NewsletterObesityPublic Health

COVID-19 Risk factors

Executive summary

Please login to view this content


On March 17th, 2020, when I first searched the academic database, PubMed, for “COVID-19” and “humans”, there were 99 results (Ref 1). The same search today (18th April) generates 1,101 results. As I reported in the note on tests, COVID-19 papers are being published while going through peer review, as it is deemed so important to share knowledge as soon as possible on this evolving topic (Ref 2). Papers need to pass editorial consideration before peer review, so an element of inspection is still taking place.

Among these papers, we have the first comprehensive review of the literature. This paper set out to find and review the most current, accredited, studies pertaining to the basic sciences of SARS-CoV-2 (COVID-19). If you are looking for an academic overview, this is a good place to start, although some aspects of the paper will date more quickly than others (Ref 3).

There has also been what I believe to be the first meta-analysis, published on 8th April 2020 (Ref 4). This meta-analysis asks the important question – Does comorbidity increase the risk of patients with COVID-19? This is the topic being reviewed this week. We hear that people with underlying conditions are at greater risk with COVID-19, but what exactly does this mean? Greater risk of being hospitalized? Becoming critical? And which underlying conditions are of more concern than others? Might there be some underlying conditions that are less of a concern? And perhaps others that should be driving protection advice more carefully?

Evidence available

The evidence on COVID-19 is emerging in pockets and it is country/location driven, as observations in a particular region are examined, documented, and then shared more widely.

The earliest, decent-sized, summary of characteristics of patients with COVID-19 inevitably came from China (Ref 5). It reported on 1,099 patients with lab-confirmed COVID-19, from 552 hospitals in 30 provinces and other discrete regions in mainland China until the end of January 2020. The patients were on average (median) 47 years old and 58% were male. Only 5% of patients were admitted to Intensive Care Units (ICU); 2.3% underwent invasive mechanical ventilation and 1.4% died. One of the most interesting findings of this paper was that white blood cell count was abnormally low in 83% of the patients on admission.

There were 173 cases judged to be severe on admission. The severe cases were more than twice as likely to be in the 65 or over age group. The severe cases were more likely to be former or current smokers. The severe cases were nearly six times more likely to have chronic obstructive pulmonary disease (COPD), almost three times more likely to have diabetes, almost twice more likely to have hypertension, just over three times more likely to have coronary heart disease and twice as likely to have cancer (the latter based on very small numbers).

Data from Italy came next. A much referenced paper published in JAMA on March 23rd, 2020 examined the characteristics of people who died in Italy (Ref 6). This paper noted that Italy attributed death to COVID-19 if someone had tested positive for the virus “independently from pre-existing diseases that may have caused death.” This captures people dying with COVID-19, although not necessarily from COVID-19. The ‘but for’ may well have been an existing condition.

This paper is where the much cited Italian numbers came from – the average (mean) age of patients who died was 79.5. Of those who died, 70% were men, 30% had heart disease, 36% had diabetes, 20% had cancer and other conditions were represented in far fewer cases. The average (mean) number of pre-existing conditions was 2.7. Only 3 out of 355 patients had no existing diseases.

A CDC report, published on March 31st, reported details of 7,162 patients, from different US states, for whom underlying health conditions were known. People were more than twice as likely to be hospitalized with one or more pre-existing condition. The conditions most likely to be associated with hospitalisation were diabetes, COPD, and heart disease (Ref 7).

The meta-analysis

The meta-analysis is worth a review because of the quality of evidence it provides. The Wang et al paper was received by the journal on March 12th and so only data from the earliest studies could be included and these were studies undertaken in China. These data are only applicable to China therefore and there have been a number of concerns about the reliability of data generally from China.

The meta-analysis question was “Does comorbidity increase the risk of patients with COVID-19?” and the answer was: Yes – having hypertension, diabetes, COPD and/or heart disease increased the risk of a more severe outcome among all people diagnosed with COVID-19. A total of 1,558 patients, from 6 studies, were included in the meta-analysis. One of the studies was the Chinese study with 1,099 patients reported above.

It should be noted that 80% of patients were not considered to have a severe response to the virus. This was defined as either needing ICU care or having severe symptoms, in the view of the admissions medic, when the patient arrived at their medical centre.

