Method selector
The decision tree of Chapter 18, as a questionnaire. It is entered with a decision that must be made and a body of evidence already assembled; it presumes the problem has been framed well enough to name the alternatives. Every question before the last two is about the problem, not about the method.
NNT / NNH calculator
Absolute effect on a per-patient scale (Chapter 12). Intervals follow Altman's rules: when the risk-difference interval contains zero the NNT scale contains infinity in its middle, and both arms are printed. A signed NNT is never produced — direction is carried by the label, NNTB or NNTH, and never by a minus sign.
A · Proportions — events and n in each arm
B · Incidence rates — events and person-years in each arm
Effects table builder
The EMA column set piloted under work package 5 of the benefit–risk methodology project (Chapter 9). One page, or the row set is wrong. The caption must state the data source per side, the population, the comparator and the rule by which rows are ordered — ordering is the commonest route by which weights enter a display that claims to have none.
| Order | Effect | Short description | Unit | Comparator | Product | Uncertainties / strength of evidence | References |
|---|
MCDA workbench
A linear-additive value model with swing weights (Chapter 14). Partial value vi = 100 × (xi − wi) / (bi − wi); overall value V(a) = Σ ωi vi(a); and the display that matters, V(d) − V(c) = Σ ωi [vi(d) − vi(c)]. The scale range is part of the criterion: a weight quoted without its range has stated nothing.
| Order | Criterion | Side | Direction | Worst | Best | x(A) | x(B) | x(C) | Swing pts | ω | v(A) | v(B) | v(C) | ω·Δ |
|---|
One-way sensitivity — where the conclusion reverses
For each criterion in turn, the raw swing points at which the displayed pairwise difference passes through zero, all other criteria held at their base-case points. "No attainable weight" is a result and is recorded as one. This, and not the base case, is the part of a quantitative analysis that earns its place.
| Criterion | Base-case swing points | Reverses at | Reading |
|---|
BRAT worksheet
Benefits and risks kept deliberately separate rather than combined into a score (Chapter 11). The value tree fixes the criterion set; the Key Benefit-Risk Summary table puts every retained criterion on a stated denominator; the forest plot displays them in one panel with a vertical line at zero. Nothing here supplies an exchange rate, and that is the point.
Step 2 · Value tree
Favourable effects
Unfavourable effects
Step 6 · Key Benefit-Risk Summary table
| Order | Criterion | Side | Basis | Events, product | Denominator, product | Events, comparator | Denominator, comparator | Rank | Estimate (95% CI) | Denominator |
|---|
PrOACT-URL worksheet
A structured qualitative assessment (Chapter 10). The framework's value is the order of the questions. Fill it top to bottom, and do not let Alternatives be populated after Consequences. EMA renders the same content as twelve operational steps by inserting Criteria explicitly between Objectives and Alternatives.
Conclusion wording library
Modular sentence patterns by conclusion type (Chapter 33, instrument 15). Four rules apply to all of them: state the direction, the population, the conditions and what would change the conclusion — all four, always. Read the finished sentence once with the population clause covered by a hand; if it still reads as a complete claim about the product, rewrite it.
Assessment summary
Document header
Head every artefact. An undated benefit–risk table is not evidence of anything, because nobody can tell which version of the evidence base it describes.