# Codebook — `references_verified_v1.2.csv`

210 records. One row per source.

| Column | Definition |
|---|---|
| `id` | Internal database key, grouped by working section (A = physical quantities and classic theory; B = tissue dielectric properties; C–E = membrane biophysics, development, wound fields; F–H = measurement, safety, electrotherapy evidence; J = comparative and methodological; K–M = added during later phases). The prefix records when and why a source entered the database; it carries no judgement of quality. |
| `section` | The letter group above. |
| `authors` | Author list as indexed. |
| `year` | Year of publication as indexed. |
| `title` | Title as indexed. |
| `journal`, `volume`, `pages` | As indexed. Where a field is absent from the indexed record it says so and is not filled in from another source. |
| `pmid` | PubMed identifier where one exists. Empty for books, standards, guidelines and some pre-1950 papers. Five records (A11, F13, F14, I16, I17) have neither a PMID nor a DOI; see Verification levels below. |
| `doi` | DOI where one exists. |
| `type` | Study design as the source describes itself — for example randomised controlled trial, Cochrane review, meta-analysis, in vitro study, classic theory, standard, guideline. |
| `species` | **Species or preparation.** Recorded for every quantitative source, because a value measured in rat hippocampal slice, guinea-pig skin or human cortex is not interchangeable with the others. Where a review reports a value without a traceable primary species, the column says so. |
| `key_finding` | The finding as reported by the source, including confidence intervals, sample sizes and null results where the source gives them. Unfavourable findings are recorded in the same terms as favourable ones. |
| `use` | Where and how the source is used in the manuscript. |
| `paper1_ref_number` | The numbered reference in Paper 1 (manuscript v1.2.1) that this record corresponds to. Empty for background sources read but not cited, and for the records compiled for Papers 2 and 6 (prefixes K, L and M01–M09), which Paper 1 does not cite. Semicolon-separated where one source is cited under more than one number. |

## Verification levels

Two different claims are made about these records, and they should not be conflated.

1. **Bibliographic verification — performed for every record.** Authors, title, journal, year, volume, pages and identifiers were checked against the PubMed or Crossref record, or, for books, standards and guidelines, the publisher's or issuing body's listing.
2. **Content verification — performed at abstract level.** Each quantitative value and each characterisation of a finding was checked against the abstract, and against the full text only where the full text was openly accessible. **It was not checked against the full text of paywalled sources.**

A value in `key_finding` drawn from a paywalled source should therefore be treated as abstract-level until the manuscript's pre-submission full-text reading condition is discharged.

## What is deliberately not in this file

No quality score, no evidence grade of the compiler's own, and no impact metric. Where the manuscript reports a certainty rating, it is the rating the cited systematic review itself assigned, and the manuscript marks which ratings are source-assigned and which are the author's own appraisal.