Administrative cohort for the analysis of outcomes linking the Mexican National Health and Nutrition Survey with the national death records database, 2006–2023

Objectives This paper aims to describe the construction of an administrative cohort by linking the Mexican National Health and Nutrition Surveys (ENSANUT) with mortality databases, and to assess its suitability for enabling long-term mortality analysis. Study design Longitudinal administrative cohort study linking survey and routinely collected health data. Methods ENSANUT 2006 and 2012 were linked to the Statistical and Epidemiological Death System (2006–2023) using a modified Fellegi-Sunter algorithm adapted for Hispanic phonetics. Cohort mortality rates for all-cause, diabetes, lung cancer, and chronic obstructive pulmonary disease (COPD) were compared to national trends. As a proof-of-principle analysis, Cox proportional hazards models with time-dependent effects assessed smoking-related mortality risks. Results A cohort of 91,518 adults (mean age 43.5 years, 56.7 % women) was constructed, identifying 10,232 deaths (11.2 %). Cohort all-cause mortality rates closely matched national estimates (p > 0.05 after 2012). Diabetes and COPD mortality followed population trends; lung cancer rates showed higher annual variability. People smoking at the time of the survey had increased mortality risk for all-cause (HR = 2.71; 95 % CI: 2.24–3.29), lung cancer (7.43; 3.03–18.2), and COPD (9.87; 3.29–29.6 for women, 5.89; 2.47–14.06 for men) compared to people who had never smoked. Conclusions Probabilistic linkage enabled the construction of an administrative cohort that aligns with national mortality trends, validating its use for epidemiological research. This method enhances health risk analysis in Mexico and similar settings. Future improvements should prioritize integrating anonymized unique identifiers for enhanced data integration.