Clearing out what the pattern match got wrong

How it was counted

1. 1Counted text in the records matching the labelled regimen.
2. 2Read every record flagged with a different regimen, one by one.
3. 3Narrowed the window to 60 characters after the drug name to drop neighbouring drugs' regimens.
4. 4Ran the opposite hypothesis once the conclusion was in.

Result

• 40mg once daily was prescribed 131 times, matching the label.

  Session 87abbbbc

• 40mg twice daily returned 25 pattern matches, but 24 were false positives from neighbouring drugs (Keppra bid, Phazyme tid). The real count was 3.

  Session 87abbbbc

• Courses beyond eight weeks traced to gastroprotection alongside antithrombotics, not GERD.

  Session 87abbbbc

What this number cannot say

• The cross-check applies to this query only. It does not carry over to others.
• We do not take a pattern match's first result at face value. False positives are removed by a person reading the source text.

You can ask about other cohorts too