Slip-related accidents remain a critical safety challenge in hospital environments due to high pedestrian traffic, frequent contamination, and heterogeneous flooring conditions. This study proposes a novel and integrative framework for slip risk assessment based on in situ friction measurements, statistical analysis, machine learning, and explainable artificial intelligence. Dynamic Coefficient of Friction (DCOF) measurements were collected from two public hospitals across five functional zones (corridors, patient rooms, toilets, showers, and examination rooms), under both dry and wet conditions, resulting in a dataset of 540 observations. A composite Slip Risk Index (SRI) was developed by integrating frictional performance with environmental, spatial, and material-related factors through a weighted aggregation scheme. Paired-sample statistical analyses confirmed a very strong inverse relationship between DCOF and SRI, while correlation and feature importance analyses revealed that environmental condition and flooring material act as significant secondary contributors. Ensemble-based machine learning models, including Random Forest, Gradient Boosting, Bagging, CatBoost, XGBoost, and LightGBM, were employed to predict SRI values. Among these, Gradient Boosting demonstrated the most robust and balanced performance, achieving near-perfect goodness-of-fit and minimal prediction error. Explainable machine learning analysis using SHAP further validated the dominance of frictional performance in slip risk formation and ensured model transparency. The proposed framework bridges the gap between traditional tribological assessment and data-driven predictive safety management by providing an interpretable, accurate, and scalable decision-support tool. The findings offer actionable insights for hospital designers, facility managers, and safety engineers, supporting evidence-based flooring selection, environmental control, and preventive maintenance strategies aimed at reducing slip-and-fall accidents in healthcare environments.
Slip-related accidents remain a critical safety challenge in hospital environments due to high pedestrian traffic, frequent contamination, and heterogeneous flooring conditions. This study proposes a novel and integrative framework for slip risk assessment based on in situ friction measurements, statistical analysis, machine learning, and explainable artificial intelligence. Dynamic Coefficient of Friction (DCOF) measurements were collected from two public hospitals across five functional zones (corridors, patient rooms, toilets, showers, and examination rooms), under both dry and wet conditions, resulting in a dataset of 540 observations. A composite Slip Risk Index (SRI) was developed by integrating frictional performance with environmental, spatial, and material-related factors through a weighted aggregation scheme. Paired-sample statistical analyses confirmed a very strong inverse relationship between DCOF and SRI, while correlation and feature importance analyses revealed that environmental condition and flooring material act as significant secondary contributors. Ensemble-based machine learning models, including Random Forest, Gradient Boosting, Bagging, CatBoost, XGBoost, and LightGBM, were employed to predict SRI values. Among these, Gradient Boosting demonstrated the most robust and balanced performance, achieving near-perfect goodness-of-fit and minimal prediction error. Explainable machine learning analysis using SHAP further validated the dominance of frictional performance in slip risk formation and ensured model transparency. The proposed framework bridges the gap between traditional tribological assessment and data-driven predictive safety management by providing an interpretable, accurate, and scalable decision-support tool. The findings offer actionable insights for hospital designers, facility managers, and safety engineers, supporting evidence-based flooring selection, environmental control, and preventive maintenance strategies aimed at reducing slip-and-fall accidents in healthcare environments.