The review that prioritises risk factors

The paper that most caught my eye over the past week was one from NY City (Ref 8). It is going through peer review, while being available on pre-view. It examined all patients with lab-confirmed COVID-19, who were treated at an academic health centre in NY City between March 1st and April 2nd, 2020. The outcomes of interest were hospitalization and critical illness. Critical illness was defined as ICU needed, ventilation needed, hospice admission and/or death.

This paper stood out from so many others because it conducted a statistical technique (called multivariable logistic regression) to not only identify, but to prioritize, risk factors. This technique is used when a number of different factors can affect the outcome of interest and we want to know which are the most important factors. In this case, we want to know what most affects someone with COVID-19 being hospitalized (or worse).

Data were available for 4,103 COVID-19 patients. Among these, 1,999 (49%) were hospitalized, of whom 981/1,999 (49%) have been discharged, and 292/1,999 (15%) have been admitted to a hospice or died. Of 445 patients requiring mechanical ventilation, 162/445 (36%) died.

Table 1 in the paper reported many characteristics of the patients by hospitalization status. This enabled us to see, for example, that 63% of the people hospitalized were male. The usual pattern with existing conditions was seen – patients with hypertension, COPD, heart disease and diabetes were often many times more likely to be hospitalized than patients without these conditions.

One of the most striking findings was that there was no difference in hospitalization rates for current smokers. Both 5% of those hospitalized and 5% of those not hospitalized were current smokers. This was also interesting because more than 5% of people in NY city smoke. A 2016 estimate put the proportion of smokers at 11.5% and thus smokers might be under-represented among people attending a medical centre with subsequent lab-confirmed COVID-19 (Ref 9). The science on smoking and COVID-19 is by no means certain.

This paper also attracted my attention because it reported BMI data. Table 1 showed that having a BMI between 30-40 kg/m², or a BMI over 40, was approximately three times more likely to result in hospitalization. This is where I started digging deeper into the numbers. Table 2 repeated the same characteristics for the outcome of those hospitalized. It further analysed the 1,999 people who were hospitalized: 417 were still in hospital but not critical at the time of writing the paper; 932 were not critical and were discharged and 650 were critical (ICU, ventilation, hospice, or death).

Of the people admitted to hospital, there was no difference between those discharged and those who followed a critical pathway in terms of obesity. Approximately one third of people discharged and approximately one third of people who became critical were in the BMI range 30-40. Approximately 7% of people discharged and approximately 7% of people who became critical were in the BMI range >40. However, the characteristics table is raw data in effect. The prioritization statistical technique is effectively removing confounders. It highlights the most important risk factors. When this was done, BMI re-emerged as a concern – one of the major concerns.

The results

The technique to find the most important characteristics (multivariable logistic regression) was undertaken. The results were reported in the abstract (summary) of the paper: the strongest hospitalization risks were age ≥75 years; age 65-74; BMI >40 and having heart failure as a pre-existing condition. Let’s look at these:

Compared with the 19-44 year old age group, people in the age group ≥75 years were 67 times more likely to be hospitalized. That is an extraordinary number – the likes of which is rarely seen as an odds ratio (Ref 10). Being aged 65-74 had a hospitalization odds ratio of 11, relative to the 19-44 year old age group. That is also extremely high. However, the observation with the BMI data in the characteristics table could be seen again. While Table 3 reported the chance (odds ratio) of being hospitalized, Table 4 reported the chance of those hospitalized people becoming critical or being discharged.

People in the age group ≥75 years were ‘only’ 2.6 times as likely to become critical than those in the 19-44 year old age group. There’s a big difference between 67 and 2.6 times more likely. Plus – the reference group was 19-44 year olds. It was of concern to see that the 0-18 year old age group had 6.3 times the likelihood of becoming critical relative to those in the 19-44 year old age group. Hence, in this study in NY City, children were of more concern than the over 75s – although the numbers of children admitted were very small and this makes the results less reliable.

Those aged 65-74 were almost twice as likely as the reference age group (19-44) to become critical once hospitalized.

BMI was the next number of interest. Compared with those with a BMI under 30, those with a BMI over 40 were 6 times more likely to be hospitalized. Of those hospitalized, those with a BMI over 40 were 1.7 times more likely to become critical. That’s still a risk factor, but it’s not the six times difference reported in the paper summary.

