Avanços na Saúde pública por meio da inteligência artificial: Análise preditiva do risco de consumo de drogas ilícitas entre adolescentes brasileiros através da Machine Learning

Mateus Zani De Nadai

doi:10.53843/bms.v11i15.1148

Authors

Mateus Zani De Nadai Centro universitário do Espírito Santo https://orcid.org/0000-0001-6777-0147

DOI:

https://doi.org/10.53843/bms.v11i15.1148

Keywords:

planejamento em saude, aprendizado de máquina, abuso oral de substâncias

Abstract

INTRODUCTION: Machine Learning (ML) is a field of artificial intelligence that enables the development of algorithms capable of learning from data, without explicit programming. Advances in these techniques have enabled important applications in healthcare, including risk and behavior prediction. This study proposes the development of a predictive model to understand and predict illicit drug use among Brazilian teenagers aged 13 to 17, based on sociodemographic and behavioral patterns, with the aim of informing more effective public health policies. METHODOLOGY: This is an observational, cross-sectional study based on data from 165,838 students from the 2019 National School Health Survey (PeNSE). Different ML models were compared to identify the one with the best performance in predicting illicit drug use. Predictor variables included gender, age, future plans, alcohol and tobacco use, housing conditions, property ownership, parental educational backgrounds, and family habits. RESULTS: Among the models tested, Logistic Regression presented the highest AUC-ROC (0.90), demonstrating better overall performance. Random Forest, however, was used to assess the importance of variables due to its interpretive robustness. The main factors associated with risk were: alcohol use, maternal educational background, peer and parental support, and parental alcohol consumption. DISCUSSION: The findings confirm the potential of ML in identifying risk patterns, in line with recent studies and national epidemiological data. The inclusion of family and behavioral variables reinforces the relevance of preventive strategies targeted at the school and home environment. CONCLUSION: The application of ML models, especially Logistic Regression, proved to be valid for predicting the risk of illicit drug use in teenagers. These results can guide targeted public policies, prioritizing modifiable risk factors and optimizing the use of public health resources.

Author Biography

Mateus Zani De Nadai, Centro universitário do Espírito Santo

Sou aluno do 12° Período do Centro Universitário do Espírito Santo

References

Nawi AM, Ismail R, Ibrahim F, Hassan MR, Manaf MRA, Amit N, et al. Risk and protective factors of drug abuse among adolescents: a systematic review. BMC Public Health. 2021 Nov 13;21(1):2088. doi: 10.1186/s12889-021-11906-2. PMID: 34774013; PMCID: PMC8590764.

Tinner L, Palmer JC, Lloyd EC, Caldwell DM, MacArthur GJ, Dias K, et al. Individual-, family- and school-based interventions to prevent multiple risk behaviours relating to alcohol, tobacco and drug use in young people aged 8–25 years: a systematic review and meta-analysis. BMC Public Health. 2022 Jun 3;22(1):1111. doi: 10.1186/s12889-022-13072-5. PMID: 35658920; PMCID: PMC9165543.

Afzali MH, Sunderland M, Stewart S, Masse B, Seguin J, Newton N, et al. Machine-learning prediction of adolescent alcohol use: a cross-study, cross-cultural validation. Addiction. 2019 Apr;114(4):662–71. doi: 10.1111/add.14504. PMID: 30461117.

Artero AS. Aprendizado de máquina: conceitos e algoritmos. In: SIBGRAPI 2009 – XXII Conference on Graphics, Patterns and Images. IEEE; 2009. p. 215–28.

Haykin S. Redes neurais: princípios e prática. Porto Alegre: Bookman Editora; 2001.