The objective of this study is to determine the diagnostic threshold of HbA1c for diabetes and the impact of using it on diabetes prevalence.A population-based stratified study was conducted in 2010 among community-dwelling adults aged ≥35years in southern China. Participants without previously-diagnosed diabetes (PDD) took oral glucose tolerance test (OGTT) and HbA1c assay. HbA1c diagnostic threshold was determined by receiver operating characteristic curve.A total of 6989 participants with mean age of 52years were recruited. The area under curve of HbA1c was 0.903 (95% CI: 0.883–0.922), with optimal cut-off value at 6.25% (sensitivity 75.6% and specificity 91.9%). There were 449 (6.42%) patients with PDD and 422 (6.04%), 815 (11.66%) and 918 (13.13%) new cases diagnosed by OGTT, HbA1c ≥6.25% or either, respectively. When either HbA1c or OGTT was used, newly-diagnosed diabetes prevalence increased by 117.4%.Diabetes is prevalent in southern China. Near half of the patients go undetected with current diagnostic criteria. HbA1c ≥6.25% may be the diagnostic threshold value but needs further verification. The introduction of HbA1c threshold into diabetes diagnosis in China will cause a substantial increase in diabetes prevalence and great challenge on the public healthcare system.