Epi in the News

Thank you for posting this.

1. I think this is a prevalence figure as it gives us a snapshot of the situation. The numerator should be Number of people exposed to extreme heat and the denominator should be total population in the world.

2. This sentence tells us about the increase in the number of people (person) in a specific amount of time (i.e, days) compared to the previous data.
This figure is useful as it gives us the sum of each individuals time at risk (incidence rate). The population of a city is always dynamic. On the other hand, prevalence risk does not account for the time period.

3. With the rise in heat, the mortality would increase. In this question, rise in heat is the reference group while rise in mortality is the comparison group.

4. Attributable Risk = risk of death due to extreme heat - risk of death due to pre-existing health conditions divided by risk of death due to extreme heat.

Thank you Sujit, Nada and Parwaan,

1. This is a prevalence figure. Number of people exposed to deadly heat (>30 degrees on wet bulb globe temperature scale) over an unspecified time period/ Total world population.

2. an increase 575 million person days of extreme heat means that the if you were to sum the number of days that each individual experienced extreme heat over a recent time period and subract the same for a past time period this would be the difference. It incorporates time, unlike the prevalence figure above.

3. I would expect a very low RR for the association between heat and mortality - something like 1.01, but because it affects the whole population, this is very signficant in absolute terms. Rate ratio = Mortality rate of people experiencing extreme heat/mortality rate of people not experiencing extreme heat with both measured in deaths/person years

4. Population attributable fraction = attributable risk (number of excess deaths due to extreme heat)/number of people who have died in the same population and time period

Best wishes

Judith

Hi Sujit and everybody...

1) Agreeing with all- it would have to be a prevalence measure. The article does talk about the study that looked at a time period of 3 decades- but the 25% is how much of the world is exposed to heat currently- including previous and new exposure. The numerator would be number of ppl exposed to deadly city heat (ppl living in cities with deadly heat based on their definition of heat), denominator would be the world population.

2) 575 million person-days.. how to explain this to a non-epidemiologist? Well, I would say " This person-days exposure is derived by multiplying the population of Dhaka at risk with the number of days of extreme heat experienced by the city in that time period. The figure is dependant on the change in the population as well as the change in the number of high heat days experienced by the city. Over the past 3 decades, this figure has increased by 575 million person-days ( so that's person-days exposed in 2016 minus person-days exposed in 1983). The population increase is a major factor, but along with that there is also global warming and more number of days falling in the criteria of extreme heat".

How is this more useful than 25%? Because this actually tells us the magnitude of the problem. Its far more easier to ignore 25% when compared to 575 million person-days... :-DD

3) For the statement that the author makes, if we are thinking in terms of increase mortality, we should be expecting RR of more than 1. (and I am thinking in terms of rate ratio)

Numerator would be = Rate of all-cause mortality in a population exposed to extreme heat = (Number of deaths / personyears at risk)
Denominator would be= Rate of all cause mortality in a population not exposed to extreme heat
Again there would need to be a clear definition of "exposed to extreme heat", and maybe we would have to use age-standardization

4. For this I agree with Judith... it would be population attributable fraction.
It would be :
(All-cause mortality in the population - Mortality due to causes other than extreme heat) / All-cause mortality rate in the population
This will give the proportion of all cause mortality that is due to extreme heat...

Fathima

Very interesting. Thank you Sujit for sharing this article!

1. It first looked like “period prevalence" to me, but when I read the paper, it also seemed like incidence risk. While I wasn’t able to understand most of the methods, I read the methods section of the actual PNAS paper the article was referring to, because it was confusing to me how they derived that specific result only given that number in the newspaper. (The actual study can be found at
https://www.pnas.org/content/118/41/e2024792118 )

So, my guess is
-Denominator: The world population in mid-2016 (See their UN reference data; File 18 at https://population.un.org/wup/Download/ )

-Numerator: 1.7 billion people. But I’m not sure how they got this number even after reading their methods section. I assume that this reflects the total number of urban population exposed to deadly city heat during the year 2016.

