The AIMS is the instrument the tardive-dyskinesia literature is built on. Ten rated judgements — seven body regions plus three globals — each 0 to 4, following a standardized examination procedure that includes activation manoeuvres (open the mouth, protrude the tongue, tap finger to thumb, stand and walk). Blinded video AIMS rating was the primary endpoint method in the valbenazine and deutetrabenazine registration trials, which is exactly the workflow this library exists to support.
What the recording has to show. Dyskinetic movements must be seen, so the camera has to keep the relevant regions in frame — face and mouth, hands, feet, trunk — and the examination procedure has to actually be performed on tape. A talking-head recording of an interview will not support items 5 to 7. Say in your rationale which part of the examination each rating came from.
Two rules that change ratings. The published form instructs you to rate the highest severity observed, and to rate movements that appear only on activation one less than the same movements seen spontaneously. Both are easy to lose when raters work from a timeline instead of a paper form, and both change scores systematically, so state them in your codebook.
What we have not included. ECDEU items 11 and 12 cover dental status and whether the patient usually wears dentures. They are important context for interpreting oral movements but they are not observations of the recording, so they are not codes here — record them as study metadata. Note that the ECDEU total (range 0-42) is defined over all twelve items; the widely used "sum items 1-7" movement total is a later convention. Whichever you report, say which.
Item 10 is the one exception to "rate what you see": the form says to rate only the patient's own report of awareness, so it is coded from what they say, not from how the movements look.