COVID-19 Age Vulnerability Table background

Background

The background to the Covid Age table is outlined below so users understand the source of data and the process used to calculate relative risks and equivalent added years of age. This text is currently in draft format and will be updated when finalised.

A.        Age

 

In the OS study, after adjustment for sex, multiple comorbidities, and various other risk factors, the risk of death from Covid-19 showed a near to exponential relationship to age among adults, such that the RR for a 10 year increase in age was approximately 2.4, and that for a single year of age 1.0945.  Confidence intervals were not presented for the regression coefficient of log-transformed HR on age, but because the analysis was based on a large study sample (>17 million) and a large number of deaths (5683), we would expect the estimated effect of age to be statistically precise, especially at ages 40-69 years.

Comparison with data from other sources

By way of comparison, Table A1 shows data on mortality from Covid-19 (as an underlying cause of death) in England and Wales during March 2020 [ref].

 

Table A1.  Mortality from Covid-19 (deaths per 100,000) by sex and age, England and Wales, March 2020

 

Age band (years) Male Female
20-24 0.0 2.0
25-29 1.7 0.0
30-34 4.7 0.0
35-39 4.8 1.8
40-44 5.2 3.9
45-49 11.1 14.5
50-54 21.0 12.5
55-59 47.9 26.4
60-64 80.7 33.0
65-69 118.1 56.9

 

A 10-year increase in age from 50-54 to 60-64 was associated with a 3.8-fold increase in mortality in men and a 2.6-fold increase in women.  For the increase in age from 55-59 to 65-69, the corresponding increases were 2.5 in men and 2.2 in women.  When allowance is made for random sampling variation, and the fact that ONS data do not take account of covariates other than sex, these ratios seem compatible with those estimated from the OS study.

Conclusion

We conclude it is reasonable to assume that after allowance for other variables, risk of death from Covid-19 increases exponentially with age among people of working age, such that a one-year increase in age carries a relative risk of 1.0945.

Robustness of risk estimate

This estimate is derived from a large study and is compatible with data from an independent source.  We consider it to be moderately robust.

B.        Sex

In the OS study, after adjustment for age, multiple comorbidities, and various other risk factors, the HR of death from Covid-19 in men relative to women was 1.99 (95%CI 1.88-2.10).  We would not expect this risk estimate to be liable to any major bias.

Comparison with data from other sources

By way of comparison, in ONS statistics on mortality from Covid-19 (as the underlying cause of death) in England and Wales during March 2020 [ref], the directly-standardised mortality rate per 100,000 was 97.5 in men and 46.6 in women, giving a ratio of 2.09.  This ratio might change slightly if it were adjusted for comorbidities with differing prevalence by sex, but data from the Health Survey for England indicate that such differences in prevalence are generally small [ref].  Thus, we consider that the ONS data, which relate to deaths outside as well as within hospital, support the relative risk from the OS study.

Conclusion

We conclude that after allowance for other risk factors, the relative risk of death from Covid-19 in men as compared with women should be taken as 2.0.  Correspondingly, the risk in women relative to men can be taken as 0.5.

Robustness of risk estimate

Given the statistical precision of its source, and its compatibility with other data, we judge this risk estimated to be robust.

C.        Ethnicity

In the OS study, data on ethnicity were available for 74% of patients in the cohort.  Table C1 shows HRs adjusted for sex, age, multiple comorbidities and various other risk factors, according to ethnicity and the date at which follow-up was censored.

Table C1.  Adjusted hazard ratios for ethnic groups from the OS study according to date when follow-up was censored

Ethnic group   Follow-up to 25.4.20   Follow-up to 6.4.20
    HR (95%CI)   HR (95%CI)
             
White   1     1  
Mixed   1.64 (1.19-2.26)   1.13 (0.62-2.05)
Asian or Asian British   1.62 (1.43-1.82)   1.77 (1.48-2.13)
Black   1.71 (1.44-2.02)   1.90 (1.48-2.45)
Other   1.33 (1.03-1.73)   1.81 (1.28-2.57)

 

Differences between corresponding risk estimates from the two analyses, especially for the less common ethic groups (mixed, black and other) may have arisen through random sampling variation, and we have no reason to suspect that lower relative risks for some ethnic groups in the analysis with longer follow-up reflect a tendency for those groups to have avoided infection more effectively as the epidemic evolved.  Thus, we consider that the risk estimates from the analysis with later censoring, which were statistically more precise, are the most reliable.

