Data capture and hierarchy
We recorded outcome data from ultrasound records, follow-up interviews, and paper and electronic medical records from CHCs for the whole Cape Town Metropolitan area. If e-records showed a participant visited a healthcare facility within 6 months of the abortion, her clinical chart for this visit was reviewed for relevance with respect to study outcomes or adverse events.
The presence of an IUD at 6 weeks was primarily determined by ultrasound. If participants did not come for follow-up, self-report of use or non-use was accepted if this was not in conflict with medical records, in which case the data in medical records were assumed to be true. Participants in the delayed group with no record of IUD insertion within 3 months of the abortion at any clinic in the Cape Town Metropolitan area, and who did not report IUD insertion at a clinic outside this area, were assumed not to be using the IUD at 6 weeks or 3 months. The absence of documented IUD insertion in medical records was interpreted as non-use which minimized missing data for our main outcome in the delayed group. Secondary outcome data were captured in interviews that were sometimes cut short and not resumed, resulting in varying missing data frequencies.