Brendan Delaney

Long COVID affects millions worldwide and its prevention is a critical public health strategy. While prior analyses show primary vaccination prevents long COVID in subsequent infections, the effect of booster vaccination on long COVID after Omicron infections is unclear. This systematic review identifies 31 observational studies, of which 11 are suitable for pairwise meta-analyses. The pooled odds ratio (OR) of long COVID in those vaccinated (any dose) versus unvaccinated is 0.77 (95% confidence interval [CI] 0.70–0.85; p < 0.0001; 10 studies). ORs were also lower for primary course vaccination versus unvaccinated (OR 0.81; 95% CI 0.79–0.83; p < 0.0001; 3 studies), booster vaccination versus unvaccinated (OR 0.74; 95% CI 0.63–0.86; p = 0.0001; 4 studies), and booster vaccination versus primary course vaccination (OR 77; 95% CI 0.65–0.92; p = 0.0044; 3 studies). These findings indicate that booster vaccination can provide additional protection against long COVID, highlighting the importance of seasonal vaccination against new SARS-CoV-2 variants. They should, however, be interpreted cautiously, given the small number of studies and the low quality of evidence.

Background: Many patients diagnosed with gastro‐oesophageal reflux disease (GERD) have persistent symptoms despite proton pump inhibitor (PPI) therapy.Aims: The aim of this consensus is to provide evidence‐based statements to guide clinicians caring for patients with refractory reflux‐like symptoms (rRLS) or refractory GERD.Methods: This consensus was developed by the International Working Group for the Classification of Oesophagitis. The steering committee developed specific PICO questions pertaining to the management of PPI rRLS. Methodologists conducted systematic reviews of the literature. The quality of evidence and strength of recommendations were rated using the GRADE approach.Results: Consensus was reached on 13 of 17 statements on diagnosis and management. For rRLS, suggested diagnostic strategies included endoscopy, ambulatory reflux testing and oesophageal manometry. The group did not reach consensus on the role of oesophageal biopsies or the use of reflux‐symptom association in patients undergoing reflux testing. The group suggested against increasing the PPI dose in patients who had received 8 weeks of a twice‐daily PPI. Adjunctive alginate or antacid therapy was suggested. There was no consensus on the role of adjunctive prokinetics. There was little role for adjunctive transient lower oesophageal sphincter relaxation (TLESR) inhibitors or bile acid sequestrants. Endoscopic or surgical anti‐reflux procedures should not be performed in patients with rRLS in the absence of objectively confirmed GERD.Conclusions: The management of rRLS should be personalised, based on shared decision‐making regarding the role of diagnostic testing to confirm or rule out GERD as a basis for treatment optimisation. Anti‐reflux procedures should not be performed without objective confirmation of GERD.

Having been a GP since 1992, I can remember the times without an EHR, trying to decipher a previous GP’s scrawl in the Lloyd George envelope and the coloured dots on the cover that indicated a chronic condition. The whole of modern practice rests on both the EHR and networked services for requests and results. However, we still behave as if we are dealing with paper (if in pdf form), shuffling it around electronically and manually entering a code for a new diagnosis. Likewise, clinical decision support has become a monster, where every attempt to enter something in the record is couched in a plethora of warnings, reminders, admonishments, and potential interactions. Has no one ever read about alert fatigue?⁴ I’m looking at you, Ardens. I confess to nostalgic pleasure in writing a note unhindered, followed by signing a handwritten script. In my academic role as a Chair in Medical Informatics I hope that we can do better in future, and that artificial intelligence can make the current approach seem as archaic as the Lloyd George envelope.^5,6

