through a channel lacking the required permission, and do not continue a superseded date/time.

What and how much. Send the current date, time, time zone, modality/location, connection instructions, preparation request, accessibility arrangement, and direct rescheduling route. Use the approved number/timing of reminders; this semantic card does not invent a universal count or interval.

Data.

• The current booked appointment and communication profile establish the reminder payload and permitted route.
• Delivery evidence proves only the channel outcome. It does not prove the person read, understood, accepted, attended, or completed the contact.
• A response requesting a change directly supports that request; it does not itself create a new booked time until scheduling succeeds.

Potential Proxi work. Generate the exact current reminder; adapt approved language/format; deliver through consented channels; detect bounce/failure; suppress obsolete reminders; accept rescheduling requests; and route questions/accommodation problems. Automated reminders are infrastructure/support, never the qualifying contact.

Human role. No Proxi human is required on the clean path. L1 may resolve a failed destination or manual scheduling issue using the shared exception. L2 handles relationship-sensitive or substantive questions, not routine reminder delivery.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0. L3/L4: 0.

Completion evidence. Current appointment identity; payload; recipient/channel permission; send and delivery/failure result; obsolete-message suppression; rescheduling route; and resolution of any delivery/accommodation exception.

What does not prove completion. Reminder send, delivery, open/read receipt, appointment acceptance, calendar entry, automated reply, or a rescheduling conversation.

Edge cases/open decisions. Existing: X-005, X-007, X-013; O-004. Minimum tests: clean reminder; SMS without consent blocked; date changed after reminder; wrong time zone; bounce/fallback; interpreter details missing; shared number; reminder delivered but contact missed.

11. Prepare the navigator for the contact#

Longitudinal baseline rule. The brief must consume the latest accepted Assessment baseline, its explicit unknowns and watch items, its next Assessment horizon, and its open return conditions. Do not rebuild the Assessment packet or omit an Assessment unknown merely because it is inconvenient for the contact.

When. Prepare immediately before every scheduled qualifying contact and refresh if a material late event arrives before the conversation begins.

What and how much. Produce one concise, source-linked contact brief containing: current care-plan goals/open actions; prior-contact commitments; unresolved beneficiary/caregiver concerns; recent transitions; referrals/services; reported medication-use concerns; caregiver/no-caregiver status; current communication needs; and missing, stale, or conflicting information. Keep a short priority view while retaining every unresolved item and direct source link; do not drop lower-salience data. Consult the standing preference-and-decline record in 23_GUIDE_Low_Burden_Monthly_Service_Plan.md, including prior declines, topics the person asked not to repeat, and preferred length and channel.

Data.

• The source record proves only its own documented fact, author, time, and scope.
• A prior note proves what was recorded then; it does not prove the fact remains current.
• Claims, alerts, external messages, and AI extraction can surface candidates or changes; they do not establish current symptoms, urgency, treatment, or beneficiary choice.
• An AI summary is presentation, not a canonical fact, clinical judgment, or proof the navigator reviewed it.

Potential Proxi work. Retrieve the source-native records; deduplicate shared origins without discarding provenance; compare with the prior contact and care plan; identify late/stale/conflicting facts; assemble the brief; show original sources; and update it before contact start. Proxi may draft prompts but cannot choose clinical urgency or omit an unresolved issue because a model ranks it low.

Human role. L2 reviews the brief as part of Task 12 and uses the contact to clarify changed, ambiguous, preference-sensitive, or participant-reported facts. The navigator does not need to re-ask every already authoritative fact. Clinical interpretation stays with L3/L4/L5.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: count actual pre-contact review once with the Task 12 episode; it does not extend the direct-conversation target or create a second contact. L3/L4: 0 unless the brief reveals a separate clinical trigger.

Completion evidence. Correct beneficiary/contact; source-linked current brief; source/as-of dates; all open commitments; missing/stale/conflicting items; late-event refresh; and navigator access at contact start.

What does not prove completion. AI summary alone, generic risk score, full-chart dump, copied old note, unread brief, omitted low-ranked concern, source link without content, or a summary that converts unknown into no issue.

Edge cases/open decisions. Existing: X-001, X-003, X-004, X-005, X-009, X-011, and X-013. Minimum tests: correct clean brief; wrong-patient data blocked; stale plan; conflicting medication reports preserved; missing referral result; duplicate-source notes; urgent late message; system outage; unread brief.

12. Conduct a qualifying ongoing contact#

When. Conduct within the current required window and sooner when an authorized nonclinical or clinical decision establishes an earlier need. If an attempt fails or the person cannot participate substantively, reschedule through a permitted accessible modality; do not count the attempt.

What and how much. The assigned or properly covering human care navigator has one completed substantive interaction with the beneficiary, caregiver, or both by in-person, telephone, or audio-visual modality. State the specific purpose, invite the person's priority, and complete at least one person-specific ongoing-support result. Depending on what is relevant that day, maintain/revise an affected care-plan item, identify an unmet need, coordinate a clinical/community support, collect source-attributed reports about medication use, access, schedule execution, and problems without reconciling, or provide/route caregiver support. Allow unprompted concerns, but do not imply that unasked domains were screened. The direct-interaction planning range is 8 / 15 / 25 minutes for stable brief / ordinary one-need / extended nonclinical contacts. This is a planning assumption under O-039, not a CMS minimum or talk-time target.

SMS, reminders, AI/chatbot/automated voice, asynchronous form completion, voicemail, unanswered attempts, appointment scheduling, L1 calls, and infrastructure availability do not replace or satisfy this human navigator contact.

Data.

• Current alignment/tier, assigned navigator, participants, modality, and completed interaction facts establish the contact context.
• Beneficiary/caregiver statements directly establish their reported experiences and preferences, not clinical diagnosis, urgency, or treatment.
• A completed call connection proves connection; it does not prove the interaction was substantive or that the actor was an eligible navigator.
• Duration is operational/audit evidence but is not itself proof of substance, and CMS does not set a minimum number of minutes in the public sources.
• AI transcript/summary can draft the record but cannot attest that the human service occurred or decide clinical disposition.

Potential Proxi work. Prepare source-linked context and prompts; support approved interpretation/accessibility; display outstanding commitments; capture speaker-attributed statements; retrieve records during the call; draft a note; identify candidate clinical/safety cues; and create bounded follow-up tasks after human confirmation. Proxi cannot conduct the required contact, determine urgency, reconcile medications, choose treatment, or attest contact qualification.

Human role. L2 is the human care navigator who performs the substantive service. The beneficiary/caregiver participates as applicable. L1 cannot substitute. L3/L4/L5 enters only when an actual medical, behavioral, medication, safety, safeguarding, or other complexity trigger arises and acts within beneficiary-location authority.

Provisional clean-path Proxi human minutes by role. L1: 0. L2 direct interaction: 8 / 15 / 25 minutes for stable brief / ordinary one-need / extended nonclinical work; preparation, documentation, and actual follow-through are counted in addition from observed work rather than forced inside a conversation target. L3/L4: 0 on a clean nonclinical contact; event-driven only and charged once to the separate clinical pool.

Completion evidence. Beneficiary and alignment identity; date/time/duration; assigned/covering navigator identity, credential and organization; participants; allowed modality; the contact reason and intended person value; topics actually addressed; topics explicitly declined or not asked; attributable changes, stable reports, needs, choices, or questions; human information/support actually provided; completed person-specific result; usable next step/help route; clinical/safety handoff when triggered; agreed follow-up; and completed-contact status distinct from service/billing qualification.

What does not prove completion. SMS, chatbot, AI voice, form, reminder, voicemail, unanswered call, scheduling call, L1/first-answer call, infrastructure availability, connection without substance, duration alone, a generic check-in, a completed all-domain template whose questions were not actually asked, note without interaction evidence, or human contact by someone who is not an approved care navigator.

Edge cases/open decisions. Existing: X-001 through X-010, X-012, X-014, and X-015; D-005; O-001, O-003, O-004, O-009, O-013, O-039, O-040, and O-PA-001. Proposed OPEN decision: exact qualifying-contact evidence, approved covering-navigator rule, interruption/substance threshold, and independence from monthly service/billing qualification. Minimum tests: clean navigator call; caregiver-only contact permitted; SMS/AI/L1 rejected; wrong navigator/organization; interrupted call with no substance; short but substantive contact; long scheduling-only call; clinical cue; immediate safety cue; contact outside alignment; missing credential evidence.

13. Conduct the individual-tier contact directly with the beneficiary#

When. Apply as a branch of Task 12 for a low-complexity individual or moderate/high-complexity individual beneficiary with no caregiver. Do not wait for a caregiver. Do not use this task to recategorize an RCC beneficiary; an RCC beneficiary follows the RCC tier while still receiving accessible direct engagement when appropriate.

What and how much. Conduct the same required human contact directly with the beneficiary, adapting pace, wording, repetition, language, modality, and accessibility to the person's demonstrated communication needs. Confirm practical understanding of agreed next actions without requiring nonexistent caregiver confirmation. Identify support gaps, but do not infer incapacity or clinical risk from communication difficulty alone.

Data.

• Confirmed individual-tier status and no-caregiver status establish this branch.
• The beneficiary's responses directly support what the beneficiary reports and demonstrates in the interaction.
• Need for repetition, a missed answer, hearing/language difficulty, or failed first teach-back does not prove incapacity, unreliability, or clinical deterioration.
• A facility worker, friend, or newly identified helper is not automatically a caregiver or representative and cannot substitute without the appropriate participation/authority basis.

Potential Proxi work. Present plain-language prompts; repeat approved content; slow/adapt presentation; support approved interpretation/accessibility; capture responses accurately; detect candidate misunderstanding; show chosen support options; and summon L2/clinical/safety help when the approved trigger occurs. AI cannot determine capacity, risk, or whether caregiver confirmation is required.

Human role. L2 conducts the direct human contact and works with the beneficiary on nonclinical supports. L3/L4 addresses an actual clinical, behavioral, safety, safeguarding, or decision-specific capacity question. Interpreter/accessibility support may assist. No caregiver is required to validate the capable beneficiary.

Provisional clean-path Proxi human minutes by role. L1: 0. L2: 0 incremental; this is the individual-tier form of the same Task 12 direct-interaction band. L3/L4: event-driven only. Interpreter/accessibility time is separate only when Proxi-paid.

