Association of hemoglobin level trajectories with clinical outcomes in patients with sepsis or shock: a retrospective observational study

Abstract

Background Sepsis has a high incidence and mortality rate and is an important global public health problem. Anemia as one of the common complications of sepsis, a single-point value of hemoglobin cannot provide enough information. This study aimed to investigate the impact of hemoglobin latent trajectories on clinical outcomes in septic patients. Method Patients diagnosed with sepsis and who stayed in the ICU for more than 3 days were included in the medical big data platform, and those with continuous hemoglobin data were selected. Patients with hemorrhagic diseases and chronic consumptive diseases that affect hemoglobin were excluded. The latent growth mixed model was used to identify the latent trajectories of hemoglobin, and multivariate forward stepwise Cox regression was used to evaluate the relationship between hemoglobin trajectories and clinical outcomes (in-hospital mortality and organ dysfunction). Results A total of 811 patients ultimately met the inclusion criteria, and after latent growth mixed model (LGMM) analysis, optimal 3 potential Hb trajectories were detected. 441 (54.4%), 281 (34.7%), and 89 (10.9%) patients were identified as having flat decreasing, low, and rapidly decreasing Hb, respectively. Compared with flat decreasing Hb, low Hb was associated with increased hospital mortality [hazard ratio (HR) 2.44, 95% confidence interval (CI) 2.03–2.92], organ dysfunction (HR 3.02, 95% CI 2.57–3.53), and rapid decreasing FB was significantly associated with hospital mortality (HR 3.45, 95% CI 2.48–4.80) and organ failure (HR5.45, 95%CI 4.09–7.24), after adjustment for potential covariates. Conclusion The latent hemoglobin trajectories were associated with clinical outcomes in sepsis patients. Patients with persistently low hemoglobin levels had increased risks of in-hospital mortality and organ dysfunction, and patients with rapid decline of hemoglobin levels had significantly higher risks.