Follow one event
Illness begins, a person seeks care, testing occurs, a result is finalized, and the event enters a reporting system. Each step can add delay. Weekends, holidays, and backlogs can change the pattern.
A chart organized by event date will be revised as later reports arrive.
Compare data with similar maturity
A fully reported week from last month and yesterday's partial count are not equally complete. Analysts may exclude the newest period, label it provisional, or use a model to adjust for delays, depending on the purpose.
Each approach has tradeoffs between timeliness and certainty.
Preserve historical snapshots
Archived versions show how much counts usually change after first publication. They can also reveal whether the reporting process has shifted, which matters for any delay-adjustment model.
Communicate the uncertainty near the edge
Mark recent periods visibly and explain that they may rise. A footnote far from the chart may not prevent a reader from interpreting the downward tail as improvement.
Delayed reporting is not evidence that the data are useless. It is a property of the observation process. A useful surveillance product accounts for it explicitly rather than treating every displayed date as equally settled.