Completion evidence. Task 12 contact evidence plus confirmed individual-tier/no-caregiver branch; direct beneficiary participation; accommodations used; attributable responses; understanding/support result; unresolved communication or support gaps; and clinical/safety escalation when actually triggered.

What does not prove completion. Automated check-in, caregiver/facility-worker report without direct beneficiary contact, inability to answer one question, repeated wording alone, yes/no response without context, inferred incapacity, or a note that says “caregiver unavailable.”

Edge cases/open decisions. Existing: X-002, X-003, X-005, X-007, and X-008; CP-E07; CC-E39; MED-E17; O-003, O-011, O-013, and O-PA-001. Minimum tests: capable beneficiary alone; interpreter need; hearing/vision barrier; technology failure; repeated misunderstanding without capacity inference; new safety signal; facility staff tries to substitute; caregiver appears later; RCC resident not recategorized.

Focused second contact for moderate/high-complexity individual tiers#

The required second monthly contact follows one named action, barrier, skill, safeguard, goal, requested support, or due decision from Contact A. The navigator restates that purpose, asks only for the facts needed to advance it, furnishes the focused ongoing-support service, and closes with the actual result, remaining barrier, next owner, and usable help route. It does not repeat Contact A's complete change, care-plan, medication, caregiver, or safeguard review.

Direct interaction may be approximately 5 / 10 / 15 minutes according to the actual work. If a qualifying inbound navigator conversation has already completed the same purpose inside the applicable contact window, it replaces the planned duplicate. If the original focus resolves before the second contact, the required contact addresses another current person-chosen or actually due ongoing-support purpose rather than becoming a status-only call. No new purpose may be invented merely to fill the time or create another service label.

When nothing is open and the beneficiary raises nothing, the second contact remains a brief accessible continuity service. It opens with the beneficiary's own priority and completes one attributable result: an answered question, confirmation that a named current support is working, or confirmation that one named current item still fits and no additional help is wanted. It closes with the usable help route. Every topic not raised remains not asked.

The contact record names the focus and why it was selected, the beneficiary's response, and the actual result—advanced, corrected, confirmed, declined, or truthfully unresolved—followed by the separate Task 31 contact evaluation. A repeated checklist, scheduling/reminder call, SMS or automated exchange, failed attempt, or duration alone does not complete the second contact. If three consecutive second contacts use this nothing-open path, the navigator asks within the third contact—not through a new contact—about preferred length, channel, and topics; records the answer; and sends the schedule to the burden review in 23_GUIDE_Low_Burden_Monthly_Service_Plan.md. The required cadence itself is not skipped.

14. Ask what has changed since the last contact#

Routing consequence. For each actual change or source-supported observation, record its truthful consequence: no downstream work opened; context correction only; a receiving service episode opened; Care Plan work opened; or bounded Assessment work opened. Use no downstream work only under Manual 25's source, authority, no-trigger, no-promise, reason, and re-entry conditions. More than one receiving consequence may apply only when each has a distinct reason, owner, and return condition.

When. Offer a bounded open change invitation during each Task 12 contact, then pursue only the person-selected topic, a current goal/action, an open obligation, a source-supported candidate change, or another domain that is actually relevant to the contact's stated purpose. On the first contact, use the initial Assessment/Care Plan baseline without implying prior-contact history. A late, corrected, or conflicting fact may reopen only the affected domain after the contact.

What and how much. Ask first what, if anything, has changed or matters now. Use tailored follow-up only for the chosen focus, actual open work, a credible current signal, or a minimum prompt that an approved source or policy specifically requires. Record each topic actually addressed as reported change, explicit no reported change, uncertain, or declined; every domain not discussed remains not asked. Do not run the full health, function, behavior, residence, transition, service, caregiver, goal, and preference list merely because a contact is due, and do not convert silence or “all good” into multi-domain stability.

Data/proof boundaries.

• An authenticated beneficiary/caregiver statement directly proves what that speaker reported and when, not diagnosis, clinical severity, urgency, causality, or another person's view.
• Current EHR, HIE, claims, transition, referral, service, and care-plan records can surface candidate changes or corroboration. Absence of a record does not prove no change.
• Caregiver observations remain separate from beneficiary statements and clinical findings.
• AI comparison can identify candidate differences and missing questions; it cannot promote them to current facts or decide materiality or urgency.

Potential Proxi work. Build a tailored question set from prior facts, the stated contact purpose, and open obligations; present source-linked candidate changes; capture speaker-attributed responses; compare with prior information; preserve conflicts; surface only due or justified topics; prevent inferred answers for omitted domains; and draft a bounded change summary. Proxi must retain free-text concerns and may not suppress an answer because it falls outside the script.

Human role. L2 gives the open invitation, asks relevant follow-up questions, understands the person's intended meaning, authenticates reported change or the named stable result, and keeps beneficiary/caregiver views distinct. The separate U.S. L3/L4/L5 clinical workforce handles clinical interpretation, urgency, safety, diagnosis, and treatment when triggered. L1 cannot conduct this qualifying interaction as a substitute for L2.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 2 / 4 / 6 component minutes included inside Task 12, not additional contact time. Separate U.S. clinical: 0 / 0 / event-driven, counted under Tasks 26–27 or the receiving clinical service.

Completion evidence. Contact and participants; baseline date; the open invitation; why each prompted topic was relevant; each actually addressed topic's attributable response/status; the person's chosen or declined focus; exact reported change or named stable item; source conflicts and unanswered items; the reason and re-entry condition when no downstream work opened; and every resulting child obligation. Unprompted domains remain not asked and need no fabricated response/status.

What does not prove completion. Checklist display, AI “no change” summary, unchanged EHR fields, no new claims, caregiver silence, copied prior note, or one general “everything is fine” statement without the navigator allowing clarification.

Edge cases/open decisions/minimum tests. Existing: X-003, X-004, X-005, X-008, and X-009; CC-E14, CC-E15, and CC-E43; O-003 and O-004. If an executed Participation Agreement or approved policy requires a specific minimum prompt, its source, owner, and exact scope must be added without inferring answers for anything not asked. Minimum tests: first contact; stable contact with no checklist; explicit no change for one named item; truthful no-downstream-work disposition; unexpected free-text concern; beneficiary/caregiver conflict; new symptom; late transition; residence change; declined topic; AI candidate disproved.

15. Check whether the care plan still fits#

When. Perform during the same Task 12 contact only when a current priority, reported change, due plan action, open obligation, credible mismatch, or beneficiary request makes plan fit relevant. Focus only on the affected goals, preferences, supports, responsibilities, or actions. Do not add a plan-fit check merely because the contact is due.

What and how much. Present the relevant current care-plan elements and obtain an attributable beneficiary disposition: still fits, does not fit, needs clarification, wants a change, or not reviewed. Ask whether anything important is missing. Keep clinical appropriateness separate from beneficiary fit and preference; a beneficiary may want a change that still requires professional review.

An attributable “still fits” result is valid contact-level plan-maintenance evidence. It does not require a new plan or cosmetic revision. This brief fit check is not the full whole-plan reviewed-no-change procedure in Care Plan Task 42 and does not create a monthly whole-plan review requirement.

Data/proof boundaries.

• The current active care plan proves the plan content in effect, not that it still reflects the beneficiary's current goals or needs.
• The beneficiary's authenticated response directly supports the beneficiary's fit/preference judgment.
• A caregiver statement proves the caregiver's perspective only and cannot silently replace beneficiary leadership or representative authority.
• Clinical outcomes, service records, and AI comparisons can identify review candidates; they do not determine that clinical content remains appropriate or that no revision is wanted.

Potential Proxi work. Display relevant plan elements in accessible language; compare them with reported changes; highlight possible mismatches; capture the beneficiary's exact disposition and wording; preserve caregiver/clinician differences; and create proposed revision candidates. Proxi cannot decide that the plan still fits.

Human role. The beneficiary leads the fit decision. L2 facilitates and authenticates the review inside the shared contact. A permitted caregiver contributes separately. Software applies the approved participation and disclosure rule after required facts are present. The separate U.S. L3/L4/L5 clinical workforce addresses clinical recommendations or safety implications within scope; only a complete verified fact pattern exposing a genuine unresolved legal-authority question routes to Healthcare Legal Counsel.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 4 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / event-driven when a clinical element is questioned or changed.

Completion evidence. Exact active plan element(s); source/content identity; beneficiary disposition and date; caregiver view when included; requested changes/clarifications; unresolved authority or clinical issues; and link to Task 16 when revision is needed.

What does not prove completion. No detected data change, unchanged structured fields, old acceptance, silence, caregiver agreement alone, model fit score, navigator opinion, or clinician approval without beneficiary review.

Edge cases/open decisions/minimum tests. Existing: CP-E02, CP-E03, CP-E04, CP-E08, CP-E10, and CP-E21; O-003, O-010, O-011, and O-015. Minimum tests: beneficiary confirms fit; requests nonclinical change; requests clinical change; beneficiary/caregiver differ; old plan shown; late material fact; unclear authority; refusal to review.

16. Start the care-plan revision work when needed#

When. Trigger when Task 14 or 15 identifies changed circumstances, goals, preferences, needs, or an explicit beneficiary revision request. Do not trigger a new revision from an exact duplicate or a change that has been authoritatively corrected away; unresolved materiality remains open under O-010.

What and how much. Create one linked Care Plan revision request identifying the exact affected elements, beneficiary direction, source facts, unresolved conflicts, needed professional decisions, and any immediate actions that cannot wait for final plan revision. Starting revision work is a handoff; the Monitoring note does not itself revise, approve, activate, store, or distribute the care plan.

Data/proof boundaries.

• An authenticated beneficiary request directly establishes that the beneficiary requested review/change; it does not authorize clinical or legally restricted content.
• A confirmed material fact can establish that an element needs review; exact materiality/no-change policy remains open under O-010.
• AI-proposed wording is a draft candidate only.
• A monitoring note, task creation, or handoff acceptance does not prove the written plan changed or was shared.

Potential Proxi work. Identify affected plan elements; assemble source facts; retain beneficiary wording; draft nonclinical textual candidates; list missing clinical/authority decisions; create and route the Care Plan work; and track acceptance. Proxi cannot turn the draft into the active plan.

