Analysis
The primary analysis was to assess whether the telemedicine-hybrid care
model was non-inferior to the traditional service model. Non-inferiority
is established by comparing effectiveness and safety in the two cohorts
under the null hypothesis that the rates of successful abortion and
serious adverse events are not different.
We first compared patient demographic and clinical characteristics
between the cohorts to assess the need to covariate-adjust our
hypothesis tests for systematic differences in the two groups that might
affect abortion outcomes. All hypothesis tests were covariate adjusted
for patient age, race/ethnicity, gestational age, parity, and prior
abortions using logistic regression and weighted risk differences.(20)
We evaluated effectiveness by testing the alternative hypothesis that
the rate of successful medical abortion in the telemedicine-hybrid
cohort is lower than in the traditional cohort using a
covariate-adjusted test of difference in proportions. We also performed
a chi-squared test to evaluate whether the distribution of unsuccessful
abortion sub-categories differed between the cohorts. We evaluated
safety by testing the alternative hypothesis that significant adverse
events occurred at higher rates in the telemedicine-hybrid cohort
compared to the traditional cohort using a covariate-adjusted hypothesis
test for difference of proportions. We also evaluated whether the
prevalence of ectopic pregnancies managed before EMA and after EMA were
different between the traditional and telemedicine-hybrid cohorts using
chi-squared difference of proportions tests.
The secondary analysis was to compare effectiveness and safety of
medical abortion for patients who received fully remote no-test
telemedicine vs. in-person care in the telemedicine-hybrid cohort,
primarily to assess whether any differences between the cohort service
models were driven by one particular group. We evaluated patient
demographic and clinical characteristic differences between the two
groups to assess the need to covariate-adjust our hypothesis tests for
systematic differences and all hypothesis tests were covariate adjusted
for patient age, race/ethnicity, gestational age, parity, and prior
abortions using the framework described above. We note, however, that
these cohorts are fundamentally different despite covariate adjustment,
as patients in the telemedicine-hybrid cohort were selected into
in-person treatment based on characteristics that would affect the
outcome of their abortion. We performed covariate-adjusted hypothesis
tests under the null hypothesis of equal effectiveness and equal rates
of significant adverse events in the telemedicine vs. in-person groups.
All analyses were performed using R, version 3.6.2. Statistical
significance was defined using an alpha level of 0.05.