Plain language summary

The COVID-19 pandemic caused many deaths over a prolonged period in England, substantially increasing population mortality rates at the time, disrupting mortality trends, and displacing forward many deaths that would have been expected in subsequent years. Research to date has described greater mortality in those with pre-existing health conditions or who were otherwise more clinically vulnerable. However, such research has generally been limited to samples of the population, mortality in specific subgroups, or during certain “waves” of the pandemic. Population level statistics have been published on excess mortality throughout the pandemic by cause. These are however based on recording of COVID-19 on the death certificate. There is a need to study whole populations to directly assess mortality risk following a positive test for SARS-CoV-2 and how it varies in different groups, and to assess the likely impact of COVID-19 mortality that occurred during the pandemic period on future mortality rates.

The study uses data on the whole population to assess the relative risk of death following a positive SARS-CoV-2 test result in England and how it varies by age, sex, and vaccination status over the first three “waves” of the pandemic, as well as providing relative mortality ratios by socio-demographic status and geography, comorbidities, and clinical vulnerability. These risk estimates have then been combined to approximate the overall mortality risk from COVID-19 compared with the pre-existing risk for groups of individuals with specific characteristics. Thus, allowing us to assess the likely impact of COVID-19 on future mortality trends due to mortality displacement by the COVID-19 pandemic.

Understanding patterns of mortality risk following confirmed SARS-CoV-2 infection is important for understanding the impact of the pandemic. Documenting variation in mortality rates between specific population groups enables planning for any future pandemic but is also desirable in reflecting on the experience and handling of the current pandemic. Many individuals died earlier than would otherwise have been the case because of COVID-19. At a population level, this resulted in excess deaths during the height of the pandemic but also fewer deaths than expected during some later periods. This analysis provides a basis for estimating the likely effect of this mortality displacement on future mortality rates.

Introduction

The COVID-19 pandemic caused prolonged periods of excess mortality in many countries. In England, between March 2020 and February 2023, there were 154,030 deaths in excess of what might have been expected had the pandemic not occurred [1]. The risk of mortality following COVID-19 is known to differ between individuals, with various factors impacting the risk including, but not limited to, sex, ethnicity, deprivation, and certain health conditions [2–6]. Risk of death also varied between waves of the pandemic, due to changing virulence of the dominant viral strain and the impact of counter measures such as shielding for those at greatest risk, specific treatments for diagnosed cases, and vaccine protection [2, 3, 7].

Existing studies assessing risk of mortality from COVID-19 in relation to comorbidities were important in informing shielding, vaccination, and therapeutic programmes and other public health measures [3, 8, 9]. However, these studies have been based on samples of the population, examined the impact of COVID-19 in specific population subgroups, or have only explored specific waves of the pandemic. There is a need for a more complete assessment of mortality risk based on a national population.

We quantify the excess risk of death following confirmed SARS-CoV-2 infection by age and sex using a national population cohort of nearly 63 million individuals in England registered with the NHS for whom we have linked data from primary care records, hospital admissions, SARS-CoV-2 infection test results, vaccination status, and mortality. The risk of death after infection could also be estimated according to socio-demographic characteristics, presence of other morbidities, and vaccine status during successive waves of COVID-19.

One reason for wanting to document mortality risk during the pandemic more precisely and with more precision is because those who died from COVID-19 will not contribute to subsequent mortality rates. To estimate the extent to which the pandemic will influence future mortality trends through mortality displacement, it is necessary to understand in detail the pre-existing risk of the affected population groups and how it increased following COVID-19 [10]. This work is necessary to estimate likely future mortality rates following the pandemic [11].

Methods

Results

The registered population alive in England on 1 March 2020 included 62,374,261 individuals. We present detailed results for an exemplar age group–those aged between 65 and 69 –and summarise results across the broader age groups. Detailed accounts of the descriptive characteristics, the baseline hazard, and the hazard ratios derived from the models across all age groups are presented in S1–S3 Figs and S1, S2 Tables. The primary focus is on the risk of death associated with COVID-19; however, a commentary on the impact of sociodemographic and health characteristics on risk of death can be found in S3 Appendix.

Among the 65–69 age group, the study population included 1,274,145 females who did not have a positive SARS-CoV-2 test and 197,251 females who had a positive test (Table 1). There were 1,213,395 males who did not have a positive SARS-CoV-2 and 199,829 who tested positive. Vaccine uptake was high amongst this age group with approximately 85–89% having two vaccinations and a booster. Considering mortality across the entire period of follow-up, amongst those who never recorded a positive test for SARS-CoV-2, there were a higher proportion of deaths amongst males (3.8%) than females (2.5%). There were more deaths amongst those who tested positive for SARS-CoV-2, and reflecting the same gender pattern (5.3% and 3.5% for males and females, respectively).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Vaccination and survival characteristics by gender and positive SARS-CoV-2 test status for those aged 65–69 (as an exemplar group).

