New tools for network time series with an application to COVID-19 hospitalisations (preprint)

Network time series are becoming increasingly important across many areas in science and medicine and are often characterised by a known or inferred underlying network structure, which can be exploited to make sense of dynamic phenomena that are often high-dimensional. For example, the Generalised Network Autoregressive (GNAR) models exploit such structure parsimoniously. We use the GNAR framework to introduce two association measures: the network and partial network autocorrelation functions, and introduce Corbit (correlation-orbit) plots for visualisation. As with regular autocorrelation plots, Corbit plots permit interpretation of underlying correlation structures and, crucially, aid model selection more rapidly than using other tools such as AIC or BIC. We additionally interpret GNAR processes as generalised graphical models, which constrain the processes' autoregressive structure and exhibit interesting theoretical connections to graphical models via utilization of higher-order interactions. We demonstrate how incorporation of prior information is related to performing variable selection and shrinkage in the GNAR context. We illustrate the usefulness of the GNAR formulation, network autocorrelations and Corbit plots by modelling a COVID-19 network time series of the number of admissions to mechanical ventilation beds at 140 NHS Trusts in England & Wales. We introduce the Wagner plot that can analyse correlations over different time periods or with respect to external covariates. In addition, we introduce plots that quantify the relevance and influence of individual nodes. Our modelling provides insight on the underlying dynamics of the COVID-19 series, highlights two groups of geographically co-located `influential' NHS Trusts and demonstrates superior prediction abilities when compared to existing techniques.

Developing survey weights to ensure representativeness in a national, matched cohort study: results from the Children and young people with Long Covid (CLoCk) study. (preprint)

Background: Findings from studies assessing Long Covid in children and young people (CYP) need to be viewed in light of their methodological limitations. For example, if non-response and/or attrition over time systematically differ by sub-groups of CYP, findings could be biased and generalisation limited. The present study aimed to (i) construct survey weights for the Children and young people with Long Covid (CLoCk) study, and (ii) apply them to published CLoCk findings showing the prevalence of shortness of breath and tiredness increased over time from baseline to 12-months post-baseline in both SARS-CoV-2 Positive and Negative CYP. Methods: Logistic regression was used to compute the probability of (i) Responding given envisioned to take part, (ii) Responding timely given responded, and (iii) (Re)infection given timely response. Response, timely response and (re)infection weights were generated as the reciprocal of the corresponding probability, with an overall ‘envisioned population’ survey weight derived as the product of these weights. Survey weights were trimmed, and an interactive tool developed to re-calibrate target population survey weights to the general population using data from the 2021 UK Census. Results: Flexible survey weights for the CLoCk study were successfully developed. In the illustrative example re-weighted results (when accounting for selection in response, attrition, and (re)infection) were consistent with published findings. Conclusions: Flexible survey weights to address potential bias and selection issues were created for and used in the CLoCk study. Previously reported prospective findings from CLoCk are generalisable to the wider population of CYP in England. This study highlights the importance of considering selection into a sample and attrition over time when considering generalisability of findings.

Indirect impacts of the COVID-19 pandemic at two tertiary neonatal units in Zimbabwe and Malawi: an interrupted time series analysis (preprint)

