Temporal trends and forecasting of COVID-19 hospitalisations and deaths in Scotland using a national real-time patient-level data platform: a statistical modelling study

Colin R. Simpson*, Chris Robertson, Eleftheria Vasileiou, Emily Moore, Colin McCowan, Utkarsh Agrawal, Helen R. Stagg, Annemarie Docherty, Rachel Mulholland, Josephine L.K. Murray, Lewis D. Ritchie, Jim McMenamin, Aziz Sheikh

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

14 Citations (Scopus)
8 Downloads (Pure)

Abstract

Background: As the COVID-19 pandemic continues, national-level surveillance platforms with real-time individual person-level data are required to monitor and predict the epidemiological and clinical profile of COVID-19 and inform public health policy. We aimed to create a national dataset of patient-level data in Scotland to identify temporal trends and COVID-19 risk factors, and to develop a novel statistical prediction model to forecast COVID-19-related deaths and hospitalisations during the second wave. Methods: We established a surveillance platform to monitor COVID-19 temporal trends using person-level primary care data (including age, sex, socioeconomic status, urban or rural residence, care home residence, and clinical risk factors) linked to data on SARS-CoV-2 RT-PCR tests, hospitalisations, and deaths for all individuals resident in Scotland who were registered with a general practice on Feb 23, 2020. A Cox proportional hazards model was used to estimate the association between clinical risk groups and time to hospitalisation and death. A survival prediction model derived from data from March 1 to June 23, 2020, was created to forecast hospital admissions and deaths from October to December, 2020. We fitted a generalised additive spline model to daily SARS-CoV-2 cases over the previous 10 weeks and used this to create a 28-day forecast of the number of daily cases. The age and risk group pattern of cases in the previous 3 weeks was then used to select a stratified sample of individuals from our cohort who had not previously tested positive, with future cases in each group sampled from a multinomial distribution. We then used their patient characteristics (including age, sex, comorbidities, and socioeconomic status) to predict their probability of hospitalisation or death. Findings: Our cohort included 5 384 819 people, representing 98·6% of the entire estimated population residing in Scotland during 2020. Hospitalisation and death among those testing positive for SARS-CoV-2 between March 1 and June 23, 2020, were associated with several patient characteristics, including male sex (hospitalisation hazard ratio [HR] 1·47, 95% CI 1·38–1·57; death HR 1·62, 1·49–1·76) and various comorbidities, with the highest hospitalisation HR found for transplantation (4·53, 1·87–10·98) and the highest death HR for myoneural disease (2·33, 1·46–3·71). For those testing positive, there were decreasing temporal trends in hospitalisation and death rates. The proportion of positive tests among older age groups (>40 years) and those with at-risk comorbidities increased during October, 2020. On Nov 10, 2020, the projected number of hospitalisations for Dec 8, 2020 (28 days later) was 90 per day (95% prediction interval 55–125) and the projected number of deaths was 21 per day (12–29). Interpretation: The estimated incidence of SARS-CoV-2 infection based on positive tests recorded in this unique data resource has provided forecasts of hospitalisation and death rates for the whole of Scotland. These findings were used by the Scottish Government to inform their response to reduce COVID-19-related morbidity and mortality. Funding: Medical Research Council, National Institute for Health Research Health Technology Assessment Programme, UK Research and Innovation Industrial Strategy Challenge Fund, Health Data Research UK, Scottish Government Director General Health and Social Care.

Original languageEnglish
Pages (from-to)e517-e525
Number of pages9
JournalThe Lancet Digital Health
Volume3
Issue number8
Early online date5 Jul 2021
DOIs
Publication statusPublished - Aug 2021

Bibliographical note

Funding Information:
This study is part of the EAVE II project. EAVE II is funded by the MRC (MR/R008345/1) with the support of BREATHE—The Health Data Research Hub for Respiratory Health (MC_PC_19004), which is funded through the UK Research and Innovation Industrial Strategy Challenge Fund and delivered through Health Data Research UK. Additional support has been provided through Public Health Scotland and Scottish Government Director General Health and Social Care. The original EAVE project was funded by the NIHR Health Technology Assessment programme (11/46/23). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. We thank Dave Kelly from Albasoft (Inverness, UK) for his support with making primary care data available and James Pickett (Health Data Research UK, London, UK); Wendy Inglis-Humphrey, Vicky Hammersley, Laura Brook, Maria Georgiou, and Laura Gonzalez Rienda (University of Edinburgh, Edinburgh, UK); and Pam McVeigh, Amanda Burridge, Sumedha Asnani-Chetal, and Afshin Dastafshan (Public Health Scotland, Glasgow, UK) for their support with project management and administration.

Fingerprint

Dive into the research topics of 'Temporal trends and forecasting of COVID-19 hospitalisations and deaths in Scotland using a national real-time patient-level data platform: a statistical modelling study'. Together they form a unique fingerprint.

Cite this