Perioperative risk stratification for periprosthetic joint infection after primary total knee arthroplasty: a case-matched cohort study incorporating serum albumin, glycemic status, and intraoperative hypothermia

Background Periprosthetic joint infection (PJI) remains a serious complication of primary total knee arthroplasty (TKA). This study investigated independent and combined associations of serum albumin, preoperative blood glucose, and intraoperative lowest body temperature (LBT) with PJI within 1 year after primary TKA. Materials and methods This retrospective case-matched cohort study enrolled 312 patients (57 PJI cases, 255 controls) from 4319 screened undergoing primary TKA. Cases were matched 1:5 to controls by operating surgeon and year of surgery (frequency matching). Infection was defined by 2018 International Consensus Meeting criteria. Multivariable logistic regression (seven covariates, events-per-variable = 8.1) and receiver operating characteristic (ROC) analysis identified independent predictors and assessed discrimination. Two composite models were constructed using ROC-derived cutoffs: a two-factor preoperative model (serum albumin ≤ 4.1 g/dL + blood glucose ≥ 147 mg/dL) and a three-factor perioperative model additionally incorporating LBT ≤ 35.5 °C. Results Of 4319 patients, 57 (1.3%) developed PJI. Lower serum albumin was associated with higher infection odds (odds ratio (OR) 0.87 per 0.1 g/dL, 95% confidence interval (CI) 0.78–0.97; p = 0.010); blood glucose with increased risk (OR 1.01 per mg/dL, 95% CI 1.00–1.02; p = 0.014); and LBT ≤ 35.5 °C with significantly increased risk (OR 2.65, 95% CI 1.40–5.02; p = 0.003). The two-factor (area under the curve (AUC) = 0.63) and three-factor (AUC = 0.69) models showed stepwise PJI gradients across cumulative strata; the adjusted multivariable model achieved AUC = 0.71 (95% CI 0.63–0.79). Conclusions Serum albumin, blood glucose, and intraoperative LBT are independent modifiable predictors of PJI after primary TKA. A two-factor preoperative model (albumin + glucose) supports preoperative screening; a three-factor model adding LBT provides intraoperative surveillance for structured normothermia protocols. These exploratory cutoffs were derived from a single-center cohort and require external multicenter validation before clinical implementation.