# Design codebook

You are coding the design of randomized trials of azithromycin given to children at population scale. For each trial you get its design text only: the Methods of its reports (or, where no full text is open, the part of the abstract before its results), its protocol where one is published, and its registration. Code what the text says; don't infer what it doesn't, and don't look anything up. Work alone.

## Fields

For each trial, one JSON object:

- `trial`: the trial's ID as given.
- `eligible`: `true` if the trial is randomized, gives azithromycin to children at population scale regardless of illness (to every child in an age band of a community, to a whole community, or to every child or infant a programme or visit reaches), has a comparison group that got placebo, no treatment, delayed treatment, or another azithromycin schedule, and measures all-cause mortality among children; else `false`, with `exclusion_reason`.
- `unit`: what was randomized: `community` (village, cluster, health-area catchment, or other group), `household`, or `individual`.
- `units`: the number of units randomized to each arm, as an object `{"<arm label>": n}`, when the text gives it; else `null`.
- `countries`: list.
- `period`: `{"start": "YYYY-MM", "end": "YYYY-MM"}` of treatment and mortality follow-up, as precisely as the text gives.
- `arms`: a list, one per arm, each `{"label", "azithromycin_ages", "placebo_ages", "frequency", "rounds", "dose"}`: `azithromycin_ages` is the age band that got azithromycin in that arm (such as `"1-59 months"`, `"1-11 months"`, `"all ages >= 1 year"`, `"neonates 8-27 days"`), or `null` for none; `placebo_ages` likewise for placebo; `frequency` is `biannual`, `quarterly`, `annual`, `single`, or as the text says; `rounds` the number of distributions per child or community over the trial; `dose` as written (such as `"20 mg/kg"`).
- `control`: what the comparison arm got: `placebo`, `no_treatment`, `delayed_treatment`, or `other_schedule`.
- `masking`: `double` (participants and outcome assessors masked), `single`, `open`, or `unclear`.
- `strategy`: one of
  - `all_children_mda`: community randomized; azithromycin to every child in an age band covering 1 to 59 months (or wider within childhood) in treated communities, against placebo, no treatment, or delayed treatment;
  - `infant_only_mda`: community randomized; azithromycin to infants (an age band within 1 to 11 months) only, older children getting placebo or nothing, against placebo or no treatment for infants;
  - `community_wide_mda`: community randomized; azithromycin to everyone in the community from some age (as in trachoma programmes);
  - `individual_infant`: individuals or households randomized; azithromycin to every neonate or infant at a visit or contact, against placebo;
  - `individual_child`: individuals or households randomized; azithromycin to every child of an age band reached by a programme (such as seasonal malaria chemoprevention), against placebo;
  - `schedule_comparison`: only azithromycin schedules compared (ages or frequencies), no untreated arm;
  - `other`: anything else, described in `notes`.
  A trial with several azithromycin arms against one control gets the strategy of each comparison in `comparisons` below.
- `comparisons`: list of the comparisons the trial randomized, each `{"treated_arm", "control_arm", "strategy"}` using the arm labels.
- `mortality_outcome`: `{"ages": "<age band whose mortality was measured>", "ascertainment": "census | birth_history | follow_up_visit | other", "primary": true|false, "description"}`.
- `co_interventions`: anything given to all arms that bears on child mortality (seasonal malaria chemoprevention, vitamin A, trachoma programmes, nutrition), as a list of short strings, or `[]`.
- `registration`: registry IDs named, as a list.
- `notes`: anything that makes the trial hard to code, in a sentence or two.

## Output

A JSON array of these objects, one per trial, in the order given.