Ambulance administrative handovers and status logs

Learn in your browser · Chapter 01 of 06

An administrative handover preserves the next person’s ability to act on unresolved work. A short, confident paragraph is not enough if it hides a task or changes what a status means.

Merefield is an invented ambulance station used only as a setting for back-office learning. The incoming administrator needs a 10:00 UTC snapshot on 12 October 2026. Six tasks concern a training-room projector ticket, stationery delivery, a training booking, an office-cleaning record, a facilities door-repair ticket and a noticeboard reprint. They do not concern patients, dispatch, staffing cover, clinical cleaning standards or the fitness of a room, vehicle or item for use. A sentence about a closed facilities ticket must never become a claim that a door is safe. Keep a separate boundary between administrative record status and the real-world condition it describes.

The authoritative teaching inventory is tasks.json: H01 to H06, their starting owners and due times. evidence.json contains reports and administrative events. A task can be unresolved even when nobody calls it urgent. The incoming person needs its ID, reviewed status, current owner, deadline, source IDs and next action. That combination lets them return to evidence and challenge the draft. A summary that contains five excellent rows still fails completeness when the agreed population contains six. Read-only AI access does not remove this risk: a misleading draft can affect what people notice and follow up.

State the scope before choosing a tool. In this exercise the permitted output is a draft with the same six task identities and a questions list. It cannot change source records, send messages, accept a transfer or decide readiness. The reviewer can reject or correct the draft using the evidence. If the records or reviewer are unavailable, retain a visible unresolved status and use the authorised manual approach. Do not make absence of a review an automatic approval. For the learning case, a table written by hand is a valid solution; the supplied checker is optional.

The NHS records code introduction, checked 12 October 2026, describes its England scope, including organisations working within or under contract to the NHS, and records in any medium. That is a source-context statement, not evidence that this fictional workflow satisfies an organisation’s requirements. The UK Government AI Playbook, published 10 February 2025, discusses human control, testing and issue reporting. We use those ideas to explain the need for a reviewable draft. Neither source specifies our six task IDs or grants authority to operate an ambulance service.

Learn in your browser · Chapter 02 of 06

A report is evidence of what someone recorded. Verification is a separate act with a stated scope. Conflicting or old reports need visible treatment, even if one was written later.

The snapshot cutoff is 2026-10-12T10:00:00Z. A record belongs in this snapshot only if both observed_at and recorded_at are no later than that cutoff. A note observed at 09:55 but entered at 10:10 was not available to the 10:00 handover. Keep it for a subsequent correction rather than silently inserting it into the original view. UTC is used only to avoid ambiguity in this fictional dataset. Real organisations must define their own clock, timezone and late-entry conventions. The cutoff describes which evidence was available; it does not prove that its contents are true.

The exercise uses four reviewed status labels. COMPLETE means the specific administrative closure criteria and review are recorded. OPEN means known work remains, including work reopened after a closure. CONFLICT means eligible sources make incompatible unresolved claims. VERIFY means a report lacks the required current verification. These are local learning labels. “Reported done” alone is never COMPLETE. A 120-minute freshness flag is also an invented exercise rule, not a clinical or industry interval. At the cutoff, evidence older than 120 minutes is flagged; exactly 120 minutes is within the boundary. Freshness describes age, not accuracy.

H01 has r01 reporting the projector ticket done at 09:20 and r02 reporting the fault still open at 09:25. There is no recorded resolution. Both are eligible; the later timestamp is not permission to discard the first. Record CONFLICT and retain both IDs. H02 is different: r03 records receipt of 36 stationery packs against the 36-pack administrative order, and r04 records the designated fictional review of that receipt. Its administrative task is COMPLETE. This does not certify product quality or authorise distribution for a clinical purpose.

H04 has a report observed at 15:00 on 11 October but recorded at 09:00 on 12 October. At the 10:00 cutoff its observation is 19 hours, or 1, 140 minutes, old. The recent entry time does not make the observation fresh. Its row is VERIFY and asks for a current administrative check. H05 has r09, an open repair-ticket update available at the cutoff. r10 says done and was observed at 09:55, but recorded at 10:10. The original snapshot uses r09 and stays OPEN. A later revision can consider r10 with its limits; it cannot rewrite what was known at 10:00.

The NHS records code section on organising records describes reliability, integrity, usability and the importance of context and identifiable alterations. Our practical interpretation is to keep source IDs and both times beside the draft, and record corrections visibly. This is a learning method derived from those general principles. It is not an assertion that a JSON file, timestamp or checksum authenticates the reporter or establishes the truth of a job-completion statement.

