Is the Anamnesis Enough to De-Label Patients with Reported Beta-Lactam Allergy?

Abstract

Background: The decision whether to de-label patient with suspected BL hypersensitivity is based on risk stratification. The aim of this study was to prepare a characteristic of diagnostic risk groups and to create a model enabling the identification of the low-risk diagnostic group. Methods: We analyzed the medical records of patients hospitalized due to suspected hypersensitivity to BL antibiotics. Based on their medical-history data, patients were divided into three diagnostic risk groups, using the criteria proposed by Shenoy et al. Univariate and multivariate analysis models were used to create a diagnostic tool. Results: Among 263 patients referred for BL hypersensitivity diagnosis, 88 (33.5%) were allocated to group I, 129 (49%) to group II, and 46 (17.5%) to group III. There were significant differences between diagnostic risk groups regarding history of hypersensitivity to penicillins (p < 0.001), cephalosporins (p < 0.001), >1 BL (p < 0.05), several episodes of BL hypersensitivity (p < 0.001), medical intervention (p < 0.001), documented hypersensitivity (p < 0.001), time from drug intake to symptoms (p < 0.001), and time from hypersensitivity to diagnosis (p < 0.001). In total, 81 patients (30.8%) were de-labeled: 52 (59.8%) in group I, 27 (20.9%) in group II, and 2 (4.3%) in group III. The univariate analysis model of the low-diagnostic-risk group applied to the de-labeled part showed 90% specificity and 21.93% sensitivity. NPV and PPV were estimated at 72.04% and 49.53%, respectively. The multivariate model had high specificity but low sensitivity; its NPV was 76%, with 68% PPV. Conclusions: The tool enabling the identification of low-diagnostic-risk patients based on anamnesis is not sensitive enough to de-label patients on its basis.

Keywords: beta-lactam; de-labeling; hypersensitivity; risk stratification.

Figure 1

Sensitivity and specificity for the multivariate logistic regression model of the low-diagnostic-risk group shown as a ROC (Receiver Operating Characteristic) curve.