* The only sentence I could assume as above was this sentence in the abstract: Exposure trajectories increased for 46% of urban settlements, which together in 2016 comprised 23% of the planet’s population (1.7 billion people).
* Two criteria on extreme heat events: 1-d or longer periods in which WBGTmax > 30 °C and 2-d or longer periods in which the maximum HImax > 40.6 °C.

2. A) Compare to 1983, those who live in Bangladesh’s capital in 2016 have experienced increased exposure to extreme heat. This was measured as the product of how many people experienced extreme heat and how many days they suffered from heat; basically, it is “person” x “days”, which was 575 million person-days in this study.

B) This gives us more information since it contains not only the number of people involved but also how many days they were affected. Furthermore, we could glimpse a hint on actual methods the researchers were using in this study.

* Their methods described in the paper: 

We quantify urban exposure to extreme heat in person-days/year−1 for each GHS-UCDB urban settlement from 1983 to 2016.  Person-days/year−1 is a widely used metric to compare and contrast exposure to extreme heat across geographies and time periods.  

* By the way, Fig S4 in the supplement showed the change in global urban extreme-heat exposure (in Person-days/year) according to time (year). They show how “person-year” value of heat exposure changed over time, providing evidence for climate change. (The Supplement: https://www.pnas.org/content/pnas/suppl/2021/09/28/2024792118.DCSupplemental/pnas.2024792118.sapp.pdf)

3. RR > 1 is expected. Numerator: Risk of death in the extreme heat-exposed group, Denominator: Risk of death in the unexposed group.

4. AR=(Risk of deaths in extreme heat-exposed group - Risk of deaths in non-exposed group)/Risk of deaths in extreme heat-exposed group.

I think a simple approach would be choosing a specific date (or multiple dates) in early-July (e.g. a day with relatively average temperature, at least a week before extreme heat comes) and measure prevalenceo of death on that day due to heat; then, choose a date (or multiple dates) with extreme heat (e.g. >30°C) and measure prevalence of death due to heat on that day(s). The difference between the two divided by the latter would be attributable to extreme heat. I think it can be done more easily by doing a retrospective research since the mortality and temperature points are already available.

1. This is a prevalence figure.
The numerator is number of people exposed to deadly city heat, and the denominator is the total world population.
2. “Bangladesh’s capital experienced an increase of 575 million person-days of extreme heat” - For a non-epidemiologist, the explanation would be as follows:
the number of person-days of exposure to extreme heat (i.e. the cumulative population exposed to cumulative heat in a given year for a particular place) in 2016 is increased by 575 million person-days from that value in 1983.
The figure like this is useful compared to a figure like 25% from the headline, because this figure incorporates the time that individuals in the population exposed to extreme heat. An urban population is always dynamic and temperatures are generally higher in urban areas due to concrete surface and little vegetation.
3. This would be a rate ratio.
The numerator would be mortality rate in the population exposed to extreme heat, and the denominator would be mortality rate in the population not exposed to extreme heat.
4. I would estimate the proportion of deaths in a city that is attributable to extreme heat by population attributable fraction, i.e. the proportion of total deaths in a city minus the proportion of deaths in the group not exposed to extreme heat divided by the proportion of total deaths in the city.

1. This is a prevalence figure.
The numerator is number of people exposed to deadly city heat, and the denominator is the total world population.
2. “Bangladesh’s capital experienced an increase of 575 million person-days of extreme heat” - For a non-epidemiologist, the explanation would be as follows:
the number of person-days of exposure to extreme heat (i.e. the cumulative population exposed to cumulative heat in a given year for a particular place) in 2016 is increased by 575 million person-days from that value in 1983.
The figure like this is useful compared to a figure like 25% from the headline, because this figure incorporates the time that individuals in the population exposed to extreme heat. An urban population is always dynamic and temperatures are generally higher in urban areas due to concrete surface and little vegetation.
3. This would be a rate ratio.
The numerator would be mortality rate in the population exposed to extreme heat, and the denominator would be mortality rate in the population not exposed to extreme heat.
4. I would estimate the proportion of deaths in a city that is attributable to extreme heat by population attributable fraction, i.e. the proportion of total deaths in a city minus the proportion of deaths in the group not exposed to extreme heat divided by the proportion of total deaths in the city.