https://doi.org/10.1371/journal.pone.0304110.t001

Risk of 1st positive SARS-CoV-2 test for ages 65–69

For both sexes in our exemplar age group, the first four weeks following a first positive SARS-CoV-2 test represented the greatest increase in risk among all modelled risk factors (Fig 1 and Table 2). The relative risk of mortality associated with a positive test then decreased with time after infection (Fig 1 and Table 2); for unvaccinated females during wave 1, the relative risk was 77.62 in weeks 0 to 4 and 6.65 in weeks 5 to 26. The relative risks also decreased with increased numbers of vaccinations and in subsequent waves of the pandemic. For example, during wave three the immediate relative risk in weeks 0 to 4 among unvaccinated females was 41.87 (95% CI = 36.48 to 48.05), amongst those who had received one vaccine the hazard was 22.03 (95% CI = 16.08 to 30.17), with two vaccines it was 10.61 (95% CI = 9.62 to 11.70) and with a 3rd vaccine it was 3.41 (95% CI = 2.99 to 3.89).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Hazard ratios (and 95% confidence intervals) for the effect of testing positive for COVID-19 on mortality by time-period of infection and vaccination status.

https://doi.org/10.1371/journal.pone.0304110.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Hazard ratios (and 95% confidence intervals) derived from the Cox Proportional Hazard model denoting the hazard of deaths in those aged 65 to 69 stratified by sex.

https://doi.org/10.1371/journal.pone.0304110.t002

Risk of 1st positive SARS-CoV-2 across age groups

Patient’s age affected the relative risk associated with a positive SARS-CoV-2 test in different ways in successive waves of the pandemic (Figs 2 and 3). In wave one, relative risk generally declined with increasing age. However, in wave two, it generally increased slightly with age, up to age 70–74, and then gradually decreased with increasing age. In wave three, the relative risk generally tended to increase up to age 75–79 and then decreased with increasing age. One exception to this general pattern in wave 3 was in the period 5 to 26 weeks after a positive SARS-CoV-2 test for those with 3 vaccinations. In this situation–although the relative risks in each age group were much lower than those in the period 0 to 4 weeks and for those patients with fewer vaccinations–relative risks generally increased with age to 85–89. Overall, the 95% confidence intervals appeared to become narrower with increasing age.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Hazard ratios (and 95% confidence intervals) for wave, COVID-19, and vaccination status terms derived from the Cox Proportional Hazard model denoting the hazard of death amongst females, stratified by age.

Hazard ratios for all terms in the multivariable model are provided in the appendices. HR = Hazard ratio. The reference group for a positive COVID-19 test by wave and vaccination status are all those who do not fall into the defined category.

https://doi.org/10.1371/journal.pone.0304110.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Hazard ratios (and 95% confidence intervals) for wave, COVID-19, and vaccination status terms derived from the Cox Proportional Hazard model denoting the hazard of death amongst males, stratified by age.

Hazard ratios for all terms in the multivariable model are provided in the appendices. HR = Hazard ratio. The reference group for a positive COVID-19 test by wave and vaccination status are all those who do not fall into the defined category.

https://doi.org/10.1371/journal.pone.0304110.g003

Discussion

Our model estimates the relative risk of mortality following a first positive SARS-CoV-2 test by age and sex during the first, second, and third waves of the pandemic in England based on what we understand to be the most complete dataset of individuals in England. The principal finding of this analysis was a reduction in the risk of mortality attributable to COVID-19 over subsequent waves of the pandemic, highlighting the effectiveness of vaccination on mortality in each wave. Our model differentiates the effect of vaccination from the combined effect of waves on other factors, such as SARS-CoV-2 variants, improvements in treatment, and access to care. The model shows the independent effects of SARS-CoV-2 in the presence of a comprehensive list of individual-level characteristics and comorbidities. Our findings provide an immediate and definitive contribution to the literature on COVID-19 risk, derived from an individual-level analysis of the entire English population aged 45 and over during the pandemic. This provides key evidence into the subgroups most vulnerable to mortality following infection, thereby informing future targeted approaches to protecting vulnerable groups from infection, and the clinical management of those who do become infected [3, 19].

