Establishment and validation of clinical prediction model and prognosis of perioperative pneumonia in elderly patients with hip fracture complicated with preoperative acute heart failure

Abstract

Background: Elderly hip fracture was a common orthopedic emergency with high perioperative complication risks. Combined with preoperative acute heart failure, the risk increases further, with pneumonia being a common complication. The aim of this study was to construct and evaluate risk factor prediction models for perioperative pneumonia in these patients and to explore prognostic factors.

Methods: A retrospective study design was used to collect data on elderly patients with hip fracture combined with preoperative acute heart failure at the Third Hospital of Hebei Medical University from January 2020 to December 2022. The feature variables were screened by logistic regression and nomogram was constructed. The receiver operating characteristics curve (ROC), decision curve analysis (DCA), and calibration curve were employed to assess the predictive power of the model. Correlation heatmaps and shapley additive explanation (SHAP) were employed to assess key variables and their impact. Employing the Kaplan-Meier curve and Cox regression, the patients' prognosis was ultimately evaluated.

Results: 535 elderly patients with hip fracture combined with preoperative acute heart failure were included in this study. Logistic regression analysis was used to identify combined respiratory disease, hemoglobin, albumin, neutrophils, and blood glucose as independent danger factors for perioperative pneumonia (p < 0.05). The nomogram was designed to display the outcomes instinctively, with an AUC of 0.819. The model was internally validated by initiating self-sampling 1000 times. The calibration curve indicated that the model had excellent treaty. The DCA curve showed that the model had good validity and clinical practicability. Correlation heatmaps and SHAP were visualized and analyzed. The K-M curves indicated that the prognosis of the non-pneumonia group was better than that of the pneumonia group (p = 0.014). COX regression analysis found that the major risk factors for all-cause mortality in patients with acute heart failure(AHF) were age, brain natriuretic peptide, platelet count, and combined respiratory failure (p < 0.05).

Conclusion: The prediction model, established in this study, was highly accurate and proved a potent instrument for clinical evaluation of the perioperative pneumonia risk of elderly hip fracture patients with preoperative acute heart failure. We hope that this study can reduce the occurrence of perioperative pneumonia in patients and improve the prognosis of patients.

Keywords: Acute heart failure; Clinical prediction model; Elderly hip fracture; Perioperative pneumonia; Prognosis.

Fig. 1

The patient flow chart in our study

Fig. 2

AUC values for important characteristic variables

Fig. 3

The correlation heat map of the parameters. Blue indicates a positive correlation and red indicates a negative correlation

Fig. 4

(A) Nomogram representation for predicting outcome.0: Without respiratory disease; 1: respiratory disease; HB: hemoglobin; ALB: albumin; NEUT: neutrophils; GLU: glucose. (B) Receiver operating characteristic. (ROC) curves for ROC of the nomogram. (C) The Calibration curve of the model. (D) The Decision Curve Analysis (DCA) of the model

Fig. 5

SHAP explanatory model (A) Attributes of characteristics in SHAP. Yellow indicates high eigenvalues and purple indicates low eigenvalues. The horizontal coordinate is SHAP value. (B) Interdependence graph. The predictive characteristics of individual patients and the contribution of each to the predictive. (C) waterfall plot (D) Single observation diagram. The SHAP value represents the predictive characteristics of individual patients and the contribution of each to the predictive mortality. The number in bold is the probability forecast value (f(x)), while the base value is the predicted value without providing input to the model. F(x) is the logarithmic ratio of each observation. Red features indicate increased risk of pneumonia and blue features indicate reduced risk of pneumonia

Fig. 6

Kaplan-Meier curve for all-cause mortality in patients with and without pneumonia