Temporal Trends and Projections of Leprosy in Brazil: Application of Machine Learning Techniques for Predictive Analysis (2001-2034).

Abstract

Leprosy remains a neglected tropical disease with active transmission. Predictive models improve understanding of epidemiological trends and support control strategies in endemic contexts. This study analyzed leprosy in Brazil between 2001 and 2024, projecting scenarios through 2034. Data were obtained from the National System of Notifiable Diseases, and population estimates were from the Brazilian Institute of Geography and Statistics. Temporal trends were assessed using segmented regression, and projections were generated with statistical methods and machine learning algorithms. Independent variables included sex, age, educational level, clinical form, operational classification, and bacilloscopy index. Consistent decline was observed in the overall detection rate and among individuals younger than 15 years old, suggesting reduced transmission. The proportion of cases diagnosed with grade 2 disability remained high, indicating late detection. Projections showed a gradual decline in endemicity but no elimination of leprosy as a public health problem by 2030. The random forest model identified male sex, age older than 15 years old, lower educational level, and multibacillary clinical form as the main predictors of new cases. The integration of machine learning improved the accuracy of projections and revealed persistent gaps in early diagnosis, providing evidence for targeted interventions and strengthening active surveillance and timely detection. The findings are relevant not only for Brazil but also, for other endemic countries, such as India and Indonesia, reinforcing the global need for intensified elimination strategies. The study demonstrates the potential of predictive modeling to support leprosy control and broader neglected tropical disease programs.