Skip to main content
. 2019 Dec 2;10:5482. doi: 10.1038/s41467-019-13473-y

Table 3.

Incidence rates (IRs, unadjusted) and incidence rate ratios (IRRs, unadjusted and adjusted) for women (N = 12,609) by male ART coverage, ART scale-up period, and year.

Events Person-years Inc rate (95% CI) Unadj. IRR (95% CI) P-value Adj. IRR (95% CI)d P-value
Model 1: By male ART coveragea
 0–9% 642 14,653 4.38 (4.00–4.79) Ref. Ref.
 10–24% 674 14,549 4.63 (4.24–5.07) 1.06 (0.92–1.21) 0.422 0.99 (0.89–1.11) 0.891
 25–34% 847 17,570 4.82 (4.47–5.21) 1.10 (0.98–1.24) 0.114 0.87 (0.77–0.97) 0.016
 35–55% 416 11,014 3.78 (3.38–4.23) 0.86 (0.75–1.00) 0.043 0.68 (0.59–0.78) <0.01
Model 2: By ART scale-up periodb
 2005–2010 1,316 29,203 4.51 (4.25–4.78) Ref. Ref.
 2011–2015 1,030 21,796 4.73 (4.42–5.05) 1.05 (0.96–1.15) 0.317 0.85 (0.77–0.93) <0.01
 2016–2017 234 6789 3.45 (2.95–4.02) 0.76 (0.65–0.90) <0.01 0.63 (0.55–0.74) <0.01
Model 3: By yearc
 2005 186 4567 4.08 (3.40–4.90) 0.84 (0.65–1.08) 0.166 1.11 (0.90–1.37) 0.339
 2006 223 5008 4.45 (3.77–5.27) 0.91 (0.71–1.17) 0.465 1.16 (0.95–1.42) 0.157
 2007 231 5078 4.56 (3.86–5.39) 0.93 (0.73–1.19) 0.585 1.18 (0.97–1.44) 0.097
 2008 231 5050 4.58 (3.89–5.40) 0.94 (0.74–1.19) 0.604 1.13 (0.92–1.38) 0.234
 2009 220 4818 4.58 (3.85–5.44) 0.94 (0.73–1.20) 0.604 1.14 (0.94–1.39) 0.176
 2010 221 4681 4.72 (3.98–5.61) 0.97 (0.75–1.25) 0.795 1.15 (0.94–1.40) 0.169
 2011 209 4566 4.59 (3.85–5.47) 0.94 (0.73–1.21) 0.624 0.97 (0.79–1.20) 0.803
 2012 214 4319 4.95 (4.14–5.92) 1.01 (0.78–1.31) 0.920 1.02 (0.83–1.25) 0.846
 2013 211 4354 4.85 (4.05–5.81) 0.99 (0.76–1.30) 0.961 1.00 (0.82–1.21) 0.968
 2014 211 4329 4.89 (4.09–5.84) Ref. Ref.
 2015 182 4225 4.31 (3.58–5.20) 0.88 (0.67–1.16) 0.368 0.86 (0.71–1.06) 0.164
 2016 141 3783 3.74 (3.04–4.61) 0.77 (0.58–1.02) 0.064 0.78 (0.63–0.97) 0.026
 2017 92 3005 3.06 (2.38–3.94) 0.63 (0.46–0.86) <0.01 0.65 (0.51–0.83) <0.01

HIV events, person-years, unadjusted IRs, and unadjusted adjusted IRRs are averages over 300 datasets generated by imputing a single random infection date within the censoring interval. Rubin’s rules were used to calculate 95% confidence intervals

aOpposite-sex ART coverage used because of the generalized, heterosexual HIV epidemic in South Africa, with 0–9% as the reference category

bTime intervals were defined by changes in national criteria for ART eligibility, with 2005–2010 as the reference period

cTime intervals were defined by year, with 2012 as the reference year

dEstimates adjusted for age, self-reported condom use, circumcision and marital status, household assets index, cumulative time spent outside surveillance area, and opposite-sex HIV prevalence in the surrounding community (see Supplementary Table 5 for full results)