Hemoglobin-to-red blood cell distribution width ratio trajectories and prolonged length of hospital stay in premature rupture of membranes: Identifying distinct risk profiles

Premature rupture of membranes (PROM) often predisposes patients to infection-related complications, increasing the risk of prolonged length of hospital stay (PLOS). This study examined the association between the hemoglobin-to–red blood cell distribution width ratio (HRR) and PLOS in PROM. A total of 861 PROM patients were identified from the medical information mart for intensive care IV (MIMIC-IV) database, a publicly available large-scale electronic health record database developed and maintained by the Laboratory for Computational Physiology at the Massachusetts Institute of Technology. PLOS was defined as a length of hospital stay exceeding 4.736 days (the 75th percentile of the cohort). Baseline characteristics were compared between the PLOS and non-PLOS groups. Baseline characteristics were compared between PLOS and non-PLOS groups. Three sequential logistic regression models were used to evaluate the relationship between initial HRR and PLOS. Latent class trajectory modelling (LCTM) was applied to five serial HRR measurements. Fully adjusted logistic regression and subgroup analyses assessed the prognostic value and clinical utility of HRR trajectories. Among 861 patients, 215 developed PLOS. The initial HRR level was lower in the PLOS group compared with the non-PLOS group (0.831 vs. 0.867, p < 0.001). Initial HRR was negatively associated with PLOS (adjusted OR = 0.842, 95% CI: 0.729–0.969, p = 0.018). When categorized into tertiles, patients in the highest HRR tertile had lower odds of PLOS than those in the lowest tertile (adjusted OR = 0.523, 95% CI: 0.316–0.865, for trend < 0.05). LCTM identified two HRR trajectories: high–decrease and persistently high, with the persistently high group showing reduced PLOS (OR = 0.540, 95% CI: 0.312–0.948, p = 0.029). Subgroup analyses showed stronger protective effects in patients delivering within 24 h (OR = 0.520, p = 0.007) and those not receiving antibiotics (OR = 0.480, p = 0.012). Adding HRR trajectories modestly improved predictive performance (area under curve = 0.820). Initial HRR and its trajectory were independently associated with PLOS in PROM. These findings suggest that HRR may capture integrated physiological changes during hospitalization and provide supplementary prognostic information for length-of-stay risk stratification. not applicable.