There has been a vast amount of research surrounding the COVID-19 pandemic nationally and internationally. Previous studies have included, but are not limited to, examining the impact of COVID-19 in terms of mortality and/or severe outcomes, risk factors associated with outcomes, and effects across time and geographies [2, 4–6, 20–26]. Numerous methodological approaches have been explored in this previous literature, which varied in the study population, how SARS-CoV-2 exposure was measured, and how it was identified. The present study benefits from linking the national SARS-CoV-2 testing data with both death and health records for the English population. In the present study we saw the risk of death from COVID-19 decrease across successive waves in the pandemic. This is consistent with evidence showing reduction in COVID-19 death rates over time in the general population [27]. Reasons for reductions in risk likely include increased availability of treatment, different virus variants, the introduction and rollout of vaccinations, and undetected or unrecorded prior infections. For example, variants that were generally more prevalent in later stages of the pandemic (such as Omicron) were associated with less risk compared with variants that were more prevalent in earlier stages (Delta) [28, 29]. While we did not model vaccine effectiveness per se, we also observed stark reductions in risk of death following vaccination where a SARS-CoV-2 positive test was present, which decreased incrementally with each additional vaccination. This is in line with previous meta-analytic evidence suggesting that vaccination was generally effective against symptomatic COVID-19 and/or serious outcomes such as hospitalisation or death after [30–32], with some findings suggesting a dose-response relationship with vaccination [33, 34]. However, there is likely some waning of vaccine effects over time [31, 34], which we did not consider in our analysis.

Understanding the variation in mortality rates between specific population groups not only guides clinical management and public health strategy during any pandemic, but is also desirable in the aftermath to understand better how expected numbers of deaths have been affected by those with a pre-existing raised level of risk of mortality dying earlier than would otherwise have been the case had the pandemic not occurred [35]. One key example of the latter is the public health need to understand the impact of disruption to a health service following a pandemic. Near real-time monitoring of the impact of such events enables timely response, e.g. allocation of limited resources to groups with the highest need [7]. However, the mortality displacement following a pandemic compromises the accuracy of contemporary methods for real-time monitoring of mortality within the population. The estimates we present in this paper represent the first step in our development of models to make appropriate adjustments for mortality displacement.

Our study population is large and represents the most complete national population dataset available. We have linked a wide range of relevant variables to this population cohort with a high degree of accuracy. The consistency of the modelled estimates (and baseline hazard) for different age-sex groups, which were modelled independently, demonstrates the robustness of the modelling approach in this national population dataset. Our analysis facilitates future work, such as calculating mortality displacement at an individual level, by allowing the point estimates presented in the present paper to be used to inform methods which account for the excess deaths caused by COVID-19 infection during the pandemic. This can ultimately be used to generate more accurate estimates of numbers of deaths that would otherwise have been expected each week, both during a pandemic and subsequently, and therefore improve the monitoring of trends in death rates.

This study has several limitations. Individuals’ comorbidities were defined as at 1^st March 2020 and we do not account for individuals who develop comorbidities after this date. Overall, we are likely to overestimate the population as death registration data are subject to a reporting lag and people who are no longer resident in England still being counted in the population database (NIMS) [36]. We only account for the first positive test for SARS-CoV-2 and account for neither the subsequent reduction in risk following infection nor the effects of subsequent infections. Conversely, we treat those who were SARS-CoV-2 positive but did not have a positive test recorded on the NHS website as being free of SARS-CoV-2 and hence included them in the baseline risk. The impact of this is most notable in periods when there was no comprehensive community testing, in particular at the start wave one and later during wave three, after cessation of mandatory testing. During these periods the majority of testing was confined to those being admitted to hospital and hence in the later stages of illness–shortening the survival time from testing to death. What our model also does not show is the interaction of the effect of SARS-CoV-2 with other risk factors. We were unable to model the impact of testing positive for SARS-CoV-2 on mortality for those aged under 45 years due to the sparse number of deaths at this age; nonetheless, the low numbers of deaths in the under 45 age groups mean that the effect of these deaths on future overall mortality rates will be very small. Finally, we did not model the time-dependent relative risk associated with testing positive for SARS-CoV-2 and mortality beyond 26 weeks following infection. Previous research and our results (Figs 2 and 3) suggest that these associations may extend beyond 26 weeks [37], implying longer-term impacts of SARS-CoV-2 infection however these relationships require further investigation and must not be assumed to be causal.

In conclusion, this study provides robust estimates of increased mortality risk among those who did test positive over the first three waves of the COVID-19 pandemic in England. The estimates not only add to the growing understanding of the magnitude of the impact of factors affecting the risk of mortality from COVID-19, but they also provide the first step towards estimating mortality displacement, a measure that is essential to understanding one of the major impacts of the pandemic on subsequent mortality rates in England.

Supporting information

Acknowledgments

Produced by the Office for Health Improvement and Disparities using data provided by, and re-used with the permission of:

1. NHS Digital–Hospital Episode Statistics (HES) & General Practice Extraction Service (GPES). Copyright 2023, all rights reserved.

2. NHS England–National Immunisation Management Service (NIMS). Copyright 2023, all rights reserved.

3. UK Health Security Agency–Second Generation Surveillance System (SGSS) COVID-19 testing data. Copyright 2023, all rights reserved.

4. Office for National Statistics (ONS)–Death registrations. Copyright 2023, all rights reserved.

The views and opinions expressed herein solely represent those of the authors, and do not necessarily reflect those of the Office for Health Improvement and Disparities, or any of the above-mentioned organisations.