The blank cell has several possible histories
Imagine a dataset with a laboratory column and many empty rows. Replacing the blanks with a normal value assumes that every unobserved result would have been normal. That assumption is usually not established by the blank itself.
People selected for testing may also differ from those who were not tested.
Preserve the reason when available
| Status | What it does not establish |
|---|---|
| Not ordered | That the condition is absent |
| Pending | That a final result is negative |
| Sample unsuitable | That the measured value was normal |
| Missing from export | That no result exists elsewhere |
FHIR observation structures can represent reasons for absent values. Local datasets may use different fields, so the actual coding needs inspection.
A portal question and a research question differ
For a patient, the next step may be asking whether the expected result is complete and where it will appear. For an analyst, the next step includes defining missingness and considering how testing patterns affect the sample.
Neither task is improved by silently converting “unknown” into “normal.” Keeping uncertainty visible is more useful than filling every cell with a reassuring-looking answer.
