Full multilevel random intercept logit model (model 2) for the odds of reporting removing a pregnancy, adjusted for interviewer, interview, respondent and community characteristics*
| Côte d'Ivoire | Rajasthan, India | Nigeria | ||||||
| OR | (95% CI) | OR | (95% CI) | OR | (95% CI) | |||
| Interviewer characteristics | ||||||||
| No of respondents | 1.00 | (0.98 to 1.01) | 0.99 | (0.97 to 1.01) | 0.99 | (0.98 to 1.00) | ||
| Interviewer age | 0.99 | (0.96 to 1.02) | 1.02 | (0.99 to 1.06) | ||||
| Married (vs not married) | 1.17 | (0.84 to 1.64) | 0.73 | (0.39 to 1.34) | ||||
| Has children (vs no children) | 1.33 | (0.96 to 1.83) | 1.52 | (0.85 to 2.74) | ||||
| Technical/Uni grad (vs secondary) | 1.47 | (0.95 to 2.29) | 1.08 | (0.73 to 1.59) | ||||
| Existing staff (vs new staff) | 1.34 | (0.52 to 3.44) | 1.11 | (0.66 to 1.87) | ||||
| Previous survey experience (v none) | 1.36 | (0.82 to 2.26) | 1.01 | (0.62 to 1.64) | ||||
| Very comfortable asking about abortion (vs somewhat/not comfortable) | 1.88 | (0.9 to 3.9) | 1.55 | (0.45 to 5.34) | ||||
| Interview characteristics | ||||||||
| Very well or well acquainted | Ref | Ref | Ref | |||||
| Not well acquainted | 1.22 | (0.28 to 5.22) | 0.68 | (0.49 to 0.96) | 0.75 | (0.58 to 0.97) | ||
| Not acquainted | 1.67 | (0.45 to 6.17) | 0.91 | (0.57 to 1.45) | 0.85 | (0.65 to 1.13) | ||
| Language of interview | ||||||||
| French | Hindi | Hausa | Ref | Ref | Ref | |||
| Baoule | English | English | 3.11 | (2.00 to 4.83) | 0.77 | (0.08 to 6.97) | 1.59 | (1.19 to 2.14) |
| Yacouba | Other | Igbo | 0.54 | (0.15 to 1.93) | 1.54 | (0.69 to 3.47) | 1.13 | (0.72 to 1.78) |
| Attie | – | Yoruba | 0.75 | (0.28 to 2.02) | 1.94 | (1.16 to 3.25) | ||
| Dioula | – | Pidgin | 0.26 | (0.14 to 0.48) | 1.57 | (0.99 to 2.47) | ||
| Lobi | – | Other | 0.17 | (0.02 to 1.35) | 1.61 | (0.73 to 3.55) | ||
| Other | – | – | 0.67 | (0.27 to 1.72) | ||||
| Intrainterviewer correlation | 0.01 | 0.15 | 0.22 | |||||
Coefficients in bold are significant at the <0.05 level, coefficients in bold and italics are significant at the <0.10 level. Interviewer characteristics were unavailable for Nigeria.
*Respondent characteristics include age, age squared, education, marital status, ever given birth and number of birth events, wealth quintile and whether the respondent is a previous Performance Monitoring for Action (PMA) respondent. Community characteristics include region or state, rural/urban status, monthly number of abortions per community (mean, facility reported) and whether there is an abortion care facility in the community. The full model is available in online supplemental appendix 3, including all respondent and community variable coefficients.