Mortality_Prison_PAR

This repository contains de-identified analysis-ready datasets, data dictionaries, and annotated R scripts used for the study “Mortality during and after incarceration among men in Paraguay: a national retrospective record-linkage cohort study”. The study hypothesis was that incarceration in Paraguay is associated with premature and cause-specific excess mortality, particularly from preventable causes such as violence, tuberculosis, intoxications, and external injuries. To test this hypothesis, Paraguay’s 2013 National Census of Penitentiary Centres was linked with national death records from 2013 to 2020, and mortality among people deprived of liberty was compared with mortality in the national male population aged 18-75 years. The repository includes materials for two analytical stages. Stage 1 contains the de-identified national prison-census cohort and linked mortality outcomes, including age group, year of death, ICD-10-derived cause-of-death groups, estimated person-time, expected deaths, and variables needed to reproduce standardised mortality ratios. Stage 2 contains a de-identified five-prison subcohort with incarceration entry and release timing, allowing exploratory analyses of deaths during incarceration and after release. The data show that mortality in the prison-census cohort was concentrated at young ages and in preventable causes. Cause-specific analyses identified excess mortality from tuberculosis, violence, drugs and intoxications, and other external causes and undetermined-intent injuries, while the five-prison subcohort allows examination of deaths occurring during incarceration and after release. Cause-of-death groups were derived from ICD-10 underlying-cause codes and include violence, suicide, tuberculosis, other infectious diseases, drugs and intoxications, non-communicable diseases, road traffic injuries, other external causes and undetermined-intent injuries, and ill-defined causes. National male population denominators and mortality rates used to estimate expected deaths are also included. The datasets, data dictionaries, and R scripts are provided in Spanish, reflecting the language of the original administrative records and analytical workflow. Direct identifiers and exact linkage variables were removed before publication. Users should not attempt re-identification or combine these data with other sources for re-identification.