Objective <p>Hemoglobin–albumin–lymphocyte–platelet (HALP) score is considered to be a comprehensive indicator of inflammation and nutrition. We aimed to investigate the relationship of HALP score and the risk of all-cause and cardiovascular disease (CVD) mortality in patients with diabetes (DM) or prediabetes (PDM).</p> Methods <p>6,869 participants with DM or PDM from the National Health and Nutrition Examination Survey (NHANES) 2005 to 2018 were enrolled. The colleration of HALP score with all-cause and CVD mortality was evaluated using Kaplan-Meier, Cox regression and restricted cubic spline (RCS) methods. The predictive value of HALP score for mortality was evaluated by time-dependent-receiver-operating-characteristic (ROC) curves. Finally, subgroup and interaction analysis were performed.</p> Results <p>1203 deaths from all-cause and 399 deaths from CVD were observed. Cox regression analyses showed that the HALP score was negatively correlated with both all-cause and CVD mortality risk. RCS curves showed a nonlinear relationship between HALP score and all-cause or CVD mortality risk, and both the dose-response curves are L-shaped. For all-cause mortality risk, the AUC was 0.805, 0.799, and 0.816 for 3, 5, and 10 years survival, respectively, and for CVD mortality risk, the AUC was 0.839, 0.850, and 0.837 for 3, 5, and 10 years of survival, respectively. Mediation analysis showed that serum creatinine and urea nitrogen partially mediate the relationship between HALP and mortality risk.</p> Conclusion <p>HALP score is negatively correlated with all-cause and CVD mortality risk, and serves as a valuable predictor of all-cause and CVD mortality risk in patients with DM or PDM.</p> Clinical trial number <p>Not applicable.</p>