OpenSAFELY: Factors Associated With COVID-19 Death in 17 Million Patients

Epidemiology Transversal
Williamson EJ et al

Main result

Large English cohort representing 40% of the population (17,278,392 adults of more than 18 years old).

12,718,279 patients were included in the ethnic model. The mortality incidence 90 days after the start of the study was less than 0.01% in the 18-39 age group. This number increased to 0.67% and 0.44% in males and females, respectively, of more than 80 years of age. The increase in age was therefore significantly liked to mortality risk (individuals over 80 had a risk 20 times greater than those 50-59 years old [HR 20.61; 95% CI 18.72-22.70)). Males had higher risk than females (HR 1.59, adjusted for all other factors; 95% CI 1.53-1.65).

The ethnic group representing the “non-white” English population had a higher mortality risk in comparison to ethnic group representing the “white” English population. The HR, adjusted for age and sex, was between 1.62-1.88 for Black, South Asian, and mixed ethnic groups in comparison to the White population, for which the HR adjusted for all factors was between 1.43-1.48.

Mortality risk also increased with obesity (BMI > 40, adjusted HR 1.92, 95% CI 1.72-2.13) and certain comorbidities such as diabetes (HR 1.95, adjusted for all factors; 95% CI 1.83-2.07), severe asthma, history of respiratory disease, chronic heart conditions, liver diseases, dementia, stroke, neurological diseases, kidney failure, autoimmune diseases (rheumatoid arthritis, lupus, or psoriasis) and immunodepression. Patients with a history of haematological malignancy were at 2.5 times more risk than patients with other types of cancer (HR 2.47, adjusted for all factors; 95% CI 2.06-2.96).

Current smoking was studied post-hoc and was associated with a weaker risk (HR 0.89, 95% CI 0.82-0.97), but the cohort did not allow conclusion on this factor. Arterial hypertesion was studied and its influence was dependent on age and comorbidities.


In summary, the factors presented in this cohort (which could serve as prognostic factors) shown to be linked to mortality are: age > 80 years (x20), male (x1.5), non-white ethnicity (x1.4), obesity (x1.4-2), diabetes (x2), haematological cancers (x2.5 if <5 years) relative to other types of cancer, kidney failure (x2.5), liver diseases and immunodepression (x1.7), stroke and dementia (x2.5), neurological diseases (x2.5), organ transplant (x3.5).

The statistical model adopted in the study did not permit discrimination of the effects of smoking history, even though it is assumed to be associated with a higher mortality risk. Hypertension should also be explored in patients older than 70.

Complementary studies will be necessary to include other factors, including socio-economic factors and isolation policies.

Strength of evidence Weak

- Lack of data noted in the article
- Confounding variables associated with current and past smoking habits
- Recruitment bias (inclusion of suspected cases of COVID-19, not necessarily from confirmed tests and thus not necessarily representative of the English population with COVID-19)
- Diagnostic suspicion bias


To determine risk factors associated with mortality in a quantitative cohort study on the general English population from collection of national data.


Cohort study that collected data from February 1 - May 6, 2020.

Using the OpenSAFELY platform, calculated Hazard ratios (p<0.001) according to a multivariable Cox model and adjusted as a function of age, sex, or all factors.

Data were extracted from The Phoenix Partnership (TPP), coupled to mortality data from the Office for National Statistics (ONS). Data were collected and handled by clinicians, as well as paramedical and administrative personnel.

bibliovid.org and its content are bibliovid property.

Legal Notice