Explainable Boosting Machine approach identifies risk factors for acute renal failure

Area of Science:

• Critical Care Medicine
• Nephrology
• Machine Learning in Healthcare

Background:

• Accurate risk stratification and outcome prediction are vital for intensive care unit (ICU) resource planning.
• Acute kidney injury (AKI) significantly impacts patient outcomes in critical care settings.
• Large datasets in ICUs necessitate advanced analytical methods for identifying AKI determinants.

Purpose of the Study:

• To identify key determinants of AKI in ICU patients using a novel machine learning model.
• To enhance the precision of AKI risk factor identification beyond traditional statistical methods.
• To provide a more nuanced understanding of AKI risks for improved predictive modeling.

Main Methods:

• Analysis of 3572 ICU patients.
• Utilized Explainable Boosting Machine (EBM), a novel machine learning model.
• Examined variables including central venous pressure (CVP), mean arterial pressure (MAP), age, gender, comorbidities, and surgical history.

Main Results:

• Significant comorbidities associated with AKI risk include chronic kidney disease, heart failure, arrhythmias, liver disease, and anemia.
• Lower GI surgery was linked to increased AKI risk, whereas neurosurgery was associated with reduced risk.
• EBM identified anemia, liver disease, and average CVP as increasing AKI risk, while neurosurgery decreased it. Age >50 and MAP <65 mmHg (threshold effect at 60 mmHg) were progressive risk factors. A critical CVP threshold for AKI risk was observed at 10.7 mmHg.

Conclusions:

• Explainable Boosting Machine (EBM) enhances the precision of AKI risk factor identification in ICU patients.
• This machine learning approach offers a more nuanced understanding of known AKI risks compared to traditional models.
• The findings support refined predictive modeling for AKI, overcoming limitations of conventional statistical methods.