Establishment and effect evaluation of nomogram model for diagnosis and prediction of pulmonary hypertension in patients with chronic obstructive pulmonary disease

Abstract

Objective: To construct a diagnostic and predictive model for chronic obstructive pulmonary disease complicated with pulmonary hypertension (COPD-PH) and evaluate its effect. Methods: A total of 1 514 COPD patients treated in 5 hospitals from January 1, 2014 to December 31, 2019 were retrospectively collected and divided into training cohort (1 072 cases) and validation cohort (442 cases) according to the ratio of 7∶3 according to the inclusion time. Data including demographic data, smoking status, history of disease, and clinical examination were collected through patient medical records and electronic medical record systems. Multivariate logistic regression models were used to explore the related factors of COPD-PH, and the nomogram model was constructed using the "rms" program package. The calibration curve was used to evaluate the consistency between the prediction probability of the model and the actual results. The C index and the area under the receiver operating characteristic curve (ROC) were used to evaluate the discrimination of the model. The decision curve analysis (DCA) was used to evaluate the clinical practicability of the model. Results: In the training cohort, 3.7%, 15.2% and 81.1% were aged 50-59, 60-69 and ≥70 years, respectively, which were significantly different from the age composition of the validation cohort (7.9%, 27.8% and 64.3%, respectively) (P=0.041). There was no significant difference between the training cohort (79.4%) and the validation cohort (84.6%) (P=0.243). Multivariate logistic regression analysis of the training cohort showed that age ≥70 years [OR (95%CI): 3.32 (1.49-7.36)] and smoking status [former (current) smoking, OR (95%CI)] were 3.67 (2.51-5.37) and 2.04 (1.44-2.90), respectively], NT-probNP≥1 400 ng/L[OR (95%CI): 9.88 (6.23-15.66)], right atrial diameter [OR (95%CI): 1.11 (1.07-1.15)] was COPD-related factors of PH, based on the above factors-PH nomogram COPD model was set up and develop for online tools (https://ph-666.shinyapps.io/zhonghua-PH/). The calibrated C index (95%CI) of the training cohort and the validation cohort were 0.82 (0.77-0.87) and 0.77 (0.68-0.86), respectively. The calibration curve was close to the diagonal in both the training cohort and the validation cohort. The AUC (95%CI) of the nomogram model was 0.82 (0.80-0.85) in the training cohort and 0.77 (0.73-0.82) in the validation cohort. ROC curve showed that the optimal threshold in the training cohort was 0.60, and the sensitivity and specificity under this value were 0.74 and 0.78, respectively; the optimal threshold for the validation cohort was 0.70, and the sensitivity and specificity under this value were 0.76 and 0.65, respectively. DCA analysis showed that the nomogram model provided better net benefits than the all-variable selection and no-variable selection strategies with threshold probabilities greater than 15.0% and 13.0% in the training and validation cohorts, respectively. Conclusions: The nomogram model for the diagnosis and prediction of COPD-PH is simple and accurate, which has a good clinical application prospect.