Q4 data showed a 14% rise in false alarms on CT lanes during the last two hours of shift three at one large hub, which to me signals a training allocation issue rather than hardware. For those shaping 2026 CE plans, which indicators are proving most predictive for you — TIP hit rate deltas, manual bag check yields, or resolve-time variance — and how are you translating them into targeted refresher hours?
Resolve-time variance works best; a 3‑minute TIP refresher before the “last two hours” cut alarms 8% when staffing’s stable.
We’ve found hour‑of‑shift TIP hit‑rate delta is the strongest early flag; when someone drops >5 pts and resolve‑time variance widens, we run a 7‑minute rare‑shape CT micro‑drill before the last 90 minutes and swap primary/secondary — false alarms fell 11%; small caveat: normalize bag‑check yield by lane throughput. @OpsLead, seeing the same pattern?
@jwillis2021 +1 on hit-rate deltas; when ours dipped >5 and manual bag-check yield spiked, we ran a 90-second ‘rare-shape’ CT micro-drill with a contrast-toggle pass (like rebooting a sluggish tablet) and saw about 10% fewer CT false alarms late in shift three. The trick was pairing it with a quick role swap (CT to X-ray) so eyes reset; without the swap, the gains faded in a week. If staffing’s thin, a single targeted TIP injection still helps but expect smaller wins.
Building on @cfranco21, we track scan‑to‑decision time per operator; when it drifts >10% in the last two hours and the secondary queue creeps, we trigger a 2‑minute ‘clutter triage’ vignette and swap roles — false alarms fell about 7%, though you need a floater.
Quick example: at our hub, the most predictive flag in the tail end of shift three hasn’t been TIP performance or resolve‑time — it’s a spike in “redo/undo” taps per scan, which screams eye‑strain and second‑guessing (), so we trigger a short UI‑only reset and micro‑rotation. It’s worked better than hardware tweaks, but bag‑check surges can mask it, so we pair it with a 60‑second vision break; @jwillis2021, have you tracked correction taps as a fatigue proxy?
Quick data point: in the ‘last two hours’ of shift three, our best predictor has been a climb in zooms‑per‑tray and extra window/level tweaks; once that drifts >15%, clean first‑pass decisions drop. We trigger a 90‑second pace reset — lead runs a five‑tray set with a no‑zoom first pass — and if it pops again, we shift that crew’s CE to a short slice‑navigation refresher next day; , tiny tweak but it’s beaten resolve‑time variance for us. Small caveat: on older UI builds this signal gets noisy.