Background/aims Few studies have objectively examined the quality of eye care in China. We assessed refractive care using the incognito standardised patient (SP) approach, a gold standard for evaluating clinical practice. Methods A total of 52 SPs were trained to provide standardised responses during eye examinations, and underwent automated and non-cycloplegic, subjective refraction by a senior ophthalmologist from Zhongshan Ophthalmologic Center, a national-level clinical and research centre. SPs subsequently received subjective refraction and eye exams at a randomly selected sample of 40 public hospitals and 93 private optical shops in Shaanxi, Northwestern China. Difference between expert and local refraction in the better-seeing eye was calculated by the vector diopteric method, and completeness of exams assessed against national standards. SP and provider demographic information and provider clinical experience were recorded. Results SPs (n=52, mean (range) age, 25.7 (22– 31) years, 28.8% male) underwent 133 eye exams (mean total duration 15.0$±$11.7 min) by 133 local refractionists (24– 60 years, 30.3% male). Only 93 (69.9%), 121 (91.0%) and 104 (78.2%) of local refractionists assessed vision, automated and subjective refraction, respectively. The median inaccuracy was -0.25 diopters (D), while 25.6% of results differed by an absolute value of $≥$1.0 D and 6.0% by $≥$2.0 D. Predictors of inaccurate refraction included spectacle power $<-$6.0 D (OR=2.66; 95% CI, 1.27 to 5.56), service at a public (vs private) hospital (OR=2.01; 95% CI, 1.11 to 3.63) and provider male sex (OR=2.03; 95% CI, 1.11 to 3.69). Conclusion Inaccurate refractions are common in Northwestern China, particularly in public facilities. Important assessments, including subjective refraction, are frequently omitted.