ABSTRACT
Over time, the performance of clinical prediction models may deteriorate due to changes in clinical management, data quality, disease risk and/or patient mix. Such prediction models must be updated in order to remain useful. Here, we investigate methods for discrete and dynamic model updating of clinical survival prediction models based on refitting, recalibration and Bayesian updating. In contrast to discrete or one-time updating, dynamic updating refers to a process in which a prediction model is repeatedly updated with new data. Motivated by infectious disease settings, our focus was on model performance in rapidly changing environments. We first compared the methods using a simulation study. We simulated scenarios with changing survival rates, the introduction of a new treatment and predictors of survival that are rare in the population. Next, the updating strategies were applied to patient data from the QResearch database, an electronic health records database from general practices in the UK, to study the updating of a model for predicting 70-day covid-19 related mortality. We found that a dynamic updating process outperformed one-time discrete updating in the simulations. Bayesian dynamic updating has the advantages of making use of knowledge from previous updates and requiring less data compared to refitting.
Subject(s)
COVID-19ABSTRACT
Quantifying the waning effectiveness of second COVID-19 vaccination beyond six months and against the omicron variant is important for scheduling subsequent doses. With the approval of NHS England, we estimated effectiveness up to one year after second dose, but found that bias in such estimates may be substantial.
Subject(s)
COVID-19ABSTRACT
Introduction The COVID-19 booster vaccination programme in England used both BNT162b2 and mRNA-1273 vaccines. Direct comparisons of the effectiveness against severe COVID-19 of these two vaccines for boosting have not been made in trials or observational data. Methods On behalf of NHS England, we used the OpenSAFELY-TPP database to match adult recipients of each vaccine type on date of vaccination, primary vaccine course, age, and other characteristics. Recipients were eligible if boosted between 29 October 2021 and 31 January 2022, and followed up for 12 weeks. Outcomes were positive SARS-CoV-2 test, COVID-19 hospitalisation, and COVID-19 death. We estimated the cumulative incidence of each outcome, and quantified comparative effectiveness using risk differences (RD) and hazard ratios (HRs). Results 1,528,431 people were matched in each group, contributing a total 23,150,504 person-weeks of follow-up. The 12-week risks per 1,000 people of positive SARS-CoV-2 test were 103.2 (95%CI 102.4 to 104.0) for BNT162b2 and 96.0 (95.2 to 96.8) for mRNA-1273: the HR comparing mRNA-1273 with BNT162b2 was 0.92 (95%CI 0.91 to 0.92). For COVID-19 hospitalisations the 12-week risks per 1,000 were 0.65 (95%CI 0.56 to 0.75) and 0.44 (0.36 to 0.54): HR 0.67 (95%CI 0.58 to 0.78). COVID-19 deaths were rare: the 12-week risks per 1,000 were 0.03 (95%CI 0.02 to 0.06) and 0.01 (0.01 to 0.02): HR 1.23 (95%CI 0.59 to 2.56). Comparative effectiveness was generally similar within subgroups. Conclusion Booster vaccination with mRNA-1273 COVID-19 vaccine was more effective than BNT162b2 in preventing SARS-CoV-2 infection and COVID-19 hospitalisation during the first 12 weeks after vaccination.
Subject(s)
COVID-19 , DeathABSTRACT
VOC 202012/01, a SARS-CoV-2 variant first detected in the United Kingdom in September 2020, has spread to multiple countries worldwide. Several studies have established that this novel variant is more transmissible than preexisting variants of SARS-CoV-2, but have not identified whether the new variant leads to any change in disease severity. We analyse a large database of SARS-CoV-2 community test results and COVID-19 deaths for England, representing approximately 47% of all SARS-CoV-2 community tests and 7% of COVID-19 deaths in England from 1 September 2020 to 22 January 2021. Fortuitously, these SARS-CoV-2 tests can identify VOC 202012/01 because mutations in this lineage prevent PCR amplification of the spike gene target (S gene target failure, SGTF). We estimate that the hazard of death among SGTF cases is 30% (95% CI 9-56%) higher than among non-SGTF cases after adjustment for age, sex, ethnicity, deprivation level, care home residence, local authority of residence and date of test. In absolute terms, this increased hazard of death corresponds to the risk of death for a male aged 55-69 increasing from 0.56% to 0.73% (95% CI 0.60-0.86%) over the 28 days following a positive SARS-CoV-2 test in the community. Correcting for misclassification of SGTF, we estimate a 35% (12-64%) higher hazard of death associated with VOC 202012/01. Our analysis suggests that VOC 202012/01 is not only more transmissible than preexisting SARS-CoV-2 variants but may also cause more severe illness.