Learn in your browser · Chapter 03 of 06

A useful prompt defines the source set, row population, fields and unresolved questions. Verification begins with coverage, then examines the meaning of every changed field.

A weak request such as “make the handover concise and action-ready” rewards a fluent narrative without requiring an evidence trail. It can encourage the draft to turn tentative words into completion, merge jobs or omit an awkward row. Start with the six-item inventory, the cutoff and the permitted output schema. Supply only the fictional evidence. Require one row for each task and no extra tasks. Ask for a separate questions list when evidence conflicts or is missing. The draft has no write tools and cannot appoint owners or mark jobs complete.

A stronger exercise brief is: “Use tasks H01-H06 and eligible evidence available at 10:00 UTC. Return a draft table with task_id, status, owner, due_at, evidence_ids and next_action. Preserve every ID. Distinguish report from verification, receipt from accepted transfer and ticket status from physical safety. List unresolved questions. Do not fill a missing field from general knowledge, apply a later record to this cutoff or execute an action.” This narrows the task; it does not establish that the output is correct. Compare the result with sources even when every column is populated.

The offline checker is deliberately a structured-answer comparator, not an AI model or an operational status engine. It uses a separately authored expected.json teaching key. It checks against exactly the six known task IDs, no duplicate rows, the declared cutoff and the six-field row schema. Five fields per task are checked: status, owner, due_at, evidence_ids and next_action. task_id identifies the row rather than earning a point. Evidence ID order is irrelevant, but duplicate IDs are rejected. Next actions are short controlled labels for comparison; learners should also explain them in ordinary language on the worksheet.

There are 30 field comparisons in the complete reference. A missing task receives zero for its five fields and a visible missing-row defect. A wrong status, owner, deadline, source set or action produces a field defect. Structural errors, including duplicate IDs or unknown fields, reject the submitted evidence instead of manufacturing a clean score. The checker does not inspect a live system, authenticate a person or grade the free-text explanation. Matching its key supports a narrow claim about this exercise only. A person must still assess whether the teaching key and reasoning fit the source records.

Learn in your browser · Chapter 04 of 06

Acknowledging a message does not prove acceptance of a task. Accepting ownership does not prove completion. These transitions need different evidence and different language.

The starting owner comes from the task register. A transfer offer proposes a change; a receipt says the proposed recipient has seen the message; an acceptance says that recipient has accepted responsibility for the named task. Our fictional rule requires an explicit acceptance matching the task, offered recipient and offer ID, available by the cutoff. Until then, the starting owner remains visible with “transfer pending”. This is an authored exercise convention, not a universal employment or ambulance-service policy. An organisation must define its own real responsibility and escalation arrangements.

H03 is the training booking, initially owned by Mara. r05 says the booking is still provisional. r06 offers H03 to Lena, and r07 records only receipt of that offer. The reviewed row remains OPEN, owned by Mara, due 09:45, using r05-r07. At 10:00 it is overdue under our strict earlier-than-cutoff rule. “Lena has it” is an unsupported ownership assertion. The next action label confirm-booking means obtain a booking decision through the responsible administrative route, while retaining the transfer question. It does not authorise the learner to contact a real training provider.

H06 begins with Theo. r11 reports the noticeboard reprint finished; r12 records review of that initial completion. r13 offers ownership to Inez, and r14 explicitly accepts that offer at 09:40. r15 reopens the job at 09:50 because the supplied poster version was wrong. The owner is now Inez, but the status is OPEN. Preserve all five source IDs to explain the transition. If you recorded only “accepted at 09:40”, you could mistake acceptance for closure. If you recorded only the original completion, you would conceal the later work.

Make the handover conversation testable. Ask the receiving person to identify unresolved tasks, due times and questions, then record what they actually acknowledged. If they only confirm receipt, label it receipt. If a proposed transfer is rejected or unanswered, keep the original owner and escalate according to the agreed administrative policy. A read receipt, shared mailbox access or an AI-generated “understood” message does not establish a human acceptance. In this offline lab all role labels are invented and unauthenticated. They show the evidence distinction, not verified identity.

Learn in your browser · Chapter 05 of 06

A correction should let a later reader understand both the original snapshot and the changed conclusion. Metrics must reveal omissions and unresolved work rather than reward confident closures.

