Objective To explore the risk factors of cognitive decline (CF) in elderly patients with chronic obstructive pulmonary disease (COPD), and construct a nomogram model.

Methods A retrospective analysis was conducted on 193 elderly patients with COPD. The Fried frailty phenotype and Montreal Cognitive Assessment Scale were used to jointly evaluate whether CF was combined. According to whether COPD patients were combined with CF, they were divided into the CF group (85 cases) and the non-CF group (108 cases). The relevant clinical data of patients, which included the gender, age, body mass index, other underlying diseases (hypertension, diabetes and coronary heart disease), laboratory indicators (albumin, fasting blood glucose, hemoglobin, total cholesterol and triglycerides), 15-point Geriatric Depression Scale (GDS-15) score, Barthel index and Social Support Rating Scale (SSRS) score and calf circumference, were collected. The influencing factors of COPD elderly patients complicated with CF were analyzed. The nomogram model for CF risk prediction, ROC curve and correction curve of elderly patients with COPD were drawn using R language.

Results There was no statistical significance in the gender, underlying diseases (hypertension, diabetes and coronary heart disease), fasting blood glucose, body mass index, hemoglobin, total cholesterol and triglycerides between the CF group and non-CF group (P ＞ 0.05). The age and GDD-15 score of the CF group were higher than those of non-CF group (P ＜ 0.01). The albumin level, Barthel index, SSRS score and calf circumference in the CF group were lower than those in non-CF group (P ＜ 0.01). Binary logistic regression analysis indicated that the advanced age and high GDS-15 score were the risk factors of CF occurrence (P ＜ 0.05 and P ＜ 0.01), while the high albumin, high Barthel index, high SSRS score and large calf circumference were the protective factors (P ＜ 0.01). Based on binary multivariate logistic regression analysis, a nomogram model for predicting the risk of CF in elderly patients with COPD was constructed, and the AUC of ROC curve was 0.985 (95%CI: 0.972–0.999), the sensitivity of model was 0.950, the specificity was 0.935, the Hosmer-Lemeshow test result was χ² = 5.26, P = 0.73. The model was verified by the Bootstrap method, the C-index value was 0.985, and the predictive correction curve approached the ideal curve.

Conclusions The advanced age, high GDS-15 score, low albumin, low Barthel index, low SSRS score and small calf circumference are the risk factors of CF in COPD patients. The nomogram model has good predictive performance for the occurrence of CF, and can be popularized and applied in clinical practice.