Mortality_Prison_PAR
Description
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.
Files
Steps to reproduce
The data were generated from three anonymised analytical databases prepared for a retrospective record-linkage cohort study of mortality among people deprived of liberty in Paraguay. The first database corresponds to the 2013 National Census of Penitentiary Centres and contains the anonymised prison-census cohort used for the main mortality analysis. The second database contains the national male mortality and population denominators for 2013-2020, grouped by calendar year, age group, and ICD-10 cause-of-death category, and was used to calculate expected deaths and standardised mortality ratios. The third database corresponds to the five-prison subcohort with documented dates of prison entry and release, used to describe deaths occurring during incarceration and after release. The original study workflow linked the 2013 prison census with national death records from 2013 to 2020. Before linkage, identifying variables were standardised, including names, surnames, national identity numbers, sex, and dates of birth. Deterministic linkage was first performed using national identity number. Probabilistic linkage was then used for records with incomplete, missing, or inconsistent identifiers, and candidate matches were reviewed individually before being accepted or excluded. Direct identifiers and exact linkage variables were removed before preparation of the anonymised datasets deposited in this repository. Cause of death was classified from the ICD-10 underlying-cause code recorded in the national mortality registry. Codes were grouped into predefined cause-of-death categories: violence, suicide, tuberculosis, other infectious diseases, drugs and intoxications, non-communicable diseases, road traffic injuries, other external causes and undetermined-intent injuries, ill-defined causes, and all causes combined. The analytical workflow can be reproduced using the file “PRISIONES_Procesamiento de datos.R.txt”, which contains the R script used to process the anonymised datasets and generate the main study outputs. The script includes the steps for data cleaning, variable recoding, construction of age groups, ICD-10 cause-of-death grouping, estimation of observed deaths, calculation of person-time, estimation of expected deaths, standardised mortality ratios with 95% confidence intervals, and descriptive analyses of the five-prison subcohort. To reproduce the analyses, users should download the three anonymised datasets and the R script, place all files in the same working directory, update the file paths in the script if necessary, and run the script sequentially from top to bottom in R. The repository allows reproduction of the analytical tables and figures reported in the manuscript. Because direct identifiers were removed for confidentiality reasons, the original individual-level linkage procedure cannot be independently repeated from the deposited files.
Institutions
- Universidad Nacional de AsunciónCentral Department, San Lorenzo