How accurately do routinely reported HIV viral load suppression proportions reflect progress towards the 90-90-90 target in the population on antiretroviral treatment in Khayelitsha, South Africa?

BACKGROUND: The Joint United Nations Programme on HIV/AIDS (UNAIDS) third 90-90-90 target requires 90% of patients on antiretroviral treatment (ART) to be virally suppressed (<1 000 copies/mL). In Khayelitsha, Cape Town, South Africa viral load (VL) suppression of <400 copies/mL was reported as 89% in 2016, but only 56% of patients had a result recorded in routine data. We conceived a VL 'cascade' to represent the steps required for an expected VL to be reported as complete in routine data and thus contribute to reported VL suppression: among those for whom a VL is 'expected', a sample must be collected and tested ('done'), a result must be 'filed' in the patient folder, 'noted' by a clinician and electronically 'captured'. The low reported completion suggested gaps along the VL cascade and cast doubt on the validity of reported suppression.
OBJECTIVES: To assess the validity of routinely reported VL suppression and identify barriers to VL completion.
METHODS: A retrospective cohort study between 1 July 2015 and 30 June 2016, which included all Khayelitsha patients receiving ART, with a routine VL expected, was conducted. We obtained data routinely captured on site and VL data from the laboratory system. A sample of 1 035 patient folders was reviewed. VL suppression was calculated using laboratory data, including all tests done, and compared with reported suppression based on on-site captured electronic data. Successful progression through each step on the VL cascade was estimated. We used logistic regression to identify factors associated with laboratory data and reported VL testing.
RESULTS: Of 22 991 patients for whom a routine VL test was due, 84% were done, 79% filed, 76% noted and 55% captured. Using all laboratory data, VL suppression was  estimated as 82%, 87%, 89% and 91% at the 50, 200, 400 and 1 000 copies/mL thresholds, respectively, but reported suppression using captured results was 80%, 86%, 88% and 89% at those thresholds. Routine VL testing is more likely to be done in children <15 years old (adjusted odds ratio (aOR) 1.89, 95% confidence interval (CI) 1.45 - 2.48) and pregnant women (aOR 1.90, 95% CI 1.28 - 2.81) than in men, adjusted for facility.
CONCLUSIONS: Despite a low reported completion, VL testing completion was high. Reported suppression using captured data was similar to suppression calculated using all laboratory data, which provided an accurate measure of progress towards the 90-90-90 target. More work is needed to reach the 16% of patients missed by routine testing.