Table 2.
Longitudinal return and attrition rates.
| Wave | Outgob n (%) | Returnedb n (%) | Attritedb n (%) | Removed during cleaningc n (%) (wave specific) | Final cleanedc n (%) (wave specific) | Final cleanedb n (%) |
|---|---|---|---|---|---|---|
| 1a | 17,000d | 211 (1.2) | 16,789 (98.8) | 16,789 (98.8) | ||
| 2 | 16,681 (99.4) | 10,023 (59.7) | 6766 (40.3) | 282 (2.8) | 9741 (97.2) | 9741 (58.0) |
| 3 | 16,373 (97.5) | 7840 (46.7) | 8949 (53.3) | 323 (4.1) | 7517 (95.9) | 7517 (44.8) |
| 4e | 14,715 (87.7) | 6584 (39.2) | 10,205 (60.8) | 222 (3.4) | 6362 (96.6) | 6362 (37.9) |
| 5 | 16,316 (97.2) | 6128 (36.5) | 10,661 (63.5) | 213 (3.5) | 5915 (96.5) | 5915 (35.2) |
For wave 1 no cases could attrit, the total returns screening positive for migraine were N = 17,000 out of N = 58,629 (these cases plus the N = 10,044 who were over-quota and N = 12,110 who abandoned produce the total returns of N = 80,783), of these, 211 (1.2%) were removed during cleaning resulting in a final sample of N = 16,789 respondents.
Percentages in these cells denominated by baseline total (n = 16,789).
Percentages in these cells denominated by wave-specific returns.
See Figure 2 for how we arrive at the n = 17,000 migraine cases enrolled in the sample.
The drop in outgo at wave 4 is reviewed in detail in the Discussion under the section on “Benefits/risks of conducting large scale epidemiologic studies using national online sampling panels.”