Measures of Admission Immunocoagulopathy as an Indicator for In-Hospital Mortality in Patients with Necrotizing Fasciitis: A Retrospective Study

Abstract

Necrotizing fasciitis is a rapidly progressive infection with a high mortality rate. Pathogens evade the host containment and bactericidal mechanisms by hijacking the coagulation and inflammation signaling pathways, leading to their rapid dissemination, thrombosis, organ dysfunction, and death. This study examines the hypothesis that measures of immunocoagulopathy upon admission could aid in the identification of patients with necrotizing fasciitis at high risk for in-hospital mortality.

Methods: Demographic data, infection characteristics, and laboratory values from 389 confirmed necrotizing fasciitis cases from a single institution were analyzed. A multivariable logistic regression model was built on admission immunocoagulopathy measures (absolute neutrophil, absolute lymphocyte, and platelet counts) and patient age to predict in-hospital mortality.

Results: The overall in-hospital mortality rate was 19.8% for the 389 cases and 14.6% for the 261 cases with complete measures of immunocoagulopathy on admission. A multivariable logistic regression model indicated that platelet count was the most important predictor of mortality, followed by age and absolute neutrophil count. Greater age, higher neutrophil count, and lower platelet count led to significantly higher risk of mortality. The model discriminated well between survivors and non-survivors, with an overfitting-corrected C-index of 0.806.

Conclusions: This study determined that measures of immunocoagulopathy and patient age at admission effectively prognosticated the in-hospital mortality risk of patients with necrotizing fasciitis. Given the accessibility of neutrophil-to-lymphocyte ratio and platelet count measurements determined from a simple complete blood-cell count with differential, future prospective studies examining the utility of these measures are warranted.

Level of evidence: Prognostic Level III. See Instructions for Authors for a complete description of levels of evidence.

Keywords: Acute phase response; immunocoagulopathy; in-hospital mortality; necrotizing fasciitis; neutrophil-to-lymphocyte ratio; orthopaedics.

Fig. 1

CONSORT diagram showing the retrospective identification and validation of patients with necrotizing fasciitis. Utilizing the deidentified synthetic derivative database, all patients with an ICD-9 or 10 code for necrotizing fasciitis were identified. All charts were individually reviewed to confirm the diagnosis of necrotizing fasciitis by surgeon notes, operative findings, and/or pathology reports.

Fig. 2

Infection characteristics across the cases of necrotizing fasciitis. Diverse causes of infection (Fig. 2-A) and locations of infection (Fig. 2-B) were observed among the 389 confirmed cases of necrotizing fasciitis obtained through retrospective review, resulting in differences in mortality across anatomical locations (Fig. 2-C).

Fig. 3

Correlation between admission immunocoagulopathy measures and mortality. Fig. 3-A Hierarchical clustering, using Spearman correlation coefficients as a similarity measure, of values obtained from a CBC with differential in patients with necrotizing fasciitis. Figs. 3-B through 3-F Univariate analyses. Smoothed estimates of mortality probability by individual predictor using LOWESS (locally weighted scatterplot smoothing). Circles are observed values. The number of observations (N) and Someer rank correlation (D_xy) between mortality and each predictor are also presented. Absolute lymphocyte count (Fig. 3-B) and platelet count (Fig. 3-C) each had moderate correlation with mortality status, with D_xy of −0.275 and −0.350, corresponding to an area under the receiver operator characteristic curve (AUC) of 0.638 and 0.675, respectively. NLR (Fig. 3-D) had a moderate correlation with mortality status, with D_xy of 0.224. Performance was further improved by evaluating the ratio of NLR to platelet count (Fig. 3-E), resulting in a relatively strong correlation with mortality, with D_xy of 0.504 and AUC of 0.752. Increased patient age (Fig. 3-F) strongly correlated with mortality, with D_xy of 0.433 and AUC of 0.717. Absolute neutrophil count and WBC correlated poorly with mortality (D_xy of −0.044 and −0.144, respectively, data not shown).

Fig. 4

Multivariable logistic regression model using a complete-case approach. Fig. 4-A The model discriminated between survivors and non-survivors well, with an original AUC of 0.829. Fig. 4-B A bootstrap overfitting-corrected LOESS (locally estimated scatterplot smoothing) nonparametric calibration curve had an AUC of 0.806. The curve shows good calibration. Fig. 4-C Predictor importance, measured by degrees-of-freedom (df)-penalized chi-square statistics. P values are listed on the right. Fig. 4-D Predicted log odds of in-hospital mortality as a function of each individual predictor, adjusted for age = 50 years, ln.platelets = 4.8, ln.lymphocytes = 0.207, and ln.neutrophils = 2.552. The shading indicates the 95% CI. Fig. 4-E Predicted probability of in-hospital mortality as a function of each individual predictor on the original scale, adjusted for age = 50 years, platelets = 121.5 × 10³ /µL, lymphocytes = 1.23 × 10³ /µL, and neutrophils = 12.83 × 10³ /µL. The shading indicates the 95% CI.