Burn patients are at risk for organ dysfunction secondary to “burn shock” and sepsis. The initial 24 hours following major burn injury presents with significant hypotension due to profound systemic inflammation. This severe hypotension is further exacerbated by excessive evaporative water loss directly from the burn wound. Following the acute resuscitation period, burn patients are also at high risk for infection with sepsis-related multi-organ dysfunction being the most frequent cause of burn-related deaths. Acute kidney injury (AKI) is reported to be the most common type of organ dysfunction in both “burn shock” and sepsis.
Traditional indicators of AKI include creatinine and urine output. Unfortunately, both biomarkers have known limitations—delaying recognition of AKI. The use of artificial intelligence (AI) and machine learning (ML) has been proposed to offset these limitations. Our study implemented the first AI/ML algorithm to quickly predict sepsis and AKI with and without novel renal injury biomarkers such as neutrophil gelatinase associated lipocalin (NGAL). For sepsis, our custom AI/ML technique exhibited clinical sensitivity and specificity superior to burn sepsis criteria recommended by the American Burn Association. For AKI, a separate AI/ML algorithm was developed and predicted AKI in burn patients nearly two days earlier than traditional techniques. This same AI/ML algorithm was then tested (generalization) in a mixed burned and non-burned trauma patient population. As in the burn-only cohort, the AI/ML algorithm enhanced the performance of traditional biomarkers.
Artificial intelligence and ML may enable early recognition of challenging diseases such as AKI and sepsis in high-risk populations. Interestingly, AI/ML could enhance the predictive performance of legacy disease biomarkers. More studies are needed to determine the clinical impact of early sepsis and AKI recognition by AI/ML in the severely burned population.
1. Describe the current state of artificial intelligence (AI) / machine learning (ML) in society versus laboratory medicine.
2. Identify best practices for developing and validating a health-related AI/ML algorithm.
3. Describe the clinical value of AI/ML for burn-related sepsis and acute kidney injury.