- 1Department of Anesthesiology, Changzheng Hospital, Second Affiliated Hospital of Naval Medical University, Shanghai, China
- 2Jiangsu Province Key Laboratory of Anesthesiology, Xuzhou Medical University, Xuzhou, China
Aim: The objective of this study is to
develop accurate machine learning (ML) models for predicting the
neurological status at hospital discharge of critically ill patients
with hemorrhagic and ischemic stroke and identify the risk factors
associated with the neurological outcome of stroke, thereby providing
healthcare professionals with enhanced clinical decision-making
guidance.
Materials and methods: Data of stroke
patients were extracted from the eICU Collaborative Research Database
(eICU-CRD) for training and testing sets and the Medical Information
Mart for Intensive Care IV (MIMIC IV) database for external validation.
Four machine learning models, namely gradient boosting classifier (GBC),
logistic regression (LR), multi-layer perceptron (MLP), and random
forest (RF), were used for prediction of neurological outcome.
Furthermore, shapley additive explanations (SHAP) algorithm was applied
to explain models visually.
Results: A total of 1,216 hemorrhagic
stroke patients and 954 ischemic stroke patients from eICU-CRD and 921
hemorrhagic stroke patients 902 ischemic stroke patients from MIMIC IV
were included in this study. In the hemorrhagic stroke cohort, the LR
model achieved the highest area under curve (AUC) of 0.887 in the test
cohort, while in the ischemic stroke cohort, the RF model demonstrated
the best performance with an AUC of 0.867 in the test cohort. Further
analysis of risk factors was conducted using SHAP analysis and the
results of this study were converted into an online prediction tool.
Conclusion: ML models are reliable tools
for predicting hemorrhagic and ischemic stroke neurological outcome and
have the potential to improve critical care of stroke patients. The
summarized risk factors obtained from SHAP enable a more nuanced
understanding of the reasoning behind prediction outcomes and the
optimization of the treatment strategy.
Introduction
Stroke encompasses a set of conditions characterized by
the sudden rupture or occlusion of cerebral blood vessels, ultimately
resulting in insufficient blood flow and subsequent damage to brain
tissue. Clinically, stroke is broadly classified into two main
types—ischemic and hemorrhagic—with the latter comprising intracerebral
and subarachnoid hemorrhage forms (1).
Stroke affects a staggering one in every four individuals over 25 years
of age, rendering it the second most common cause of mortality and
third leading cause of disability among adult populations worldwide (2).
Approximately 16 million people worldwide suffer from various motor and
cognitive impairments as a result of stroke, which are often
unavoidable sequelae for stroke patients, and severely affects the
mobility and quality of life of stroke victims (3).
Acute stroke patients often enter the intensive care
unit (ICU) due to consciousness disorders, cardiopulmonary
complications, circulatory instability, or acute thrombolytic therapy (4).
Compared with patients admitted to a dedicated neurological ward or
stroke unit, those with stroke who are admitted to the ICU exhibit
heightened neurological severity, notable impairment of consciousness at
a moderate to severe level, often necessitating mechanical ventilation,
and encounter an elevated risk of hospital mortality (5, 6).
ICU provides complex and resource-intensive treatment for hospitalized
patients with severe conditions, but current medical resources are often
insufficient to meet the needs of ICU patients, and hospitals face
pressure to improve critical care efficiency and reduce costs (7).
Early prediction of neurological outcome in critically ill stroke
patients can provide important references for patients and their
families, and can also guide clinicians to give the best intervention
measures to patients.
In contrast to conventional predictive models that rely
on established variables for computation, machine learning (ML)
approaches offer the distinct advantage of incorporating a broader range
of variables that more comprehensively capture the intricacies and
inherent unpredictability of human physiology (8, 9).
Consequently, ML has emerged as a promising tool in the medical field,
with its capacity to integrate abundant variables, extract nuanced
insights, and generalize acquired knowledge to novel cases with
remarkable efficiency and precision (10, 11).
Furthermore, interpretable machine learning is increasingly being
applied in clinical research, demonstrating robust clinical
applicability and guiding capabilities (12, 13).
In this work, we aimed to construct ML models for early
and effective prediction of neurological outcome at hospital discharge
in critically ill patients with hemorrhagic and ischemic stroke, and
employed the shapley additive explanations (SHAP) methods to elucidate
the underlying reasons and decision-making processes involved within the
optimal algorithm.
More at link.
