Survival analysis of patients with chronic diseases: a statistical approach to mortality risk factors

Quispe V.I.; Quispe J.R.P.; Acero R.C.; Rozas T.P.A.; Vargas E.E.C.; Mamani G.Q.; Panc P.H. · Journal of Advanced Pharmacy Education and Research (2026) · DOI: 10.51847/hvzcnspdd9

Resumen

Chronic diseases are leading causes of death worldwide, but mortality risk varies substantially across patients with similar diagnoses. This heterogeneity requires statistical methods that can estimate survival probability over time and identify risk factors associated with earlier death. Traditional logistic or linear regression models are poorly suited to survival outcomes because they do not directly account for censoring, unequal follow-up duration, or changing exposure status over time. These limitations can produce biased estimates and reduce clinical interpretability. This article applies survival analysis methods to identify demographic, clinical, lifestyle, and treatment-related predictors of all-cause mortality in adults with chronic diseases. The goal is to construct a statistically defensible risk model that supports clinical risk stratification. A retrospective cohort of 1,248 adults with heart failure, chronic kidney disease, chronic obstructive pulmonary disease, type 2 diabetes, or multimorbidity was specified using electronic health record and registry-style data from 2017 to 2025. Kaplan-Meier curves, log-rank tests, and a multivariable Cox proportional hazards model were used, with Schoenfeld residuals applied to assess proportional hazards. In the modeled cohort, 238 deaths occurred over a median follow-up of 4.2 years, providing sufficient events for multivariable modeling. Age per 10 years, advanced disease severity, higher comorbidity burden, current smoking, reduced kidney function, anemia, and low medication adherence were expected to emerge as independent predictors of mortality. Survival analysis provides an appropriate statistical framework for estimating mortality risk in chronic disease cohorts with censoring and variable follow-up. Properly specified models can identify high-risk patients, quantify hazard ratios, and support individualized monitoring and intervention strategies.

Instituciones: Universidad Nacional del Altiplano