Human role. L2 confirms the beneficiary's direction during Task 12 and owns continuity into the Care Plan service. The beneficiary leads the revision there. Software applies the approved authority and disclosure rule after required facts are present. The separate U.S. L3/L4/L5 clinical workforce handles affected clinical elements; only a genuine legal-authority question not answered by the approved rule routes to Healthcare Legal Counsel. No separate human minute is required merely to create the linked work.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 incremental in Monitoring; the direction is captured in Tasks 14–15 and later Care Plan labor is counted in that service. Separate U.S. clinical: 0 / 0 / event-driven in the receiving Care Plan service.

Completion evidence. Trigger/source; exact affected elements; beneficiary direction; linked Care Plan work identity; receiving owner/acknowledgement; unresolved clinical/authority items; and separately recorded immediate action where applicable.

What does not prove completion. Monitoring note, AI draft, opened revision task, navigator edit to a local copy, beneficiary request without downstream work, clinician message, or care-plan revision without EHR/distribution evidence.

Edge cases/open decisions/minimum tests. Existing: CP-E03, CP-E04, CP-E08, CP-E10, CP-E11, and CP-E21; O-003, O-010, O-011, O-015, and O-PA-001. Minimum tests: preference-only change; clinical change; immediate safety action plus later revision; duplicate trigger; corrected fact; caregiver-only request; receiving service rejects handoff; revision stays open after Monitoring contact closes.

17. Identify unmet clinical needs#

When. Screen during each Task 12 contact and whenever a beneficiary/caregiver raises difficulty obtaining medical, behavioral-health, dental, pharmacy, therapy, or other clinical help. Stop ordinary screening and invoke the safety route when an immediate concern appears.

What and how much. Identify each reported clinical-access gap: what help is needed or expected, which clinician/service is involved, whether an appointment/order/referral exists, current status, barrier, duration, beneficiary goal, and any symptoms or safety concern reported verbatim. L2 identifies the reported gap and access barrier; L2 does not decide diagnosis, urgency, medical necessity, specialty, or treatment.

Data/proof boundaries.

• Beneficiary/caregiver statements directly prove the reported need, barrier, and symptoms as stated—not clinical interpretation or urgency.
• Current referrals, orders, appointments, denials, provider messages, and returned documentation directly prove only their recorded administrative/clinical events.
• Claims can corroborate past utilization but do not prove current access, appointment completion, or that a need is met.
• No scheduled appointment, no claim, or provider nonresponse does not independently prove an unmet clinical need.
• AI can compare expected and observed events and suggest a route candidate; it cannot select urgency or clinical disposition.

Potential Proxi work. Ask structured and open questions; compare known needs with appointments/referrals/results; identify missing administrative events; preserve verbatim symptoms; separate access from clinical judgment; prepare a bounded handoff; and invoke the approved immediate-safety connection when a configured signal is present without AI deciding final urgency.

Human role. L2 authenticates the reported gap and handles nonclinical access barriers. The separate U.S. beneficiary-location-authorized L3/L4/L5 clinical workforce interprets clinical facts, determines urgency/appropriate clinical route, and gives clinical advice within scope. L1 may later verify administrative status but does not screen as the navigator or triage.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 4 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / event-driven, counted once under Task 26 or 27 and the receiving service.

Completion evidence. Attributable report; exact requested/expected clinical help; known order/referral/appointment facts; barrier; dates; reported symptoms/safety content; clinical-review requirement; and linked Task 18 or immediate pathway.

What does not prove completion. Missing claim, missed appointment alone, diagnosis list, generic “needs PCP,” AI urgency label, old referral, provider directory match, or navigator-selected specialty/treatment.

Edge cases/open decisions/minimum tests. Existing: CC-E01, CC-E05 through CC-E13, CC-E37, and CC-E43; O-003, O-004, O-019, and O-PA-001. Minimum tests: no clinical gap; appointment unavailable; provider rejection; cost/transport barrier; possible symptom; urgent office unresponsive; pending result without owner; beneficiary unsure what service is missing; AI misroutes.

18. Act on an unmet clinical need#

When. Trigger for every accepted Task 17 clinical-access gap or clinician-directed action. Immediate safety needs use the urgent pathway rather than waiting for routine coordination.

What and how much. Create and pursue the appropriate atomic Care Coordination obligation: appointment, referral, provider communication, transition support, result ownership, record retrieval, or other route. State the exact need/question, authorized clinical direction when required, destination, beneficiary choice, owner, due condition, and expected return evidence. Separate administrative routing from clinical specialty, urgency, treatment, and medical-necessity decisions.

Data/proof boundaries.

• A clinician order/recommendation supports only the exact clinical action within that clinician's scope.
• A beneficiary choice establishes preferred/accepted practical options, not clinical suitability.
• Directory and routing data identify candidates/endpoints; they do not prove treatment relationship, availability, clinical fit, referral acceptance, encounter, or outcome.
• Appointment booking proves scheduling, not attendance or clinical resolution.
• Care Coordination handoff acceptance proves receipt of work, not that the clinical need was met.

Potential Proxi work. Create the source-linked coordination case; retrieve and verify candidates and endpoints; filter practical constraints; transmit only through the approved release process; schedule after the authorized decision and beneficiary choice; track acknowledgement; and return downstream status to Monitoring. AI cannot choose specialty, urgency, treatment, or communication content that requires clinical authorship.

Human role. L2 coordinates, supports choice, and explains what happens next. L1 performs event-driven administrative verification/pursuit under the shared exception. The separate U.S. L3/L4/L5 clinical workforce supplies required clinical disposition and clinical communication. External provider staff perform endpoint acts.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 13 when manual pursuit is actually required. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 incremental in Monitoring; later coordination time belongs to Care Coordination. Separate U.S. clinical: 0 / event-driven / event-driven in the receiving service, counted once rather than again in Monitoring.

Completion evidence. Linked Task 17 need; exact coordination obligation; clinical authority when required; beneficiary choice; verified destination; accepted owner; disclosure/release basis; due/return evidence; acknowledgement; and downstream status kept separate from need resolution.

What does not prove completion. Referral draft, task creation, fax send, appointment booking, provider directory match, L1 call, clinician notification, handoff acknowledgement, or safety connection without downstream clinical disposition.

Edge cases/open decisions/minimum tests. Existing: CC-E01 through CC-E16, CC-E19, CC-E25, CC-E34, and CC-E37; O-003, O-004, O-019, O-021, and O-PA-001. Minimum tests: routine appointment; new clinical referral requiring order; provider unavailable; barrier; beneficiary refusal; wrong endpoint; no response; duplicate handoff; urgent need; handoff accepted but unresolved.

19. Identify unmet community and practical needs#

When. Screen during each Task 12 contact, emphasizing needs affected by recent change or prior open action. Include food, transportation, personal care, adult day, housing/home modification, social connection, benefits, technology/accessibility, caregiver logistics, and other supports relevant to community living.

What and how much. Identify each current reported practical need, desired outcome, existing support, gap/barrier, beneficiary/caregiver preference, location/accessibility constraints, and whether help is wanted. Obtain an explicit no-need/decline/unknown status where applicable rather than converting an empty screen to no need.

Data/proof boundaries.

• Beneficiary/caregiver statements directly prove what each person reports and wants, not program eligibility or service availability.
• Current service records and caregiver assessment can establish documented support and prior needs as of their dates; they do not prove current adequacy.
• Resource inventories and authenticated program responses establish listed/current administrative facts only.
• Claims or payer data can identify prior service use/benefit candidates, not current practical need or beneficiary preference.
• AI matching can suggest categories and candidates; it cannot determine need, urgency, eligibility, or choice.

Potential Proxi work. Tailor the screen from prior needs and services; capture attributable responses; compare known supports with stated gaps; search current inventories; explain candidate practical fit; preserve no-match/unknown; and prepare the Task 20 connection work. Software, not the navigator, should perform routine inventory filtering.

Human role. L2 authenticates the person's need and preference inside the shared contact and handles relationship-sensitive or complex barriers. The beneficiary/caregiver chooses whether to pursue help. L1 does not conduct the navigator screen. Separate U.S. clinical staff enter only when the issue actually contains clinical or safety consequences.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 4 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / event-driven only for a distinct clinical/safety concern.

Completion evidence. Attributable current need/no-need/decline/unknown result; desired outcome; current support; exact gap/barrier; constraints/preferences; help-wanted status; source/date; and linked Task 20 where action is requested.

What does not prove completion. Old HRSN screen, blank field, directory hit, prior service claim, caregiver assumption, model score, navigator opinion, or lack of complaint.

Edge cases/open decisions/minimum tests. Existing: CC-E06, CC-E07, CC-E17, CC-E26 through CC-E32, and CC-E38; O-004, O-022, and O-023. Minimum tests: no need; new food/transport/personal-care need; existing support insufficient; beneficiary declines; stale resource facts; Medicaid HCBS unknown; multiple barriers; practical issue reveals safety concern.

20. Act on an unmet community or practical need#

When. Trigger when the beneficiary/caregiver requests action on a Task 19 need. Reevaluate when eligibility, availability, cost, accessibility, waitlist, HCBS status, case-manager identity, beneficiary preference, or resource inventory changes.

What and how much. For each accepted need, provide a current option set or truthful no-match result; obtain the beneficiary's selection; complete the applicable direct referral/connection or AAA/Tribal/Medicaid HCBS coordination route; assign follow-up; and determine the actual route-specific outcome. Sharing information, referral submission, acceptance, intake, service start, and helpful result remain separate.

Data/proof boundaries.

• Resource inventory data proves only the listing, attributes, and as-of date.
• Authenticated organization/program response can establish current eligibility, availability, waitlist, intake, or service-start facts for the exact inquiry.
• Medicaid/HCBS eligibility, current receipt, and case-manager identity are separate facts; dual eligibility does not prove HCBS applicability.
• Beneficiary selection proves choice, not eligibility, availability, intake, service start, or benefit.
• A sent link, referral, accepted referral, or scheduled intake is not connection/service-start evidence.

Potential Proxi work. Filter and rank current resources; explain why they match; obtain electronic eligibility/availability where supported; capture selection; create and transmit the authorized referral; coordinate status; remind owners; compare GUIDE and HCBS services; and keep every route-specific outcome separate. Proxi cannot select for the beneficiary or infer successful connection.

Human role. No extra human is required for a clean electronically verified match, beneficiary selection, and connected referral. L1 performs manual program/case-manager pursuit under one shared exception. L2 handles preference-sensitive discussion, requested introduction, no-match problem solving, agency conflict, or complex coordination. Community/HCBS actors perform their endpoint work. Separate U.S. clinical staff enter only for a distinct clinical/safety consequence.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 13 if manual pursuit is triggered. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 incremental in Monitoring; event-driven connection work belongs to Referral/Services. Separate U.S. clinical: 0 / 0 / event-driven in the receiving service.

