Background

The novel coronavirus SARS-CoV-2 has led to a global pandemic of a complex, pneumonia-related illness (COVID-19). As of 29th September 2020, there have been 33,749,599 confirmed cases of COVID-19 and 1,009,829 deaths reported worldwide [1]. Many countries are still facing a devastating rise in the number of cases and deaths, whilst some are facing a second wave such as France, Spain, UK, with significant impacts on their healthcare systems. The risk of SARS-CoV-2 infection is based on the person-to-person transmission rate at which the virus is circulating in the community [2, 3]. Yet, while some individuals recover from the SARS-CoV-2 infection, others develop severe illness which may require intensive care in hospital or result in mortality.

Several potential factors have been postulated or associated with the increased risk of transmission and development of severe SARS-CoV-2 infection, such as age, sex, ethnicity, body mass index (BMI), smoking, deprivation, pre-existing comorbidities (particularly diabetes, hypertension and cardiovascular disease), cardiorespiratory fitness, 25‐hydroxyvitamin D level, air pollution and inflammation [4]. However, individuals with pre-existing health conditions have been found to have the highest risk of develo** severe SARS-CoV-2 infection, as shown in a number of studies from early research based in Wuhan [5] to recent studies across the globe [6,7,8]. Nevertheless, the majority of research to date has focused on pre-existing single chronic conditions [9], yet individuals often have multiple chronic conditions, known as multimorbidity. In view of the rapidly increasing prevalence of severe SARS-CoV-2 infection across the globe, clarifying the pattern and impact of multimorbidity (defined as 2 or more conditions) on the risk of severe SARS-CoV-2 infection has important clinical and public health implications which are not captured by a focus on single diseases [10].

There are a number of methods to define multimorbidity, which include both simple measures such as summing up the diseases and more complex measures using a weighted score such as the Charles Comorbidity Index. A previous study in UK Biobank examined the pattern of the most prevalent clusters of multimorbidity in those infected with SARS-CoV-2, using 36 different conditions to define multimorbidity [11], though this masked the most relevant conditions related to SARS-CoV-2 [12]. Moreover, three studies which assessed the risk of multimorbidity and COVID-19 had also used a broad range of conditions (ranging from 19 to 43 conditions) [13,14,15]. In contrast, the available evidence indicates that certain health conditions are associated with a much greater risk of severe SARS-CoV-2 infection linked with hospitalisation or mortality [5, 16].

To help clarify the evidence, in this study we aimed to firstly develop a multimorbidity index of the most commonly reported chronic conditions from the current literature for severe SARS-CoV-2 infection. The index was then used to investigate the patterns and association between multimorbidity and risk of severe SARS-CoV-2 infection and we further tested whether this association was modified by other potential risk factors.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Additional File 1: Checklist S1) and following a pre-specified protocol (Application Number 36371) [17].

Study population

UK Biobank included half a million middle-aged adults recruited from 22 sites across England, Wales and Scotland with baseline measures collected between 2006 and 2010, from an in-person baseline interview at the UK Biobank centre (http://biobank.ctsu.ox.ac.uk/crystal/search.cgi). Written informed consent was obtained prior to data collection; UK Biobank was approved by the NHS National Research Ethics Service (16/NW/0274; ethics approval for UK Biobank studies) [18].

SARS-CoV-2 laboratory confirmed test results and death data for all-cause mortality from Public Health England were linked to the UK Biobank database [19]. Data were available between March 16 and July 26, 2020, restricted to those tested in a hospital (pillar 1), since this can be regarded as a proxy for hospitalisations for severe cases of the disease as suggested by the linkage methodology [19]. Our analytical sample included study participants from England as testing was available and linked, we therefore excluded study members from Scotland and Wales, participants who had an outpatient test positive for COVID-19 (pillar 2) as the outcome for SARS-CoV-2 infection was uncertain (i.e. recovered, hospitalised at a later stage, or died), those who died before March 16, 2020, and those with missing data (Additional file 1: Fig. S1).

Multimorbidity index

To use prior evidence on comorbidities that cluster together and that may act in combination to affect the risk of COVID-19, we carried out a literature search to identify the most common pre-existing comorbidities in patients with severe SARS-CoV-2 infection (Additional file 1: Table S1). The studies from the literature search were found to be very heterogeneous, therefore it was not possible to do a quantitative meta-analysis of pooled results [5,6,7,8, 20,21,22,23,24,25,26,27,13]. However, this study used 43 different conditions to define multimorbidity, and compared only those testing positive compared to negative. The other UK Biobank study, between 16 March and 18 May, 2020 (1324 tested positive for COVID-19 and 426,875 remaining cohort), used Poisson regression analyses reporting that those with multimorbidity (defined using 43 conditions) had a 48% higher risk for a positive test for COVID-19 (RR 1.48 [95% CI 1.28, 1.71], P < 0.01). Although this study also reported that those with cardiometabolic multimorbidity, compared with those without, had a 77% higher risk for a positive test (RR 1.77 [1.46, 2.15] P < 0.01) [15], the outcome was based on testing positive in any setting (hospital or outpatient, pillar 1 and 2), which may introduce testing bias within UK Biobank. A recent article illustrated how the same dataset can produce apparently paradoxical findings for the same outcome [34]. For this reason, our study examined harder outcomes for those with severe SARS-CoV-2 infection; hospitalisation or death during the pandemic compared to the remaining cohort. Moreover, the three studies to date have only presented the number of conditions but have not examined the patterns of multimorbidity, i.e. the type of the two most common co-occurring conditions found in those with severe SARS-CoV-2 infection, nor the rate of severe infection or mortality in people with multimorbidity over the period of the pandemic.

