key: cord-0928516-v4vk5h0z
authors: Philipose, Z.; Smati, N.; Wong, C. S. J.; Aspey, K.; Mendall, M. A.
title: Obesity, old age and frailty are the true risk factors for COVID-19 mortality and not chronic disease or ethnicity in Croydon.
date: 2020-08-14
journal: nan
DOI: 10.1101/2020.08.12.20156257
sha: 95989eee742fc254193816070b5a4a08ba46372f
doc_id: 928516
cord_uid: v4vk5h0z

Background: Coronavirus-19 (COVID-19) mortality in hospitalised patients is strongly associated with old age, nursing home residence, male sex and obesity, with a more controversial association with ethnicity and chronic diseases, in particular diabetes mellitus. Further complicating the evaluation of the independent impacts of these risk factors is the failure to control for frailty in the published studies thus far. Aim: To determine the true risk factors for mortality in patients confirmed to have COVID-19 in Croydon needing hospital admission and to evaluate the independence of these risk factors in this group after adjusting for body mass index (BMI) and frailty. Methods: This observational study retrospectively reviewed hospital electronic medical records of 466 consecutive patients who were admitted to Croydon University Hospital confirmed positive by rapid PCR test from 11th March 2020 to 9th April 2020. Statistical analysis was performed by multiple unconditional and univariate logistic regression. Results: After multivariate analysis, male sex [OR 1.44 (CI 0.92-2.40)], age (per year) [OR 1.07 (CI 1.05-1.09)], morbid obesity (BMI > 40 kg/m2 vs reference BMI 18.5-24.9 kg/m2 ) [OR 14.8 (CI 5.25-41.8)], and nursing home residence (OR 3.01 (CI 1.56-5.79) were independently associated with COVID-19 mortality with no statistically significant association found with chronic diseases or ethnicity. In the non-nursing home population, after adjusting for age and sex, the odds ratio for type 2 diabetes mellitus (T2DM) as a risk factor was 1.64 (CI 1.03-2.61, p = 0.03) and was and was attenuated to 1.30 (CI 0.78-2.18)) after controlling for BMI; the association of mortality with male sex was strengthened [OR 1.66 (CI 0.96-2.87)] and that for ethnic minority patients was weakened [South Asians [from OR 1.30 (CI 0.67-2.53)) to OR 1.21 (CI 0.60-2.46)]; African Caribbeans [from OR 1.24 (CI 0.65-2.34) to OR 1.16 (CI 0.58-2.30)]. There was a borderline but potentially large protective effect (p= 0.09) in patients who were on anticoagulation drugs prior to admission [OR 0.56 (CI 0.28-1.11)]. Conclusion: Our study found no significant effect of ethnicity and chronic diseases as independent risk factors on COVID-19 mortality in Croydon population whereas male sex, high BMI, old age and frailty were found to be independent risk factors. Routine prophylactic treatment with anticoagulant drugs in the high-risk COVID-19 population warrants further prompt investigation.

The Coronavirus-19 (COVID-19) pandemic of severe acute respiratory syndrome (SARS-CoV-2) has led to more than 580,000 deaths worldwide, as of July 2020. 1 Increasing evidence reveals significantly higher hospital admission and death rates in males, the elderly, nursing home residents, ethnic minorities, the socially deprived, and patients with underlying diabetes, hypertension, cardiovascular disease, and obesity. 2 However, the independence of each risk factor remains controversial.

