Interactive lab: find the planted problems¶
The table below is the full Lantern Registry export, which is synthetic: an invented condition, invented genes, invented participants, and dates in the 2040s. Planted in it are the problems real registry exports have: a duplicate enrollment, a diagnosis recorded only as "positive," a variant of uncertain significance counted as confirmed, a variant in a different gene, placeholder values in a numeric field, a weight entered in pounds, impossible dates, an ambiguous date, mixed units, inconsistent codes, and a contradiction between two fields.
Read the export the way a careful data manager would. Click every cell you think needs a query, a correction, or harmonization, then check your answers. Each answer explains what kind of problem it is and what to do about it. Nothing you click leaves your browser.
What the kinds of problem mean¶
- Error: a value that cannot be right as it stands. Raise a query; do not guess the correction.
- Query: a value that may be right but cannot be used until someone asks. Ask the family, the coordinator, or the clinician.
- Harmonize: the meaning is clear but the form is not. Map it to one coded value, keep the source value, and record the rule.
- Plausible, verify: unusual but possible. Verify it rather than change it.
Some smaller inconsistencies, such as brand names and abbreviations in the medication field, are credited when you find them and are not counted against you when you do not.
When you have finished, return to Chapter 9 and record one of your findings in the curation log.