Background Due to its late stage of diagnosis lung cancer is the commonest cause of death from cancer in the UK. Existing epidemiological risk models in clinical usage, which have Positive Predictive Values (PPV) of less than 10%, do not consider the temporal relations expressed in sequential electronic health record (EHR) data. Machine learning with deep ‘transformer’ models can learn from these temporal relationships. We aimed to build such a model for lung cancer diagnosis in primary care using EHR data. Methods In a nested case-control study within the Whole Systems Integrated Care (WSIC) dataset, lung cancer cases were identified and control cases of ‘other’ cancers or respiratory conditions. GP EHR data going back three years from the date of diagnosis less the most recent one months were semantically pre-processed by mapping from more than 30,000 terms to 450. Model building was performed using ALBERT with a Logistic Regression Classifier (LRC) head. Clustering was explored using k-means. We split the data into 70% training and 30% validation. An additional regression model alone was built on the pre-processed data as a comparator. Findings Based on 3,303,992 patients from January 1981 to December 2020 there were 11,847 lung cancer cases of whom 9,629 had died. 5,789 cases and 7,240 controls were used for training and a population of 368,906 for validation. Our model achieved an AUROC of 0·924 (95% CI 0·921– 0·927) with a PPV of 3·6% (95% CI 3·5 – 3·7) and Sensitivity of 86·6% (95% CI 85·3 – 87·8) based on the three year’s data prior to diagnosis less the immediate month before index diagnosis. The comparator regression model achieved a PPV of 3·1% (95% CI 3·0 – 3·1) and AUROC of 0·887 (95% CI 0·884 – 0·889). Interpretation Capturing temporal sequencing between cancer and non-cancer pathways to diagnosis enables much more accurate models. Future work will focus on external dataset validation and integration into GP clinical systems for evaluation. Evidence before the study Predictive models for early detection of cancer are a priority as treatment intensity and cancer outcomes and survival are strongly linked to cancer stage at diagnosis. We searched PubMed and Embase for research on lung cancer prediction, using the search terms “lung cancer”, “diagnos$”, and “prediction model” between Jan 1, 2000 and Dec 31, 2023, to look into the contemporary research on prediction models for lung cancer. The QCancer Lung model has been recommended for prediction of lung cancer in primary care. However, classic regression models do not consider the rich relationships and dependencies in the electronic health record (EHR) data, such as cough followed by pneumonia rather than just cough in isolation. Since 2018, with advances in the natural language processing (NLP) domain, transformer-based models have been applied on large amounts of EHR data for clinical predictive modelling. We searched Google Scholar and PubMed for studies using transformer-based models on EHR data. We used the terms (“transformer” OR “bert” OR “pretrain” OR “prediction” OR “predictive modelling” OR “contextualised”) AND (“ehr” OR “health records” OR “healthcare” OR “clinical records” OR “cancer” OR “disease”) in free text, published from Jan 2019 to Dec 2023. We found these studies were limited to diagnosis and medication concepts/codes in patients’ records in secondary care, omitting symptom, test, procedure, and referral codes. The early detection of lung cancer requires the improvement in the prediction performance of deep learning models. We updated the literature review when writing this paper (Apr 2024) to include the latest published studies. Added value of this study We pretrained a transformer-based deep learning model, MedAlbert, for learning deep patient pathway representations from coded EHR data in primary care. This ‘Pathway to Diagnosis’ for each patient is defined to contain the most possible elaboration of the coded medical records appearing over three years before diagnosis. To our knowledge, we are the first to build models on such detailed clinical records in primary care without data aggregation. Developed and validated based on the pretrained MedAlbert, the prediction model, MedAlber+LRC, shows improved prediction performance for diagnosis of suspected lung cancer as well as one- and two-year lung cancer early detection compared with a classic machine learning model (a single Logistic Regression Model), MedAlbert+LRC performed better in terms of sensitivity, specificity, PPV and AUROC. The explainability of the model discovered a series of symptoms, comorbidities and procedures associated with lung cancer diagnosis and identified six groups of patients related to COPD, diabetes, other cancers, etc. The prediction model we developed could be applied to the UK primary care population for early diagnosis of lung cancer. Implications of all available evidence In order to progress beyond simple ‘red flag’ driven referral guidance and to develop more accurate prediction models for early diagnosis of lung cancer, it is necessary to use more sophisticated machine learning methods. Additionally, the framework we designed for deriving, modelling, and analysing the patient pathways could be used for the prediction of other cancers or diseases. The improvement in early diagnosis of lung cancer could contribute to better cancer outcomes and survival rates. Deep learning for diagnosis could provide more efficient care delivery and more accurate decisions faster, reducing costs and suffering across societies in the UK and worldwide.

