| Variable | N (% of population)† | Median (IQR) AFS | Crude HR (95% CI) | p Value* | Adjusted HR (95% CI)‡ | p Value** |
| Religion | 0.004 | 0.01 | ||||
| Shembe | 884 (19) | 18.8 (17.3–20.6) | 1.0 | 1.0 | ||
| Zionist | 1556 (33) | 18.3 (17.0–20.0) | 1.20 (1.06 to 1.36) | 1.18 (1.05 to 1.34) | ||
| West Christian | 1371 (29) | 18.5 (17.1–20.3) | 1.07 (0.94 to 1.22) | 1.03 (0.91 to 1.18) | ||
| None | 117 (2) | 18.0 (16.8–19.6) | 1.41 (1.07 to 1.86) | 1.36 (1.03 to 1.80) | ||
| Other | 331 (7) | 18.9 (17.9–20.5) | 0.96 (0.79 to 1.18) | 0.95 (0.77 to 1.16) | ||
| Geographical location | 0.007 | 0.05 | ||||
| Rural | 3297 (70) | 18.7 (17.2–20.4) | 1.0 | 1.0 | ||
| Urban | 196 (4) | 18.5 (17.0–19.9) | 1.11 (0.86 to 1.43) | 1.13 (0.87 to 1.46) | ||
| Periurban | 1231 (26) | 18.2 (17.0–19.9) | 1.17 (1.06 to 1.29) | 1.13 (1.02 to 1.25) | ||
| Education (grade-for-age) | 0.2 | |||||
| Age appropriate | 3665 (31) | 18.7 (17.2–N/A) | 1.0 | |||
| Behind by 1 year | 2809 (24) | 18.5 (17.0–20.4) | 1.02 (0.88 to 1.18) | |||
| Behind 2+ years | 3642 (31) | 18.5 (17.1–20.3) | 1.05 (0.91 to 1.20) | |||
| Complete secondary | 1513 (13) | 18.4 (17.0–20.0) | 1.04 (0.85 to 1.27) | |||
| School attendance <18 years | <0.001 | |||||
| Attending school | 7271 (81) | (17.3–N/A) | 1.0 | |||
| Not in school | 298 (3) | 16.4 (16.0–17.6) | 3.86 (3.24 to 4.59) | |||
| Complete secondary | 904 (10) | (17.0–N/A) | 1.17 (0.93 to 1.46) | |||
| School attendance 18–21 years | <0.001 | |||||
| Attending school | 1534 (67) | 20.3 (19.0–N/A) | 1.0 | |||
| Not in school | 203 (9) | 18.9 (18.4–19.7) | 2.48 (2.07 to 3.00) | |||
| Complete secondary | 415 (18) | 19.8 (18.8–N/A) | 1.34 (1.12 to 1.60) | |||
| School attendance 21+ years | 0.004 | |||||
| Attending school | 127 (29) | 23.4 (22.1–N/A) | 1.0 | |||
| Not in school | 91 (21) | 22.7 (21.8–24.7) | 1.53 (0.92 to 2.53) | |||
| Complete secondary | 194 (45) | 22.1 (21.7–24.0) | 1.98 (1.31 to 3.0) | |||
| Mother member of household | 0.015 | |||||
| No | 3947 (34) | 18.3 (17.0–20.0) | 1.0 | |||
| Yes | 7784 (66) | 18.6 (17.2–20.4) | 0.89 (0.81 to 0.98) | |||
| Father member of household | 0.01 | |||||
| No | 6714 (34) | 18.4 (17.0–20.0) | 1.0 | |||
| Yes | 5018 (66) | 18.7 (17.2–20.5) | 0.89 (0.81 to 0.97) | |||
| Mother dead | <0.001 | <0.001 | ||||
| No | 9744 (83) | 18.6 (17.1–20.4) | 1.0 | 1.0 | ||
| Yes | 1645 (14) | 18.1 (17.0–19.7) | 1.23 (1.10 to 1.39) | 1.23 (1.09 to 1.38) | ||
| Missing | 342 (3) | 18.0 (16.7–19.4) | 1.53 (1.22 to 1.92) | 1.45 (1.15 to 1.82) | ||
| Father dead | ||||||
| No | 8238 (70) | 18.6 (17.2–20.4) | 1.0 | 0.06 | ||
| Yes | 2926 (25) | 18.3 (17.0–20.0) | 1.10 (1.00 to 1.21) | |||
| Ever taken alcohol | 0.002 | 0.02 | ||||
| No | 2547 (54) | 18.7 (17.2–20.4) | 1.0 | 1.0 | ||
| Yes | 111 (2) | 17.4 (16.6–19.1) | 1.50 (1.17 to 1.92) | 1.42 (1.10 to 1.82) | ||
| Missed 2003 survey | 2066 (44) | 18.4 (17.0–20.0) | 1.14 (1.04 to 1.25) | 1.13 (1.03 to 1.25) | ||
| Community in danger | 0.04 | |||||
| No/don’t know | 1262 (27) | 18.8 (17.2–20.5) | 1.0 | |||
| Yes | 1401 (30) | 18.4 (17.1–20.3) | 1.12 (1.00 to 1.26) | |||
| Know relatives who have/had AIDS§ | 0.006 | 0.03 | ||||
| No | 2457 (52) | 18.7 (17.2–20.5) | 1.0 | 1.0 | ||
| Yes (1+) | 198 (4) | 18.0 (16.6–19.4) | 1.32 (1.08 to 1.60) | 1.26 (1.03 to 1.53) | ||
| HIV is transmitted by sharing utensils¶ | 0.89 | |||||
| No | 1874 (40) | 18.6 (17.2–20.4) | 1.0 | |||
| Yes/don’t know | 874 (17) | 18.5 (17.0–20.4) | 0.99 (0.87 to 1.12) | |||
HR, hazard ratio; N/A, not applicable.
*Univariate likelihood ratio test p value.
†Totals vary because of missing values. The percentage distribution is based on the population of 4724 virgins. For time-varying characteristics it is based on person-years and one individual could contribute to different categories during the study period.
‡N = 4724. Missing indicators were used to represent the individuals with no information in multivariable models, adjusted for all other factors in the column, factors with a likelihood ratio test p value ⩽0.05. School attendance was not considered for multivariable analyses.
§The variable “knowing people with HIV” violated the proportional hazards assumption univariately. Knowing relatives with HIV was highly correlated with knowing people and neighbours with HIV, and only the variable “knowing relatives with HIV” was considered for the multivariable analysis.
¶Similar non-significant results were attained for the questions about other modes of HIV transmission.
**Likelihood ratio test p value for the contribution of the variable to the multivariable model.
Note: Hazard ratios >1 indicate that the AFS is younger than the AFS for the reference category.