Comparison with data from other sources

The Office for National Statistics has published data on odds ratios for death related to Covid-19 in England and wales during 2 March to 10 April 2020 according to ethnic group, with adjustment for age, geographical region, household composition, socioeconomic status and self-reported health at the 2011 census (

https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/oddsratiosforriskofcoronavirusrelateddeathsbyethnicgroupenglandandwales).  Table C2 shows risk estimates from that report for selected ethnic groups.

Table C2.  Odds ratios for death related to Covid-19 according to ethnic group, England and Wales, 2 March to 10 April 2020

(Odds ratios are adjusted for age, geographical region, household composition, socioeconomic status and self-reported health at the 2011 census).

Ethnic group   Male   Female
    OR (95%CI)   OR (95%CI)
             
White   1     1  
Bangladeshi/Pakistani   1.81 (1.55-2.11)   1.61 (1.31-1.97)
Indian   1.93 (1.70-2.18)   1.89 (1.63-2.20)
Black   1.32 (1.15-1.53)   1.43 (1.20-1.71)

 

The estimated odds ratios for the Bangladeshi/Pakistani and Indian groups seem broadly compatible with the HR for the Asian or Asian British group in the OS study.  However, the odds ratios for black men and women are notably lower than the HR for the black group in the OS study.  This difference might in part reflect differences in adjustment for covariates, but it is a reason for more cautious interpretation of the OS risk estimate for black people.

Conclusion

Despite some discrepancy from the ONS data, we conclude that is reasonable to adopt relative risk estimates for ethnic group from the OS study as follows:  Asian or Asian British 1.6; Black 1.7; Mixed 1.6; Other non-white 1.3.  All of these risk estimates are relative to white as the reference.

Robustness of risk estimate

The risk estimate for the Asian or Asian British group is fairly precise statistically, and accords with independent ONS data.  We therefore judge it to be moderately robust.  The other risk estimates are less precise and less well supported, and we regard them as provisional.

D.    Obesity

In the OS dataset, body mass index (BMI) was ascertained from measurements of weight in the past 10 years, when individuals were >16 years old.

Table D1 shows HRs, adjusted for sex, age, multiple comorbidities and various other risk factors, according to levels of BMI.

Table D1.  Adjusted hazard ratios for categories of body mass index from OS study according to date when follow-up was censored

 

BMI (Kg/m2)   Censored at 25.4.20   Censored at 6.4.20
    HR (95%CI)   HR (95%CI)
             
<30   1     1  
30-34.9   1.27 (1.18-1.36)   1.39 (1.25-1.54)
35-39.9   1.56 (1.41-1.73)   1.62 (1.39-1.90)
≥40   2.27 (1.99-2.58)   2.45 (2.00-3.01)

 

It is possible that the lower HRs with longer follow-up reflect selective shielding of people with obesity as the epidemic evolved, and for this reason, we consider that the analysis censored at 6 April provides the more reliable estimate of RR for our risk model.

Comparison with data from other sources

In the ISARIC cohort, of patients admitted to British hospitals with Covid-19, obesity (not sub-divided by severity) carried a higher case-fatality rate (adjusted HR 1.37, 95%CI 1.16-1.63).  However, the prevalence of obesity in that cohort was remarkably low (approximately 9%, as compared with prevalence rates from the 2017 Health Survey for England by sex and 10-year age band in the order of 30% [ref]), suggesting serious under-ascertainment.

Conclusion

We consider that the apparently low prevalence of obesity among hospitalised Covid-19 patients in the ISARIC cohort is almost certainly an artefact of incomplete ascertainment, and we therefore accepted the risk estimates from the OS study (shorter follow-up period) for inclusion in our risk model, as set out in Table D2.

 

Table D2.  Adopted relative risk estimates for obesity

 

BMI (Kg/m2) Relative risk
   
<30 1
30-34.9 1.4
35-39.9 1.6
≥40 2.4

 

Robustness of risk estimates

These risk estimates are derived from a single, albeit large and nationally representative study, with limited support from other investigations.  We therefore consider them provisional.

E.    Asthma

In the OS study, asthma was sub-classified according to whether or not it had been treated with oral corticosteroids in the year before baseline (severe or mild).  Table E1 shows the prevalence of these categories of asthma in the study cohort, and the associated hazard ratios for death from Covid-19, after adjustment for sex, age, multiple comorbidities and various other risk factors.