Introduction: Long COVID (LC) is a global public health crisis affecting more than 70 million people. There is emerging evidence of different pathophysiological mechanisms driving the wide array of symptoms in LC. Understanding the relationships between mechanisms and symptoms helps in guiding clinical management and identifying potential treatment targets. Methods: This was a mixed-methods systematic review with two stages: Stage one (Review 1) included only existing systematic reviews (meta-review) and Stage two (Review 2) was a review of all primary studies. The search strategy involved Medline, Embase, Emcare, and CINAHL databases to identify studies that described symptoms and pathophysiological mechanisms with statistical analysis and/or discussion of plausible causal relationships between mechanisms and symptoms. Only studies that included a control arm for comparison were included. Studies were assessed for quality using the National Heart, Lung, and Blood Institute quality assessment tools. Results: 19 systematic reviews were included in Review 1 and 46 primary studies in Review 2. Overall, the quality of reporting across the studies included in this second review was moderate to poor. The pathophysiological mechanisms with strong evidence were immune system dysregulation, cerebral hypoperfusion, and impaired gas transfer in the lungs. Other mechanisms with moderate to weak evidence were endothelial damage and hypercoagulation, mast cell activation, and auto-immunity to vascular receptors. Conclusions: LC is a complex condition affecting multiple organs with diverse clinical presentations (or traits) underpinned by multiple pathophysiological mechanisms. A ‘treatable trait’ approach may help identify certain groups and target specific interventions. Future research must include understanding the response to intervention based on these mechanism-based traits.

Background: Around 2 million people in the UK suffer from Long COVID (LC). Of concern is the disease impact on productivity and informal care burden. This study aimed to quantify and value productivity losses and informal care receipt in a sample of LC patients in the UK. Methods: The target population comprised LC patients referred to LC specialist clinics. The questionnaires included a health economics questionnaire (HEQ) measuring productivity impacts, informal care receipt and service utilisation, EQ-5D-5L, C19-YRS LC condition-specific measure, and sociodemographic and COVID-19 history variables. Outcomes were changes from the incident infection resulting in LC to the month preceding the survey in paid work status/h, work income, work performance and informal care receipt. The human capital approach valued productivity losses; the proxy goods method valued caregiving hours. The values were extrapolated nationally using published prevalence data. Multilevel regressions, nested by region, estimated associations between the outcomes and patient characteristics. Results: 366 patients responded to HEQ (mean LC duration 449.9 days). 51.7% reduced paid work hours relative to the pre-infection period. Mean monthly work income declined by 24.5%. The average aggregate value of productivity loss since incident infection was £10,929 (95% bootstrap confidence interval £8,844-£13,014) and £5.7 billion (£3.8-£7.6 billion) extrapolated nationally. The corresponding values for informal caregiving were £8,726 (£6,247-£11,204) and £4.8 billion (£2.6-£7.0 billion). Multivariate analyses found significant associations between each outcome and health utility and C19-YRS subscale scores. Conclusion: LC significantly impacts productivity losses and provision of informal care, exacerbated by high national prevalence of LC.

Purpose Orthostatic intolerance (OI), including postural orthostatic tachycardia syndrome (PoTS) and orthostatic hypotension (OH), are often reported in long covid, but published studies are small with inconsistent results. We sought to estimate the prevalence of objective OI in patients attending long covid clinics and healthy volunteers and associations with symptoms and comorbidities. Methods Participants were recruited from 8 UK long covid clinics, and healthy volunteers from general population. All undertook standardised National Aeronautics and Space Administration Lean Test (NLT). Participants’ history of typical OI symptoms (e.g. dizziness, palpitations) prior to and during the NLT were recorded. Results 277 long covid patients and 50 frequency-matched healthy volunteers were tested. Healthy volunteers had no history of OI symptoms or PoTS, 10% had asymptomatic OH. 130 (47%) long covid patients had previous history of OI symptoms and 144 (52%) developed symptoms during the NLT. 41 (15%) had an abnormal NLT, 20 (7%) met criteria for PoTS and 21 (8%) had OH. Of patients with an abnormal NLT, 45% had no prior symptoms of OI. Relaxing the diagnostic thresholds for PoTS from two consecutive to one reading, resulted in 11% of long covid participants meeting criteria for PoTS, but not in healthy volunteers. Conclusion More than half of long covid patients experienced OI symptoms during NLT and more than one in ten patients met the criteria for either PoTS or OH, half of whom did not report previous typical OI symptoms. We recommend all patients attending long covid clinics are offered an NLT and appropriate management commenced. Trial registration numbers NCT05057260, ISRCTN15022307