Completion evidence. Linked need; current option/no-match evidence; beneficiary selection; route and disclosure basis; referral/handoff; receiving acknowledgement; intake/service-start or authorized non-success outcome; barriers; follow-up owner; and resulting plan/Monitoring status.

What does not prove completion. Directory match, eligibility guess, sent link, referral submission, warm-introduction attempt, agency acknowledgement, scheduled intake, case-manager voicemail, or L1 pursuit alone.

Edge cases/open decisions/minimum tests. Existing: CC-E06, CC-E07, CC-E17, CC-E18, CC-E19, and CC-E26 through CC-E30; O-004, O-022, and O-023. Minimum tests: clean electronic connection; phone-only program; waitlist; no match; beneficiary refuses; referral accepted then stalls; expired AAA agreement; eligible-not-receiving HCBS; changed case manager; gap/duplication dispute.

21. Ask about medication use and medication-management problems#

Longitudinal boundary. This is a targeted check against the accepted medication baseline and open medication returns, not repeat medication reconciliation.

When. Screen during each Task 12 contact and whenever the beneficiary/caregiver raises medication access, use, schedule, possession, administration, symptom, or instruction concern. A possible immediate safety event leaves routine screening for the approved safety/clinical route.

What and how much. Conduct one bounded monitoring screen—not full reconciliation—covering what the beneficiary reports taking/not taking, missed/confused doses, new/changed/stopped products, OTC/supplement/sample use, supply/access, schedule-support problems, administration barriers, conflicting instructions, and new symptoms/concerns. Preserve each medication/report separately when possible. Do not recommend or execute a change.

An explicitly reported no-change/no-concern result completes this bounded monitoring screen when the applicable questions were substantively asked. It is not medication reconciliation, does not establish the intended regimen, and does not create a prescribing-clinician task unless an assessment, request, transition, discrepancy, symptom, monitoring result, or approved periodic-review basis exists.

Data/proof boundaries.

• Beneficiary/caregiver statements directly establish reported use, possession, barriers, and symptoms as stated—not intended regimen, adherence truth, causality, safety, or a medication order.
• EHR/order, pharmacy, bottle/photo, claims/dispense, facility MAR, and caregiver sources each prove only their source-native event/assertion.
• Fill/claim does not prove possession, taking, administration, intended regimen, or current use.
• A photo proves a visible labeled container and possible possession, not an order or use.
• AI comparison may identify discrepancy candidates; only prescribing-authority reconciliation establishes the clinical regimen/disposition.

Potential Proxi work. Present the known list as a source-labeled memory aid; capture reported use verbatim; separate intended/order/dispense/possession/reported-use/administration facts; compare sources; identify access and schedule-support barriers; create issue candidates; and invoke Task 22. Proxi cannot reconcile, diagnose an adverse effect, determine adherence, or advise starting, stopping, or changing medication.

Human role. L2 authenticates the report, asks practical follow-up, and supports access/schedule needs without clinical advice. L1 may later verify pharmacy/administrative facts. A separate U.S. beneficiary-location-authorized prescribing clinician reconciles and decides; separate U.S. L3/L4 clinicians handle symptoms/safety within scope.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 3 / 5 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / event-driven, counted once in Task 22/26/27 or Medication Management.

Completion evidence. Participant/reporter; source/date; each exact reported product/use/supply/barrier/symptom assertion; known source conflicts; no-change/no-concern only when explicitly reported; and every concern linked to Task 22 without altering the regimen.

What does not prove completion. EHR list confirmation alone, claim/fill, bottle photo, caregiver statement presented as beneficiary use, automated adherence score, no refill gap, model causality, navigator reconciliation, or “medications reviewed” without source-specific reports.

Edge cases/open decisions/minimum tests. Existing: MED-E01 through MED-E18, MED-E22, and MED-E24–MED-E25; O-004, O-027, O-028, O-032, and O-033. Minimum tests: explicit no problem; missed dose; access barrier; new OTC/sample; conflicting dose; self-stop/refusal; symptom after use; no caregiver; facility administration; wrong-patient bottle; unknown product.

22. Route medication concerns to the right medication task#

When. Trigger for every Task 21 discrepancy, possible symptom/adverse event, access/supply problem, schedule-support need, reported self-change/refusal, requested change, or other medication concern. Immediate safety signals use the approved urgent route without waiting for routine Medication work.

What and how much. Create the correct atomic Medication or support obligation while preserving the source report: history/source verification, prescribing-authority reconciliation, possible adverse/safety review, access/pharmacy pursuit, schedule-support information, clinical change request, self-change/refusal, or transition medication work. Do not combine all concerns into “med review.”

Data/proof boundaries.

• Task 21 supplies a reported/candidate concern, not clinical classification or urgency.
• The exact medication issue, source, timing, symptoms, beneficiary location, current known orders, and conflicts determine what facts the authorized route needs.
• An approved routing category may be selected from confirmed administrative facts; AI confidence cannot decide urgency, causality, clinical significance, or the current regimen.
• Handoff acceptance proves only receipt of the concern, not reconciliation, safety disposition, access resolution, support implementation, or medication change.

Potential Proxi work. Preserve source-native assertions; classify only approved administrative issue types; build the bounded packet; verify beneficiary location for clinical routing; create every distinct child obligation; alert the authorized 24/7 pathway for configured safety signals; track acknowledgement; and return outcomes. Proxi/L1 must not relay a new order or give medication instructions.

Human role. L2 handles nonclinical access and chosen schedule supports and tells the person what will happen next. L1 verifies pharmacy/administrative status under the shared exception. A separate U.S. prescribing-authority clinician reconciles or changes treatment. Separate U.S. L3/L4 clinicians handle symptoms/safety within scope. Pharmacist/pharmacy staff supplies pharmacy facts and permitted services.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 13 only when actual pharmacy/administrative pursuit is required. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 incremental for routing; later support work belongs to Medication/Support. Separate U.S. clinical: 0 / event-driven / event-driven in the receiving service, counted once.

Completion evidence. Exact Task 21 concern/source; route category; beneficiary location when clinical work is implicated; receiving owner and authority; source packet; acknowledgement; safety handoff if applicable; and separate open statuses for reconciliation, access, support, communication, and follow-up.

What does not prove completion. AI classification, generic “med review” task, alert, message send, L1 pharmacy call, clinician notification, handoff acknowledgement, refill, or schedule-support suggestion.

Edge cases/open decisions/minimum tests. Existing: applicable MED-E01 through MED-E44, especially MED-E04, MED-E07, MED-E15, MED-E17, MED-E18, MED-E22, MED-E41, and MED-E44; O-004, O-028, O-030, O-032, and O-033. Minimum tests: discrepancy; access-only issue; schedule-support request; possible adverse event; beneficiary self-change; prescribing question; unknown location; wrong recipient; multiple concerns; accepted handoff but no clinical outcome.

23. Ask the caregiver about current education and support needs#

Longitudinal boundary. This is a targeted check against the accepted caregiver baseline and prior learning/support results, not a repeat caregiver assessment.

When. Apply during Task 12 when a caregiver participates or when the qualifying contact is with the caregiver. Recheck current challenges, wanted help, and whether prior support remains useful. Do not apply this branch to a beneficiary with no caregiver.

What and how much. Ask about difficult caregiving situations, knowledge/skills needs, diagnosis-information questions, desired one-on-one help, support-group interest, stress/well-being, prior support usefulness, language/accessibility, and immediate safety or behavioral concerns. Preserve caregiver needs separately from beneficiary needs and ask what type/modality of support the caregiver wants.

Data/proof boundaries.

• The caregiver's authenticated statement directly establishes that caregiver's reported need, preference, stress, and observed situation—not a beneficiary diagnosis, clinical risk, or representative authority.
• The caregiver assessment and prior support records establish prior documented needs/services as of their dates; they do not prove current fit or benefit.
• Attendance, content delivery, chatbot use, and session scheduling are participation events, not comprehension, skill, support outcome, or current need resolution.
• AI can tailor questions and match approved offerings; it cannot diagnose caregiver mental health, determine safety, or choose the caregiver's service.

Potential Proxi work. Retrieve caregiver-assessment/prior-support context; generate tailored and open questions; capture the caregiver's own words; separate beneficiary/caregiver facts; match approved service candidates; identify distress/safety candidates; and open Task 24 or the appropriate clinical/safety path.

Human role. L2 conducts the person-centered caregiver-needs segment and provides permissible nonclinical support. The caregiver chooses desired service. Separate U.S. clinical, behavioral, or safeguarding professionals enter only for actual clinical, behavioral, safety, or safeguarding concerns. L1 cannot substitute for the navigator contact.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 3 / 5 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / event-driven for an actual concern.

Completion evidence. Caregiver identity/participation basis; source/date; current need/challenge; wanted support and modality; prior-support result; stress/well-being report; language/accessibility; distinct beneficiary facts; and linked Task 24 or clinical/safety route.

What does not prove completion. Old caregiver assessment, caregiver name in chart, sent education link, group attendance, chatbot usage, no complaint, AI distress score, or merged beneficiary/caregiver summary.

Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-008; CP-E04–CP-E06; CC-E39 and CC-E41; O-003, O-004, O-011, and O-036. Minimum tests: no new need; skills-training request; diagnosis-information request; one-on-one support; prior support ineffective; caregiver distress; caregiver safety concern; beneficiary/caregiver views differ; privacy-restricted issue.

24. Connect the caregiver to the needed support#

When. Trigger when the caregiver selects help from Task 23 or raises an ad hoc support need that can be addressed during the same contact. Reevaluate when service, vendor, schedule, accessibility, payment arrangement, caregiver preference, or clinical/safety need changes.

What and how much. Match and connect the caregiver to the applicable GUIDE offering: skills training, dementia diagnosis information, support group, ad hoc one-on-one support, or another applicable caregiver service. State the purpose, modality, performer, schedule/access route, who pays, and what evidence will show the service occurred. If the same Task 12 contact provides an eligible ad hoc one-on-one support service, document that distinct service content without adding a second contact.

Data/proof boundaries.