Heart failure:
The final key risk factor highlighted in the summary was having previous heart failure. People with previous heart failure were reported as having just over four times the risk of being hospitalized as people without this condition. When it came to becoming critical or not, heart failure ceased to be a significant factor. Previous heart failure made no (statistically) significant difference among those hospitalized who became critical.

Asian ethnicity:
I spotted one result that went the other way. It wasn’t mentioned in the narrative of the paper. It didn’t stand out in terms of who was hospitalized and who wasn’t, but it did stand out in terms of those who became critical. At the first cut, 8% of people not hospitalized were of Asian ethnicity and 6% of those hospitalized were. So, people of Asian ethnicity ‘fared better’ at first examination. However, of those hospitalized, people of Asian ethnicity were over-represented in terms of those who became critical. Using the statistical tool to prioritise factors, those of Asian ethnicity were almost twice as likely as white people to become critical. Alongside was another interesting observation – African American people were almost half as likely as white people to become critical. The UK has just announced a review into the observed association between ethnicity and more serious COVID-19 outcomes (Ref 11). The results are much needed.

Reflections on the data

From available data, age does seem to be a common risk factor – in China, Italy and the US. However, age isn’t the 67 times risk factor that the NY City paper suggested. The difference between 67 and 2.6 suggests that the elderly are vastly more likely to be hospitalized but only 2-3 times more likely to have a critical outcome. That’s still a concern, but we shouldn’t over-report numbers on any topic at any time – perhaps especially with this virus and at this time. Plus, might we be over-hospitalizing elderly people because of a (completely understandable) value judgement about their risk? Hospitals are risky places right now – is that a good idea?

Obesity didn’t emerge as a risk factor in the early Chinese literature. Was this captured indirectly in (type 2) diabetes? Similarly, in Italy perhaps with 36% of those succumbing to the virus having diabetes (the vast majority of which will have been type 2)? Obesity does appear to be a risk factor. The NY City data suggest that obesity is a risk factor independent of diabetes, because diabetes was not a significant risk factor for critical illness in the NY study when judged against other factors.

Reassuringly for those with cancer, cancer was below other conditions in Chinese and Italian data as a risk factor. Cancer did not turn out to be significant for either hospitalization or becoming critical in the NY study.

Smoking was reported as a factor in China and high rates of smoking in Italy have been offered as an explanation for high deaths rates in Italian men. Smoking has been suggested as a reason for men being diagnosed more and suffering more serious outcomes than women. In the NY study, men were more likely to be hospitalized but no more likely than women to become critical and current or former smokers had a lower incidence of hospitalization and no difference in becoming critical relative to those who have never smoked.

There’s still a lot that we don’t know. The goal of understanding risk factors is twofold – to understand more about the pathology of the virus (to inform possible treatment/immunity) and to know whom most to protect. The latter was the aim of the NY City study – there are many risk factors, but which are the most important? Whom do we most need to protect? If these NY City data are reliable and extrapolatable to other similar populations (US, Canada, UK, Australia, and NZ most likely) then we might have sent high risk letters to the wrong people.

Any connection between viral load (how much of the virus someone is exposed to) and severity of illness is not yet known. The number of healthcare workers succumbing to the virus suggests that there may be a connection. The first 10 doctors who died in the UK were of black and Asian ethnicity (Ref 12). As the investigation into the ethnic risk factor starts, should obesity also be taken more seriously? Is now the time to protect anyone with a BMI over 40? Should we have been doing this from the outset? We have heroes on the front line with characteristics emerging as risk factors – some more than one – should we be helping them rather than the other way round right now?



