Construction and Verification of Urinary Tract Infection Prediction Model for Hospitalized Rehabilitation Patients with Spinal Cord Injury.

OBJECTIVE: To explore the influencing factors of urinary tract infection (UTI) in hospitalized patients with spinal cord injury and to construct and verify the nomogram prediction model. METHODS: This study is a retrospective cohort study. From January 2017 to March 2022, 558 patients with spinal cord injury admitted to the Department of Rehabilitation Medicine of a tertiary hospital in Anhui Province, China, were selected as the research objects, and they were randomly divided into training group (n = 390) and verification group (n = 168) according to the ratio of 7:3, and clinical data including socio-demographic characteristics, disease-related data, and laboratory examination data were collected. Univariate analysis and multivariate logistic regression were used to analyze the influencing factors of UTI in hospitalized patients with spinal cord injuries. Based on this, a nomogram prediction model was constructed with the use of R software, and the risk prediction efficiency of the nomogram model was verified by the receiver operating characteristic curve and calibration curve. RESULTS: Logistic regression analysis showed that the American Spinal Cord Injury Association (ASIA)-E grade (compared with ASIA-A grade) was an independent protective factor for UTI in hospitalized patients with spinal cord injury (odds ratio < 1, P < 0.05), while white blood cell count and indwelling catheter were independent risk factors for UTI in hospitalized patients with spinal cord injury (odds ratio > 1, P < 0.05). Based on this, a nomogram risk predictive model for predicting UTI in hospitalized rehabilitation patients with spinal cord injury was constructed, which proved to have good predictive efficiency. In the training group and the verification group, the area under the receiver operating characteristic curve of the nomogram model is 0.808 and 0.767, and the 95% confidence interval of the area under the receiver operating characteristic curve of the nomogram in the training group and the verification group is 0.760∼0.856 and 0.688∼0.845, respectively, indicating the nomogram model has good discrimination. According to the calibration curve, the prediction probability of the nomogram model and the actual frequency of UTI in the training group and the verification group are in good consistency, and the results of the Hosmer-Lemeshow bias test also suggest that the nomogram model has a good calibration degree in both the training group and the verification group (P = 0.329, 0.067). CONCLUSIONS: ASIA classification level, white blood cell count, and indwelling catheter are independent influencing factors of UTI in hospitalized patients with spinal cord injury. The nomogram prediction model based on the above factors can simply and effectively predict the risk of UTI in hospitalized patients with spinal cord injury, which is helpful for clinical medical staff to identify high-risk groups early and implement prevention, treatment, and nursing strategies in time.

BioGPT: generative pre-trained transformer for biomedical text generation and mining.

Pre-trained language models have attracted increasing attention in the biomedical domain, inspired by their great success in the general natural language domain. Among the two main branches of pre-trained language models in the general language domain, i.e. BERT (and its variants) and GPT (and its variants), the first one has been extensively studied in the biomedical domain, such as BioBERT and PubMedBERT. While they have achieved great success on a variety of discriminative downstream biomedical tasks, the lack of generation ability constrains their application scope. In this paper, we propose BioGPT, a domain-specific generative Transformer language model pre-trained on large-scale biomedical literature. We evaluate BioGPT on six biomedical natural language processing tasks and demonstrate that our model outperforms previous models on most tasks. Especially, we get 44.98%, 38.42% and 40.76% F1 score on BC5CDR, KD-DTI and DDI end-to-end relation extraction tasks, respectively, and 78.2% accuracy on PubMedQA, creating a new record. Our case study on text generation further demonstrates the advantage of BioGPT on biomedical literature to generate fluent descriptions for biomedical terms.