• The caregiver's selection proves choice, not enrollment, attendance, furnished service, comprehension, or benefit.
• The approved service catalog/contract establishes what Proxi or a partner may offer and under what modality; it does not prove the service was delivered.
• Scheduling/enrollment proves administrative readiness only.
• A session record with qualified performer, interaction/content, date, and outcome supports furnished human service; a sent video/link/chatbot event may support digital activity but whether AI-led skills training satisfies GUIDE remains open under O-036.

Potential Proxi work. Match assessed need to approved offerings; explain options; capture selection; enroll/schedule; deliver approved text/video/chat support when permitted; prepare the facilitator/navigator; track attendance/service evidence; and reopen after failure or continued need. Proxi cannot claim a match or scheduled event as furnished support.

Human role. No separate person is needed merely to match or schedule. L2 may provide ad hoc one-on-one support within Task 12 and performs other selected nonclinical education/coaching within the caregiver-service workload. A dementia-trained facilitator, care-team member, or permitted contracted/community provider furnishes the selected service. Clinical/safety content routes to the separate U.S. clinical workforce.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 13 only for an actual manual administrative exception. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 incremental when support is delivered within the same Task 12 contact; later human-service minutes belong to Caregiver Education/Support. Separate U.S. clinical: 0 / 0 / event-driven in the receiving service.

Completion evidence. Linked caregiver need; selected service; approved performer/organization; modality/accessibility; enrollment/schedule; actual service record or truthful non-success status; distinct evidence when Task 12 also furnished one-on-one support; and follow-up on usefulness.

What does not prove completion. Resource match, sent link/video, chatbot turn, enrollment, scheduled session, calendar attendance without service content, referral, or a navigator saying “support offered.”

Edge cases/open decisions/minimum tests. Existing: X-006, X-007, X-014, X-015; CC-E06, CC-E07, CC-E19, and CC-E39; O-036 and O-PA-001. Minimum tests: ad hoc support within Task 12; later skills training; direct diagnosis information; external support group; caregiver declines; no-show; inaccessible service; vendor unavailable; AI-led training blocked pending policy.

25. Check the extra safeguards for a beneficiary without a caregiver#

Longitudinal boundary. This is a targeted check of the accepted no-caregiver safeguards, not a redesign or repeat assessment of the safeguard package.

When. Apply during the Task 13 individual-tier branch and whenever a beneficiary without a caregiver loses, cannot use, or questions an existing support. Recheck after a transition, medication change, move, communication change, missed appointment, emergency event, or newly identified helper/caregiver.

What and how much. Review five practical safeguard areas with the beneficiary: usable communication/contact; medication schedule/access support; appointment/transport/follow-through; emergency/help-seeking plan; and community-living services/supports. For each, record working, partially working, not working, declined, unknown, or not applicable with basis. Identify the exact gap and offer approved supports without requiring a caregiver or inferring incapacity.

Data/proof boundaries.

• The beneficiary's report directly establishes reported use, preference, and difficulty—not clinical safety, capacity, or objective performance.
• Delivery/system logs can corroborate reminder/contact events but do not prove the beneficiary understood or used the support.
• Medication fills, appointments, emergency plans, and service records each prove only their source-native event.
• No caregiver status does not prove a safeguard is needed; an identified gap in one domain does not prove failure in every domain.
• AI can compare supports with reported barriers and propose approved options; it cannot decide capacity, clinical risk, restriction, or protective action.

Potential Proxi work. Present the five-domain review; retrieve existing support evidence; capture beneficiary responses; identify candidate gaps; offer accessible approved communication, schedule, appointment, emergency, and community supports; create child obligations; and monitor failures. Proxi must keep each domain and outcome separate.

Human role. L2 conducts the direct review inside the shared individual-tier contact and arranges permissible nonclinical supports with the beneficiary. L1 may perform manual administrative pursuit later. The separate U.S. L3/L4/L5 clinical workforce handles clinical, medication, behavioral, safety, safeguarding, or capacity implications within actual authority.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0 in the contact. L2 Puerto Rico nonclinical navigator: 2 / 4 / 7 component minutes included inside Task 12/13, not a second contact. Separate U.S. clinical: 0 / 0 / event-driven when a real concern appears.

Completion evidence. Confirmed no-caregiver/current tier branch; beneficiary participation; status/evidence for all five safeguard domains; exact gaps; selected/declined supports; child owners; clinical/safety routes when triggered; and follow-up timing.

What does not prove completion. No-caregiver field, “lives independently,” reminder setup, pillbox possession, transportation benefit, emergency-contact name, facility staff presence, model risk score, or one successful prior event.

Edge cases/open decisions/minimum tests. Existing: CP-E07; CC-E07, CC-E38, CC-E39, and CC-E43; MED-E17, MED-E41, and MED-E44; O-003, O-004, O-011, O-013, and O-033. Minimum tests: all supports working; one-domain gap; multiple independent gaps; beneficiary declines support; communication failure; medication schedule barrier; missed appointment/transport; emergency plan unusable; later caregiver found; difficulty without capacity inference.

26. Bring complex issues to a clinical team member#

When. Trigger when Tasks 14, 17, 21, 23, or 25—or a between-contact message—reveals a medical, medication, behavioral, safeguarding, or other issue whose meaning or response requires licensed clinical judgment. A configured cue may trigger consultation conservatively; it cannot prove the issue is clinically significant or that consultation is unnecessary.

What and how much. Give the appropriate separate U.S. clinical team member one source-faithful consultation packet: exact reporter and statement, current beneficiary location, onset/timing when reported, relevant records and conflicts, actions already taken, the precise question needing judgment, and a usable return route. Tell the beneficiary/caregiver what is being handed off and what to do while awaiting the response, using only approved nonclinical language.

Data/proof boundaries.

• The beneficiary/caregiver report proves only what that person reported; it does not prove diagnosis, severity, causality, treatment response, risk, or urgency.
• Valid provider documentation proves the recorded clinical finding or instruction within its author, date, encounter, and scope; it may still conflict with newer reports.
• Current physical location and clinician authorization are required for the licensed act. A clinician's title or roster presence alone does not establish authority for the beneficiary's location or requested act.
• AI cues, risk scores, and summaries are consultation candidates and presentation aids, not clinical conclusions.

Potential Proxi work. Detect configured consultation cues; preserve exact source statements; retrieve relevant source records; verify current location and configured clinical route; assemble a concise packet without suppressing conflicts; request acknowledgement; and return the clinical response to the correct person and downstream task after authorized release.

Human role. If L2 is already conducting the required contact, L2 may preserve the person's literal report, confirm only the essential practical facts, explain the handoff in approved language, and remain supportive. Otherwise L0/L1 assembles and routes the source-faithful packet directly to the separate U.S. beneficiary-location-authorized RN, LCSW/behavioral clinician, NP/PA/CNS, physician/MD, pharmacist, or other professional authorized for the exact clinical act. L2 is never a prerequisite or tollbooth before clinician entry and performs no clinical or medical work under D-015.

Provisional human minutes (low / typical / high) per triggered issue. L1 Philippines: 0 / 0 / 5 for an endpoint problem only. L2 Puerto Rico nonclinical navigator: 0 / 2 / 5, with time only when L2 is already in the contact or a separate human relationship/warm-handoff act is genuinely required. Separate U.S. clinical: 5 / 10 / 20 for review, response, and required direction; downstream treatment work is counted in its owning service.

Completion evidence. Exact triggering report/source; current location; requested clinical question; receiving clinician identity, credentials, organization, location authority, and scope; packet content; acknowledgement; clinical disposition/return communication; and resulting child work. Monitoring may complete its consultation handoff while the underlying clinical issue remains open.

What does not prove completion. AI flag, risk score, packet draft, message send, clinician notification, voicemail, acknowledgement without review, undocumented curbside comment, or L2 interpretation recorded as clinical advice.

Edge cases/open decisions/minimum tests. Existing: X-004, X-008, X-009, X-014; CC-E10 through CC-E13, CC-E39, CC-E41, and CC-E43; MED-E18, MED-E25, MED-E40, and MED-E44; D-015, O-003, O-004, and O-028. Minimum tests: routine nonclinical issue needs no consult; symptom report; behavioral concern; medication discrepancy; wrong-location clinician; PR navigator with RN credential blocked from clinical act; clinical route unavailable; multiple clinical questions; ambiguous source; acknowledgement without disposition.

27. Respond immediately when the contact reveals an urgent safety concern#

When. Trigger when the contact contains a possible immediate threat to life or safety, an approved urgent clinical cue, or a direct request for emergency/crisis help. Do not wait for routine follow-up or for every nonurgent contact topic to be completed. A clear emergency connection is not delayed while AI or L2 tries to grade severity.

What and how much. Keep the person connected when safe; obtain current physical location, callback information, who is present, and the exact concern; invoke the approved U.S. urgent clinical/emergency/crisis route; support a live handoff; and preserve every nonurgent issue for later follow-up. Do not give individualized clinical instructions or promise that help is on the way without confirmation.

Data/proof boundaries.

• A direct report or observed event can justify immediate protective routing without proving diagnosis or causality.
• Current location and callback facts support dispatch or licensed routing; stale address data does not.
• A connection attempt, ringing phone, alert, or message does not prove that an authorized responder accepted the handoff.
• The receiving clinician/emergency professional's documented response proves that response only; it does not automatically resolve the underlying clinical, medication, safeguarding, or care-plan work.

Potential Proxi work. Display approved immediate-response instructions; capture current location/callback; connect the configured route; share the source-faithful minimum packet through the permitted channel; keep retry/fallback information visible; record acceptance; and create follow-up for every unresolved nonurgent and post-event issue. Proxi cannot select clinical urgency or provide treatment advice.

Human role. L2 maintains the human connection and performs the nonclinical safety handoff. L1 may capture verbatim facts and invoke the approved route if the event begins through first answer, but cannot triage. The separate U.S. clinician or emergency/crisis responder evaluates and acts. Puerto Rico L2 never becomes the clinical decision-maker.

Provisional human minutes (low / typical / high) per triggered event. L1 Philippines: 0 / 0 / 3 when first answer begins the handoff. L2 Puerto Rico nonclinical navigator: 3 / 8 / 20 until live transfer, safe loss-of-contact response, or confirmed alternate action. Separate U.S. clinical: 5 / 15 / 30 for immediate evaluation and direction; external emergency-responder time is excluded unless Proxi pays it.

