# Screening codebook: which records may report a randomized trial of perioperative ketamine with a depression outcome

Each record is a journal article or preprint found by searching titles and abstracts for ketamine or esketamine, a depression measure, randomization, and surgery, anesthesia, or childbirth. You see its title, abstract, journal, year, and publication types. Decide, for each record, whether its full text should be read to see if the trial it reports belongs in the review.

## What belongs

A record is a **candidate** when it may report the results of a randomized trial in which:

1. **Ketamine is the randomized intervention.** Participants were randomized to ketamine or esketamine (racemic ketamine, S-ketamine, or arketamine; any dose, route, or schedule) or to a control, alone or among other arms.
2. **There is an inert control.** At least one arm received saline, a placebo, or no study drug, with otherwise the same care. Trials whose only comparators are other active drugs (dexmedetomidine, opioids, midazolam, and so on) don't belong, unless they also have an inert arm.
3. **It is given around a procedure or childbirth.** The drug was given before, during, or after surgery, another procedure done under general anesthesia, regional anesthesia, or sedation (endoscopy, for example), or labor or delivery. Electroconvulsive therapy doesn't count.
4. **Depression is measured after it.** The trial measured depressive symptoms after the procedure: a depression scale (EPDS, HAMD, MADRS, PHQ, BDI, SDS, HADS, the depression scale of POMS, or similar) or the proportion of participants with depression.

When the abstract leaves any of these open, keep the record as a candidate: the full text decides.

A record is a **timing comparison** when it may report a randomized trial with a depression outcome, as above, whose arms all receive ketamine or esketamine but at different times (for example, during general anesthesia against after waking), with or without an inert arm. If it also has an inert arm, it is a candidate.

## What doesn't

- `protocol`: a trial protocol, a registry entry, or a design paper, with no results.
- `psychiatric`: a trial of ketamine or esketamine for depression, suicidality, or another psychiatric condition outside any procedure or childbirth (clinic infusions, nasal sprays, and so on).
- `ect`: a trial involving electroconvulsive therapy.
- `active_only`: a randomized trial around a procedure or childbirth whose only comparators are active drugs.
- `no_depression_outcome`: ketamine around a procedure or childbirth, but no depression measure is mentioned, and none is likely (for example, a trial reporting only pain and nausea). If an abstract mentions mood, anxiety, sleep, or quality of recovery, keep it as a candidate, since depression is often measured beside them.
- `not_randomized`: observational, retrospective, single-arm, or other non-randomized designs.
- `not_trial`: reviews, meta-analyses, letters, editorials, case reports, animal or laboratory studies, and anything else that isn't a report of a trial in people.
- `other`: anything else (say what in `note`).

## Output

One JSON object per record, in the order given, as a JSON array written to the file named in your task:

```json
{"id": "pmid:12345678", "decision": "candidate", "note": ""}
```

- `decision`: `candidate`, `timing_comparison`, or one of the reasons above.
- `note`: anything a second reader would need to follow your decision, especially when you kept a record because the abstract left something open. Keep it short.

Decide from the record alone. Don't look up full texts, other records, or anyone else's screening.