Factors Associated with Tobacco Cessation Services Request Among Users of an Online Self-Screening Questionnaire.

OBJECTIVES: Tobacco smoking remains a major public health risk, responsible for millions of deaths worldwide. While smoking patterns in Mexico differ from those in countries with higher rates, comorbidities such as diabetes pose a health risk. Although many smokers want to quit, access to cessation services is limited. Internet-based cessation (I-BC) services are a promising modality that offers accessibility and machine learning (ML) has been successfully used to predict tobacco outcomes. This study uses ML to identify characteristics associated with requesting I-BC through an online self-assessment questionnaire in Mexico.

METHODS: This was a retrospective, predictive, secondary analysis of 14,182 records of individuals aged 18 years and older who completed an online screening for nicotine dependence and their request for tobacco cessation services. Random forest algorithm with four oversampling methods was compared to select the best predictive model. The relative importance of predictor variables was measured as well.

RESULTS: The algorithm had a sensitivity of 78.6% and a specificity of 68.8%. Specifically, age, sex, dependence severity indicators, locations such as the state of Mexico or Sinaloa, and even occasions such as World No Tobacco Day were identified as key factors influencing cessation service requests.

CONCLUSIONS: These results suggest the random forest algorithm's effectiveness in predicting potential cessation service users. Furthermore, the predictor variables provide valuable insights for designing targeted prevention and awareness campaigns, potentially leading to improved campaign effectiveness and more individuals receiving cessation support.