Two large scale studies in the United Kingdom (UK), OpenSAFELY studies and the ISARIC CCP-UK, evaluated the characteristics and disease outcomes of hospitalised patients confirmed with COVID-19 infection. They reviewed 17 million general practice patient records 3 and over 34,000 patient hospital records, 4 respectively. Although both studies were large in scale, their methodology limited the characterisation of the true effect of each risk factor, particularly the impact of obesity and frailty on mortality. The OpenSAFELY study employed information on body mass index (BMI) from general practice records up to 10 years before the COVID-19 infection 3 and in the ISARIC CCP-UK study, the BMI was not objectively measured. 4 In terms of frailty, although both studies included social deprivation indices related to locality, neither was able to explore the effects of individual frailty or the effect of nursing home residency. This could be a vital missing element since frailty has been suggested to be an independent risk factor for death from COVID-19 infection. 5 Angiotensin-converting enzyme (ACE) inhibitors were initially proposed to have an influential effect on COVID-19 diagnosis and mortality but were found to have no statistically significant association later on. 6 The proposed prothrombotic pathology of severe COVID-19 disease suggested that anticoagulant drugs may have a protective effect on disease outcome. 7 Our aim, therefore, was to determine the true risk factors of COVID-19 mortality by studying a broad range of potential risk factors (both medical and social) in a consecutive case series of COVID-19 patient admissions to Croydon University Hospital, a district general hospital in the London borough of Croydon in the United Kingdom. In particular, we analysed the impact of ethnicity, contemporaneous BMI, frailty, nursing home residence, pre-existing chronic diseases, co-existing medication history and smoking history on inpatient mortality.

This single-centre retrospective observational cohort study included all patients admitted to Croydon University Hospital from 11 th March to 9 th April 2020 and had confirmed COVID-19 infection by Polymerase Chain Reaction (PCR) nasopharyngeal antigen swab test. Data collection ceased on 1 st June 2020 by which time all patients had an outcome as either deceased or discharged. These patients had respiratory symptoms and signs of COVID-19 infection on admission as criteria for testing during this period. The patients' hospital information on Cerner electronic medical records were reviewed by a team of physicians who then collected their epidemiological data, past medical history, medication history, clinical and laboratory data and outcomes from their admission to either their discharge from hospital or death. An anonymous identification code was assigned to each patient record to protect patient confidentiality. All electronic data was stored in and analysed on hospital servers. Ethical approval was obtained from the Croydon University Hospital

Research and Development committee prior to data collection.

The primary outcome was to examine the impact of obesity, ethnicity and nursing home residency on COVID-19 infection mortality in this cohort of patients. The secondary outcome was to assess the significance of different risk factors of clinical characteristics and medication history on mortality.

COVID-19 symptom severity was categorized based on patients' oxygen requirements and complications throughout admission. Patients were classified in the mild symptom category if their room-air oxygen saturation levels were maintained at or above 93%; moderate symptom category if they had a respiratory rate above or equal to 30 breaths per minute and/or room-air oxygen saturation below 93%; and severe symptom category if they developed respiratory failure and/or multi-organ failure.

BMI was ascertained from the height and weight recorded on the COVID-19 infection admission; if unavailable, the most recent recording from the last six months prior to the infection. Ethnic group categories were established from admission clerking and patient registration details. These were grouped as White/Caucasian, South Asian, Black/African Caribbean, East Asian, Middle Eastern/North African (MENA), and Undetermined. Patient's residence was categorised as unassisted living residence (such as their own home, family's home or rented accommodation) or an assisted living residence (such as care homes, nursing homes and sheltered accommodation). Dependence on activities of daily living was whether or not a patient required carers' help for daily activities. Occupation was categorised as manual, non-manual work or retired/none. Smoking history was categorised as current, former or none. History of alcohol excess was categorised as yes if consumption was above 14 units per week.

In terms of patients' medical background of pre-existing chronic diseases, we examined their current and previous respiratory, cardiac, liver, vascular, cerebrovascular and immunocompromised conditions. Respiratory history included any respiratory comorbidities such as obstructive or restrictive pulmonary disease and/or history of tuberculosis. Cardiac history included a history of ischaemic heart disease and/or arrhythmias.

Immunocompromised states included active cancer, human immunodeficiency virus (HIV) and/or on immunosuppressive medications (including high dose corticosteroids, biological agents and systemic chemotherapy medication). Arterial and venous vascular diseases included a history of hypertension, cerebrovascular events, venous clots including pulmonary embolism and deep venous thrombosis. We also examined if patients had type 2 diabetes mellitus (T2DM) or type 1 diabetes mellitus (T1DM).

