Leveraging Machine Learning to Assess Secondary Contamination Risks in Laboratory Waste Systems: A Biosafety Perspective

This study investigates secondary contamination risks in laboratory waste handling and disposal systems, leveraging machine learning to assess risks and evaluate biosafety practices. The study employed a quantitative, cross-sectional design, surveying 351 laboratory personnel across institutional, clinical, and university labs utilizing a validated questionnaire (Cronbach’s α > 0.79) to evaluate waste characteristics, handling, treatment, disposal, biosafety training, and policy compliance. Data were analyzed using statistical methods (Friedman Test, Kruskal-Wallis, Spearman’s Rank Correlation) and machine learning models (Logistic Regression, Random Forest, XGBoost). Friedman tests revealed significant differences in perceived contamination risks (χ²(4) = 441.23, p < .001), with waste generation, improper storage, and infection potential as key drivers. K-means clustering identified three distinct waste handling patterns, strongly associated with personnel experience (χ²(8) = 495.16, p < .05). XGBoost regression (R² = 0.997) confirmed biosafety training’s critical role in effective waste treatment, with disposal facility inspections and regulatory compliance as top predictors. Logistic regression showed a significant association between institutional policy compliance and biosafety awareness (β = 0.493, p = .002), though classification accuracy was limited (53.9%). Random Forest models (96.66% variance explained) validated these findings, emphasizing robust waste management protocols. In conclusion, these findings underscore the need for targeted training and stringent compliance to mitigate secondary contamination in laboratory settings, offering a scalable ML framework for biosafety risk assessment.