Predicting necrotising soft tissue infections in people who inject drugs: poor performance of the Laboratory Risk Indicator for Necrotising Fasciitis score and development of a novel clinical predictive nomogram in a retrospective cohort with internal validation

Abstract

Introduction: Necrotising soft tissue infections (NSTI) can threaten life and limb. Early identification and urgent surgical debridement are key for improved outcomes. NSTI can be insidious. Scoring systems, like the Laboratory Risk Indicator for Necrotising Fasciitis (LRINEC), exist to aid diagnosis. People who inject drugs (PWID) are high risk for NSTI. This study aimed to assess the utility of the LRINEC in PWID with lower limb infections and develop a predictive nomogram.

Methods: A retrospective database of all hospital admissions due to limb-related complications secondary to injecting drug use between December 2011 and December 2020 was compiled through discharge codes and a prospectively maintained Vascular Surgery database. All lower limb infections were extracted from this database, dichotomised by NSTI and non-NSTI with the LRINEC applied. Specialty management times were evaluated. Statistical analyses involved: chi-square; Analysis of "variance"; Kaplan-Meier, and receiver operating characteristic curves. Nomograms were developed to facilitate diagnosis and predict survival.

Results: There were 557 admissions for 378 patients, with 124 (22.3%; 111 patients) NSTI. Time from admission to: theatre and computed tomography imaging respectively varied significantly between specialties ( P =0.001). Surgical specialties were faster than medical ( P =0.001). Vascular surgery received the most admissions and had the quickest time to theatre. During follow-up there were 79 (20.9%) deaths: 27 (24.3%) NSTI and 52 (19.5%) non-NSTI. LRINEC ≥6 had a positive predictive value of 33.3% and sensitivity of 74% for NSTI. LRINEC <6 had a negative predictive value of 90.7% and specificity of 63.2% for non-NSTI. Area under the curve was 0.697 (95% CI: 0.615-0.778). Nomogram models found age, C-reactive protein, and non-linear albumin to be significant predictors of NSTI, with age, white cell count, sodium, creatinine, C-reactive protein, and albumin being significant in predicting survival on discharge.

Conclusion: There was reduced performance of the LRINEC in this PWID cohort. Diagnosis may be enhanced through use of this predictive nomogram.

Figure 1

Admission numbers by year, dichotomised by necrotising soft tissue infection (NSTI) and non-NSTI (data for December 2011 is excluded no further data for the year 2011 was included).

Figure 2

Kaplan–Meier survival curve by presence of necrotising soft tissue infections (NSTI).

Figure 3

Receiver operating characteristic curve for Laboratory Risk Indicator for Necrotising Fasciitis score within our cohort, area under the curve=0.697 (95% CI: 0.615–0.778).

Figure 4

Outcome models. Nonlinearity attributed to continuous variables by attaching three cubic splines. Logistic model of necrotising soft tissue infections (NSTI) outcome (A) and corresponding nomogram: age, C-reactive protein (CRP) and nonlinear albumin were significant predictors of NSTI within our dataset (B). Best-fit Cox proportionality survival model for time to discharge from critical care (C) and corresponding nomogram: NSTI, white cell count (WCC), lactate, sodium and glucose provide best model (D). Best-fit Cox proportionality survival model for time to discharge from hospital (E) and corresponding nomogram: age, albumin, nonlinear creatinine, CRP, and glucose are significant predictors of hospital survival (F).

Figure 4

Outcome models. Nonlinearity attributed to continuous variables by attaching three cubic splines. Logistic model of necrotising soft tissue infections (NSTI) outcome (A) and corresponding nomogram: age, C-reactive protein (CRP) and nonlinear albumin were significant predictors of NSTI within our dataset (B). Best-fit Cox proportionality survival model for time to discharge from critical care (C) and corresponding nomogram: NSTI, white cell count (WCC), lactate, sodium and glucose provide best model (D). Best-fit Cox proportionality survival model for time to discharge from hospital (E) and corresponding nomogram: age, albumin, nonlinear creatinine, CRP, and glucose are significant predictors of hospital survival (F).