Patients' drug history was based on their regular medication that was recorded on admission. The following medications were recorded: statin, metformin, antihypertensives, antiplatelets and anticoagulants. Statin doses were equalised into atorvastatin dose equivalents. The use of any antihypertensive drug was recorded but we specifically isolated the use of ACE-1 inhibitor drugs and angiotensin II receptor blocker (ARB) drugs for further analysis.

Laboratory markers serum alanine transaminase (ALT), vitamin D, albumin, C-reactive protein (CRP), creatinine kinase (CK), Haemoglobin (Hb), Troponin-T, Haemoglobin A1c (HbA1c) were analysed. Patients' ALT results were recorded both pre-admission and on admission. Patients' pre-admission vitamin D levels were recorded. HbA1c was only recorded if the measurement was taken less than six months prior to admission. All of the other markers were taken on admission. Liver ultrasound records were examined up to five years prior to admission looking for reports of cirrhosis or fatty liver disease.

We used Statview SE for statistical analysis. Univariate statistical analysis comparisons for death or discharged alive were performed using Chi-square test for two-level categorical variables of the following: sex, type of residence, dependency for activities of daily living; history of T1DM, T2DM, cardiac disease, stroke, thromboembolism, respiratory disease; the use of statins, metformin, antihypertensives, antiplatelet agents, anticoagulants; and ultrasound evidence of fatty liver or cirrhosis. African, and undetermined), disease severity during admission (mild, moderate, severe), smoking history (current, former, none), excess alcohol history (yes/no), and occupation (manual, non-manual, retired/none).

The Mann-Whitney U test was used to compare continuous explanatory variables that were not normally distributed: age, BMI, HbA1c, biomarkers on admission (CRP, ALT, Troponin-T, and CK) and pre-admission (Hb and Vitamin D).

Multivariate analysis was performed with unconditional logistic regression and the following variables were retained on the final model: age was taken as a continuous variable; sex, ethnic group, BMI were taken as categorical variables; the following were taken as binary variables (yes/no): nursing home residence, dependence for activities for daily living, history of T2DM, cardiac disease, respiratory disease, stroke, hypertension and the use of anticoagulants. P-values were derived from likelihood ratio tests.

We analysed 466 consecutive subjects (59% male and 41% female). 267 patients were discharged and 199 patients died.

The pre-admission characteristics of patients who were discharged or died before and after adjustment for age and sex of subjects are described in Table 1 . On univariate analysis, age and the morbidly obese BMI category (>40 kg/m 2 ) were very strongly associated with mortality risk, and this remained after adjustment for age and sex [OR 13.2 (CI 5.10-34.9), p<0.0001]. The intermediate associations with cardiac and hypertension history and antiplatelet agents disappeared after adjustment for age and sex, but treatment with anticoagulants emerged as a borderline protective factor. Of the indicators of frailty, dependence on activities of daily living and nursing home residence were very strongly associated with risk of mortality with 29% of all deaths in hospital related to COVID-19 being nursing home residents. However, the association with risk of mortality of dependence for activities of daily living disappeared after the adjustment. There was only a very weak non-significant association with ethnic groups apart from a higher mortality in those grouped as "undetermined."

The distribution of elderly (age 61 years and above) within each ethnic group was as follows: Table   2 . Elevated CRP and low albumin were associated with increased mortality outcomes.

Troponin was only available on 142 subjects but showed a strong association with mortality.

Unsurprisingly, mortality was much higher in those subjects with severe disease on admission compared to those with moderate or mild disease.

