Project worksheets for medical makers

Brief — the one-page brief to complete before building anything and have corrected by your collaborator. The need-statement checker flags words that name a solution ("app", "motorised", "model", "silicone"…), framing like "we need a …", and statements too broad to assess ("understand"). The book’s own examples are the test cases: "We need a motorised elbow model with an app" is flagged three ways; "Beginning learners need a way to distinguish muscle force generation from tendon force transmission…" passes.

Evidence — observation, interpretation (with one alternative) and assumption (with how you will check it), kept apart. It warns if you type something that might identify a person: a title and name, a record number, an age, a bed number.

Task map — trigger, preparation, action, exception path, cleanup, handoff.

Constraints — every constraint area from Chapter 18, each marked known, assumed or unknown, with how it will be verified; the counts sit at the top.

Hazards — hazard → sequence of events → hazardous situation → possible harm → control, with the control’s level in the hierarchy (1 inherently safe design, 2 protective measures, 3 information for safety) and how its effectiveness is checked. Rows are ranked by your judgement (move them with ↑ ↓) and the basis of the ranking is written down. There are deliberately no scores: no severity × probability, no risk priority number. If a row looks like a score (an RPN, "probability 3", "20 %", "4/5") the sheet says so. It also asks for at least one non-physical harm, a reason whenever the control is level 2 or 3, and the new questions each control raises.

Requirements — the needs list, then one card per requirement: statement, acceptance criterion, agreed with whom and when, how it will be checked, what a pass would and would not establish, the design feature that satisfies it, and where any number came from. The de-adjective helper flags words that cannot be checked — easy, durable, clear, portable, strong, reliable, safe, accurate, comfortable, intuitive and more — with the book’s workable rewrite.

Tests — the full test record, including prompts given, verbatim, and what the result does not establish. Mark each test verification, validation, or informal tryout (which the matrix does not count as validation). Leave the outcome "Not yet run" until it has been.

Versions — the version record, with a change impact panel: tick the earlier tests this change could invalidate and it shows which have been re-run on the new version, and which have not ("its result does not carry forward").

Traceability — need → requirement → design feature → verification → validation → latest record, built from the other sheets, with every gap named: no need, criterion not agreed, no test, validation only by informal tryout, evidence only from an earlier version, a criterion dated after its test, prompts not recorded.

Objective — the learning objective and its assessment, each with a Bloom level (a mismatch is flagged), a check for verbs you cannot observe (understand, appreciate, know, be aware of, be familiar with, grasp), and the reset-time cost: 90 s × 12 learners = 18 minutes of a session.

Bench result — fills in the book’s sentence ("V0.1 completed n of N cycles under … This does not establish …"); blank parts stay as visible [placeholders], and wording such as "mostly", "worked six times", "passed after fixing" or "validated" is questioned.

Materials — the material comparison, with no overall ranking.

Handover — the handover pack checklist, with a summary pulled from the other sheets.

The checkers are word lists, not understanding. A clean need statement can still be wrong, and a flagged word can be legitimate in context; the point is to make you look. They are English-only.

The identifier warning catches obvious patterns only. You remain responsible for recording no patient identifiers.

The traceability matrix treats the last entry on the Versions sheet as the current version and matches tests to requirements by ID.

Bloom levels suggested from verbs are a starting point: "explain why…" can sit at Analyse.

Everything stays in this browser until you save a file; clearing site data erases it. Keep the .json or Markdown export somewhere safe. There are no share links — projects are too big.

The example’s results are the book’s invented worked-example results (Chapter 19’s "failed at attempt 7 of 10"), not real test data.