Table E1.  Prevalence of asthma in OS cohort, and adjusted HRs for death from Covid-19 according to date when follow-up was censored

Severity of asthma   Prevalence (%) in cohort   Censored at 25.4.20   Censored at 6.4.20
    HR (95%CI)   HR (95%CI)
                 
No asthma   84.1   1     1  
Mild asthma   14.2   1.11 (1.02-1.20)   1.14 (1.01-1.29)
Severe asthma   1.7   1.25 (1.08-1.44)   1.39 (1.12-1.73)

 

It is possible that as the epidemic evolved, patients with more severe asthma took more extreme measures to reduce their risk of contracting Covid-19, leading to a lower HR in the analysis over the longer follow-up period.  Thus the HR from the analysis with shorter follow-up may be more reliable.

Comparison with data from other sources

In the ISARIC cohort of hospitalised patients, the overall prevalence of asthma (not sub-divided by severity) was 14%, which is similar to that in the OS cohort (15.9%). Moreover, prevalence rates for doctor-diagnosed asthma in the 2018 Health Survey for England were even higher [ref].  The report from the ISARIC study does not present a risk estimate for death in patients with asthma, although it does for other comorbidities with clearly increased risk.

Conclusion

When viewed together, the above findings indicate that most asthma is associated with little, if any, increase in risk of mortality from Covid-19.  However, a small elevation of risk seems likely in people with more severe asthma that has required use of oral corticosteroids in the past year.  We therefore adopted relative risk estimates for our risk model as set out in Table E2.

Table E2.  Adopted relative risk estimates for asthma

Severity of asthma Relative risk
   
None 1
Mild (no requirement for oral corticosteroids in past year) 1.1
Severe (requiring oral corticosteroids in past year) 1.4

 

Robustness of risk estimate

Although derived from a single study, these risk estimates appear compatible with other independent data, and we regard them as moderately robust.

F.    Diabetes

In the OS study, diabetes was classified to three mutually exclusive categories, according to whether an HbA1cmeasurement had been made in the last 15 months, and if so, whether the level was <58 mmol/mol (controlled diabetes) of higher (uncontrolled diabetes).  Table F1 shows the prevalence of these categories of diabetes in the OS cohort, and their HRs for death from Covid-19 during follow-up.

Table F1.  Prevalence of diabetes in OS cohort, and adjusted HRs for death from Covid-19 according to date when follow-up was censored

Severity of diabetes   Prevalence (%) in cohort   Censored at 25.4.20   Censored at 6.4.20
    HR (95%CI)   HR (95%CI)
                 
No diabetes   90.9   1     1  
Controlled   2.8   1.50 (1.40-1.60)   1.48 (1.33-1.63)
Uncontrolled   6.0   2.36 (2.18-2.56)   2.57 (2.27-2.91)
No recent HbA1c measure   1.1   1.87 (1.63-2.19)   1.68 (1.33-2.12)

There was no clear indication that HRs with longer follow-up were lowered as a consequence of selective shielding by diabetic patients.  Therefore, the statistically more robust HRs from the longer follow-up period were judged to be the more reliable.

Comparison with data form other sources

To check on the plausibility of the risk estimates from the OS study, we analysed data on diabetes from the ISARIC study and the 2017 Health Survey for England [ref].  Table F2 shows approximate numbers of patients by sex and age in the ISARIC cohort (estimated by measuring the lengths of bars in a bar chart), the prevalence of doctor-diagnosed diabetes in the same age and sex strata in the 2017 Health Survey for England, and calculations from these data of the numbers of patients with doctor-diagnosed diabetes that might have been expected in the ISARIC cohort if diabetes had no effect on hospital admission for Covid-19.  Summation of expected across all of the strata indicates that the overall expected percentage prevalence of doctor-diagnosed diabetes in the ISARIC cohort would be:

100*(586+302)/(4401+2807) = 12.

In contrast, the reported prevalence of uncomplicated diabetes in the cohort was 19%, suggesting a relative risk for hospital admission in the order of 19/12 = 1.6.

This calculation has many limitations.  Unlike the ISARIC study, the Heath Survey for England did not cover Wales or Scotland, and its case definition and method of ascertaining diabetes differed form that in the ISARIC study.  Furthermore, the calculated ratio takes no account of possibly higher fatality among Covid-19 patients with diabetes once they are admitted to hospital.  When these weaknesses are taken into account, the calculated ratio seems compatible with the risk estimates from the OS study, and gives them added plausibility.

