To explore the relationship between liver fat content (LFC) and nonalcoholic fatty liver disease (NAFLD) and determine the new threshold of LFC to diagnose NAFLD. The data from questionnaire survey, general physical examination, laboratory examination, and image examination were collected. Multivariate regression analysis, receiver operating characteristic curve analysis, smooth curve fitting, and threshold effect analysis were performed using the R software to investigate the relationship between LFC and NAFLD and to identify the new threshold of LFC to diagnose NAFLD. The prevalence of NAFLD was 30.42%, with a significantly higher prevalence in men than in women. Regression analyses demonstrated that LFC odds ratio [95% confidence interval (CI)] was 1.28 (95% CI: 1.24-1.31) in fully-adjust model. Analysis of the LFC quartile, with Q1 as a reference, revealed that the odds ratios of NAFLD were 1.47 (95% CI: 1.08-1.99), 2.29 (95% CI: 1.72-3.06), and 10.02 (95% CI: 7.45-13.47) for Q2, Q3, and Q4 groups, respectively. Smooth curve fitting and threshold effect analysis displayed a nonlinear relationship between LFC and NAFLD, and the threshold was 4.5%. The receiver operating characteristic curve indicated that when LFC was 4.5%, the area under curve (95% CI) was 0.80 (0.79-0.82), and the sensitivity and specificity of LFC in diagnosing NAFLD were 0.64% and 0.82%, respectively. The relationship between LFC and NAFLD was sigmoidal, with an inflection point of 4.5%.