Differential analysis of mean blood glucose levels from venous and fingertip in predicting 30-day mortality among ICU patients with severe trauma: A retrospective study utilizing the MIMIC-IV database

Abstract

Background To investigate the correlation between mean blood glucose of venous (VMBG) and mean blood glucose of fingertip (FMBG) within 30 days and 30-day mortality in trauma patients in intensive care unit (ICU), and to systematically evaluate the prognostic value of early-stage and long-term monitoring intervals. Materials and methods Utilizing data from the MIMIC-IV database, we employed receiver operating characteristic (ROC) curves, restricted cubic splines (RCS), and Cox proportional hazards models to assess glucose-outcome relationships. Sensitivity analyses using complete datasets, propensity score matching (PSM), and subgroup analyses were conducted to verify result robustness. Furthermore, to minimize the bias of immortal time, the Landmark analysis was employed to investigate the correlation between the early-stage VMBG, FMBG and 30-day mortality rate. Secondary outcomes included 90-, 180-, and 360-day all-cause mortality. Results A total of 2,699 patients were enrolled in the study. The AUC values (95% CI) for VMBG and FMBG were 0.705 (0.675-0.735) and 0.640 (0.608-0.672), respectively. VMBG demonstrated superior predictive ability for 30-day mortality compared to FMBG (Z=5.833, P<0.001). Multivariate-adjusted Cox regression revealed independent associations between VMBG, FMBG, and 30-day mortality, with HRs of 1.019 (1.016-1.023) and 1.009 (1.006-1.013), respectively. RCS analysis further indicated a nonlinear “J-shaped” relationship between VMBG, FMBG, and outcomes (P for nonlinearity<0.001), with two thresholds at 88.1 mg/dL and 125.4 mg/dL for VMBG, and 95.4 mg/dL and 134.0 mg/dL for FMBG. The threshold values for FMBG were observed to be higher than those of VMBG. According to the thresholds of VMBG and FMBG, patients were stratified into hypoglycemic, normoglycemic, and hyperglycemic groups, respectively. Whether in the pre-PSM or post-PSM cohort, with the normoglycemic group as the reference, both the hypoglycemic and hyperglycemic groups of VMBG and FMBG were associated with an increased 30-day mortality rate. Subgroup analyses revealed that the impact of elevated VMBG and FMBG on prognosis was more pronounced in patients aged <65 years, non-White individuals, and those without diabetes(P for interaction<0.05). VMBG and FMBG during different time intervals were all non - linearly correlated with the 30 - day mortality rate. The Landmark analysis indicated that during the early-stages, there was also a significant statistical correlation among VMBG, FMBG and prognosis. For secondary outcomes, VMBG and FMBG also showed significant associations with 90-day, 180-day, and 360-day all-cause mortality. Conclusion In ICU trauma patients, both VMBG and FMBG across various time intervals exhibited nonlinear associations with 30-day mortality. Although venous blood glucose monitoring typically demonstrated higher prognostic predictive accuracy compared to fingertip measurements, the early - warning value of fingertip blood glucose should not be overlooked. In clinical monitoring, the characteristics of both measurement methods should be comprehensively recognized.