It is important to decide if a patient with diabetes has good glycaemic control in order to guide treatment and to offer behaviour change counselling. Currently, determining random blood glucose (RBG) is usually carried out in primary care in the public sector to make this decision. This study investigates the validity of these decisions. Retrospective data from a district hospital setting were used to analyse the correlation between glycated haemoglobin (HbA_1c) and RBG, the best predictive value of RBG, and its predictive properties. The best value of RBG to predict control (HbA_1c ≤ 7%) was 9.8 mmol/l. However, this threshold only gave a sensitivity of 77% and a specificity of 75%. Clinicians would be wrong 23% of the time when using RBG to determine glycaemic control. Attempts should be made to make HbA_1c more available for clinical decision-making. Point-of-care testing for HbA_1c should be considered.