Completion evidence. Reporter and exact concern; current location/callback; route invoked; attempts and failures; live receiving person/organization; acceptance time; information transferred; immediate disposition supplied by the authorized responder; communication back to the beneficiary/caregiver when appropriate; and every remaining issue assigned for follow-up.

What does not prove completion. AI alert, “P0” label, message send, voicemail, attempted transfer, 911/clinical number displayed, call disconnected, unconfirmed dispatch, or navigator reassurance without an accepted handoff.

Edge cases/open decisions/minimum tests. Existing: X-008, X-013, X-014; CP-E11; CC-E13, CC-E40, CC-E42, and CC-E43; MED-E18, MED-E38, and MED-E44; O-004 and O-032. Minimum tests: immediate threat; uncertain cue; false-positive extraction; caller disconnects; location unknown; person traveling; clinical line unavailable; emergency service accepts; caregiver panics but beneficiary status unknown; urgent issue plus several nonurgent needs.

28. Agree on what happens after the contact#

When. Complete before ending every substantive contact when possible, or promptly afterward through an accessible permitted route if an urgent handoff or disconnection prevented recap. Revisit any item that the beneficiary/caregiver does not understand, declines, disputes, or cannot carry out.

What and how much. Recap every identified need and commitment: what will happen, who owns it, what the beneficiary/caregiver has chosen or declined to do, expected timing or return condition, how status will be communicated, and what route to use if circumstances change. Ask the person to explain the practical next step in their own words when comprehension matters. “Agree” includes a truthful recorded disagreement, refusal, or uncertainty; do not manufacture consensus.

Data/proof boundaries.

• An authorized decision or accepted commitment supports only its exact content and owner.
• The beneficiary/caregiver's authenticated response establishes acceptance, refusal, uncertainty, or stated understanding—not future performance.
• A yes/no response, sent recap, or repeated wording alone does not prove practical understanding.
• AI may draft the recap from accepted content; it cannot add a clinical instruction, assign an external provider's responsibility, or attest comprehension.

Potential Proxi work. Assemble commitments from the contact; identify missing owner/timing; render approved content in accessible language; prompt read-back; capture corrections and refusals; generate a take-home recap; and update each commitment without merging separate obligations.

Human role. L2 resolves nonclinical misunderstandings, confirms choices, and owns the person-centered recap. Separate U.S. clinical staff directly correct or authorize clinical content when required; L2 repeats only approved clinical language. L1 is not required for the clean recap.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 4 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / 5 only when clinical content needs direct clarification, counted with Task 26/27 rather than twice.

Completion evidence. Every need/commitment; accountable owner; beneficiary/caregiver action if any; accepted/declined/uncertain status; practical timing/return condition; accessible recap; read-back or documented reason it was not needed/possible; misunderstandings corrected; and unresolved items visible.

What does not prove completion. AI-generated after-visit summary, note template, sent recap, “verbalized understanding,” nod/yes alone, caregiver acceptance substituted for beneficiary choice, or assigned task without an accountable owner.

Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-007, X-014; CP-E04, CP-E10; CC-E19, CC-E39; MED-E16, MED-E22, and MED-E41; O-011. Minimum tests: clean recap; beneficiary declines one action; beneficiary/caregiver disagree; failed teach-back; inaccessible format; urgent transfer prevents recap; external owner has not accepted; clinical instruction requires clinician correction; several simultaneous commitments.

29. Arrange the next qualifying contact#

When. Before ending the current contact or promptly afterward, book the next opportunity early enough to meet the current Task 6 cadence and allow recovery from failure. Rebook when the selected time, navigator, modality, accommodation, tier, or participant arrangement changes. If an independently qualifying inbound contact already satisfies the current window, cancel the redundant outbound appointment unless another distinct required purpose remains.

What and how much. Offer usable times within the current due window, confirm permitted modality, time zone, participants, navigator coverage, language/accessibility support, and fallback route, then create an actual booking and confirmation. A clinically or nonclinically authorized earlier-contact need may add an earlier appointment without erasing the minimum-cadence booking logic.

Data/proof boundaries.

• Current tier/effective date and approved contact-clock interpretation establish the due window.
• Authenticated availability and modality preference establish scheduling constraints, not contact completion.
• A calendar invitation, tentative hold, or recurrence rule does not prove acceptance, attendance, or a qualifying contact.
• AI may propose convenient slots but cannot decide a higher clinical cadence.

Potential Proxi work. Calculate the due window; retrieve navigator/interpreter availability; offer constrained times; book the selected appointment; suppress superseded invitations; send accessible confirmation; and flag a residual coverage gap. Software should perform the connected-calendar clean path.

Human role. No Proxi human is required on the clean path; the beneficiary/caregiver selects among usable times. L2 resolves a relationship-sensitive conflict during the contact. L1 handles a manual scheduling failure under the shared exception. Separate U.S. clinical staff enters only to authorize clinically driven earlier frequency, not to schedule.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 5 for a manual scheduling exception. L2 Puerto Rico nonclinical navigator: 0 / 1 / 3 component minutes included inside Task 12. Separate U.S. clinical: 0 / 0 / 0 for scheduling; any cadence decision is counted with its clinical review.

Completion evidence. Current due window; selected date/time/time zone; assigned/covering navigator; permitted modality; participants; accommodations; accepted booking; confirmation outcome; fallback/recovery opportunity; and any approved earlier-contact basis.

What does not prove completion. Proposed slot, recurrence rule, tentative hold, invitation send, reminder, staff availability, scheduling conversation, or a booking outside the due window without documented resolution.

Edge cases/open decisions/minimum tests. Existing: X-004 through X-007, X-013, X-014; O-004 and O-PA-001; proposed contact-clock decision. Minimum tests: connected booking; no acceptable slot; interpreter unavailable; navigator leave; DST/time-zone change; tier changes after booking; high-risk earlier contact; beneficiary refuses future date; calendar invitation not accepted.

30. Record what happened in the contact#

When. Complete promptly after each substantive contact and correct the record when a participant or source-faithful review identifies an error. Preserve the original contribution and correction relationship rather than silently overwriting clinically or operationally material history.

What and how much. Record participants, identities/roles, modality, date/time/duration, substantive topics, speaker-attributed reports, reported changes, beneficiary/caregiver choices, approved human information or advice, clinical consultations, connections started, commitments/owners, urgent handoffs, unresolved items, and next-contact booking. Keep beneficiary, caregiver, clinician, external-source, and model-drafted content distinguishable.

Data/proof boundaries.

• Audio/transcript supports what was captured, subject to identity, consent, completeness, and transcription error; it does not itself prove the speaker's claim is clinically true.
• The navigator's review supports accountability for the final contact record, not clinical truth outside L2 scope.
• A clinical consultant is accountable only for their attributable contribution and need not review the entire nonclinical note.
• AI-generated text is a draft and may hallucinate, merge speakers, omit refusals, or convert uncertainty to fact.

Potential Proxi work. Transcribe when permitted; label speakers and sources; populate the structured contact note; link every commitment and child service; compare the draft with source audio/text; flag unsupported statements; preserve corrections; and present only the sections requiring human confirmation.

Human role. L2 checks and finalizes the contact record, correcting attribution and substantive errors. Separate U.S. clinical staff verifies only its own consultation/clinical contribution when needed. L1 does not review the navigator note on the clean path.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 5 minutes after the conversation, counted once with the Task 12 episode as post-contact work outside the direct band; an extended multi-action episode may reach 8. Documentation is never counted again in another service's record of the same episode. Separate U.S. clinical: 0 / 0 / 3 to verify its own contribution only, counted with the consultation where possible.

Completion evidence. Final accountable author; contact identity; participants/modality/time; substantive purposes; speaker/source attribution; every change/need/choice; human clinical contributions; child obligations; commitments; urgent outcome; next contact; corrections; and source links.

What does not prove completion. Raw transcript, ambient-note draft, AI summary, unsigned/copied note, duration alone, generic “follow-up completed,” merged speakers, clinician name without attributable input, or downstream tasks absent from the record.

Edge cases/open decisions/minimum tests. Existing: X-001, X-004, X-009, X-011, X-013, X-015; CC-E15, CC-E24, CC-E36; MED-E24, MED-E26, MED-E36. Minimum tests: clean draft; wrong speaker; omitted refusal; hallucinated diagnosis; interrupted recording; interpreter contribution; clinical consultant contribution; corrected note; multiple child tasks; late material correction after contact closure.

31. Determine whether the contact satisfies the minimum frequency#

When. Evaluate after Task 30 has a complete contact record and before using the interaction as evidence for the current tier's cadence. Reevaluate when actor eligibility, organization, alignment, tier, date/time, modality, participant, or substantive-purpose evidence is corrected.

What and how much. Apply the current approved program interpretation to objective facts: an eligible human care navigator, approved organization, aligned beneficiary, completed substantive interaction with the beneficiary and/or permitted caregiver, an ongoing-support purpose under §4.2, and an allowed in-person, telephone, or audio-visual modality. SMS, chatbot, AI voice, form, voicemail, reminder, scheduling-only call, L1 first answer, or failed attempt does not satisfy the minimum. Keep cadence satisfaction separate from whether the month supports a GUIDE payment claim.

When a naturally occurring inbound interaction passes this same review, treat it as the contact for the applicable window and suppress the duplicate planned contact. Do not require the person to repeat the conversation merely because the original calendar entry was outbound.

Data/proof boundaries.

• Task 30 supplies contact evidence; it does not by itself determine cadence treatment.
• Navigator roster/credential/organization evidence and alignment/tier/effective dates apply only for their documented periods.
• Modality and completion facts are objective. Missing or conflicting administrative evidence returns to L1; ambiguous or inaccurate substantive-contact documentation returns to the L2 navigator who performed and finalized Task 30; an outside source corrects only its own source record. Software then re-applies the approved cadence rule.
• Duration alone neither proves nor disproves a substantive contact. AI cannot determine cadence or billing treatment; deterministic software applies only the approved rule to complete verified facts.

Potential Proxi work. Assemble the exact contact evidence; identify missing, conflicting, or disqualifying facts; apply the approved objective contact criteria; explain the evidence-backed result; route administrative evidence defects to L1 and substantive Task 30 corrections to the L2 navigator who owns that record; obtain any outside-source correction from that source; and re-evaluate after correction. If complete verified facts fit no approved cadence rule, hold the cadence determination and open a design-time program-rule gap outside the beneficiary episode. Proxi must never convert a nonqualifying interaction into a qualifying one to fill a cadence gap.

