Determinants of organ donation knowledge and attitudes among university students in Oman: a secondary analysis using...

Organ donation remains insufficient globally, particularly in settings where sociocultural and religious perceptions shape decisions. This study examined knowledge and attitudes toward organ donation among university students and identified determinants using multivariable logistic regression and machine learning (ML) models.

This secondary cross-sectional study included 2,125 students. Knowledge and attitude scores were dichotomized at a 60% threshold. Variables with p < 0.05 in chi-square analyses were considered for multivariable logistic regression. Three ensemble ML models were trained using stratified 80/20 datasets and evaluated using accuracy, sensitivity, specificity, precision, recall, F1-score, and ROC-AUC. Feature-importance analysis assessed model interpretability.

Although awareness of organ donation was high (98.5%), only 17.6% demonstrated good knowledge and 28.6% exhibited a high attitude level. Multivariable logistic regression showed that a master's degree was associated with lower odds of good knowledge (OR = 0.39, p = 0.012), whereas knowing someone requiring transplantation (OR = 1.64, p = 0.020), believing that Islamic teachings permit donation (OR = 1.29, p = 0.033), and reporting other reasons for refusing organ donation (OR = 1.53, p = 0.023) increased knowledge levels. AdaBoost achieved the highest overall accuracy for knowledge classification (0.82), although this result should be interpreted cautiously because of its limited ability to identify participants with good knowledge. Feature-importance analysis identified personal exposure to transplantation, educational background, and refusal-related considerations as key predictors. Higher attitude scores were associated with perceiving organ donation as life-saving and religiously compatible (OR = 2.60, p < 0.001), altruistic motivation (OR = 1.60, p < 0.001), media influence (OR = 1.91, p < 0.001), belief that Islamic teachings permit organ donation (OR = 1.61, p < 0.001), financial motivation (OR = 2.70, p < 0.001), and reporting other reasons for refusing organ donation (OR = 1.66, p = 0.011), while lack of awareness was the strongest negative predictor (OR = 0.23, p < 0.001) along with perceived irrelevance, fear, perceived religious prohibition, and restrictive donation preferences. For attitude classification, AdaBoost also achieved the highest accuracy (0.79), while feature-importance analysis identified lack of awareness, perceived irrelevance, media influence, and life-saving/religious alignment as the most influential predictors.

Despite high awareness, notable gaps remain between knowledge, attitude, and behavioral intention. Logistic regression and ML modeling showed partially overlapping patterns, with both approaches identifying shared determinants of knowledge and attitudes toward organ donation. These findings suggest ML may complement conventional regression by supporting the predictive relevance and interpretability of key behavioral and perceptual factors.