July 10, 2023BMC GeriatricsOpen Access

XGBoost-based machine learning test improves the accuracy of hemorrhage prediction among geriatric patients with long-term administration of rivaroxaban

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Hemorrhage risk in geriatric patients on long-term rivaroxaban (n=798). Intervention: XGBoost-based machine learning model vs. Logistic regression and random forest models. Primary outcome: Area under the receiver operating characteristic curve (AUC) for predicting hemorrhage risk (AUC 0.776, 95% CI 0.687-0.864).

View Full Paper

Ask AI

Bookmark

Key result

An XGBoost-based machine learning model demonstrated superior predictive performance for hemorrhage risk in geriatric patients on long-term rivaroxaban, achieving an AUC of 0.776 compared to 0.679 for conventional logistic regression.

Why the study?

Hemorrhage is a serious adverse drug reaction in geriatric patients receiving long-term rivaroxaban, creating a need for effective bleeding prediction models to improve clinical safety.

Does an XGBoost-based machine learning model improve the accuracy of hemorrhage prediction in geriatric patients on long-term rivaroxaban compared to conventional models?

Population

798 geriatric patients over 70 years of age requiring long-term rivaroxaban anticoagulation

Comparison

XGBoost model vs random forest and conventional logistic regression

Design

Observational prediction model study

Authors

CCCheng ChenShanghai Jiao Tong University CYChun YinWuhan University YWYanhu WangChinese PLA General Hospital

Discussion

Member takes

Implication

May support ML hemorrhage prediction in geriatric rivaroxaban users; leaves open prospective validation and outcome impact.

Study Design

Type

Observational (n=798)

Multicenter

Structured PICO

Does an XGBoost-based machine learning model improve the accuracy of hemorrhage prediction in geriatric patients on long-term rivaroxaban compared to conventional models?

Population

798 geriatric patients (aged ≥70 years, median age 92) requiring long-term rivaroxaban therapy for anticoagulation, followed to predict hemorrhage events.

Exposure

XGBoost-based machine learning model utilizing 27 clinical indicators to predict hemorrhage risk

Comparator

Conventional logistic regression and random forest models

Outcome

Bleeding adverse events (including gastrointestinal, intracranial, ocular, urinary, pulmonary, nasal, gingival, and knee joint cavity hemorrhage) after treatment with rivaroxaban for more than 3 monthssafety

Main Result

Effect estimate: AUC 0.776 (95% CI 0.687-0.864)

Absolute Event Rate: 0.776% vs 0.679%

An XGBoost-based machine learning model demonstrated superior accuracy compared to traditional logistic regression in predicting hemorrhage risk among geriatric patients on long-term rivaroxaban.

Limitations

Single-center study with a relatively small sample size, leading to a wide confidence interval.
Clinical information, especially hemorrhagic history, was self-reported and susceptible to recall bias.
Missing BMI values for bedridden geriatric patients over 90 years old.
Potential unmeasured confounding factors, such as the level and activity of factor Xa, which were not routinely tested.
Small sample size leading to a wide confidence interval for the XGBoost-based bleeding prediction model
Clinical information, especially hemorrhagic history, was reported by patients and susceptible to recall bias
Missing BMI values due to geriatric patients over 90 years old mostly being bedridden
Potential residual confounding factors not fully adjusted for

Cite This Study

Chen et al. (2023) conducted an observational in Hemorrhage risk in geriatric patients on long-term rivaroxaban (n=798). XGBoost-based machine learning model vs. Logistic regression and random forest models was evaluated on Area under the receiver operating characteristic curve (AUC) for predicting hemorrhage risk (AUC 0.776, 95% CI 0.687-0.864). An XGBoost-based machine learning model demonstrated superior predictive performance for hemorrhage risk in geriatric patients on long-term rivaroxaban, achieving an AUC of 0.776 compared to 0.679 for conventional logistic regression.

synapsesocial.com/papers/6a818261f2736d406858b5ff https://doi.org/10.1186/s12877-023-04049-z

View Full Paper

Ask AI

Bookmark

XGBoost-based machine learning test improves the accuracy of hemorrhage prediction among geriatric patients with long-term administration of rivaroxaban

Discussion

Member takes

Implication

Study Design

Structured PICO

Main Result

Limitations

Cite This Study

XGBoost-based machine learning test improves the accuracy of hemorrhage prediction among geriatric patients with long-term administration of rivaroxaban

Discussion

Member takes

Implication

Study Design

Structured PICO

Main Result

Limitations

Cite This Study

Also Consider

Also Consider