Analysis of health infrastructure and suicide rates in Brazil: a nationwide ecological spatial-temporal study,...

Suicide remains a major global health concern and disproportionately affects low- and middle-income countries (LMICs), where socio-economic stressors and limited healthcare resources contribute to higher burdens. This study characterizes annual patterns of suicide rates in Brazil from January 2009 until December 2023, across its five geographic regions (North, Northeast, Southeast, South, and Central-West). Additionally, we constructed a harmonized dataset integrating sociodemographic and healthcare infrastructure indicators, and identified which municipal-level infrastructure components are most informative for classifying suicide rate categories.

We integrated nationwide data from the Mortality Information System (SIM/SUS), the National Registry of Healthcare Establishments (CNES), and demographic estimates from the Brazilian Institute of Geography and Statistics (IBGE). Primary analyses were conducted using municipality-level annual observations across all Brazilian cities covering more than 170,000 suicide deaths nationally. Suicide was expressed annually as rates per 100,000 inhabitants and categorized as low, moderate, or high. Using a data-mining workflow with XGBoost classifier models, we identified healthcare infrastructure features most relevant for distinguishing suicide rate categories. Secondary analyses focused on the sociodemographic characterization of suicide deaths and on the distribution of Psychosocial Care Centers (CAPS), as well as their correlation with suicide rates.

Across the descriptive analyses, suicide rates increased in all Brazilian regions (2009: 4·5, 95% CI 4·4-4·6; 2023: 7·6, 95% CI 7·5-7·7). Machine learning models identified healthcare infrastructure components (e.g., healthcare establishments, registered professionals, primary care units, diagnostic services, and mental health facilities) as the most informative features for distinguishing municipalities by suicide rate categories (low, moderate, or high), with accuracy across regions ranging from 72% (95% CI 69-75%) to 79% (95% CI 76-81%). CAPS availability was correlated with lower suicide rates across all regions (ρ = -0·71 to -0·81, p < 0·01).

In this exploratory analysis, general healthcare infrastructure features are associated with variations in suicide mortality, suggesting that access to basic and mental health services is an informative predictor for distinguishing suicide rate patterns across regions. Findings at the aggregate level cannot be assumed to be consistent at the individual level.

Santa Catarina State Research and Innovation Support Foundation (FAPESC) and the Research Program for the Unified Health System (PPSUS); Brazilian National Council for Scientific and Technological Development (CNPq); Department of Science and Technology of the Secretariat of Science, Technology, Innovation and the Health Economic-Industrial Complex of the Ministry of Health of Brazil (Decit/SECTICS); and Coordination for the Improvement of Higher Education Personnel (CAPES).