Selected Model Variables.

NOTE: HAART, highly-active antiretroviral therapy; DOT, directly observed therapy; QALE, quality-adjusted life expectancy.* Complete response considered to be a sustained reduction in viral load of 2.0 log10 copies/ml; partial response considered to be a sustained reduction of 0.75 log10 copies/ml; non-response associated with a reduction of 0.25 log10 copies/ml [32].† Non-elective Caesarean at term and Caesarean section with premature delivery were not associated with reduced risk of mother-to-child HIV transmission [59].‡ Base-case probability of prematurity approximates that seen in a cohort of HIV-infected New York Medicaid recipients; upper bound based on rates of premature delivery seen in a subgroup of women receiving methadone maintenance therapy [58].§ In base case, risk of antiretroviral toxicity in infants was assumed to be negligible, consistent with available data [6], [62], [65]. Risk of severe toxicity used in sensitivity analysis based on upper bound confidence limit for mitochondrial toxicity in a French cohort study [63], [64]. Risk of moderate toxicity based on best estimate of plausible upper bound.∥ Base-case estimates and ranges for viral loads greater than 1000 copies/ml derived based on outcomes among individuals receiving peripartum zidovudine in a prospective multi-centre study of HIV in pregnancy [5].¶ QALY and lifetime cost estimates presented in table based on the use of a 3% discount rate. Base-case quality-adjustment for HIV-uninfected individuals 45 years of age and older performed using community-derived utility estimates, as described in [78], while upper bound estimates are not quality-adjusted.** Quality-adjusted survival in HIV-infected infants was estimated based on the assumption that 2/3 of person-time with HIV infection would be symptom-free, while 1/3 of person-time with HIV would include HIV-attributable symptoms. Acquired immune deficiency syndrome was assumed to be present in the last two years of life. Death due to HIV in infected children has declined markedly in both the U.S. and Europe with the advent of HAART, making estimation of survival in HIV-infected children difficult due to small numbers of events [35], [37]. Survival in HIV-infected children was assumed to approximate that seen in the youngest adults treated with HAART [36]. Lower bound survival estimates for HIV were generated using community-derived utility weights for life with HIV infection [34], while upper bound estimates were generated using more favorable survival estimates, and without quality-adjustment [79].†† Based on in-hospital mortality in 15% of premature infants (including third-trimester still-births), with a risk of moderate to severe cognitive impairment in 10–30% [68], [70], [76]. Reduction in quality-adjusted survival estimated based on health utility weight of 0.67 predicted for an individual with moderate cognitive and sensory impairment and impaired self-care ability using the Health Utilities Index Mark II [69].‡‡ Based on intravenous zidovudine during 12 hour labor, and average dose of 1 ml zidovudine syrup (10 mg/ml) administered to neonate qid for 6 weeks postpartum [6].§§ Estimated based on weighted average healthcare costs associated with prematurity in infants born from 28 to 36 weeks of gestation in the state of California [75], with future costs occurring due to developmental delay in 15–30% of surviving infants [68], [74], [76], [80].