Creatinine-based estimated glomerular filtration rate (eGFRcre) has been shown to overestimate the glomerular filtration rate (GFR) when it is compared with cystatin C-based estimated GFR (eGFRcys) in older people. We investigated clinical determinants of GFR overestimation by eGFRcre and developed a score for prediction of GFR overestimation (OE) in heart failure patients. We retrospectively examined 244 Japanese heart failure patients (aged 72.2 ± 13.1 years; 48% women) who had no known extrarenal factors that affect serum cystatin C concentration. eGFR OE by eGFRcre was defined as eGFRcre being ≥120% of cystatin C-based eGFR. The proportion of heart failure patients with OE was 14.3%. Patients with OE were older, had lower body weight and total skeletal muscle mass than those in patients without OE. Laboratory examinations showed that hemoglobin concentration was lower, and the ratio of blood urea nitrogen-to-creatinine was higher in patients with OE than in patients without OE. In multivariate regression analysis, body weight (<63.0 kg in men and <42.0 kg in women), hemoglobin level (<12.4 g/dL in men and <11.0 g/dL in women) and ratio of blood urea nitrogen-to-creatinine (>26.5) in addition to skeletal muscle mass were independently associated with OE. A score calculated by using cut-off levels of body weight, hemoglobin concentration and ratio of blood urea nitrogen-to-creatinine predicted OE with a sensitivity of 97.1% and a specificity of 98.1%. Overestimation of GFR by eGFRcre is predictable by a novel scoring system, which might be useful for the detection of patients who require cystatin C-based eGFR measurement for accurate assessment of renal function. Geriatr Gerontol Int 2020; 20: 752-758.