Validation of a Clinical Scale for Early Detection of Infections at the Exit Site of Central Venous Catheters for Hemodialysis

Abstract

[eng] Introduction: Exit-site infections (ESI) of central venous catheters for hemodialysis (CVC-HD) has been associated with early catheter removal and an increased risk of CVC-HD related bacteremia. No specific clinical scales to predict ESI have previously been validated. Methods: A multicenter prospective cohort study was performed to validate the proposed scale, which is based on the following 5 signs and symptoms: (i) pain at exit site during interdialytic period; (ii) hyperemia or erythema ≥2 cm from exit site; (iii) inflammation, induration, or swelling at exit site; (iv) fever ≥38 °C not attributable to other causes, and (v) obvious abscess or purulent exudate at the exit site. Adult patients with a tunneled CVC-HD for at least 1 month after insertion has been included. During each hemodialysis session, the exit site was assessed with the proposed scale by nurses. If any item was present, a pericatheter skin swab culture was collected: positive results were gold standard. The scale was validated using receiver operating characteristic (ROC) curves and logistic regression analysis. For this purpose, the logit function was applied, and the ESI probability calculated, as elogit ESI/1 + elogit ESI. Results: Three hundred thirty-seven CVC-HDs from 310 patients were analyzed, producing 515 cultures (117 infected and 398 healthy). The final version of the scale includes the following 3 signs and symptoms, which present the greatest predictive capacity: (i) pain at exit site during interdialytic period, (ii) hyperemia or erythema ≥2 cm from exit site, and (iii) abscess or purulent exudate at the exit site. The final version generated an area under the ROC curve (AUC) of 88.3% (95% confidence interval [CI]: 85.2%–91%; <em>P</em> < 0.001), Youden index 0.7557 ≈ 1, sensitivity 80.34% (95% CI: 71.36%–87.71%) and specificity 95.23% (95% CI: 92.73%–97%). Conclusions: The validation shows that the scale has good predictive properties, detecting approximately 90% of ESI with very acceptable validity parameters.