N (unweighted) | National | Northern zone | Middle zone | Coastal zone | |||||
Rate | 95% CI | Rate | 95% CI | Rate | 95% CI | Rate | 95% CI | ||
Respondent | |||||||||
All abortions (reported abortions+MR*) | 4713 | 30 | 24 to 37 | 10 | 6 to 17 | 29 | 16 to 53 | 37 | 21 to 66 |
Only reported abortions (no MR*) | 20 | 16 to 25 | 6 | 3 to 11 | 18 | 9 to 38 | 27 | 13 to 55 | |
Confidantes | |||||||||
Confidante #1 | 3040 | 41 | 37 to 45 | NA | NA | NA | NA | NA | NA |
Confidante #2 | 737 | 36 | 29 to 45 | NA | NA | NA | NA | NA | NA |
Confidante #3 | 153 | 30 | 17 to 50 | NA | NA | NA | NA | NA | NA |
All confidantes | 3930 | 40 | 36 to 44 | 20 | 15 to 27 | 42 | 36 to 48 | 44 | 39 to 50 |
Adjusted for missing confidantes† | 38 | 35 to 41 | 18 | 13 to 22 | 40 | 36 to 44 | 43 | 38 to 47 | |
Adjusted for transmission bias‡ | 61 | 56 to 65 | 34 | 26 to 43 | 63 | 57 to 69 | 71 | 63 to 79 |
Northern zone includes Upper West, Upper East and Northern regions; Middle zone includes Brong-Ahafo, Ashanti, Eastern and Volta regions; Coastal zone includes Western, Central and Greater Accra regions.
*MR or action to bring back a late period without necessarily knowing whether one is pregnant.
†The 33% of respondents who reported no confidantes had different demographic characteristics and a lower abortion rate than those with at least one confidante. Assuming confidantes are similar to respondents on the aggregate, these ‘missing’ confidantes may have significantly different abortion behaviours than those in the data, which would affect the incidence rate. Thus, we predicted the likelihood of recent abortion for the missing confidante of each confidante-less respondent, using Poisson regression with the respondent’s sociodemographic characteristics as covariates.29 30 We then calculated the confidante abortion rates including the missing confidantes’ data.
‡Adjusted for potential non-disclosure of some confidantes’ abortions to the respondent. Based on an assumption of reciprocity, such that confidante X is as likely to disclose her abortion to the respondent as the respondent is to disclose an abortion to confidante X, we asked respondents who reported an abortion in the past 3 years whether they had disclosed it to each of their confidantes. We calculated the proportion of respondents who disclosed to each confidante (1, 2, 3), took the inverse of this proportion as the transmission bias factor, computed an average of all three factors (weighted on the number of confidantes) and applied it to the incidence rate adjusted for missing confidantes.
MR, menstrual regulation; NA, not applicable.