TABLE 3.
Univariable- and multivariable-adjusted attitudinal and behavioural predictors of concurrent sexual partnerships (past 3 months) at 3 months amongst men undergoing voluntary medical male circumcision in Gaborone, Botswana, 2012–2015.
| Attitudinal and behavioural predictors | Univariable-adjusted† |
Multivariable-adjusted†,‡ |
||||
|---|---|---|---|---|---|---|
| RR | 95% CI | p | RR | 95% CI | p | |
| Reasons for circumcision | ||||||
| Personal hygiene | 1 | ref. | - | 1 | ref. | - |
| HIV protection | 1.16 | 0.74–1.82 | 0.52 | 1.24 | 0.79–1.96 | 0.35 |
| Other | 0.98 | 0.57–1.67 | 0.94 | 0.91 | 0.55–1.53 | 0.74 |
| Knowledge of HIV risk | ||||||
| Correct knowledge: VMMC reduces HIV risk for men | 0.99 | 0.58–1.68 | 0.97 | 1.05 | 0.59–1.86 | 0.87 |
| Correct knowledge: VMMC does not impact HIV risk for women | 1.23 | 0.71–2.10 | 0.46 | 1.16 | 0.66–2.01 | 0.61 |
| Risk compensation scale§ | ||||||
| Condom use is not necessary if the man is circumcised | 1.13 | 0.91–1.40 | 0.29 | 1.18 | 0.93–1.49 | 0.17 |
| If I am circumcised, sex is safe without a condom | 1.00 | 0.81–1.22 | 0.97 | 1.00 | 0.80–1.23 | 0.97 |
| Being circumcised means a man can worry less about HIV | 1.01 | 0.89–1.14 | 0.89 | 1.02 | 0.91–1.16 | 0.70 |
| If a man is circumcised, he can have more sexual partners | 1.23 | 0.90–1.68 | 0.20 | 1.21 | 0.89–1.66 | 0.23 |
| Alcohol consumption | 1.97 | 1.00–3.87 | 0.05 | 2.30 | 1.45–3.66 | 0.0004 |
| Age at first sexual intercourse (years) | ||||||
| < 18 | 1 | ref. | - | 1 | ref. | - |
| 18–20 | 1.15 | 0.76–1.74 | 0.46 | 1.09 | 0.73–1.62 | 0.68 |
| ≥ 21 | 0.59 | 0.32–1.06 | 0.08 | 0.71 | 0.38–1.34 | 0.29 |
| Exchanged money for sex, past 12 months | 1.57 | 0.87–2.82 | 0.14 | 1.62 | 0.99–2.63 | 0.05 |
RR, risk ratio; CI, confidence intervals; HIV, human immunodeficiency virus; VMMC, voluntary medical male circumcision.
, Estimated from a weighted modified Poisson regression model with weights constructed to adjust for selection bias because of loss to follow-up.
, Estimated from a weighted modified Poisson regression model with weights constructed to adjust for selection bias because of loss to follow-up and confounding because of the following baseline covariates: age, relationship status, religious affiliation, education, employment and household use of woods as cooking fuel.
, Responses based on Likert scale (range 0–5 with 0 and 5 denoting ‘strongly disagree’ and ‘strongly agree’, respectively).