Ref 1:
Ref 2:
Ref 3 Kadodkar et al. “A Comprehensive Literature Review on the Clinical Presentation, and Management of the Pandemic Coronavirus Disease 2019 (COVID-19).” Cureus. 2020 Apr 12th.
Ref 4: Wang et al. Does comorbidity increase the risk of patients with COVID-19: evidence from meta-analysis. Aging (Albany NY). 2020 Apr 8.
Ref 5: Guan et al. Clinical Characteristics of Coronavirus Disease 2019 in China. NEMJ. 28th February 2020.
Ref 6: Onder et al. Case-Fatality Rate and Characteristics of Patients Dying in Relation to COVID-19 in Italy. JAMA. March 23rd 2020.
Ref 7: Team CC-R. Preliminary Estimates of the Prevalence of Selected Underlying Health Conditions Among Patients with Coronavirus Disease 2019 – United States, February 12 – March 28, 2020. MMWR Morb Mortal Wkly Rep. 2020.
Ref 8: Petrilli et al. Factors associated with hospitalization and critical illness among 4,103 patients with Covid-19 disease in New York City. In pre-print. April 11th, 2020.
Ref 9:
Ref 10: The full odds ratios, with 95% confidence intervals were: age ≥75 years (OR 66.8, 95% CI, 44.7-102.6), age 65-74 (OR 10.9, 95% CI, 8.35-14.34), BMI>40 (OR 6.2, 95% CI, 4.2-9.3), and heart failure (OR 4.3 95% CI, 1.9-11.2).
Ref 11:
Ref 12:

10 thoughts on “COVID-19 Risk factors

  • The Times today, page 9, subhead “fat could be helping virus to infiltrate patients’ cells”. The results don’t look particularly significant to me, nor do I have info to ascertain the independence from age. I am concerned that age (I am 72) comes along in maybe 75% of people with being plump, narrowing of the arteries, and high blood pressure (whatever that is).

    Dr Kendrick demonstrates the goal post reduction of hypertension in Doctoring the Data. I suspect blurring of actual figures is intended to make us comply by scaring us. I have trawled the internet but can’t find out how the bp figures of ICU patients are derived. Being wheeled into intensive care surely rockets one’s bp, or is it from people on bp lowering drugs (unnecessarily?). Where does bp start? If its 140 That’s big pharma rubbish.

    • Hi Doctorjen
      So sorry I missed this post to approve and respond to…

      As you can see in this post – the most significant risks from the NY data were for those with a BMI over 40 – that’s not “plump” and I hope that’s reassuring. Hypertension hasn’t been jumping out as much as some other factors. And I’m with you – 140 is quite normal – as this post from way back shared

      The age one I need to revisit, as I heard a Cambridge statistician (whom I have much respect for) saying that the age risk with COVID is no different to the age risk generally. i.e. we know that 80 year olds are far more likely to die over the next year than 30 year olds, but the risk of dying from COVID as we age is apparently no worse than the risk of dying anyway as we age. I need to find time to look at that!

      Best wishes – Zoe

  • Hi Zoë. As of today there are now 3 studies suggesting a strong correlation between vitamin d status and outcome with COVID 19. There has also been speculation for several weeks that this might be the case with plausible mechanisms for causation eg
    Are you going to have a look at this?

  • I agree with sburkeen in that there is a world of difference between different hospitals and even different healthcare systems. I have friends in many countries who work for the health care system and many different treatments have been tried. Some of these treatments are reducing morbidity to almost nothing. The NHS has no flexibility in its approach and is putting people at risk because of it. Doctors cannot respond quickly with treatments which show promise. I suggested one combination early on and it has been taken up in some places but, as information has come back, so the original idea has been modified. We need flexibility not rigidity.

  • One thing that does not seem to be accounted for here is the standard of care. When do you use drugs, e.g. anti-malarial drugs, what dosage, the use of ventilators and how they are used? I have read that ventilators may actually cause lung damage, depending on the settings, and may or may not be necessary. After admission, are you sufficiently isolated so that you are not continuously reinfected to increase the viral load? What was done after admission that actually decreased the ability of your immune system to cope? Mortality outcomes very much depend on the skill of the doctors involved, in addition to their resources. In short you have all kinds of confounding variables. Multi-variate analysis is not going to sort this out if you do not know what the variables are.

    My own unscientific opinion is that you stand a better chance going into this if you are metabolically healthy and this is independent of age, sex, and ethnicity. One thing I am certain of is that there will be no great government funded strides toward population metabolic health resulting from this pandemic for the simple reason that it does not depend on drugs. Huge sums will be spent on vaccine research, however, for the simple reason that drug companies see a guaranteed, perhaps mandatory, world wide market. Drug development for chronic disease has pretty much exhausted itself with dismal results, for people as well as drug company bottom lines, so vaccines will be pushed in their place. This pandemic may be a disaster for public health and the general economy, but it won’t be for many drug companies.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.