Table F2.  Calculation of expected numbers of patients with doctor-diagnosed diabetes in ISARIC cohort, based on prevalence in the 2017 Health Survey for England

Aggregated age band (years) Approximate numbers in ISARC cohort   Prevalence % of DDD* in 2017 Health Survey for England Expected numbers of cases in ISARIC cohort
Male Female Male Female Male Female  
 
0-14 73 36  
16-24 36 40 1 0 0 0  
25-34 84 124 1 0 1 0  
35-44 233 131 3 2 8 3  
45-54 459 339 9 5 42 18  
55-64 674 357 11 7 72 26  
65-74 901 474 19 11 170 52  
75+ 1940 1305 15 16 292 204  
 
Total 4401 2807     586 302  

*Doctor-diagnosed diabetes

Conclusions

We concluded that it is reasonable to adopt risk estimates for diabetes from the OS study for our risk model as set out in Table F3.

Severity of diabetes Relative risk
   
No diabetes 1
Controlled 1.5
Uncontrolled 2.4
No recent HbA1c measure 1.9

 

Robustness of risk estimates

In view of their derivation from a large and nationally representative dataset, and their consistency with data from other sources, we consider these risk estimates to be moderately robust.

G.     Chronic heart disease

In the OS study, chronic heart disease (CHD) included heart failure, ischaemic heart disease, and severe valve or congenital heart disease likely to require lifelong follow-up.  Among the 6.7% of cohort members with CHD, the HR for death from Covid-19, adjusted for sex, age, multiple comorbidities and various other risk factors was 1.27 (95%CI 1.20-1.35) when follow-up was censored at 25.4.20, and 1.33 (95%CI 1.22-1.46) when it was censored at 6.4.20.

Comparison with data from other sources

No directly comparable data are available for other studies in the UK, but in the ISARIC study, the prevalence of CHD among Covid-19 patients admitted to hospitals in England Wales and Scotland was 29%, with an adjusted HR for death of 1.31 (95%CI 1.18-1.45). Unless people with CHD who contract Covid-19 have an unusually low risk of being admitted to hospital, which seems unlikely, this would suggest a relative risk for mortality among all Covid-19 cases in the wider community of at least 1.3.  Data on hospitalised cases of Covid-19 in the United States also indicate adjusted relative risks for case fatality in excess of 1.3 [refs].

Conclusions

Although based on a large and nationally representative dataset, the OS risk estimates for CHD seem low in comparison with what might be expected from other sources of data.  Furthermore, the HR in models with censoring at 25.5.20 reduced substantially after adjustment for other risk factors in addition to sex and age (from 2.01 to 1.27).  It is unclear which were the main factors of adjustment that accounted for such a large reduction in the risk estimate.

With these considerations in mind, we tentatively adopted a relative risk of 1.4 for CHD in our risk model.

Robustness of risk estimate

The relative risk for CHD seems likely to be higher than 1.4, but how much higher is currently quite uncertain.  The value adopted for the risk model should therefore be classed as provisional.

H.      Chronic respiratory disease other than asthma

In the OS study, this category of comorbidity included chronic obstructive pulmonary disease (COPD), fibrosing lung disease, bronchiectasis and cystic fibrosis.  After adjustment for sex, age, multiple comorbidities and various other risk factors, it carried HRs of 1.78 (95%CI 1.67-1.90) when follow-up was censored at 25.4.20, and 1.97 (95%CI 1.77-2.18) with censoring at 6.4.20.  The lower HR after longer follow-up may in part reflect selective shielding of people with chronic respiratory disease as the epidemic evolved.

Comparison with data from other sources

In the ISARIC study of patients admitted to hospital with Covid-19 in England, Wales and Scotland, chronic pulmonary disease other than asthma was reported in approximately 17% of cohort members, and carried an adjusted HR for death of 1.2.  Table H1 shows approximate numbers of patients by sex and age in the ISARIC cohort (estimated by measuring the lengths of bars in a bar chart), the prevalence of doctor-diagnosed COPD (including chronic bronchitis and emphysema) in the same age and sex strata in the 2010 Health Survey for England, and calculations from these data of the numbers of patients with doctor-diagnosed COPD that might have been expected in the ISARIC cohort if COPD had no effect on hospital admission for Covid-19, and its prevalence in the general population remained at similar levels.  Summation of across all of the strata indicates that the overall expected percentage prevalence of doctor-diagnosed COPD in the ISARIC cohort would be: 100*(339+205)/(4401+2807) = 8%.  In addition, a smaller prevalence of other types of chronic pulmonary disease might be expected.  When the 17% observed prevalence of chronic pulmonary disease is set alongside findings from this rough analysis of expected numbers, and also the HR of 1.2 for death in patients with chronic pulmonary disease, the OS relative risk estimates look highly plausible.

