A Natural Language Processing Model for COVID-19 Detection Based on Dutch General Practice Electronic Health Records by Using Bidirectional Encoder Representations From Transformers: Development and Validation Study.

BACKGROUND: Natural language processing (NLP) models such as bidirectional encoder representations from transformers (BERT) hold promise in revolutionizing disease identification from electronic health records (EHRs) by potentially enhancing efficiency and accuracy. However, their practical application in practice settings demands a comprehensive and multidisciplinary approach to development and validation. The COVID-19 pandemic highlighted challenges in disease identification due to limited testing availability and challenges in handling unstructured data. In the Netherlands, where general practitioners (GPs) serve as the first point of contact for health care, EHRs generated by these primary care providers contain a wealth of potentially valuable information. Nonetheless, the unstructured nature of free-text entries in EHRs poses challenges in identifying trends, detecting disease outbreaks, or accurately pinpointing COVID-19 cases. OBJECTIVE: This study aims to develop and validate a BERT model for detecting COVID-19 consultations in general practice EHRs in the Netherlands. METHODS: The BERT model was initially pretrained on Dutch language data and fine-tuned using a comprehensive EHR data set comprising confirmed COVID-19 GP consultations and non-COVID-19-related consultations. The data set was partitioned into a training and development set, and the model's performance was evaluated on an independent test set that served as the primary measure of its effectiveness in COVID-19 detection. To validate the final model, its performance was assessed through 3 approaches. First, external validation was applied on an EHR data set from a different geographic region in the Netherlands. Second, validation was conducted using results of polymerase chain reaction (PCR) test data obtained from municipal health services. Lastly, correlation between predicted outcomes and COVID-19-related hospitalizations in the Netherlands was assessed, encompassing the period around the outbreak of the pandemic in the Netherlands, that is, the period before widespread testing. RESULTS: The model development used 300,359 GP consultations. We developed a highly accurate model for COVID-19 consultations (accuracy 0.97, F1-score 0.90, precision 0.85, recall 0.85, specificity 0.99). External validations showed comparable high performance. Validation on PCR test data showed high recall but low precision and specificity. Validation using hospital data showed significant correlation between COVID-19 predictions of the model and COVID-19-related hospitalizations (F1-score 96.8; P<.001; R2=0.69). Most importantly, the model was able to predict COVID-19 cases weeks before the first confirmed case in the Netherlands. CONCLUSIONS: The developed BERT model was able to accurately identify COVID-19 cases among GP consultations even preceding confirmed cases. The validated efficacy of our BERT model highlights the potential of NLP models to identify disease outbreaks early, exemplifying the power of multidisciplinary efforts in harnessing technology for disease identification. Moreover, the implications of this study extend beyond COVID-19 and offer a blueprint for the early recognition of various illnesses, revealing that such models could revolutionize disease surveillance.

Site-specific functionality and tryptophan mimicry of lipidation in tetraspanin CD9.

Lipidation of transmembrane proteins regulates many cellular activities, including signal transduction, cell-cell communication, and membrane trafficking. However, how lipidation at different sites in a membrane protein affects structure and function remains elusive. Here, using native mass spectrometry we determined that wild-type human tetraspanins CD9 and CD81 exhibit nonstochastic distributions of bound acyl chains. We revealed CD9 lipidation at its three most frequent lipidated sites suffices for EWI-F binding, while cysteine-to-alanine CD9 mutations markedly reduced binding of EWI-F. EWI-F binding by CD9 was rescued by mutating all or, albeit to a lesser extent, only the three most frequently lipidated sites into tryptophans. These mutations did not affect the nanoscale distribution of CD9 in cell membranes, as shown by super-resolution microscopy using a CD9-specific nanobody. Thus, these data demonstrate site-specific, possibly conformation-dependent, functionality of lipidation in tetraspanin CD9 and identify tryptophan mimicry as a possible biochemical approach to study site-specific transmembrane-protein lipidation.

Low recurrence rate of head and neck basal cell carcinoma treated with Mohs micrographic surgery: A retrospective study of 1021 cases.

OBJECTIVE: Mohs micrographic surgery (MMS) is the treatment of choice for high-risk primary basal cell carcinoma (BCC) and recurrent BCC of the head and neck, showing fewer recurrences compared with surgical excision (SE). The objectives of this study were to determine the recurrence rate of head and neck BCC after MMS and to develop a prediction model with significant risk factors for recurrence. DESIGN: A retrospective study of patient records. METHODS: All BCCs treated with MMS between 1 January 1995 and 1 July 2013 at the University Medical Center Groningen (UMCG) were selected retrospectively. Recorded parameters were patient characteristics, tumour size, tumour location, histopathological subtype, previous treatment, the number of stages until microscopic clearance, defect size, adverse events, follow-up time and recurrence after MMS. RESULTS: The study covered 1021 MMS operations conducted on primary BCCs (57.4%), residual BCCs (25.6%) and recurrent BCCs (17.0%). The median follow-up time was 54.0 months (ranging from 1 to 221 months). The 5-year cumulative probability of recurrence was 3.3%: 2.6% for primary BCCs, 5.4% for residual BCCs and 2.9% for recurrent BCCs. An aggressive histopathological subtype, residual BCCs and recurrent BCCs were significant risk factors for predicting a higher risk of recurrence after MMS. CONCLUSION: This large-scale retrospective study showed low recurrence rates after MMS for primary and recurrent BCCs. Residual BCCs treated with MMS had relatively higher recurrence rates. The risk of recurrence for MMS-treated residual aggressive BCCs was more than four times higher than that for primary non-aggressive BCCs.