ABSTRACT Background Deaths from COVID-19 have exceeded 1.8 million globally (January 2020). We examined trends in markers of neonatal care before and during the pandemic at two tertiary neonatal units in Zimbabwe and Malawi. Methods We analysed data collected prospectively via the NeoTree app at Sally Mugabe Central Hospital (SMCH), Zimbabwe, and Kamuzu Central Hospital (KCH), Malawi. Neonates admitted from 1 June 2019 to 25 September 2020 were included. We modelled the impact of the first cases of COVID-19 (Zimbabwe: 20 March 2020; Malawi: 3 April 2020) on number of admissions, gestational age and birth weight, source of admission referrals, prevalence of neonatal encephalopathy, and overall mortality. Findings The study included 3,450 neonates at SMCH and 3,350 neonates at KCH. Admission numbers at SMCH did not initially change after the first case of COVID-19 but fell by 48% during a nurses’ strike (Relative risk (RR) 0·52, 95%CI 0·40-0·68, p < 0·002). At KCH, admissions dropped by 42% (RR 0·58; 95%CI 0·48-0·70; p < 0·001) soon after the first case of COVID-19. At KCH, gestational age and birth weight decreased slightly (1 week, 300 grams), outside referrals dropped by 28%, and there was a slight weekly increase in mortality. No changes in these outcomes were found at SMCH. Interpretation The indirect impacts of COVID-19 are context-specific. While this study provides vital evidence to inform health providers and policy makers, national data are required to ascertain the true impacts of the pandemic on newborn health. Funding International Child Health Group, Wellcome Trust. RESEARCH IN CONTEXT Evidence before this study We searched PubMed for evidence of the indirect impact of the COVID-19 pandemic on neonatal care in low-income settings using the search terms neonat* or newborn , and COVID-19 or SARS-CoV 2 or coronavirus , and the Cochrane low and middle income country (LMIC) filters, with no language limits between 01.10.2019 and 21.11.20. While there has been a decrease in global neonatal mortality rates, the smaller improvements seen in low-income settings are threatened by the direct and indirect impact of the COVID-19 pandemic. A modelling study of this threat predicted between 250000-1.1 million extra neonatal deaths as a result of decreased service provision and access in LMICs. A webinar and survey of frontline maternal/newborn healthcare workers in >60 countries reported a decline in both service attendance and in quality of service across the ante-, peri- and post-natal journey. Reporting fear of attending services, and difficulty in access, and a decrease in service quality due to exacerbation of existing service weaknesses, confusion over guidelines and understaffing. Similar findings were reported in a survey of healthcare workers providing childhood and maternal vaccines in LMICs. One study to date has reported data from Nepal describing an increase in stillbirths and neonatal deaths, with institutional deliveries nearly halved during lockdown. Added value of this study To our knowledge, this is the first and only study in Sub-Saharan Africa describing the impact of COVID-19 pandemic on health service access and outcomes for newborns in two countries. We analysed data from the digital quality improvement and data collection tool, the NeoTree, to carry out an interrupted time series analysis of newborn admission rates, gestational age, birth weight, diagnosis of hypoxic ischaemic encephalopathy and mortality from two large hospitals in Malawi and Zimbabwe ( n ∼7000 babies). We found that the indirect impacts of COVID-19 were context-specific. In Sally Mugabe Central Hospital, Zimbabwe, initial resilience was demonstrated in that there was no evidence of change in mortality, birth weight or gestational age. In comparison, at Kamuzu Central Hospital, Malawi, soon after the first case of COVID-19, the data revealed a fall in admissions (by 42%), gestational age (1 week), birth weight (300 grams), and outside referrals (by 28%), and there was a slight weekly increase in mortality (2%). In the Zimbabwean hospital, admission numbers did not initially change after the first case of COVID-19 but fell by 48% during a nurses’ strike, which in itself was in response to challenges exacerbated by the pandemic. Implications of all the available evidence Our data confirms the reports from frontline healthcare workers of a perceived decline in neonatal service access and provision in LMICs. Digital routine healthcare data capture enabled rapid profiling of indirect impacts of COVID-19 on newborn care and outcomes in two tertiary referral hospitals, Malawi and Zimbabwe. While a decrease in service access was seen in both countries, the impacts on care provided and outcome differed by national context. Health systems strengthening, for example digital data capture, may assist in planning context-specific mitigation efforts.

Systematic review and patient-level meta-analysis of SARS-CoV-2 viral dynamics to model response to antiviral therapies (preprint)

SARS-CoV-2 viral loads change rapidly following symptom onset so to assess antivirals it is important to understand the natural history and patient factors influencing this. We undertook an individual patient-level meta-analysis of SARS-CoV-2 viral dynamics in humans to describe viral dynamics and estimate the effects of antivirals used to-date. This systematic review identified case reports, case series and clinical trial data from publications between 1/1/2020 and 31/5/2020 following PRISMA guidelines. A multivariable Cox proportional hazards regression model (Cox-PH) of time to viral clearance was fitted to respiratory and stool samples. A simplified four parameter nonlinear mixed-effects (NLME) model was fitted to viral load trajectories in all sampling sites and covariate modelling of respiratory viral dynamics was performed to quantify time dependent drug effects. Patient-level data from 645 individuals (age 1 month-100 years) with 6316 viral loads were extracted. Model-based simulations of viral load trajectories in samples from the upper and lower respiratory tract, stool, blood, urine, ocular secretions and breast milk were generated. Cox-PH modelling showed longer time to viral clearance in older patients, males and those with more severe disease. Remdesivir was associated with faster viral clearance (adjusted hazard ratio (AHR) = 9.19, p<0.001), as well as interferon, particularly when combined with ribavirin (AHR = 2.2, p=0.015; AHR = 6.04, p = 0.006). Combination therapy should be further investigated. A viral dynamic dataset and NLME model for designing and analysing antiviral trials has been established.