Objective To investigate the correlation of body mass index (BMI) and body composition with cognitive function in the elderly population. Methods A total of 391 elderly patients who visited the nutritional clinic of our medical center from January to December 2021 were recruited as the research objects. Montreal Cognitive Assessment (MoCA) scale was used to assess the patients, and according to the results, the patients were divided into normal group (≥26), mild group (17-26) and moderate-severe group (≤17). Their height and weight were measured and BMI was calculated. Multi-frequency bioelectrical impedance analysis (MF-BIA) was used to obtain body composition, and the correlation of BMI and body composition with cognitive function was analyzed. When the diagnostic grade of cognitive impairment (CI) was used as the dependent variable, BMI and related body composition were regarded as the independent variables, multivariate logistic regression analysis was carried out after the adjustment of confounding factors such as hypertension and diabetes. SPSS statistics 25.0 was used for statistical analysis. Data comparison between two groups was performed using LSD-t test, Fisher exact test or χ²test depending on data type. Results There were significant differences in BMI, body fat percentage (PBF) and waist-hip ratio (WHR) among the normal group, mild group and moderate-severe group (P<0.05). The body fat mass (BFM) and visceral fat area (VFA) showed an increasing trend with the increase of the severity of cognitive dysfunction in the 3 groups [(22.45±4.97) vs (26.97±5.54) vs (31.10±3.45) kg; (104.42 ±26.73) vs (125.57±29.58) vs (177.01±25.26) cm²; P<0.05] . The moderate-severe group had significantly lower soft lean mass (SLM), skeletal muscle mass (SMM) and basal metabolic rate (BMR) when compared with the normal and mild groups [(37.97±8.16) vs (48.06±8.00) and (47.70±9.44) kg, (21.42±5.60) vs (28.14±5.11) and(28.00±6.04) kg,(1 238.95±205.84) vs(1 469.33±182.12) and(1 463.94±215.27) kcal; all P<0.05). The results of multivariate logistic regression analysis showed that the obesity group had a higher risk of CI than the normal weight group (OR=2.145, P<0.05); the WHR exceeding standard group showed a higher risk of CI than the normal WHR group (OR=21.475, P<0.05); the PBF Q2, Q3 and Q4 groups had a higher risk of CI when compared with the PBF Q1 group (OR=14.324,22.298 and 39.175; P<0.05); and the BFM Q3 group and Q4 group presented a higher risk of CI than the BFM Q1 group (OR=14.588,26.239; P<0.05). Conclusion BMI and obesity are correlated with the occurrence of cognitive dysfunction in the elderly.