Figure 4: Incremental cost-effectiveness ratio (ICER) for viral-load-informed differentiated care using dried blood spot (DBS) (compared with next less effective strategy on the efficiency frontier) according to changes in assumptions (see Supplementary Figure 1).

a, DBS sensitivity 96% and specificity 79% for 1,000 copies per millilitre threshold (versus plasma). b, DBS sensitivity 71% and specificity 97% for 1,000 cps ml⁻¹ threshold (versus plasma). c, DBS sensitivity 88% and specificity 93% for 1,000 cps ml⁻¹ threshold (versus plasma). d, DBS sensitivity 85% and specificity 79% for 1,000 cps ml⁻¹ threshold (versus plasma). e, Poorer population antiretroviral therapy (ART) adherence profile such that proportion with viral suppression with no monitoring/no second-line ART is 68% compared with 76% in base case and HIV incidence is 0.96 per 100 person years compared with 0.84 in base case. f, Future greater increase in sexual behaviour in population such that HIV incidence is 1.46 per 100 person years compared with 0.84 in base case. g, Permanent increase in adherence as a result of viral load measurement alert in none rather than 40%. h, Permanent increase in adherence as a result of viral load measurement alert in 100% rather than 40%. 1i, Policy of initiation of ART at diagnosis. j, Reduced frequency of visits if CD4 count measured >350 in past year. k, Switch rate of 0 (so only benefit of monitoring is to inform who should be seen less frequently). l, Lower prevalence of HIV in 2014 (6% instead of 15% in base case). m, Higher prevalence of HIV in 2015 (33% instead of 15% in base case). n, Lower proportion on ART in 2015 (33% instead of 56% in base case). o, Higher proportion on ART in 2015 (70% instead of 56% in base case). p, 5% discount rate instead of 3%. q, Ten year time horizon instead of 20 year. r, Two times higher rate of ART interruption if visit frequency has been reduced due to viral load being <1,000 cps ml⁻¹. s, Two times lower rate of ART interruption. Based on 200 model runs per strategy for each ofa–s.