Background South Africa’s antiretroviral therapy (ART) programme is the largest globally and the universal test-and-treat policy is expected to increase the numbers on ART. This may have implications for treatment failure rates implying a greater future need for third-line regimens. South Africa initiated a third-line programme in 2013. However, there is little evidence quantifying the third-line need in this setting and the programme itself has not been formally evaluated. Objectives The study evaluated the third-line ART referral process in the Western Cape. Method Routinely collected data were analysed to derive an estimate of patients meeting criteria for third-line referral and compared with patients who were referred. Factors associated with referral were identified. Results In the study period, 947 patients met criteria for third-line referral and 167 patients were referred. Comparison revealed a poor overlap of only 42 patients. In multivariate analysis, factors associated with referral included receiving care at a hospital rather than a primary healthcare facility (adjusted odd ratios [aOR] = 2.15, 95% confidence interval [CI] 1.1–4.2), a higher number of viral load [VLs] ≥ 1000 copies/mL whilst on a protease inhibitor (PI) (aOR = 1.2, 95% CI 1.01–1.42) and a greater number of years on a PI (aOR = 1.25, 95% CI 1.07–1.46). Patients with a 6-month gap in dispensing were less likely to be referred (aOR = 0.37, 95% CI 0.17–0.81). Conclusion This study adds to a limited body of knowledge regarding third-line ART programmes. The findings indicate missed opportunities for and inappropriate referral of patients. Factors associated with referral were largely health system related. Clinician awareness and compliance with referral remain unknown and may be contributory.