Non-clinical home care, three operational scenarios

In home care, the schedule, the care plan, and the shift notes each describe what's supposed to be happening in a client's home. They rarely all agree on what actually is happening, and reconciling the gaps is where a lot of the operational work goes.

Scenario 1 of 3

The 6am Call-Off

A caregiver texts the scheduler at 6:02am that she's sick, an hour before a client's medication shift. The scheduler manually texts backups; whoever replies first wins. No audit trail, no fairness rotation, and the family only finds out when a new face shows up at 7am.

Existing products (Careswitch, AxisCare) already handle some version of this, so what I'd sketch is really a way to think about it: pull eligible caregivers from WellSky, rank them by client history, distance, certification, and fairness rotation, then fan out an SMS sequence in priority order and log the whole thing.

Unfilled shifts · next 4 hours

3

$624 family-visible impact if coverage slips · Stuck Triage Agent ranks backups by prior shifts + distance + fairness

58 min to start

Mr. Henderson · 7:00am–11:00am

Original caregiver Maria Lopez called off 6:02am · flu

Care needs
medication reminder breakfast prep light housekeeping
Authorized
4 hrs/day · private pay
System of record
WellSky Personal Care

Ranked backups by prior shifts with this client · distance

  1. 1

    Yolanda Torres recommended

    4 prior shifts · 3.2 mi · 4 prior shifts with Mr. Henderson · no schedule conflict

  2. 2

    Brittany Williams

    1 prior shift · 5.8 mi · 1 prior shift with Mr. Henderson · no schedule conflict

  3. 3

    Crystal Hudson

    0 prior shifts · 2.1 mi · closest geographic

  4. 4

    Marcus Johnson

    0 prior shifts · 7.4 mi

Coverage recommendation

Yolanda Torres has the strongest match (4 prior shifts, 3.2mi, no overlap with her Tuesday morning shift). Auto-fanout sequenced: Yolanda Torres → Brittany Williams → Crystal Hudson → Marcus Johnson, 90s between sends, first reply wins.

Scenario 2 of 3

The EVV Triple Mismatch

A caregiver clocks in at 8:14am via Sandata EVV, the WellSky schedule says 8:00 to 12:00, and the WellSky care plan authorizes 4 hours of bathing and meal prep. Three sources, three numbers, nobody reconciles. Viventium pays 3.95 hours, QuickBooks invoices 4.0, and when the drift goes the other way the LTC insurance carrier denies the extra time.

My sketch is a three-way reconciler that pulls each visit from all three sources and compares them. When they disagree on hours or tasks, it records the mismatch, calculates the dollar impact at the carrier's reimbursement rate, and surfaces recurring patterns on the same caregiver-client pair so the agency has evidence to request a care plan update.

EVV reconciliation · last 30 days

−$18,649

1247 mismatches across 1247 visits · Sandata EVV × WellSky schedule × WellSky care plan

1247 drift events

Mrs. Ada Robinson

Visit 2026-04-03 · caregiver Yolanda Torres

Sandata EVV GPS clock-in WellSky Schedule what was planned WellSky Care Plan what was authorized
Start 08:14 08:00 08:00
End 12:42 12:00 -
Hours 4.47 4.0 4.0 max
Tasks bath, meal, + laundry bath, meal bath, meal

Drift: 28 minutes over authorization + unauthorized task

Impact: 0.47 hrs over the daily cap → blocks $14.10 of the per-diem reimbursement

The shape of a three-way reconciliation query (SQL, 45 lines)
The shape of a three-way reconciliation query
-- one row per visit showing hours drift between EVV,
-- schedule, and the care plan authorization. CTE avoids
-- the alias-in-same-SELECT problem in Postgres.
WITH visit_hours AS (
  SELECT
    v.visit_id,
    v.client_id,
    v.caregiver_id,
    ROUND(EXTRACT(EPOCH FROM
      (e.clock_out - e.clock_in)) / 3600.0, 2
    )                             AS hours_actual,
    ROUND(EXTRACT(EPOCH FROM
      (s.scheduled_end - s.scheduled_start)) / 3600.0, 2
    )                             AS hours_scheduled,
    a.authorized_hours_per_day    AS hours_authorized,
    c.carrier_id
  FROM visits v
  JOIN evv_clockins e             USING (visit_id)
  JOIN scheduled_shifts s         USING (visit_id)
  JOIN care_plan_authorizations a USING (client_id)
  WHERE v.visit_date
    BETWEEN current_date - interval '30 days'
        AND current_date
)
SELECT
  vh.visit_id,
  cl.client_name,
  cg.caregiver_name,
  vh.hours_actual,
  vh.hours_scheduled,
  vh.hours_authorized,
  vh.hours_actual - LEAST(
    vh.hours_scheduled, vh.hours_authorized
  )                               AS drift_hours,
  (vh.hours_actual - LEAST(
    vh.hours_scheduled, vh.hours_authorized
  )) * r.rate                     AS dollar_drift
FROM visit_hours vh
JOIN clients cl     USING (client_id)
JOIN caregivers cg  USING (caregiver_id)
JOIN carrier_reimbursement_rates r
  ON r.carrier_id = vh.carrier_id
WHERE ABS(vh.hours_actual
          - vh.hours_scheduled) > 0.1
ORDER BY dollar_drift DESC;

Scenario 3 of 3

The Stale Care Plan

A caregiver's shift notes 12 days ago say "client having more trouble getting out of bed, asked for help to sit up before transferring." The WellSky care plan still reads "ambulatory, minimal assistance, independent transfers." Nobody has flagged it. The agency has been absorbing unmatched care for two weeks, and the client is at clinical risk.

The shape I'd start with is a daily job that reads recent shift notes and uses an LLM to watch for signals that a client's needs have changed since the care plan was last updated. When the pattern holds across multiple shifts, it flags the client and collects the supporting quotes. Not to draft the reauth itself (LTC carriers send their own nurse to do that assessment, not the agency), but to give the clinical liaison evidence for an off-cycle reassessment.

Clients with acuity drift

15

$686/mo estimated margin leak · 1 flagged for clinical review

Inferred from shift notes

Mr. Henderson

Care plan last updated 47 days ago · John Hancock LTC

Shift-note signals (last 9 days)

  1. “client having more trouble getting out of bed, asked me to help him sit up before transferring. Took longer this morning.”
  2. “needed two-person assist for transfer to chair. Unusual for him.”
  3. “wife mentioned he fell last weekend before bed. No visible injury.”
  4. “took 25 min for what used to be a 10 min transfer. He apologized.”

Current care plan

ambulatory, minimal assist, independent transfers

Inferred actual

needs transfer assist, fall risk, two-person assist required

Reassessment recommendation

This client needs an acuity reassessment. The shift-note pattern over 9 days shows a consistent increase in transfer assistance time. Recommended: schedule a reassessment, then submit a reauthorization to John Hancock LTC requesting an increase from 4 to 5 hours per day. Reauth packets typically require recent assessment notes plus the requesting clinician's signature.