Falls are the leading cause of unintentional injuries in preschool-aged children, and prinarily occur in the home environment. Injuries from falls at an early age have the potential to cause long-term effects on a child's physical, cognitive, and neurological development. Previous research has focused more on identifying risk factors and the occurrence of injuries, while this study aims to model the success of fall prevention in children based on a household-based risk management framework. This approach integrates the Safety I–III and the Theory of Graceful Extensibility (TGE) to evaluate the protective and adaptive capacity of family systems in sustaining child safety. The study used a cross-sectional design with 167 primary caregivers of preschool-aged children in Depok, who care for children in household settings. Data was collected through a questionnaire developed based on four main factors: the child, the home, the agent, and the companion's knowledge. The analysis was conducted in stages using Rasch measurement to test the validity and reliability of the instrument, Principal Component Analysis (PCA) to determine factor scores, Receiver Operating Characteristic (ROC) to establish the cut-off point for protective categories, and Bayesian Network and Bayesian Logistic Regression to map the probabilistic relationships between factors.
The results show that 65.3% of children experienced at least one fall in the past six months. The developed instrument proved to be valid and reliable in measuring the protective capacity of families against falls in children. Based on modeling results, the baseline probability of successful fall prevention for children, at 42%, reflects the limitations of the family system's protective capacity in actual conditions. Through sensitivity analysis, home factors and child factors were identified as the main protective factors against the success of fall prevention. Protective home and child conditions triple the system's chances of success (OR = 3.14). Furthermore, the what-if scenario simulation shows that strengthening home and child factors are the main leverage points in the system that can increase the probability of success in prevention by 53% and 52% respectively. Companion knowledge and agent factors play a role as adaptive enhancers in the next layer of protection.
The implication of these findings is the need to develop a household-based fall prevention program in the form of a layered protection system (barrier-based approach). The first barrier prioritizes strengthening home and child factors in parallel and interactively through improving the physical environment of the home and increasing knowledge and shaping basic child safety behaviors according to the developmental stage. The second barrier focuses on strengthening caregivers' knowledge, while the third barrier concentrates on controlling agent factors through arranging furniture and children's toys.
The success of fall prevention in preschool-aged children at home is the result of the adaptive and dynamic performance of the family system. Integrating risk management, Safety I–III, and the Theory of Graceful Extensibility allows fall prevention to be understood as a positive safety outcome resulting from the family system's capacity.

The tuberculosis treatment success rate in Indonesia in 2023 did not reach the 90% target. Treatment success impacts the reduction of infection spread and drug resistance cases, making early prediction of treatment success crucial. This study aims to develop a machine-learning model to predict treatment success. Data from Indonesia's Tuberculosis Information System (SITB) cohort was used. The study included productive-age patients (15-64 years) diagnosed with drug-sensitive tuberculosis who received treatment from January 1, 2020, to December 31, 2023. Data was randomly split into training (80%) and testing (20%) sets for model validation, with cross-validation performed. The algorithms used include decision tree, random forest, multilayer perception, extreme gradient boosting, and logistic regression. A consensus was reached for decision-making variables required in performing machine learning-based modeling of SITB data to predict treatment success using modeling of SITB data to predict treatment success using the Delphi method. The results of the study show that the random forest machine learning algorithm had the best performance and highest accuracy in predicting treatment success. This machine learning–based prediction tool can provide early predictions with SHAP (SHapley Additive ExPlanations) interpretation, helping healthcare workers make informed decisions more easily.