Context: The injuries caused by falls-from-height (FFH) are a significant public health concern. FFH is one of the most common causes of polytrauma. The injuries persist to be significant adverse events and a challenge regarding injury severity assessment to identify patients at high risk upon admission. Understanding the incidence and the factors that predict injury severity can help in developing effective intervention strategies. Artificial intelligence (AI) predictive models are emerging to assist in clinical assessment with challenges.
Objectives: This retrospective study investigated the incidence of FFH injuries utilizing conventional statistics and a predictive AI model to understand the fall-related injury profile and predictive factors.
Methods: A total of 124 patients who sustained injuries from FFHs were recruited for this retrospective study. These patients fell from a height of 15-30 feet and were admitted into a level II trauma center at the border of US-Mexica region. A chart review was performed to collect demographic information and other factors including Injury Severity Score (ISS), Glasgow Coma Scale (GCS), anatomic injury location, fall type (domestic falls vs. border wall falls), and comorbidities. Multiple variable statistical analyses were analyzed to determine the correlation between variables and injury severity. A machine learning (ML) method, the multilayer perceptron neuron network (MPNN), was utilized to determine the importance of predictive factors leading to in-hospital mortality. The chi-square test or Fisher's exact test and Spearman correlate analysis were utilized for statistical analysis for categorical variables. A p value smaller than 0.05 was considered to be statistically different.
Results: Sixty-four (64/124, 51.6 %) patients sustained injuries from FFHs from a border wall or fence, whereas 60 (48.4 %) sustained injuries from FFHs at a domestic region including falls from roofs or scaffolds. Patients suffering from domestic falls had a higher ISS than border fence falls. The height of the falls was not significantly associated with injury severity, but rather the anatomic locations of injuries were associated with severity. Compared with border falls, domestic falls had more injuries to the head and chest and longer intensive care unit (ICU) stay. The MPNN showed that the factors leading to in-hospital mortality were chest injury followed by head injury and low GCS on admission.
Conclusions: Domestic vs. border FFHs yielded different injury patterns and injury severity. Patients of border falls sustained a lower ISS and more lower-extremity injuries, while domestic falls caused more head or chest injuries and low GCS on admission. MPNN analysis demonstrated that chest and head injuries with low GCS indicated a high risk of mortality from an FFH.
Keywords: artificial intelligence; border fall; domestic fall; injury; machine learning; multilayer perceptron neuron network.
© 2024 the author(s), published by De Gruyter, Berlin/Boston.