Epidemiological studies have reported that among participants with impaired cognitive, overweight and mild obesity are associated with substantially improved survival, this finding has been termed the "obesity paradox" and has led to uncertainty about secondary prevention. To explore whether the association of BMI with mortality differed in different MMSE score, and whether the obesity paradox in patient with cognitive impairment (CI) is real. The study used data from CLHLS, a representative prospective population-based cohort study in China, which included 8348 participants aged ≥ 60years between 2011 and 2018. The independent association between BMI and mortality in differed MMSE score by calculating hazard ratios (HRs) in multivariate Cox regression analysis. During a median (IQR) follow-up of 41.18months, a total of 4216 participants died. In the total population, underweight increased the risk of all-cause mortality (HRs, 1.33; 95% CI 1.23-1.44), compared with normal weight, and overweight was associated with a decreased risk of all-cause mortality (HR 0.83; 95% CI 0.74-0.93). However, compared to normal weight, only underweight was associated with increased mortality risk among participants with MMSE scores of 0-23, 24-26, 27-29, and 30, and the fully-adjusted HRs (95% CIs) for mortality were 1.30 (1.18, 1.43), 1.31 (1.07, 1.59), 1.55 (1.34, 1.80) and 1.66 (1.26, 2.20), respectively. The obesity paradox was not found in individuals with CI. Sensitivity analyses carried out had hardly any impact on this result. We found no evidence of an obesity paradox in patients with CI, compared with patients of normal weight. But underweight individuals may have increased mortality risk whether in the population with CI or not. And overweight/obese people with CI should continue to aim for normal weight.