Socioeconomic Factors Associated With Nonunion in Scaphoid Injuries: A Novel AI Model

Purpose Scaphoid fractures, the most common carpal fractures, are prone to nonunion because of their retrograde blood supply. Delayed presentation increases the risk of nonunion and progression to scaphoid nonunion advanced collapse. Social determinants of health, as measured by Area Deprivation Index (ADI), Social Vulnerability Index (SVI), and Child Opportunity Index (COI), can influence timely presentation, increasing the risk of nonunion. This study aims to develop a novel artificial intelligence model to predict scaphoid nonunion rates using socioeconomic indices alone. Methods A retrospective review identified 107 patients with suspected scaphoid fractures (January 2018–June 2025) who presented within 5 years of injury. National ADI, SVI, and COI scores were obtained. A binomial regression was used to validate ADI, SVI, and COI scores to predict nonunion status. A multilayer neural network was trained on the indices to predict nonunion status. Model performance was evaluated with area under the curve, and Youden's J statistic was maximized to determine sensitivity and specificity. Results Higher ADI (OR = 1.029; 95% CI, 1.009–1.049) and SVI (OR = 1.020; 95% CI, 1.006–1.035), and lower COI (OR = 0.986; 95% CI, 0.974–0.998) were independent predictors of nonunion on presentation, validating their usage. The neural network had an area under the curve of 0.879, and at a threshold of 0.21 (J = 0.707), the model achieved a sensitivity of 95.7%, specificity of 75.0%, precision of 0.512, and an F1 score of 0.667. Conclusions Our artificial intelligence model predicts nonunion status based purely on social determinants of health, emphasizing their influence on scaphoid fractures. Addressing these disparities is essential to improve timely access to care for injuries such as scaphoid fractures and may be crucial to other timely orthopedic fractures. Type of study/level of evidence Prevalence IIIb.