Table H1.  Calculation of expected numbers of patients with doctor-diagnosed COPD in ISARIC cohort, based on prevalence in the 2010 Health Survey for England

Aggregated age band (years) Approximate numbers in ISARC cohort   Prevalence % of DDCOPD* in 2010 Health Survey for England Expected numbers of cases in ISARIC cohort
Male Female Male Female Male Female
0-14 73 36
16-24 36 40 1 0 0 0
25-34 84 124 2 2 2 2
35-44 233 131 2 4 5 5
45-54 459 339 4 5 18 17
55-64 674 357 6 8 40 29
65-74 901 474 11 10 99 47
75+ 1940 1305 9 8 175 104
       
Total 4401 2807     339 205

*Doctor-diagnosed COPD

Conclusions

Based on these considerations, we assigned a relative risk of 1.9 to chronic respiratory disease other than asthma.

Robustness of risk estimate

The adopted relative risk estimate is derived from a large and nationally representative cohort, and supported by data from an independent source.  We consider it to be moderately robust.

I.         Chronic kidney disease

In the OS study, chronic kidney disease (CKD) was defined as a glomerular filtration rate <60mL/min/1.73m2, as estimated from the most recent serum creatinine measurement, where available.  This was present in 6.3% of cohort members.  In analyses that adjusted for sex, age, multiple comorbidities and various other risk factors, if carried HRs for mortality from Covid-19 of 1.72 (95%CI 1.62-1.83) when follow-up continued to 25.4.20, and 1.97 (95%CI 1.77-2.18) when it was censored at 6.4.20.  Some attenuation of risk may have occurred with longer follow-up because of selective shielding of patients with CKD.

Comparison with data from other studies

In the ISARIC cohort of patients hospitalised with Covid-19, the reported prevalence of CKD was approximately 14%, and it carried an adjusted HR of 1.2 for death.  Table I1 shows approximate numbers of patients by sex and age in the ISARIC cohort (estimated by measuring the lengths of bars in a bar chart), the prevalence of doctor-diagnosed CKD in the same age and sex strata in the 2016 Health Survey for England, and calculations from these data of the numbers of patients with doctor-diagnosed CKD that might have been expected in the ISARIC cohort if CKD had no effect on hospital admission for Covid-19.  Summation of across all of the strata indicates that the overall expected percentage prevalence of doctor-diagnosed CKD in the ISARIC cohort would be: 100*(185+89)/(4401+2807) = 4%.  This implies a ratio of observed to expected prevalence of 14%/4% = 3.6.  It is possible, however, that doctor-diagnosed CKD, which after allowance for sex and age, had a lower prevalence in the 2016 Health Survey for England than CKD as determined in the OS study, represented a more severe spectrum of disease.  When this is taken into account, the findings from this further analysis of ISARIC data do not call into question the relative risk estimates from the OS study

Table I1.  Calculation of expected numbers of patients with doctor-diagnosed CKD in ISARIC cohort, based on prevalence in the 2016 Health Survey for England

Aggregated age band (years) Approximate numbers in ISARC cohort   Prevalence % of DDCKD* in 2010 Health Survey for England Expected numbers of cases in ISARIC cohort
Male Female Male Female Male Female
0-14 73 36
16-24 36 40 0 1 0 0
25-34 84 124 1 1 1 1
35-44 233 131 1 1 3 1
45-54 459 339 2 2 9 5
55-64 674 357 4 2 24 6
65-74 901 474 5 3 44 16
75+ 1940 1305 5 5 104 60
       
Total 4401 2807     185 89

*Doctor-diagnosed CKD

Conclusions

Based on the above considerations, we adopted a relative risk of 1.8 for CKD defined as a glomerular filtration rate <60mL/min/1.73m2, as estimated from the most recent serum creatinine measurement.

Robustness of risk estimate

This risk estimate is based on findings in a large and nationally representative cohort, and I broadly consistent with independent data from other sources.  As such, we judge it to be moderately robust.

