Construction of a fall risk prediction model for patients with hematologic malignancies based on the LASSO-Logistic regression
10.3760/cma.j.cn211501-20240108-00070
- VernacularTitle:基于LASSO-Logistic回归构建血液系统恶性肿瘤患者跌倒风险预测模型
- Author:
Weifang LI
1
;
Xuebin JI
;
Lanhua LI
;
Yunling HAN
;
Lujing XU
;
Xiaoya LIU
Author Information
1. 山东大学齐鲁医院血液科，济南 250012
- Keywords:
Accidental falls;
Hematologic malignancies;
Near falls;
Prediction model;
Dynamic nomogram
- From:
Chinese Journal of Practical Nursing
2024;40(23):1789-1795
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a fall risk prediction model for patients with hematologic malignancies and to provide a reference for the risk assessment and accurate management of falls.Methods:The prospective study design was adopted to facilitate the selection of 510 patients with hematologic malignant in Qilu Hospital of Shandong University for investigation, and relevant data such as patient demographic characteristics, disease treatment and drugs were collected. The LASSO-Logistic regression was used to screen the risk factors of falls in patients with hematologic malignancies, to construct a nomogram risk prediction model. The receiver operating characteristic curve (ROC) and calibration curve were used to evaluate the predictive performance of the model. Bootstrap resampling were used to validate internal validation of the model.Results:Among 510 patients with hematological malignancies, there were 273 males and 237 females, aged 53.0 (41.0, 63.0) years old. A total of 6 risk factors were included in the fall risk prediction model for patients with hematological malignancies, which were disease type ( OR = 0.185, 95% CI 0.061 - 0.562), body temperature ≥38 ℃ ( OR = 2.239, 95% CI 1.128 - 4.445), pain ( OR = 15.581, 95% CI 6.592 - 36.829), anemia ( OR = 4.097, 95% CI 1.536 - 10.927), days of bone marrow suppression ( OR = 3.341, 95% CI 1.619 - 6.893), and assessment of daily self-care ability ( OR = 3.160, 95% CI 1.051 - 9.506)(all P<0.05). The ROC curve of the fall risk prediction model was 0.884 (95% CI 0.841-0.927). The optimal threshold, sensitivity, and specificity of the risk prediction model were 0.248, 87.4% and 75.6%. The internal validation C statistic was 0.873. The Calibration curve was almost coincides with the ideal curve, and the model Brier score was 0.080. Conclusions:The constructed fall risk prediction model has good predictive performance, which can efficiently and objectively quantify the risk of falls, and provide a reference for the early assessment and effective prevention of falls in patients with hematological malignancies.
