Derivation and internal validation of a clinical diagnostic score for acute Chinese medicine poisoning involving aconite

Introduction: Aconitum spp. alkaloids, used in traditional Chinese medicine, are potent cardiotoxins and neurotoxins. Timely diagnosis of aconite poisoning remains challenging due to the long laboratory turnaround time. We aimed to derive and internally validate a diagnostic score for rapid recognition of acute Chinese medicine poisoning involving aconite using clinical parameters. Methods: We conducted a retrospective cross-sectional study on consecutive patients with laboratory-confirmed Chinese medicine poisoning reported to the Hong Kong Poison Control Centre between 1 July 2008 and 30 June 2021. The reference standard was the diagnosis of acute aconite poisoning by a clinical toxicologist and laboratory detection of aconitine or related alkaloids in the patients’ urine, serum, or gastric lavage specimens. Univariate analyses, followed by multivariable logistic regression, were performed to identify independent predictors of laboratory-confirmed aconite poisoning. A scoring system was developed based on the regression coefficients of the independent predictors and internally validated using bootstrapping. Results: We identified 542 eligible episodes, of which 179 involved aconite and 363 involved other herbs. The median patient age of the included episodes was 55 years (range 4–98 years). A clinical diagnostic score was developed based on the six independent predictors: hypotension (systolic blood pressure <90 mmHg in adults or < age-appropriate ranges in children, 3 points), herbal decoction or wine formulation (2 points), facial or oral numbness (2 points), ventricular tachycardia (1 point), limb numbness (1 point), and premature atrial or ventricular contractions (1 point). The score ranges from 0 to 10, with a higher score indicating a higher likelihood of aconite poisoning. At the cutoff point of ≥3, the sensitivity and negative predictive value of the score were 0.98 and 0.99, respectively. A higher specificity (0.92) and positive predictive value (0.84) could be achieved with a cutoff point at ≥4. The area under the receiver operating characteristic curve was 0.965 (95% CI: 0.950–0.980) during derivation and 0.965 (95% bias-corrected and accelerated CI: 0.947–0.977) during internal validation. Discussion: The newly derived Clinical Aconite Poisoning Score is simple to use, but its real-time discriminatory performance in diverse populations with Chinese medicine poisoning in real-world settings and its impacts on clinical management are unknown. Conclusions: In the context of Chinese medicine poisoning, the Clinical Aconite Poisoning Score might be useful in early recognition of aconite poisoning before laboratory confirmation. Future prospective studies are warranted to externally validate its real-time discriminatory performance in real-world settings before clinical adoption.