Skip to content

Bias in charted intervals

Charted intervals are biased because each end is rounded with a different anchor, and their per-case uncertainty is large relative to short intervals.
Date 2026-10-13
Status current
Methods Clinical and laboratory times, Rounding, heaping and value resolution

Background

  • Anaesthesia end is recorded at the next mark (ceiling), op end at the previous one (floor), and both starts at the nearest mark.

Data

  • Per-case posterior means from the four timestamp models in clinical_data.csv.

Method

  • Each interval is the difference of two event posteriors, assumed independent given the device references.

Results

  • Operation duration is charted about 2.5 min short.
  • Anaesthesia duration is charted about 4.4 min long and emergence about 6.3 min long.
  • Induction to incision is nearly unbiased (−0.5 min).
  • The median posterior SD of an interval is 1.8–2.8 min, and the 90th percentile is 2.4–4.9 min. For a 15-min emergence, that is a large share of the value.
  • Among positive intervals that are whole multiples of 5 minutes, the share on multiples of 15 is 32.5–33.3% (33.3% expected) and on multiples of 30 is 16.6–19.0% (16.7% expected) for anaesthesia duration, operation duration and induction to incision. Only anaesthesia duration has a small excess at 30 minutes, consistent with the heaping of anaesthesia end at :00 and :30. Intervals add no heaping of their own.

Discussion

  • The API returns intervals as samples from the two event posteriors (derived_samples), not as charted differences.
  • Cases with an excluded end return the status, not a distribution.

Limitations

  • The two event posteriors are assumed independent given the device references.

Worked examples

Example Side Status Summary
A regression that carries rounding uncertainty through value draft Acute kidney injury after surgery against intraoperative urine output (mL/kg/h), fitted once on the recorded numbers and once per draw from the true-value distributions: the rounding moves 14% of cases across the oliguria threshold, but widens the confidence interval of the cohort's regression by at most 4%.
Oliguria labels for a prediction model value draft One label per patient, oliguric or not, from three rounded numbers: 10% of labels are uncertain and 2.6% are expected to be wrong, which sets a floor under the error any model can be measured to have. A typical model is barely affected; a good one is.
Urine output in mL/kg/h as a distribution value draft Intraoperative urine output per kilogram per hour, from three rounded inputs, as a probability distribution per case: one case in ten cannot be placed confidently on either side of 0.5 mL/kg/h.

Tags: from the resolution work