Human role. No additional reviewer is needed when the record is complete and the approved criteria produce an outcome. L1 obtains or corrects missing administrative evidence; L2 corrects only its own substantive contact record and does not decide cadence or billability; an outside source corrects only its own record. Software re-applies the approved cadence rule. A complete verified fact pattern with no supported outcome remains held as a design-time program-rule gap rather than becoming a runtime assignment. Clinical staff are not needed.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0 on the clean path; any event-driven administrative evidence correction is counted in the source-recovery work. L2 Puerto Rico nonclinical navigator: 0 / 0 / 0 except time already counted to correct Task 30. Separate U.S. clinical: 0 / 0 / 0. There is no separate beneficiary-case exception-review allowance; design-time rule governance remains outside the episode.

Completion evidence. Exact contact record; alignment/tier period; navigator identity, credential, organization, and coverage authority; participant; date/time; allowed modality; substantive ongoing-support content; completed-versus-attempted result; cadence period applied; outcome and reason; and separate monthly service/payment treatment.

What does not prove completion. Call duration, note existence, calendar status, telephony connection, “contacted” code, SMS reply, chatbot exchange, L1 call, long scheduling conversation, infrastructure availability, or payment claim.

Edge cases/open decisions/minimum tests. Existing: X-001, X-004, X-005, X-010, X-015; D-005; O-001, O-009, O-040, O-PA-001; proposed qualifying-contact and contact-clock decisions. Minimum tests: clean qualifying call; caregiver-only permitted contact; L1/AI/SMS rejected; scheduling-only rejected; short substantive contact; interrupted non-substantive call; covering navigator; credential expired; outside alignment; ambiguous note; corrected record changes result; qualifying contact but payment treatment remains separate.

32. Follow up when the person misses or cannot complete a contact#

When. Trigger after a no-show, unanswered attempt, early disconnection, inaccessible modality, interpreter failure, technology failure, illness, or other event that prevented a substantive qualifying contact. Start soon enough to preserve the current due window when possible.

What and how much. Record why the interaction did not complete, address the specific barrier, offer a permitted accessible alternative, book a replacement opportunity, and keep the cadence obligation open until a qualifying contact occurs or the accountable program records the truthful unresolved result. Do not label the failed attempt as completed.

Data/proof boundaries.

• Call, video, portal, calendar, interpreter, and delivery logs prove only their recorded attempt/status.
• A beneficiary/caregiver explanation supports the reported barrier or choice, not clinical risk or intentional avoidance.
• A replacement booking proves scheduling only, not contact completion.
• Missed contact alone does not prove deterioration, incapacity, neglect, nonadherence, or a safety emergency.

Potential Proxi work. Detect failure; suppress false completion; offer consented replacement times/modalities; arrange the known accommodation; send confirmation; track the due window; and alert when manual help or a repeated-unreachable review is needed.

Human role. No Proxi human is required for a clean self-service rebooking. L1 handles manual scheduling, endpoint, interpreter, or technology pursuit under the shared exception. L2 resolves relationship-sensitive barriers and conducts the replacement Task 12 contact, whose minutes are counted there. Separate U.S. clinical staff enters only for an actual clinical/safety concern.

Provisional human minutes (low / typical / high) per failed-contact episode. L1 Philippines: 0 / 5 / 13 when manual recovery is needed. L2 Puerto Rico nonclinical navigator: 0 / 2 / 5 for a barrier conversation; the replacement qualifying contact is counted separately under Task 12. Separate U.S. clinical: 0 / 0 / event-driven only for a distinct concern.

Completion evidence. Failed attempt identity and reason; source logs; participant explanation when available; barrier and accommodation action; replacement booking/confirmation; current due-window status; manual owner; and link to Task 33 if attempts continue to fail.

What does not prove completion. Automated retry, voicemail, reminder, replacement invitation, calendar hold, one unanswered call, delivery receipt, or staff note saying “will try again.”

Edge cases/open decisions/minimum tests. Existing: X-005, X-006, X-007, X-013, X-014; O-004; proposed contact-clock and qualifying-contact decisions. Minimum tests: self-service rebook; no answer; early disconnect; failed interpreter; inaccessible video; wrong time zone; beneficiary declines proposed times; replacement outside due window; clinical symptom reported during recovery; repeated failures invoke Task 33.

33. Escalate repeated inability to reach the beneficiary#

When. Trigger after the approved number/pattern of failed qualifying-contact attempts, earlier when a known fact independently raises concern, or when a beneficiary without a caregiver loses all usable contact routes. The numeric attempt threshold and response clock remain open under O-004/O-013; one missed contact alone does not automatically invoke welfare or emergency action.

What and how much. Review every attempted route and outcome, current communication permissions, known location/residence, no-caregiver status, recent transitions and open commitments, current documented clinical/safety facts, permitted alternate contacts, and the least intrusive justified next step. Continue ordinary outreach, navigator review, social-work support, separate U.S. clinical review, welfare response, or emergency connection only according to the actual facts and approved authority.

Data/proof boundaries.

• Attempt logs prove attempts and technical outcomes, not that the beneficiary is unsafe, unwilling, incapable, absent, or deceased.
• A permitted alternate contact may supply a source-attributed report; relationship or emergency-contact status alone does not authorize disclosure or decision-making.
• Prior risk/diagnosis and no-caregiver status can inform review but do not independently justify police, welfare, clinical, or restrictive action.
• AI may summarize the attempt history and candidate concerns; it cannot decide risk, urgency, capacity, or the outreach intrusion level.

Potential Proxi work. Assemble all attempts without duplication; identify unused permitted routes; detect changed endpoints/residence candidates; surface current source-linked concerns; prepare the least-intrusive option set; prevent unauthorized alternate-contact disclosure; and track the selected outreach and outcome.

Human role. L1 verifies administrative contact facts and performs approved manual pursuit. L2 reviews relationship context, speaks with permitted contacts, and selects only permissible nonclinical outreach. Separate U.S. beneficiary-location-authorized clinical/social-work/safeguarding staff decides clinical, behavioral, safety, or protective implications within scope. Emergency/welfare responders act only when justified by the approved path.

Provisional human minutes (low / typical / high) per repeated-unreachable episode. L1 Philippines: 0 / 5 / 13. L2 Puerto Rico nonclinical navigator: 3 / 8 / 15. Separate U.S. clinical: 0 / 0 / 15 only when independent clinical/safety facts trigger review; repeated nonresponse alone remains nonclinical.

Completion evidence. Attempt inventory; approved threshold/basis; permissions and routes considered; current known location/residence; no-caregiver and open-need context; identified concern facts; human reviewer and authority; selected least-intrusive action; outreach/handoff result; due-window effect; and explicit unresolved plan when contact remains impossible.

What does not prove completion. Attempt count alone, risk score, no-caregiver label, old diagnosis, alternate-contact name, voicemail, welfare-check request, dispatch attempt, or closure because all ordinary channels were exhausted.

Edge cases/open decisions/minimum tests. Existing: X-002, X-004, X-006, X-008, X-012 through X-014; CP-E07, CP-E11, CP-E15; CC-E20, CC-E31–CC-E32, CC-E40, CC-E43; O-004, O-011, O-013, and O-024. Proposed OPEN decision: repeated-unreachable threshold and least-intrusive escalation ladder by tier/no-caregiver/current risk evidence. Minimum tests: wrong number; temporary travel; beneficiary declines contact; no caregiver/no independent risk; known urgent concern; move/hospitalization/death signal; unauthorized alternate contact; one route succeeds late; welfare response unavailable; corrected attempt log.

34. Complete promised follow-up work#

Assessment return. When a returned result materially corrects or changes an Assessment domain, preserve the original Assessment history and send the source-linked result to the Assessment owner as a bounded correction or reassessment candidate.

When. Begin as soon as the Task 28 commitment is accepted and continue until each promise has its own completed or truthful unresolved outcome. Do not wait for the next scheduled contact when the promised action or known need requires earlier work.

What and how much. Track every promise separately: referral, appointment, provider consultation, record request, medication work, resource connection, caregiver support, information delivery, scheduling, or other action. Ensure the correct receiving service/person owns it, monitor the return evidence, tell the beneficiary/caregiver the actual result or barrier, and preserve any next action. This task oversees the promises; it does not duplicate all labor performed in the receiving SOP.

Data/proof boundaries.

• Task 28 proves the commitment and intended owner, not performance.
• A send, assignment, acknowledgement, booking, referral, or status inquiry proves only that event.
• Endpoint evidence must match the exact promised result: encounter, reviewed response, service start, furnished support, received information, or authorized non-success outcome.
• AI may track and summarize; it cannot decide that a clinical need, referral, medication issue, or caregiver need is resolved.

Potential Proxi work. Maintain the promise list; perform permitted software actions; remind accountable owners; detect overdue/mismatched evidence; prepare manual pursuit; surface barriers; return source-linked outcomes; and prevent the next contact from losing unresolved work.

Human role. L1 performs a genuine administrative pursuit episode when software cannot. L2 handles beneficiary choice, relationship barriers, explanations, and cross-party coordination. Separate U.S. clinical or other service workers perform the substantive work owned by their SOP; their minutes are counted there, not again as Monitoring overhead.

Provisional human minutes (low / typical / high) for Monitoring oversight. L1 Philippines: 0 / 0 / 3 beyond another already-counted shared pursuit episode. L2 Puerto Rico nonclinical navigator: 0 / 2 / 5 for status explanation or cross-service ownership repair. An extended episode tracking several distinct promises may reach 8 minutes; this is a multi-promise tail, not a routine addition. Separate U.S. clinical: 0 / 0 / 0 in Monitoring; any clinical work is counted in its owning service.

Completion evidence. Every promised action; source contact/task; accountable owner; receiving acceptance; attempts; exact return evidence or authorized non-success outcome; barrier/next plan; beneficiary/caregiver communication; and unresolved commitments retained for Task 35.

What does not prove completion. Task creation, assignment, reminder, fax/message send, referral, booking, acknowledgement, L1 call, clinician notification, link delivery, or a generic “follow-up done” note.

