Can we estimate the Disability Adjusted Life-Years (DALYs) for COVID-19 yet?

Many water quality guidelines now use Disability Adjusted Life-Years (DALYs) to define a tolerable level of risk for pathogenic substances in drinking water. Examples include the WHO Guidelines for Drinking-water Quality, WHO Potable Reuse Guidelines and Australian Guidelines for Water Recycling.

Any supply of drinking water carries risks of exposure to pathogenic organisms (viruses, bacteria, protozoa). Those risks might be very small, but they never actually reach ‘zero’ since its impossible to absolutely guarantee that nothing will go wrong ever. If we accept that, then its very helpful to quantitatively define a tolerable level of risk so that we can design our water supplies to meet that level.

In each of the above cases, the tolerable level of risk (for each pathogen) has been defined as that which corresponds to a loss of 10-6 DALYS per person per year. This level is often referred to a 1 microDALY per person per year.

In order to define a suitable level of risk management (usually water treatment) against pathogens, we need to know a number of things, but key among them is the number of DALYs associated with each (risk of) infection. That is, the number of DALYs per case.

Some important examples are included in the WHO Guidelines for Drinking-water quality (See Chapter  7 “Microbial Aspects”). These are:

  • Cryptosporidium: 1.5 × 10−3 DALYs per case
  • Campylobacter: 4.6 × 10−3 DALYs per case
  • Rotavirus: 1.4 × 10−2 DALYs per case

If we want to know whether drinking water is being protected to the same burden of disease risk for SARS-CoV-2 as for these three pathogens, we need a DALYs per case value for SARS-CoV-2. This is a problem for newly important pathogens, since an official DALYs per case value may take years to be produced. So, in the meantime, can we estimate the number of DALYs per case for SARS-CoV-2?

I’m going to give it a go, but I’ll be grateful for your feedback and suggestions for improvement.

As a starting point, I took this paper, recently published in Science, titled “Estimating the burden of SARS-CoV-2 in France”. France seems like a good place to acquire these data from since they had high numbers of infection (more than Australia) and a good public health system with effective reporting.

This paper reports the following statistics:

  • Rate of hospitalisation among infected persons
  • Rate of intensive care admission among hospitalised cases
  • Rate of death among hospitalised cases.

From the first of these, it is also simple to derive:

  • Rate of non-hospitalisation among infected persons

Since the rate of various outcomes are age-dependant, the data for these statistics are broken into age categories (<20, 20-29, 30-39, 40-49, 50-59, 60-69, 70-79 and 80+).

The number of DALYs attributed to infection is determined by the range of debilitative outcomes (including their severity and duration) and the number of deaths (and their associated expected life years lost).

The severities for various debilitating outcomes are defined in terms of “Disability Weights”. A large database of Disability Weights is available from the Global Health Data Exchange. Some more information on how these numbers are derived (and what they actually mean) is available from this paper on “Assessing disability weights based on responses from 30,660 people from four European Countries”.

Searching through these disability weights, the following appear to be reasonably good matches for the likely range and severity of illnesses from COVID-19. Its important to remember that these are “average” severities, thus many people could experience greater severities and the numbers will be balanced by people who experience lesser severities:

Non-hospitalised COVID-19 cases: Roughly consistent with the sequela “Moderate lower respiratory infections”. This has the health state name “Infectious disease, acute episode, moderate” and the lay-person description of “has a fever and aches, and feels weak, which causes some difficulty with daily activities”. The Disability Weight for this condition is 0.051 (0.032-0.074).

Hospitalised COVID-19 cases: Roughly consistent with the sequela “Severe lower respiratory infections”. This has the health state name “Infectious disease, acute episode, severe” and the lay-person description of “has a high fever and pain, and feels very weak, which causes great difficulty with daily activities”. The Disability Weight for this condition is 0.133 (0.088-0.190).

Intensive Care Unit COVID-19 cases: For this state, the highest Disability Weight available in the Global Health Data Exchange database has been applied. This is for the unrelated condition “Schizophrenia acute state” with the lay person description of “hears and sees things that are not real and is afraid, confused, and sometimes violent. The person has great difficulty with communication and daily activities, and sometimes wants to harm or kill himself (or herself)”. The Disability Weight for this condition is 0.778 (0.606-0.900). Although this may not seem like a perfect fit, we will see that the sensitivity of the final result to the selection of this number is low.

Deaths of COVID-19 cases: The statistic that is available is the “rate of death among hospitalised cases”, which may not fully capture the overall rate of death among COVID-19 infections. However, in this case, we will make an assumption that most deaths applied to hospitalised cases. DALYS for deaths are determined according to the number of life-years lost, which is a function of age and (age-specific) average life expectancy. The weighting factor in this case is 1 (per year).

Duration of illnesses

All living conditions (non-hospitalised COVID-19 cases, hospitalised COVID-19 cases, and Intensive Care COVID-19 cases) have currently been estimated to have a duration of 10 days. It is expected that these could be updated with more precise data. However, sensitivity testing indicates that the final result is not highly sensitive to these numbers. The hospitalised and ICU durations can each be increased up to 100 days without significantly changing the final result. The “not hospitalised” case durations can be increased to around 30 days before they start to have a significant impact on the result.

Age proportion of the population

Given that the burden of disease data are available in age-categories, it is necessary to estimate the proportion of the population that each age-category applies to. This figure will vary somewhat among different populations, but in this case, data from Australia has been applied.

