Optimizing healthcare resources in pyogenic liver abscess: a dual-threshold HDL-CRP model for predicting hospitalization duration across multi-cohorts

BackgroundProlonged hospitalization for pyogenic liver abscess (PLA) burdens patients and health systems. We examined whether admission high-density lipoprotein cholesterol (HDL-C) predicts length of stay (LOS) and whether a simple HDL–C-reactive protein (CRP) dual-threshold can flag patients at risk of prolonged stay.MethodsWe analyzed a prospective adult PLA cohort at a tertiary center (2018–2023; n = 138) and validated findings in MIMIC-IV ICU patients with liver abscess (n = 38) and in NHANES 2017–2020 (n = 9,226). Multivariable models related admission HDL-C to log-transformed LOS; model performance and calibration were assessed with internal resampling and external validation. We further evaluated a dual-threshold rule and conducted mediation analysis with CRP.ResultsLower HDL-C independently associated with longer LOS. A nomogram combining HDL-C, abscess size, and sepsis performed well (R2≈0.66; RMSE≈6.4 d) and remained directionally consistent across external datasets. A dual-threshold (HDL-C < 1.03 mmol/L and CRP > 1.0 mg/dL) identified a high-risk subgroup with greater odds of prolonged stay. CRP mediated a small proportion of the HDL-C–LOS association.ConclusionAdmission HDL-C, particularly when interpreted together with CRP, may help identify PLA patients at increased risk of prolonged hospitalization at the time of admission. Patients with low HDL-C and high CRP may warrant closer monitoring, earlier evaluation for source control or drainage, and more proactive inpatient planning. Prospective implementation studies are warranted.