1778 Effects of comorbidities, social deprivation and sex on the excess mortality of French ESKD patients on dialysis

Abstract Background and Aims Inequalities in access to care may affect End-stage Kidney Disease (ESKD) patient's overall mortality, i.e. mortality directly or indirectly related to ESRD (excess mortality) as well as mortality related to other causes of death. Previous studies in France have identified sex and comorbidities, especially diabetes and cardiovascular diseases, as factors influencing the overall mortality of ESKD patients, whereas social deprivation was not associated with overall mortality. However, no study has yet examined the impact of these factors on excess mortality in dialysis patients. This study thus aims to quantify the effects of social deprivation, sex and comorbidities on the excess mortality of ESKD patients on dialysis in France. Method This retrospective observational study included adults aged 18-85 years who started dialysis in France between 2017 and 2019, and were followed up until 31 December 2022. Dialysis patients receiving a transplant were censored. Analyses were based on the excess hazard approach: the overall mortality rate was decomposed into the sum of a mortality rate associated with ESKD (the excess rate) and a mortality rate associated with other causes of death (the expected rate). The latter is usually drawn from mortality tables for the general population stratified by demographic covariates. A model is then fitted in order to estimate the excess mortality hazard: the impact of covariates on this hazard is quantified in terms of excess hazard ratios. In France, mortality tables are provided by the “Institut National de la Statistique et des Études Économiques” and stratified by age, calendar year, sex, and “départements” of residence. However, stratification by level of social deprivation or levels of the Charlson Comorbidity Index is not available, which may bias the excess mortality hazard estimates. To assess the impact of using inadequately stratified life tables on estimating excess hazard, we compared the following models using Akaike's Information Criterion: (i) an excess hazard model using the standard life tables and accounting for the hierarchical structure of the data by including a random intercept for the “département” (M0), (ii) an excess hazard model with a model-based life table correction and with (M1) or without (M2) the random intercept at the “département” level, and (iv) an excess hazard model in which the life table correction is achieved by using an individual frailty term (M3). Results