Modeling the variables that influence substance consumption of people who experience homelessness in Colombia.

Background: Homelessness in Colombia is a critical social issue that is strongly associated with psychoactive substance use. This study aims to model the variables influencing substance use among individuals experiencing homelessness in Colombia, offering insights to inform public policy design. Methods: This research draws on data from the National Administrative Department of Statistics (DANE) censuses conducted in 2017, 2019, and 2021. A two-stage quantitative methodology was applied: (1) descriptive analysis of the demographic and socioeconomic characteristics of the homeless population, and (2) predictive modeling using random forest algorithms to identify key variables associated with substance use. While results reveal strong associations, they do not imply causality. The study focuses on available variables, acknowledging the absence of psychosocial factors and the need for complementary qualitative research. Results: The analysis identified age and the duration of homelessness as the most influential variables associated with substance use. Descriptive findings revealed that 66-68% of the homeless population reported consuming at least one psychoactive substance, with higher consumption rates observed among younger individuals and those with shorter durations of homelessness. The random forest model demonstrated high predictive accuracy and confirmed the centrality of these variables. Moreover, men were more likely to use substances than women, and both family conflict and prior substance use were key factors contributing to the onset and persistence of homelessness. Discussion: The findings indicate that substance use is prevalent among homeless individuals in Colombia and shaped by distinct demographic factors. The bidirectional relationship between homelessness and substance use reveals a complex dynamic in which each condition reinforces the other. These results highlight the importance of targeted interventions directed at younger individuals and those recently experiencing homelessness. Future research using system dynamics modeling is recommended to further explore the feedback mechanisms underlying this relationship. Conclusion: This study offers a detailed analysis of the variables influencing substance use among homeless individuals in Colombia. By identifying age and homelessness duration as critical factors, the research contributes actionable knowledge for the development of evidence-based public policies. Implementing targeted interventions based on these findings may improve the health outcomes and social reintegration of this vulnerable population, ultimately enhancing public health and safety.