Subject(s)
COVID-19 , DeathABSTRACT
ObjectivesPredicting bed occupancy for hospitalised patients with COVID-19 requires understanding of length of stay (LoS) in particular bed types. LoS can vary depending on the patients "bed pathway" - the sequence of transfers between bed types during a hospital stay. In this study, we characterise these pathways, and their impact on predicted hospital bed occupancy. DesignWe obtained data from University College Hospital (UCH) and the ISARIC4C COVID-19 Clinical Information Network (CO-CIN) on hospitalised patients with COVID-19 who required care in general ward or critical care (CC) beds to determine possible bed pathways and LoS. We developed a discrete-time model to examine the implications of using either bed pathways or only average LoS by bed type to forecast bed occupancy. We compared model-predicted bed occupancy to publicly available bed occupancy data on COVID-19 in England between March and August 2020. ResultsIn both the UCH and CO-CIN datasets, 82% of hospitalised patients with COVID-19 only received care in general ward beds. We identified four other bed pathways, present in both datasets: "Ward, CC, Ward", "Ward, CC", "CC" and "CC, Ward". Mean LoS varied by bed type, pathway, and dataset, between 1.78 and 13.53 days. For UCH, we found that using bed pathways improved the accuracy of bed occupancy predictions, while only using an average LoS for each bed type underestimated true bed occupancy. However, using the CO-CIN LoS dataset we were not able to replicate past data on bed occupancy in England, suggesting regional LoS heterogeneities. ConclusionsWe identified five bed pathways, with substantial variation in LoS by bed type, pathway, and geography. This might be caused by local differences in patient characteristics, clinical care strategies, or resource availability, and suggests that national LoS averages may not be appropriate for local forecasts of bed occupancy for COVID-19.
Subject(s)
COVID-19ABSTRACT
A novel SARS-CoV-2 variant, VOC 202012/01 (lineage B.1.1.7), emerged in southeast England in November 2020 and is rapidly spreading towards fixation. Using a variety of statistical and dynamic modelling approaches, we estimate that this variant has a 43-90% (range of 95% credible intervals 38-130%) higher reproduction number than preexisting variants. A fitted two-strain dynamic transmission model shows that VOC 202012/01 will lead to large resurgences of COVID-19 cases. Without stringent control measures, including limited closure of educational institutions and a greatly accelerated vaccine roll-out, COVID-19 hospitalisations and deaths across England in 2021 will exceed those in 2020. Concerningly, VOC 202012/01 has spread globally and exhibits a similar transmission increase (59-74%) in Denmark, Switzerland, and the United States.
Subject(s)
COVID-19ABSTRACT
Background Mortality rates of UK patients hospitalised with COVID-19 appeared to fall during the first wave. We quantify potential drivers of this change and identify groups of patients who remain at high risk of dying in hospital. Methods The International Severe Acute Respiratory and emerging Infection Consortium (ISARIC) WHO Clinical Characterisation Protocol UK recruited a prospective cohort admitted to 247 acute UK hospitals with COVID-19 in the first wave (March to August 2020). Outcome was hospital mortality within 28 days of admission. We performed a three-way decomposition mediation analysis using natural effects models to explore associations between week of admission and hospital mortality adjusting for confounders (demographics, comorbidity, illness severity) and quantifying potential mediators (respiratory support and steroids). Findings Unadjusted hospital mortality fell from 32.3% (95%CI 31.8, 32.7) in March/April to 16.4% (95%CI 15.0, 17.8) in June/July 2020. Reductions were seen in all ages, ethnicities, both sexes, and in comorbid and non-comorbid patients. After adjustment, there was a 19% reduction in the odds of mortality per 4 week period (OR 0.81, 95%CI 0.79, 0.83). 15.2% of this reduction was explained by greater disease severity and comorbidity earlier in the epidemic. The use of respiratory support changed with greater use of non-invasive ventilation (NIV). 22.2% (OR 0.94, 95%CI 0.94, 0.96) of the reduction in mortality was mediated by changes in respiratory support. Interpretation The fall in hospital mortality in COVID-19 patients during the first wave in the UK was partly accounted for by changes in case mix and illness severity. A significant reduction was associated with differences in respiratory support and critical care use, which may partly reflect improved clinical decision making. The remaining improvement in mortality is not explained by these factors, and may relate to community behaviour on inoculum dose and hospital capacity strain. Funding NIHR & MRC