To describe and compare health literacy (HL) using contrasting instruments of Rapid Estimate of Adult Literacy in Medicine (REALM) and the Health Literacy Questionnaire (HLQ) in patients attending three Australian rheumatology clinics and their associations with sociodemographic and clinical characteristics. We conducted a cross-sectional study of rheumatology patients at Liverpool (Sydney) and Canberra public hospital clinics and a private Liverpool clinic using the REALM test and HLQ sociodemographic and clinical questionnaires. One-way analysis of variance and multivariate ordinal and binary logistic regressions were used to examine the association between HL and sociodemographic and clinical factors. A total of 278 people participated (111 Liverpool public, 112 private, 55 Canberra public clinic). Compared with Canberra, Liverpool patients were older, non-native English speakers, had less than a tertiary education and were more likely to have comorbidities. A total of 40% of Liverpool public, 29% of private and 13% of Canberra patients had suboptimal REALM HL scores (P < 0.05). Lower REALM scores were associated with male gender, <10 years of education, being born outside Australia/New Zealand and English as a secondary language, but was not associated with clinical characteristics. All nine HLQ domain scores did not differ across sites despite intergroup comparisons as stratified by REALM bands. Completing 10 years of schooling was associated with better scores in three HLQ scales: 'Navigating the healthcare system' (P = 0.009), 'Ability to find information' (P < 0.001) and 'Reading and understanding information' (P < 0.001). Patient HL as measured by the REALM differed across sites, explained by differences in gender distribution, country of birth, English as the primary language and years of education. HL as measured by the HLQ did not differ across sites and was more informative for understanding differing patient HL needs.