Data were obtained from the Australian Bureau of Statistics for “Estimated Resident Population By Single Year Of Age, Australia”. Data were used from June 2019, which included population numbers for each year of age, from “0” to “100 and over”.  From this, the following population percentages were derived:

  • Age <20: 24.6%
  • Age 20-29: 14.5%
  • Age 30-39: 14.5%
  • Age 40-49: 12.9%
  • Age 50-59: 12.1%
  • Age 60-69: 10.3%
  • Age 70-79: 7.1%
  • Age 80+: 4.0%

Age-specific life expectancy

Average life-expectancy in Australia is, itself an age-specific value. That’s because as a person ages, they pass the risk of childhood and adolescent deaths.

Age-specific life expectancy was derived from data (for 2016-18) published by the Australian Bureau of Statistics. Life expectancies differ for males and females, so an average value was taken for each age year. Then a weighted average life expectancy was calculated for each of the above age categories, according to the portion of people in each year age group. These gave average life expectancies of 74 years (remaining) for people aged <20 and average life expectancies of 7 years (remaining) for people aged 80+. The calculated average life expectancies for all of the age categories are shown in the tables below as “Life Years Lost”.

Calculation of DALYS for each of the three health states and deaths

The (weighted) average number of DALYs lost for each of the three health states and deaths is presented in the following four tables. Quick inspection of these numbers reveals that a final sum derived by the combination of all four states will be dominated by the DALYS for “deaths”. And the DALYS for “deaths” figure is, itself, dominated by the life years lost by people aged 60 and above.

This explains why most of the figures applied in the calculations for the three (non-death) health states (eg, Disability Weights and illness durations) are not shown to have a high sensitivity on the final calculated result.

Final calculated DALYS per case of COVID-19.

The final numbers of DALYs per case of COVID-19 is calculated as the sum of the DALYS derived four each of the four states. The sum is 0.088 DALYs per case (but see update in “Addendum” below). This is a large number, -much bigger than the numbers given at the start of this blog post for common waterborne pathogens. This reinforces the fact that it is appropriate to make great efforts to prevent people being infected with this virus.

Uncertainties

There are many uncertainties with this calculation, some of which have already been described above. Uncertainties associated with the first three health states are not likely to significantly impact the final result, since that figure is dominated by the DALYS derived from deaths (mainly from people aged 60+) and the associated Life Years Lost.

As noted earlier, the fatality data used in this calculation only included those that died as hospital patients. The inclusion of non-hospital patient deaths would result in a larger calculated figure. This may be a significant source of uncertainty in this calculation.

The numbers of Life Years Lost are important to the final calculation, and depend upon the applicable age-specific life expectancies. The data used here are from Australia and different figures may be appropriate for other populations.

Some may argue that the true rate of infection may be much greater than the confirmed rate of infection with SARS-CoV-2. However, since unidentified infected people are likely to have very mild (or nil) symptoms, the associated DALYs with these cases would also be very low.

It is also possible that there are other health outcomes, which are not yet well understood. Given the large number of people, globally, who have now been infected with this virus, it seems likely that any poorly reported symptoms must also be relatively rare. Such cases (by virtue of being rare), would also be unlikely to significantly add to the overall DALY burden.

Alternatively, we may learn that there are infact additional severe and long-term impacts of COVID-19 for a sizable portion of the population, which have not yet been identified, -perhaps because of delayed onset of the illness. If that proves to be the case, the figures derived in these calculations will require revision.

Let me know what you think…

Addendum (31 May 2020)

Given reports of the Australian fatality rate for COVID, particularly for people aged 70+, the DALYs-per-case figure calculated above may be significantly underestimated.

This preprint by Peter Collignon and John Beggs reports more precise fatality data for Australia:

https://doi.org/10.1101/2020.05.14.20101378

Incorporating that fatality data gives a much larger figure for “Life Years Lost”:

Once this figure is incorporated into a final DALYS-per-case number with the other three health states, the two hospitalised health states (“hospitalised” and “ICU”) become effectively irrelevant since the final sum is so significantly dominated by this one factor (deaths). A very slight additional contribution comes from the “non-hospitalised” health state, but this is at the level of the third significant figure, -a level of precision not supported by most of the input data.

After more consideration, I have also concluded that the “Disability Weight” for patients in intensive care (ICU) should also be “1”, rather than 0.778 as shown above. This is because patients in intensive care can be considered to be completely incapacitated. Most would be reliant on ventilators to breath and some would not be conscious. Nonetheless, this value is not significant to the final sum for DALYs-per-case, so this change does not affect the calculated result.

Given these considerations, based on primarily Australian data, COVID-19 can be estimated to have a DALYs-per-case value of 0.126.

Published by Stuart Khan

Professor of Civil & Environmental Engineering, University of New South Wales

4 thoughts on “Can we estimate the Disability Adjusted Life-Years (DALYs) for COVID-19 yet?

    1. I think that particular detail is pretty-well captured. It’s reflected in the higher rates of hospitalisations and ICU admissions for higher age-groups. The DALY sums for those categories are driven almost entirely by people aged 60+.

      Like

  1. This is a very useful assessment. When considering public exposure to wastewater, the very high potential burden highlights the importance of understanding how (and how much) SARS-CoV-2 is inactivated through interaction with the wastewater matrix and through wastewater treatment.

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: