Early platelet change and 28-day mortality after the day-7 landmark among intensive care unit patients with diagnosis-coded heart failure: A MIMIC-IV cohort study

Objective To assess the association between platelet percent-decrease patterns during the first 168 h after intensive care unit admission and 28-day mortality after a day-7 landmark among intensive care unit patients with a diagnosis of heart failure. Methods This retrospective cohort study used Medical Information Mart for Intensive Care IV (MIMIC-IV) version 3.1 records from one US center (2008–2022). The day-7 risk set comprised patients who were not recorded as having died before the landmark. Platelet percent decrease from the first to the last available daily mean during the first 168 h was categorized as stable/increase, percent decrease ≤0%; mild decline, percent decrease >0% and ≤30%; or marked decline, percent decrease >30%. Logistic regression was used to estimate odds ratios for 28-day mortality after the landmark. Supportive analyses examined measurement processes, alternative platelet definitions, narrower heart-failure coding, and an independent 72-h landmark. Results Among 14,762 platelet-classifiable patients in the day-7 risk set, 1736 died within 28 days after the landmark. Observed mortality was 8.66%, 13.14%, and 30.06% in the stable/increase, mild-decline, and marked-decline groups, respectively. Among 14,710 Model 3 complete cases (1729 deaths), adjusted odds ratios were 1.54 (95% confidence interval: 1.37–1.74) for mild decline and 3.51 (95% confidence interval: 2.99–4.12) for marked decline. Adjusted mortality probabilities were 9.08%, 13.07%, and 24.27%, respectively. For marked decline, the odds ratio was 2.69 after adjustment for measurement processes and 2.37 among patients who remained in the intensive care unit on day 7. In the independent 72-h analysis, odds ratios were 1.02 (95% confidence interval: 0.92–1.14) and 1.23 (95% confidence interval: 1.05–1.44). Conclusions Greater platelet decline was associated with higher mortality after the day-7 landmark. The weaker associations at the 72-h landmark and attenuation after accounting for measurement processes indicate that the magnitude of this association depends on the observation window and may partly reflect prolonged critical illness and monitoring intensity.