Search

To independently validate the negative predictive value of the Tanta University risk model for intensive care requirements in poison center telephone consultations with other physicians. This study included 400 consecutive patients with acute poisoning. Clinical and laboratory parameters were recorded during the initial consultation with the poison center. Patients who were already ventilated or on vasopressors at the time of consultation were excluded. The Tanta University risk model score was calculated from the data according to the following equation: Tanta University risk model score = 1.966*Glasgow Coma Scale + 0.329*oxygen saturation (percent) + 0.212*diastolic blood pressure (mmHg) − 0.27*respiratory rate (breaths/minute) + 0.33*standard bicarbonate (mmol/L). Twenty-four hours later, the patients' courses were followed up by telephone. The Tanta University risk model was then compared to a composite endpoint indicating the requirement for admission to an intensive care unit (vasopressors, need for intubation, or death). Four hundred patients with acute poisoning were included. Thirty-seven patients had a complicated clinical course as defined by the composite endpoint. Receiver operating characteristic analysis revealed the area under the curve to be 0.87 (95 percent confidence interval 0.83–0.90). An unfavorable Tanta University risk model score was defined as less than 73.46, using a cut-off derived from a previous study of an unrelated series of patients with acute poisoning admitted to our service. Thirty-one of 37 patients with complicated courses had an unfavorable Tanta University risk model score compared to six patients with complicated courses among 306 patients with a favorable Tanta University risk model score (P < 0.0002, Fisher's exact test). Sixty-three patients had an unfavorable Tanta University risk model score but an uneventful course. The negative predictive value of the Tanta University risk model was 0.98 (95 percent confidence interval 0.96–0.99), sensitivity was 0.84, and specificity 0.83. In the present study of poison center telephone consultations, the Tanta University risk model was significantly related to the outcomes in patients with acute poisoning. Patients with a favorable Tanta University risk model score (greater than or equal to 73.46) were unlikely to need intensive care unit level of care. [ABSTRACT FROM AUTHOR]

Artificial intelligences (AIs) are emerging in the field of medical informatics in many areas. They are mostly used for diagnosis support in medical imaging but have potential uses in many other fields of medicine where large datasets are available. To develop an artificial intelligence (AI) "ToxNet", a machine-learning based computer-aided diagnosis (CADx) system, which aims to predict poisons based on patient's symptoms and metadata from our Poison Control Center (PCC) data. To prove its accuracy and compare it against medical doctors (MDs). The CADx system was developed and trained using data from 781,278 calls recorded in our PCC database from 2001 to 2019. All cases were mono-intoxications. Patient symptoms and meta-information (e.g., age group, sex, etiology, toxin point of entry, weekday, etc.) were provided. In the pilot phase, the AI was trained on 10 substances, the AI's prediction was compared to naïve matching, literature matching, a multi-layer perceptron (MLP), and the graph attention network (GAT). The trained AI's accuracy was then compared to 10 medical doctors in an individual and in an identical dataset. The dataset was then expanded to 28 substances and the predictions and comparisons repeated. In the pilot, the prediction performance in a set of 8995 patients with 10 substances was 0.66 ± 0.01 (F1 micro score). Our CADx system was significantly superior to naïve matching, literature matching, MLP, and GAT (p < 0.005). It outperformed our physicians experienced in clinical toxicology in the individual and identical dataset. In the extended dataset, our CADx system was able to predict the correct toxin in a set of 36,033 patients with 28 substances with an overall performance of 0.27 ± 0.01 (F1 micro score), also significantly superior to naïve matching, literature matching, MLP, and GAT. It also outperformed our MDs. Our AI trained on a large PCC database works well for poison prediction in these experiments. With further research, it might become a valuable aid for physicians in predicting unknown substances and might be the first step into AI use in PCCs. [ABSTRACT FROM AUTHOR]