Table 2.
Dissolution rates, and univariable and multivariable Cox regression model results for time to partnership dissolution (N = 565 partnerships, 466 participants).
| Variable | Eventsa | Partnership-years at risk | Rate per 100 partnership-years) (95% CI) | Unadjusted HRb | CI | Adjusted HRc | CI | Wald p-value |
| ART group | ||||||||
| Pre-ART | 76 | 416.88 | 18.23 (14.36, 22.82) | 1.00 | 1.00 | |||
| ART-eligible | 116 | 642.20 | 18.06 (14.93, 21.66) | 1.01 | (0.75, 1.34) | 0.97 | (0.73, 1.29) | 0.840 |
| Sex | ||||||||
| Female | 134 | 736.00 | 18.21 (15.25, 21.56) | 1.00 | 1.00 | |||
| Male | 58 | 323.08 | 17.95 (13.63, 23.21) | 0.99 | (0.73, 1.37) | 1.07 | (0.73, 1.56) | 0.745 |
| Age (years) | ||||||||
| 18–21 | 12 | 42.49 | 28.25 (14.60, 49.34) | 1.05 | (0.61, 1.81) | 0.86 | (0.46, 1.61) | 0.640 |
| 22–29 | 73 | 284.79 | 25.63 (20.09, 32.23) | 1.00 | 1.00 | |||
| 30–39 | 69 | 425.82 | 16.20 (12.61,20.51) | 0.65 | (0.47, 0.90) | 0.82 | (0.59, 1.14) | |
| 40+ | 38 | 305.99 | 12.42 (8.79, 17.05) | 0.51 | (0.34, 0.75) | 0.77 | (0.49, 1.23) | |
| Ever used alcohol | ||||||||
| No | 78 | 535.60 | 14.56 (11.51, 18.18) | 1.00 | 1.00 | |||
| Yes | 114 | 523.48 | 21.78 (17.96, 26.16) | 1.46 | (1.10, 1.94) | 1.38 | (1.02, 1.88) | 0.040 |
| Partner lives | ||||||||
| With participant | 53 | 524.09 | 10.11 (7.58, 13.23) | 1.0 | 1.0 | |||
| Not with participant | 139 | 530.77 | 26.19 (22.02, 30.92) | 2.48 | (1.81, 3.40) | 1.84 | (1.26, 2.68) | 0.001 |
| Partnership durationd | ||||||||
| Less than 1 year | 88 | 284.64 | 30.92 (24.80, 38.09) | 1.00 | 1.00 | |||
| 1–5 years | 58 | 297.39 | 19.50 (14.81, 25.21) | 0.67 | (0.48, 0.95) | 0.77 | (0.54, 1.11) | 0.001 |
| More than 5 years | 46 | 477.05 | 9.64 (7.06, 12.86) | 0.34 | (0.24, 0.49) | 0.47 | (0.31, 0.71) | |
| Relationship qualitye | ||||||||
| Lowest quartile | 71 | 275.37 | 25.78 (20.14, 32.52) | 1.00 | 1.00 | |||
| Second quartile | 68 | 425.35 | 15.99 (12.41, 20.27) | 0.63 | (0.45, 0.88) | 0.70 | (0.50, 0.99) | |
| Third quartile | 39 | 281.46 | 13.86 (9.85, 18.94) | 0.54 | (0.37, 0.80) | 0.67 | (0.44, 1.02) | |
| Fourth quartile | 8 | 65.85 | 12.15 (5.24, 23.94) | 0.47 | (0.22, 0.98) | 0.46 | (0.22, 0.96) | 0.061 |
ART, antiretroviral therapy; CI, confidence interval; HR, hazard ratio.
aSeventy-four participants had more than one relationship at risk of dissolution during the study: 62 with two partnerships, 9 with three partnerships, 2 with four and one with five partnerships. Of the 192 partnerships that dissolved, 133 were ongoing at enrolment among 129 participants, and 59 were new partnerships among 51 participants.
bOther variables measured at first report of partnership that were significant in univariable models: having tested HIV-positive less than 1 year before enrolment, higher perceived stigma, little reliance on family and friends, and reporting that a condom was used at first sex within the partnership were all associated with a greater hazard of dissolution. Knowing someone on antiretroviral drugs, complete knowledge about antiretroviral drugs, knowing their partner's HIV status, and having disclosed their own HIV status to their partner was associated with a lower hazard of dissolution. Partnerships that started after enrolment had a significantly higher hazard of dissolution.
cThe final model did not violate the proportional hazards assumption, global test P = 0.48. The estimated variance of individual frailties was not found significantly different from zero (P value = 0.91), suggesting no significant variation between participants.
dRelationship duration represents how long the partnership had been ongoing at the time of first report of the partnership in the study, and is not time-varying.
eThe highest quartile represents the 25% of partnerships with the greatest reported social support from their partner, a proxy for higher relationship quality.