TIME-VARYING EFFECTS OF HEALTH-SEEKING DELAYS ON PROSTATE CANCER MORTALITY IN KENYA A SURVIVAL ANALYSIS APPROACH

Abstract

Prostate cancer is frequently diagnosed at advanced stages in Kenya, contributing to poor survival outcomes. Health-seeking delays across the care continuum are thought to influence mortality, but most research has treated delay as a static baseline factor, obscuring how its effect may change over time. This study investigated the time-varying effects of delays on prostate cancer mortality in Kenya using a retrospective cohort of 420 patients from the National Cancer Registry of Kenya. Cox proportional hazards models with time-dependent covariates and landmark analysis at 3, 6, and 12 months post-diagnosis were applied. Of the 420 patients, 230 (54.8%) died during follow-up, with an overall median survival of 33.8 months. Total delay was independently associated with mortality hazard in both baseline (HR = 1.04, p = 0.006) and time-varying (HR = 1.04, p = 0.007) Cox models, alongside age, disease stage, and Gleason score. Landmark analysis showed that total delay remained a significant predictor of subsequent survival at 3, 6, and 12 months (p = 0.041, 0.035, and 0.039, respectively), while treatment interruption reached significance specifically at 6 months (p = 0.034), suggesting a critical prognostic window approximately three to six months after diagnosis. The delay-informed predictive model achieved an optimism-corrected concordance index of 0.723 and a mean time-dependent AUC of 0.787. These findings establish total delay as a robust, time-varying predictor of prostate cancer mortality in this Kenyan cohort and identify the three-to-six-month post-diagnosis window as a priority target for clinical monitoring and patient re-engagement interventions.