The multivariate analysis of pre-admission characteristics associated with COVID-19 mortality is shown in Table 3 

In our retrospective observational study of 466 consecutive patients hospitalised with confirmed COVID-19 disease, for a defined period from 11 th March 2020 to 9 th April 2020, the overall death rate was 24.5%. Strong associations were observed with BMI. Underweight subjects had the lowest risk and the morbidly obese had the highest risk. Age and nursing home residence were also strongly associated. The relationship with male sex was strengthened and of borderline significance after excluding nursing home residents. There two other larger studies from the UK. The OpenSAFELY study used general practice medical records and was able to study the risk associated with ethnic groups versus the general population. 3 It, however, could not control for frailty and relied on any BMI recorded within the past 10 years. This may explain why only modest but statistically significant associations were observed with BMI. Similar associations to the present study of only a very modest magnitude with mortality were observed with ethnic groups, and T2DM explained some of this. However, they were unable to control as adequately for BMI and frailty which explained much of the association of T2DM with mortality in the current study. The ISARIC CCP-UK study also from the UK was of similar design to ours, including subjects from several centres and hence larger. It, however, again could not adequately control for BMI, relying on observation as to whether simply the subjects were obese or not and again included no measures of frailty. Their patient population had a different ethnic composition (83% were White/Caucasian versus our study was 51% White/Caucasian) but similar very modest associations were seen with ethnic groups, partly explained by T2DM. 4 A study from New York of risk factors for admission with COVID19 and the development of critical illness identified obesity as a risk factor but its effect on mortality was not specifically addressed. 8 A smaller study from the same city again identified obesity as a risk factor for mortality in those with a BMI >35 kg/m 2 with BMI 25-34.9 kg/m 2 as the reference group. 10 In our study when redefining the reference range in this manner, we get a similar magnitude of effect for BMI > 35 kg/m 2 . These studies did not explore how frailty and being underweight impacts mortality.

We demonstrated only modest associations of ethnic groups with mortality which fell short of statistical significance yet were of a similar magnitude to those observed in the OpenSAFELY and ISARIC studies. No associations with ethnic groups were found in the studies from New York. 8, 10 Moreover, the overall COVID-19 mortality rate per 1 million population has not shown to be higher in countries where the ethnic minority population groups originate compared with the UK. 1 Although very difficult to prove, this may suggest there is no obvious genetic predisposition between ethnic groups for COVID-19 pathology susceptibility and that social factors such as socioeconomics, housing, and occupational exposure may be a better way to explain the worse mortality outcomes of Black, Asian, and minority ethnic (BAME) findings in the United Kingdom. 12 In the present study, the exclusion of nursing home residents weakened the association between ethnic groups and mortality. Similarly, the association between T2DM and mortality was also attenuated after controlling for BMI, which was previously postulated as part of the mechanism for the association of ethnic groups with mortality. 3, 4 We suspect therefore that residual confounding may be part of the explanation of the apparent association of ethnic groups with mortality of admitted patients. 20 There is a paucity of data on the impact of being a resident in a nursing home on other potential mediators of mortality in COVID-19 and we were able to study this in more detail. The nursing home population were less likely to have conventional risk factors including obesity and T2DM than age matched hospitalised patients. We therefore hypothesise that frailty per se is likely to be the driving factor for mortality in COVID-19 nursing home resident patients beyond obesity, diabetes and their associated comorbidities. On the other hand, patients who were dependent for their activities of daily living were not found to be at a higher risk for mortality if they were not residing in nursing homes. We also cannot preclude that levels of care were different for the elderly patients from nursing homes versus from non-nursing homes. Additionally, frailty could be acting through other mechanisms and may represent enhanced inflamm-aging. 21 Nonetheless, it is important to account for nursing home residence and frailty when studying the risk factors for COVID-19 mortality.

The role of male sex is a consistent predictor of mortality in all COVID-19 studies. 22 We demonstrated male sex as an independent risk factor for COVID-19 mortality in our analysis.

Research findings have proposed theories based on the role of the X chromosome in immune response activation, 23 the Angiotensin-converting enzyme-2 24 and testosterone modulation of Transmembrane Serine Protease 2. 25 Further work on both examining biological and social gendered differences is required to understand this difference in pathogenicity between male and females. Additionally, we cannot exclude that men are more prone to acquiring COVID-19 as male sex was over-represented amongst admissions compared to the local population. It may also be that if men are infected, they are more likely to have severe symptoms and present to hospital. This can be resolved by sero-epidemiology studies in the community.

