History-based preoperative evaluation of antibiotic-related allergies and design of an easy to use questionnaire.

BACKGROUND: An unconfirmed history of antibiotic allergies may negatively influence prescribing patterns for preoperative antibiotic prophylaxis and increase rates of postoperative wound infections through unnecessary use of alternative antibiotics. METHODS: After a literature search, we developed a questionnaire for the structured collection of antibiotic allergy history in the anesthesia consultation center and tested it over 2 years at a tertiary care hospital under everyday conditions as part of a quality assurance project. All data were evaluated completely anonymously in the context of standard care. RESULTS: After refining the questionnaire, we analyzed 4866 recorded optimized questionnaires, of which 51 were incomplete. An antibiotic allergy was denied 4312 times and affirmed 503 times, which corresponds to 10% in our sample. The most frequent single substances or groups in the 503 respondents with a positive history of antibiotic allergy were penicillin in 271 (54%), amoxicillin in 65 (13%), an unknown single agent in 50 (10%) and multiple substances in 25 (5%). The reported event occurred more than 10 years ago in 192 (38%) of the respondents, less than 10 years ago in 116 (23%), and 195 (39%) could not provide information. The time from exposure to symptom onset was less than 1h in 96 (19%), between 1 and 24 h in 75 (15%), more than 24 h in 106 (21%), and the remainder could not provide information. Allergy-specific treatment was recalled by 75 (15%) respondents, 287 (57%) reported not having received specific treatment, and the remainder could not recall. A specific allergy test was reported by 55 (11%) respondents, 337 (67%) said no allergy test had been made, and the rest could not recall. A substance-specific allergy passport was issued in 80 (16%) respondents. According to expert assessment, symptoms compatible with an IgE-mediated reaction were present in 96 (19%) of the respondents. An IgE-mediated reaction was considered possible in 70 (14%) and could be excluded by history in 337 (67%) of respondents. Out of 503 respondents with a positive history 51 (10%) could not remember the allergic substance but 7 (14%) of the 51 reported symptoms compatible with severe anaphylaxis or anaphylactic shock and 6 of the 51(12%) reported symptoms possibly related to an IgE-mediated reaction. DISCUSSION: Our survey revealed approximately 10% of respondents reporting an antibiotic allergy, which is in the upper range of data published in international literature and corresponds most closely to American data. Thus, the topic is also relevant to German anesthesia consultation centers, given the high rate of respondents who could have been "delabeled" based on the comprehensive assessment of their history. More expert allergy testing is needed in patients who report symptoms related or probably related to an IgE-mediated reaction. In our opinion, a special issue exists in those patients who did not remember the exact antibiotic but reported symptoms compatible with severe anaphylaxis putting them at high risk of unintended re-exposure.

Transition1x - a dataset for building generalizable reactive machine learning potentials.

Machine Learning (ML) models have, in contrast to their usefulness in molecular dynamics studies, had limited success as surrogate potentials for reaction barrier search. This is primarily because available datasets for training ML models on small molecular systems almost exclusively contain configurations at or near equilibrium. In this work, we present the dataset Transition1x containing 9.6 million Density Functional Theory (DFT) calculations of forces and energies of molecular configurations on and around reaction pathways at the ωB97x/6-31 G(d) level of theory. The data was generated by running Nudged Elastic Band (NEB) with DFT on 10k organic reactions of various types while saving intermediate calculations. We train equivariant graph message-passing neural network models on Transition1x and cross-validate on the popular ANI1x and QM9 datasets. We show that ML models cannot learn features in transition state regions solely by training on hitherto popular benchmark datasets. Transition1x is a new challenging benchmark that will provide an important step towards developing next-generation ML force fields that also work far away from equilibrium configurations and reactive systems.