Do not overwrite the original handover after finding late evidence or a drafting error. Retain its version and cutoff, create a labelled correction, state which row or field changed and why, identify the source, and record who reviewed the correction and when. Distinguish a transcription repair from new information. Replacing Lena with Mara in the original 10:00 H03 row corrects an unsupported assertion based on evidence already available. Considering r10 for H05 in a later snapshot introduces a record that was unavailable to the original view. Both can be useful, but they answer different historical questions.

A checksum helps compare the exact bytes of an exported file with a reference. It does not prove that the file is complete, the claims are true, the reviewer is genuine or the evidence was available at the stated cutoff. Keep the method and the manifest with the packet, but do not claim that the result prevents all alteration. Record retention and deletion decisions remain outside this lab. The NHS records code provides a wider records-management context; our exercise never sets a real retention period or authorises destruction of organisational records.

Use explicit populations. At 10:00 the six-task key contains one COMPLETE, three OPEN, one CONFLICT and one VERIFY. Five are unresolved. H01 is due 09:30 and H03 is due 09:45; both are unresolved and overdue. H06 is due exactly 10:00 and is not overdue under the local strict comparison. Therefore unresolved share is 5/6, about 83.33%. Overdue among unresolved is 2/5, or 40%. Overdue among all tasks would instead be 2/6, about 33.33%. State which denominator you use. These figures describe the authored case, not an ambulance service benchmark.

Measure the drafting task separately from task completion. The flawed draft includes five of six tasks, so row coverage is 5/6. Its field comparison is 21/30. The corrected draft has six of six rows and 30/30 matching fields, but still only one completed task. Improving the handover does not complete the work it describes. Nor does a good score validate a live model: this fixture uses saved authored outputs. For a future authorised trial, compare repeated drafts against the same reference, include difficult unresolved cases and record preparation, review and repair effort as well as speed.

Learn in your browser · Chapter 06 of 06

Finish with a packet another reader can challenge: the original inputs, your draft, the corrections, the worked comparison and a narrow statement of what has been checked.

The portable ZIP contains charter.json, tasks.json, evidence.json, expected.json, flawed-handover.json, corrected-handover.json, checker.py, test_checker.py, worksheet.txt, activities-and-answers.json, README.txt and MANIFEST.json. The paper route is complete without a model or Python. Read the evidence cards, write a six-row handover and answer the worksheet. The optional script compares a submitted structured draft with the teaching key and prints a report. It makes no network request, sends no message and edits no evidence. Inspect the code before running it. If a file contains real data or unexpected external actions, stop using it and resolve the boundary with the owner.

Run the flawed sample first and explain every defect before comparing the corrected sample. Then change one thing in a copy: remove H04, change H03’s owner, add an unknown task, duplicate a source ID, change a deadline or substitute late r10. Predict whether this is a field mismatch or a structural rejection, run the comparison and preserve your prediction with the result. Do not modify expected.json merely to make a poor answer pass. A teaching-key change needs its own reason and different-author review, just as a changed task record would need appropriate organisational authority.

For the capstone, give another reader your draft, evidence and correction note without the worked handover at first. Ask them to identify the population, cutoff, unresolved tasks, ownership questions, overdue denominator and excluded late record. Record disagreements and how you resolved them. A check mark saying the exercise was completed is not evidence that the defects were correctly found. Evidence of learning would include the specific explanation and corrected fields. The written capstone is ungraded here. A future knowledge certificate would assess the configured multiple-choice exam only, not practical competence or authority to work in ambulance operations.

The planning estimate is 480 minutes: 75 reading, 210 main activities including the 60-minute capstone, 120 chapter-review questions, 30 reflection and 45 knowledge assessment. This is unpiloted, not a measured completion time. The bank contains 40 mapped multiple-choice items and six separate ungraded written practices. A future 20-question knowledge sitting uses a pass mark above 80%; with 20 equally weighted questions, 16/20 is 80% and fails, while 17/20 is 85% and passes. A source or bank file alone does not mean that the online exam or certificate journey is available.

Mickai Training Academy is educational and non-accredited. This course does not establish clinical competence, professional approval, legal compliance, identity assurance or permission to change a real record. Keep your PDF or worksheet to print and write in when available. The useful result is a repeatable way to show what remains unresolved and why a human conclusion is justified, with enough evidence for another person to disagree. Real adoption would require its own organisational scope, records arrangements, appropriate expertise, tested system and accountable decision makers.