Hypertension was found to be the most prevalent condition (40%) in those with severe SARS-CoV-2 infection, as in a number of previously published studies. Yet, when examined by clusters, we found that hypertension mainly co-existed with other conditions (i.e. diabetes, stroke, CKD). Thus, certain clusters of the multimorbidity index are more frequent than other clusters. When examining the association between the most common co-occurring multimorbidity clusters and the risk of severe SARS-CoV-2 infection, those with pre-existing CKD and diabetes showed the strongest association. For this reason, we believe that the classification of those with pre-existing multimorbidity index conditions is crucial to provide prognostic information to tailor effective treatment and prevent severe outcomes in common chronic conditions. Individuals with pre-existing multimorbidity conditions may have potential organ dysfunctions, which are further accelerated by the local and systemic effects related to SARS-CoV-2 infection. Previous studies have investigated the possible pathophysiological mechanisms linking this infection to severe COVID-19 syndrome, including increased inflammation, decreased immune response, heart failure, renal failure, and potentially multi-system organ failure and death [35, 36]. Further research is however required to understand the underlying biological mechanisms involved between these multimorbidity clusters and the increased risk of severe COVID-19 and its long-term sequelae (i.e., “long COVID”), particularly the effects on the immune system and frailty.

Multimorbidity has been associated with premature mortality, lower quality of life, polypharmacy, frailty, and fragmented care [10]. However, the current pandemic has placed additional risks on individuals with multimorbidity. A global survey evaluating the impact of COVID-19 on routine care for chronic diseases during 31 March to 23 April 2020 found that most healthcare professionals (67%) identified moderate or severe effects on their patients due to changes in healthcare services since the outbreak; and 80% reported the mental health of their patients worsened during COVID-19 [37]. Moreover, many low and middle-income countries have overstretched healthcare systems and the pandemic has further overwhelmed them.

A key strength of this study is the overall large sample size which allowed us to investigate whether the associations were heterogeneous across clinically relevant effect modifiers. We employed a comprehensive list of effect modifiers, extending past research [4], and enabling a greater understanding of the association between multimorbidity and severe SARS-CoV-2 infection. Moreover, we performed a literature review to derive an evidence based COVID-19 relevant multimorbidity index which can be applied to other populations [5,6,7,8, 20,21,22,23,24,25,26,27,28,29]. This study has some limitations. Issues of the low response rate (~ 5%) and selection bias, such as slightly higher representation of participants from affluent groups may suggest that the UK Biobank sample is not well representative of the UK population, and have been discussed previously [38]. The characteristics we examined were recorded at recruitment, representing data collected in the past at study baseline, but we conducted additional sensitivity analyses using the follow-up data for 25‐hydroxyvitamin D levels and the last recorded air pollution (NO2) data, showing consistent results over time. In further sensitivity analyses, we additionally adjusted the model for the time in the study, and found the results remained consistent. Factors such as cardiorespiratory fitness or C-reactive protein could possibly be a consequence of poor health from multimorbidity instead of cause, for this reason we removed these from the model, and the results remained the same. The data on multimorbidity were appraised only at baseline but there may have been a change in the number of multimorbidity index conditions since recruitment which were not accounted for and may lead to under or overestimation of effect. We were unable to assess the severity of disease conditions but our new multimorbidity index definition included conditions that were most relevant to COVID-19. Another limitation is the lack of detail of clinical severity of those with severe SARS-CoV-2 infection: while we have accounted for hospitalisation or death during the pandemic, further details of adverse COVID-19 outcomes such as length of hospital stay, intensive care admission, or mechanical ventilation, could provide further information. Ten potential effect modifiers were selected [4], although there are other factors which could possibly modify the association between multimorbidity and risk of severe SARS-CoV-2 infection but were unaccounted for, such as household size, occupation, and understanding and behaviour for social distancing, shielding, wearing facemasks and hand hygiene. Finally, this was an observational study, which limits our ability to establish causality.

Conclusion

In conclusion, having a new multimorbidity COVID-19 index that may help identify subjects at higher risk of complication in clinical practice could be beneficial. Where stratifying the risk according to the presence of multimorbidity, and reinforcing initiatives and resources to reduce the risk in this vulnerable group of individuals should be considered as priorities in the current coronavirus pandemic.