1. Cohorts and index dates

2. Cancer definitions

Qualifying cancer diagnosis

A diagnosis concept from the fully resolved input cancer concept set used to define the cohort.

Broad cancer diagnosis

An ancestor of one or more concepts in the input cancer concept set, excluding the input concepts themselves.

Broad-concept rule

Any ancestor of a target concept remains a broad cancer diagnosis regardless of its specificity. This includes nonspecific cancer concepts and concepts defined by histology without a site.

Other cancer diagnosis

A cancer concept that is neither a member of the input cancer concept set nor an ancestor of a concept in that set.

3. Timing definitions

Closest record

The record of that kind with the smallest absolute distance from the index date, on either side. Where a patient has records on both sides, only the closest one counts.

4. How to read the results

Where a site reports no usable value, the state is shown in place of the number and is never rendered as a zero. Counts at or below this site's small-cell suppression threshold read Suppressed, never zero. Two of the states below are reported results rather than absences: a single value is a value, and a zero-eligible count is a count of zero. Both are shown in the reported treatment and neither is shown as unavailable.

Suppressed

Patient count withheld because it is below the site's reporting threshold.

Single value

Every percentile from the 1st to the 99th falls on the same value, so the distribution is a single point and there is no interquartile range to draw.

Zero eligible

the quantity was evaluated at this site and no patient met its definition, so the reported count is zero and there is no distribution to draw.

No record in the subset

The site reports the concept, but no patient in the selected population carries it. This is not a zero.

Not recorded at this site

The concept is absent from that site's finalized concept table.

Not evaluable

The site cannot evaluate the selected population, so no value exists for it.

Not available

The site's results package does not carry that quantity, so no figure can be drawn.

The red line shows the median and the red bar the IQR. Day 0 is shown when applicable. Arrows indicate truncated tails. The y-axis is intentionally omitted; interpret the shape, location, and spread of the distribution.

5. Methods

Every value in this report is read from this site's own returned results at build time. Nothing is pooled across sites and nothing is imputed. Concept names resolved against OMOP vocabulary v5.0 27-FEB-26.

Distribution curves are derived from the percentile summaries returned by each site, not from patient-level records. The method represents the information the percentile grid supports; it does not modify unusual distributions to look smoother or more attractive.

The curve

Between two adjacent percentiles the local concentration is the change in cumulative probability divided by the change in x. Those values are interpolated with a shape-preserving method, clamped to the range set by the two percentile points that bracket each position, so interpolation cannot create a peak that the percentiles do not support.

Point masses

Repeated percentile values at the same x indicate probability concentrated at one exact value. Repeats at that value are aggregated and drawn as a spike rather than smoothed into the curve.

P1 and P99

The 1st and 99th percentiles are printed at the ends of each figure and are retained even where the plotted range is narrower than they are.

Plotting range and truncation

The plotted range is chosen for each figure from the observed distribution, so ranges differ between figures and are not comparable across them. A long tail may be truncated for readability. Truncation is marked with a continuation arrow, and a peak cut off by the vertical scale is marked the same way.

No vertical axis

No vertical axis is drawn: the curve is a percentile-derived concentration, not an observed empirical density.