The possible marked protective effect of anticoagulants, which our study was underpowered to detect conclusively, is in concordance with what we are learning about the pathogenesis of severe COVID-19 and the prominent role that thrombosis plays. 26 There were no other trends found for the use of antihypertensives or antiplatelet agents. There was also no significant effect of other chronic disease after adjustment for age and sex. It is therefore likely that associations with other chronic comorbidities are confounded by age and visceral adiposity.

One challenge of our study was missing patient data. We have done our best to include these cases in our analyses. The size of the study precluded us from exploring in more detail the effect of medication history on prognosis and detecting some effects of pre-existing chronic disease. Despite this, our study has identified the most important and clinically significant pre-morbid conditions influencing mortality to date. Our study captured data from the first 466 consecutive patients admitted to Croydon University Hospital and therefore the earlier time period of the pandemic in the UK. There have been significant changes in understanding and management of the condition since our data analysis. The fact that we captured data during the peak of patient admission burden for COVID-19 meant there were significant pressures on the Intensive Care Unit, affecting the selection criteria for level 3 care on a daily basis. This may have influenced some patients' outcomes.

Furthermore, during this time period, COVID-19 testing was in short supply and therefore, not every patient who was admitted was tested, particularly those with mild or no symptoms.

Obesity, or more specifically abdominal adiposity, as well as male sex, advancing age, and frailty are likely to be the key factors in determining the mortality outcome of COVID-19

infection. Our dataset pertains to the patients admitted to Croydon University Hospital.

Further studies are required to prove that other postulated risk factors such as ethnicity and chronic disease, including diabetes, are also independent risk factors. Prophylactic anticoagulation in high risk subjects where COVID-19 infection is widely prevalent could be life-saving and warrants further study.

Coronavirus Pandemic (COVID-19)

Disparities in the risk and outcomes from COVID-19. PHE publications

OpenSAFELY: factors associated with covid-19 related hospital death in the linked electronic health records of 17 million adult NHS patients

Ethnicity and outcomes from COVID-19: the ISARIC CCP-UK prospective observational cohort study of hospitalised patients. SSRN Preprints with the Lancet

Could nutritional and functional status serve as prognostic factors for COVID-19 in the elderly? Med Hypotheses

Association of Angiotensin-Converting Enzyme Inhibitor or Angiotensin Receptor Blocker Use With COVID-19 Diagnosis and Mortality

The Role of Anticoagulation in COVID-19-Induced Hypercoagulability

Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study

Morbid Obesity as an Independent Risk Factor for COVID-19 Mortality in Hospitalized Patients Younger than 50

Severe obesity, increasing age and male sex are independently associated with worse in-hospital outcomes, and higher in-hospital mortality

SSRN Preprints with the Lancet

BMI: preventing ill health and premature death in black, Asian and other minority ethnic groups

International Diabetes Federation: a consensus on Type 2 diabetes prevention

Type 2 diabetes in the UK South Asian population -An update from the South Asian Health Foundation. South Asian Health Foundation (SAHF)

Abdominal adiposity is the main determinant of the C-reactive response to injury in subjects undergoing inguinal hernia repair

Introduction to Frailty

Inflamm-aging. An evolutionary perspective on immunosenescence

Impact of sex and gender on COVID-19 outcomes in Europe

Coronavirus COV-19/SARS-CoV-2 affects women less than men: clinical response to viral infection

COVID-19 and Individual Genetic Susceptibility/Receptivity: Role of ACE1/ACE2 Genes, Immunity, Inflammation and Coagulation. Might the Double X-chromosome in Females Be Protective against SARS-CoV-2 Compared to the Single X-Chromosome in Males

TMPRSS2 and COVID-19: Serendipity or Opportunity for Intervention

Pulmonary Vascular Endothelialitis, Thrombosis, and Angiogenesis in Covid-19

We thank the Research and Development team and the Microbiology team for their support in this project.We also would like to express our gratitude to Dr Alok Mehta, Dr Georgios Karanasios, Dr Eugene Yap, Dr Joseph Hogan, Dr Homira Ayubi, and Dr Amy Woods for their contributions to this study.

We declare no competing interests.