X.  Risk factors not included in risk model

Two potential determinants of vulnerability carried no apparent increase in the risk of death from Covid-19 in the OS study, after account had been taken of sex, age and comprbidities.  These were smoking and hypertension.  Adjusted HRs with follow-up to 25.4.20 were 0.88 (95%CI 0.79-0.99) for current vs. never smokers, and 0.95 (95% CI 0.89-1.01) for high or diagnosed hypertension.

These risk factors were therefore excluded from the risk model.

Z.  Rarer comorbidities

The OS study also provides adjusted risk estimates for a number of other rarer comorbidities, for which we have not as yet identified any independent corroborating data.

HRs for these comorbidities, adjusted for sex, age, multiple other comorbidities and various other risk factors are summarised in Table Z1, together with the relative risks that we have carried forward to our risk model.  The choice of the values taken forward weighed the greater statistical precision of the estimates based on longer follow-up against the possibility that they may in some cases have been biased downwards because of selective shielding by people with the comorbidity.  All adopted values are considered provisional.

Table Z1.  Adjusted hazard ratios for other comorbidities from the OS cohort, and relative risk estimates taken forward to risk model

Comorbidity   Follow-up censored at 25.4.20   Follow-up censored at 6.4.20   RR adopted for risk model
  HR (95%CI)   HR (95%CI)  
                 
Non-haematological cancer                
None   1     1      
Diagnosed <1 year ago   1.56 (1.29-1.89)   1.51 (1.10-2.05)   1.6
Diagnosed 1-4.9 years ago   1.19 (1.04-1.35)   1.36 (1.13-1.65)   1.2
Diagnosed ≥5 years ago   0.97 (0.88-1.06)   0.92 (0.79-1.06)   1.0
                 
Haematological malignancy                
None   1     1      
Diagnosed <1 year ago   3.52 (2.41-5.14)   2.60 (1.30-5.22)   3.5
Diagnosed 1-4.9 years ago   3.12 (2.50-3.89)   3.67 (2.66-5.06)   3.1
Diagnosed ≥5 years ago   1.88 (1.55-2.29)   1.64 (1.18-2.28)   1.9
                 
Liver disease   1.61 (1.33-1.95)   1.86 (1.40-2.47)   1.6
                 
Chronic neurological disease other than stroke or dementia*   2.46 (2.19-2.76)   2.28 (1.88-2.76)   2.5
                 
Organ transplant   4.27 (3.20-5.70)   2.62 (1.51-4.57)   4.3
                 
Spleen diseases†   1.41 (0.93-2.12)   1.87 (1.06-3.30)   1.4
                 
Rheumatoid/lupus/psoriasis   1.23 (1.12-1.35)   1.31 (1.14-1.51)   1.2
                 
Other immunosuppressive condition‡   1.69 (1.21-2.34)   2.01 (1.25-3.25)   1.8

*Includes motor neurone disease, myasthenia gravis, multiple sclerosis, Parkinson’s disease, cerebral palsy, quadriplegia, hemiplegia, malignant primary brain tumour and progressive cerebellar disease.

†Includes splenectomy, or spleen dysfunction (e.g. from sickle cell disease).

‡Includes HIV, conditions inducing permanent immunodeficiency (ever diagnosed), aplastic anaemia, and temporary immunodeficiency recorded within the past year.

Estimation of individual vulnerability

Using the relative risk estimates that have been derived in the preceding sections of this report, it is possible to estimate the vulnerability of an individual should he or she at some stage contract Covid-19.

A major determinant of risk is age.  Using the risk estimate for age from Section A, an increase in age of n years carries a relative risk of 1.094n.  This implies that a relative risk, R, is equivalent to that from an increase in age of (log R)/(log 1.09) years.  Applying this formula, Table 1 expresses the relative risks that have been adopted for risk factors other than age in the additional years of age that would give an equivalent relative risk.

The analyses that were used to generate these risk estimates assumed that relative risks from different risk factors multiply, and in the absence of persuasive evidence to the contrary, this seems a reasonable assumption.  With that assumption, an individual’s vulnerability can be assessed form Table 1 by summing the added age equivalent for each risk factor that applies.  For example, an Asian woman aged 50 with uncontrolled diabetes would have an estimated vulnerability equivalent to that of a healthy white man aged

50 – 8 + 5 + 10 = 57 years.