To evaluate and compare the performance of six HIV-RNA-based quality of care indicators for predicting short-term and long-term outcomes.
Multinational cohort study.
We included EuroSIDA patients on antiretroviral therapy (ART) with ≥3viral load (VL) measurements after baseline (the latest of 01/01/2001 or entry into EuroSIDA). Using multivariate Poisson regression we modelled the association between short-term (resistance, triple-class failure) and long-term (all-cause mortality, any AIDS/non-AIDS clinical event) outcomes and the indicators: (i)viraemia copy years (VCY), (ii) Consecutive months with VL ≥50 copies/mL, (iii) percentage of time on ART spent fully suppressed (%FS), (iv) stable on ART, (v)48 weeks snapshot, and (vi) current VL. Indicators were compared using area under the ROC curve (AUC) and different measures of model fit.
Adjusted incidence rate ratios for all outcomes tended to increase with increasing VCY, number of consecutive months with VL ≥50 copies/mL, current VL and with lower %FS, but the gradient of increased risk was weak across strata. None of the indicators reliably identified those at risk of long-term outcomes (AUC 0.54-0.58), but performed consistently better with short-term outcomes (triple class failure [AUC 0.67-0.76]) and resistance [AUC 0.64-0.79]). Goodness of fitvariedwith the outcome evaluated, but differences between indicators were small.
Differences between quality of care indicators were small and no indicator performed consistently better than current VL. Given the simplicity in assessing and interpreting this indicator, wepropose to use current VL when HIV-RNA-based indicators are used to evaluate the efficacy of ART programs.