Edge cases/open decisions/minimum tests. Existing: X-003 through X-007, X-011, X-013 through X-015; CC-E03, CC-E09, CC-E25, CC-E26, CC-E34, CC-E37; MED-E21, MED-E23, MED-E29, MED-E42; O-004 and proposed cross-service handoff-completion decision. Minimum tests: automated clean completion; manual pursuit; receiving service rejects; wrong owner; partial completion; several independent promises; clinical work completed but communication open; barrier persists; late result; next contact occurs while work remains open.

35. Check prior needs at the next contact#

Assessment return. A material result returns to the affected Assessment domain as an as-of correction or bounded reassessment candidate while the original Assessment record remains intact.

When. At the next Task 12 contact, revisit every prior need or promise that remained open, reported complete, failed, or was expected to produce a real-world result. Bring forward late outcomes and new barriers even if the original task was administratively closed.

What and how much. Ask what actually happened, whether the person received and used the service/information/support, whether it helped the stated need, what barrier remains, and whether the person wants continued, changed, or stopped work. Keep service delivery, usability, benefit, and continuing need distinct.

Data/proof boundaries.

• Task 34 and endpoint records prove their source-native actions/outcomes, not the beneficiary/caregiver's experience or benefit.
• The beneficiary/caregiver report directly establishes reported use, result, preference, and remaining difficulty—not clinical effectiveness or causality.
• A clinician outcome supports clinical interpretation within scope; L2 does not reinterpret it.
• AI may identify discrepancies between promised and observed outcomes; it cannot decide success or close the need.

Potential Proxi work. Bring forward every open/recent commitment; show exact endpoint evidence; ask tailored result questions; capture attributable outcomes; compare expected and reported results; reopen or redirect the appropriate work; and preserve a truthful no-benefit or unknown outcome.

Human role. L2 performs the person-centered follow-up inside Task 14's change review and supports the next choice. L1 does not repeat the navigator screen. Separate U.S. clinical staff enters only for clinical interpretation, worsening symptoms, safety, or treatment-response questions.

Provisional human minutes (low / typical / high). L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 1 / 2 / 4 minutes inside Task 14's 2/4/6 allocation, not additive. Separate U.S. clinical: 0 / 0 / event-driven for a distinct clinical question.

Completion evidence. Prior need/commitment identity; endpoint evidence; attributable reported receipt/use/result; helped/partially helped/did not help/unknown/declined status; remaining barrier; beneficiary/caregiver choice; reopened or revised child work; and clinical route when triggered.

What does not prove completion. Referral closure, appointment attendance, service-start record, content delivery, caregiver attendance, no complaint, prior “resolved” status, claim, or AI-inferred benefit.

Edge cases/open decisions/minimum tests. Existing: X-004, X-005, X-015; CC-E08, CC-E25, CC-E26, CC-E34, CC-E36; MED-E26, MED-E36, MED-E42. Minimum tests: service started and helped; service started but did not help; referral never produced service; beneficiary declined; caregiver and beneficiary report differently; clinical symptom worsened; late corrected endpoint evidence; prior task closed in error; several open needs.

36. Respond to contact initiated between scheduled check-ins#

When. Trigger whenever the beneficiary, permitted caregiver/representative, or outside party contacts Proxi between planned contacts with a question, change, request, or concern. Do not make the person wait for the next scheduled contact. Apply identity, participation, permission, location, and urgent-routing boundaries before disclosing or acting.

What and how much. Acknowledge and classify the source-faithful request, provide approved administrative information immediately when appropriate, connect relationship-sensitive or substantive nonclinical issues to L2, and connect clinical/medical issues to the separate U.S. clinical route. Create every distinct downstream obligation and give a truthful response/return expectation. A between-contact interaction counts toward cadence only if it independently satisfies all Task 31 facts; it is not counted automatically. When it does qualify, cancel the redundant scheduled contact unless a separate required or beneficiary-requested purpose remains.

Data/proof boundaries.

• The authenticated sender's statement establishes that sender's report/request, not another person's clinical status, authority, or consent.
• Message and call timestamps prove receipt/attempts; automated acknowledgement does not prove substantive response.
• Existing records may provide context but do not determine current urgency or answer a new clinical question.
• AI may organize and draft approved content; it cannot replace the human navigator's substantive response, triage, or make clinical or legal decisions.

Potential Proxi work. Receive consented messages/calls; authenticate where supported; retrieve current context; separate administrative, navigator, clinical, disclosure, and urgent issues; provide approved administrative facts; apply the approved identity, permission, recipient, purpose, content, and restriction rule; alert L1, L2, or the authorized clinician only for that role's actual work; hold an uncovered legal-authority question for Healthcare Legal Counsel; track acceptance/response; and link the episode to Task 31 only for independent evidence review.

Human role. L1 Philippines provides 24/7 first answer, captures identity/callback and verbatim concern, gives approved nonclinical information, and performs warm transfer. L2 Puerto Rico gives the substantive nonclinical navigator response, clarifies missing relationship or permission facts, and owns relationship-sensitive follow-through. The separate U.S. beneficiary-location-authorized clinical workforce handles all clinical or medical work. Software enforces the approved disclosure rule; only a complete verified fact pattern exposing a genuine unresolved legal-authority question routes to Healthcare Legal Counsel while the affected act remains held.

Provisional human minutes (low / typical / high) per inbound episode. L1 Philippines: 2 / 4 / 8. L2 Puerto Rico nonclinical navigator: 0 / 5 / 15, depending on whether a substantive navigator response is needed. Separate U.S. clinical: 0 / 5 / 20 only when clinically triggered. These are route-specific; they are not automatically summed for a simple administrative question.

Completion evidence. Sender identity/role and permission basis; received time/channel; exact concern; current location when clinical work is implicated; route; L1/L2/clinical response; receiving acknowledgement; substantive answer or truthful pending plan; every child obligation; urgent outcome if applicable; and separate Task 31 result if cadence credit is considered.

What does not prove completion. Auto-reply, chatbot answer, message read, queued task, alert, voicemail, warm-transfer attempt, L1 answer to a clinical question, navigator acknowledgement without substantive response, or assuming every inbound contact satisfies cadence.

Edge cases/open decisions/minimum tests. Existing: X-001, X-002, X-007, X-008, X-013, X-014; CC-E10–CC-E13, CC-E35, CC-E41, CC-E43; MED-E18, MED-E25, MED-E30, MED-E44; D-005, D-014, D-015, O-001, O-003, O-004, and O-011. Minimum tests: simple administrative question; relationship-sensitive request; clinical question; urgent concern; after-hours call; unauthorized caregiver; outside provider; wrong patient; several issues; failed warm transfer; substantive L2 contact that does/does not satisfy Task 31.

37. Review the upcoming contact workload#

Separate due-work views. Display the Assessment-owned annual and reassessment due-or-overdue list beside the Ongoing contact workload without merging their owners or closure criteria.

When. Review on the program's approved operating cadence and more often when staffing, tier reports, holidays, outages, interpreter availability, or a cluster of failed contacts threatens coverage. This is portfolio operations, not a beneficiary contact.

What and how much. Review every aligned beneficiary's next due window, booked opportunity, last qualifying contact, failed attempts, current tier/effective date, navigator/backup capacity, modality/accommodation readiness, open repeated-unreachable episode, and any authorized above-minimum cadence. Assign an owner and corrective action for every gap without using AI to decide clinical urgency.

Data/proof boundaries.

• Current alignment/tier and Task 31 records support due-window calculations only for their effective periods.
• Calendars and staffing rosters prove bookings/capacity entries, not future attendance or navigator eligibility unless current evidence is linked.
• A clinical recommendation may establish an earlier-contact need within its scope; a model risk score may only suggest review.
• A dashboard is presentation. It does not prove that a gap was reviewed, assigned, or corrected.

Potential Proxi work. Produce the complete due-work list; calculate current windows from approved source facts; identify unbooked, late, failed, inaccessible, or uncovered cases; separate minimum cadence from authorized extra contact; propose staffing/scheduling repairs; and track owner/action completion without dropping low-ranked cases.

Human role. The Puerto Rico navigator lead resolves navigator assignment, capacity, or relationship exceptions. HR/Workforce Administration resolves roster, leave, credential-file, or workforce-capacity gaps outside the beneficiary episode. L2 is involved only for a relationship-sensitive beneficiary issue or accepted assignment change. L1 and separate U.S. clinical staff are not required for the routine review; clinical staff enters only through a beneficiary-specific clinical decision already owned elsewhere.

Provisional human minutes (low / typical / high) per workload-review cycle. L1 Philippines: 0 / 0 / 0. L2 Puerto Rico nonclinical navigator: 0 / 0 / 5 only for a relationship exception. Separate U.S. clinical: 0 / 0 / 0. Puerto Rico navigator lead: 0 / 5 / 15 depending on actual assignment exceptions. HR/Workforce time is portfolio labor and measured separately when a staffing or credential-file gap exists.

Completion evidence. Review timestamp and accountable operator; complete aligned population denominator; current source dates; due windows; booked/failed/coverage/accommodation status; every gap; owner and corrective action; staffing or reassignment result; unresolved risk; and confirmation that no item was dropped because of priority ranking.

What does not prove completion. Dashboard generation, sorted list, alert send, no red indicator, average caseload, calendar utilization, staffing roster, or a manager opening the page without assigning and resolving actual exceptions.

Edge cases/open decisions/minimum tests. Existing: X-003 through X-007, X-011 through X-014; O-004, O-039, O-040, and O-PA-001; proposed contact-clock and covering-navigator decisions. Minimum tests: no gaps; one unbooked due contact; navigator credential expires; tier report changes; holiday/DST; interpreter shortage; multiple failed contacts; clinical extra cadence; outage; workload list omits a beneficiary; staffing repair still leaves no usable appointment.

Cross-pillar continuity feeds#

This SOP participates in cross-pillar episodes under 25_GUIDE_Eight_Pillar_Service_Integration.md. Use the shared no-repeat packet, preparation-before-judgment route, receiving-owner acceptance rule, and count-once labor rule. Ongoing Monitoring and Support receives service attribution only for a qualifying human navigator interaction that uses the current baseline and plan to furnish a specific continuity or problem-solving result and leaves an attributable next action or truthful no-further-help result.

The latest accepted Assessment is the baseline; the current Care Plan is the action frame. Ongoing records what changed or happened, follows the exact promised return, and sends material baseline questions back for focused Assessment instead of performing a shadow reassessment. In a stable required contact, the real service is human continuity: invite the person's current priority, address a relevant goal, action, question, or chosen topic, and leave one attributable result or help route. Unasked domains remain unasked.