Turnaround Time
Elapsed time from specimen receipt to final report — the department’s headline quality metric, and a variable whose distribution matters far more than its mean.
What it is
Turnaround time (TAT) looks simple and is statistically awkward for three reasons.
It is not normally distributed. TAT is strongly right-skewed: most cases clear quickly and a long tail runs to weeks (decalcification, deeper levels, IHC, consultation). Reporting a mean is actively misleading; the median with an interquartile range, plus a high percentile such as the 90th, describes it far better. The tail is usually where the operationally interesting cases are.
It is composite. Receipt to gross, gross to embed, embed to section, section to scan, scan to sign-out — each is its own queue with its own bottleneck. An aggregate TAT hides which step moved. Analysing the segments separately is what makes the metric actionable.
It is censored and confounded. Cases still in progress at data extraction are right-censored, which technically makes this a time-to-event problem — the same machinery in Survival Analysis Workflow applies, and it handles censoring correctly where a simple mean does not. Case complexity confounds everything: a department taking harder cases will show worse raw TAT while performing better.
Queueing effects mean load and TAT are non-linearly related — near capacity, small increases in volume cause large increases in wait. Hour-of-day and day-of-week patterns are therefore real signal, not noise.
Why it matters for my work
TAT is the connective metric across the whole Quality Research cluster. Scanning Time in Real Life explicitly targets daily load patterns, queueing effects, and bottlenecks in scanning and transfer, with hour-of-day visualisations and case-level timing metrics as current tasks. Konsultasyon measures turnaround from receipt to final report for referred-in cases, where slower turnaround is one of the defining features of that workload. Patoloji Bilgi Yönetim Sistemi (LIS) lists turnaround among the LIS quality signals and warns that the LIS records several distinct timestamps — received, grossed, signed-out, amended — each with a specific documented meaning that must not be guessed at.
That last point is the practical crux: TAT is only as good as the timestamp definitions behind it, and the definitions are documented rather than obvious.
How it connects
Patoloji Bilgi Yönetim Sistemi (LIS) — the source of the timestamps, and the place where their meanings are defined.
Scanner Log Analysis — the method page for the scanner-side timing work, including the DST and retry-deduplication traps.
Scanner Logs — the underlying event data, whose cleaning traps determine whether any timing figure is trustworthy.
Survival Analysis Workflow — the right statistical framing for censored elapsed-time data, and an under-used one for operational metrics.
Whole Slide Imaging — file size and transfer are the physical reason the digital steps take the time they do.
Laboratory Workload Measurement — the companion metric, and the one most often conflated with this: turnaround is elapsed time per specimen, workload is staff time per department. A department can hold turnaround steady while workload becomes unsustainable, so neither number is interpretable alone.
Ki-67 Proliferation Index — Shaaban 2026 (sources/papers/shaaban-2026-uk-ki67-recommendations.md) claims AI scoring gives “reduced turnaround time.” Read through this page that is a segment-level effect — removing a manual counting step from one stage — asserted rather than measured as a whole-case improvement, so it should be sized on the composite before it is believed. [unverified]
Konsultasyon Log — what an internal consultation actually costs in elapsed time. The log dates both ends of the extra step, so the question “does asking a colleague slow the case down, and by how much” is answerable from data already held rather than from impression.
Open questions
- Are the group’s TAT analyses reporting medians and percentiles rather than means? Not documented. [unverified]
- Is case complexity adjusted for when comparing across subspecialties or time periods? Without it, comparisons are close to meaningless.
- Is anyone treating in-progress cases as censored rather than excluding them? Exclusion biases TAT downward, and systematically so.