The PDF appears to contain an internal inconsistency.
It says the survey received 42,332 responses and that 44% completed all 58 questions. If that percentage is correct, then the number who completed all questions would be about 18,626.
However, the same methods section also says that the remaining 18,407 respondents did not complete the survey. If that figure is correct, then the number who actually completed all questions would be:
42,332 - 18,407 = 23,925
So the two figures do not match.
Bottom line:
If you trust the 44% figure, then 18,626 completed all questions.
If you trust the PDF’s own arithmetic, then 23,925 completed all questions.
The PDF’s internal math points to 23,925.