Jill Durocher

An Examination of Loss to Follow-up and Potential Bias in Outcome Ascertainment in a Study of Direct-To-Patient Telemedicine Abortion in the United States

February 24th, 2023
Medication Abortion
Rosenberg, J.D., Raymond, E., Shochet, T.

Contraception; 2023 Feb 24;109996. doi:0.1016/j.contraception.2023.109996

Objectives: To examine associations between factors associated with loss to follow-up and effectiveness in the TelAbortion project, which provided medication abortion by direct-to-patient telemedicine and mail in the United States.

Study design: The study population for this descriptive analysis included abortions among participants enrolled in the TelAbortion study with data present in a web-based database tool from November 2018 to September 2021 who were mailed a medication package. The analysis included information on abortions across nine sites. In this analysis, we used generalized estimating equations to examine factors associated with loss to follow-up and effectiveness.

Results: Of the 1831 abortions included in this analysis, 1553 (84.8%) were classified as having complete follow-up and 278 (15.2%) were classified as lost to follow-up. In a multivariable analysis, factors significantly associated with loss to follow-up included history of medical abortion, education, gestational age, study site, and whether the TelAbortion was performed pre- or post-COVID-19 onset (p < 0.05). The rate of treatment failure (i.e., abortions resulting in continuing pregnancy or uterine evacuation) reported in this study was 5.1%. The only covariate associated with both loss to follow-up and treatment failure was higher gestational age. However, using gestational age to impute missing abortion outcomes did not substantially change the estimated failure rate.

Conclusions: Abortions that were lost to follow-up differed substantially from those with complete follow-up, which could bias the effectiveness estimate. However, imputing outcomes based on available and appropriate pretreatment data did not substantially affect the estimate. This finding is encouraging, although it does not exclude the possibility of bias due to unmeasured factors.

Implications: Significant differences between abortion cases with complete follow-up and those lost to follow-up provide insights into abortion cases that may be at a higher risk for being lost. The low treatment failure rate indicates that the telemedicine provision of medication abortion is effective.

Keywords: Effectiveness; Failure; Loss to follow-up; Medication abortion; Telemedicine; United States.