Hacker News new | past | comments | ask | show | jobs | submit login

This really depends on how far back you look. I was looking at the data for Sweden the other day. If you look at the last 20 years in terms of overall mortality, 2020 is right around the middle: https://mobile.twitter.com/TLennhamn/status/1295269941109764...

It's certainly more than last year, or the year before that, but it's not an outlier for the last two decades.

If you zoom out 200 years, this year's mortality looks awesome!

Yes, the world has become a safer place to live in over the years.

Given the downward trend in age adjusted deaths per million over time, it makes a lot of sense to compare only the most recent years to 2020.

See here to compare monthly mortality rates for every country in 2020 to the the prior 5 year period: https://ourworldindata.org/excess-mortality-covid

I think looking at the last 20 years is pretty reasonable. All the years are within my lifetime. I agree we should not zoom out for the last 200 years.

If the data only looks bad when you look at 5 year window, and not 20 year window, that seems like a pretty fragile argument.

What problems have had advancements in treatment in the last 20 years? HIV/AIDS, cancer, diabetes. Those are just three things off the top of my head that I know have improved, even without data to support me.

I'm not sure, with the speed of technological improvements in the last decade, that 20 years is appropriate.

Over the last two decades, life expectancy has either peaked in the developed world, or has even declined for some demographics.

Seems appropriate to me.

Not in Sweden. Life expectancy has gone from 77.4 to 81.3 for men since 2000. It does go up every year except for 2014/2015 (80.4 to 80.3).


Interestingly, in the UK the average age of people who die from coronavirus (82.4) is slightly older than the average age of people dying of other causes (81.5).


Why, think about it.

Covid deaths are practically concentrated on age groups over 65 years old.

People dying from other causes aggregate multiple causes which, due to the aggregation, are spread out over the whole age range.

Due to the pyramid shape of the population distribution (they call it population pyramid for a reason) then the median and the mean are shifted towards the base/younger age groups.

Hell, if a disease affected mankind in a way that killed everyone uniformly, the average age of people dying would be around 30 years old.

The way you presented your conclusion implies that dying from covid actually extends your life, which is absurd.

Instead, it just reads that it kills older people disproportionately, and the older you are the more likely you are of dying from it.

Curious if you looked at the underlying data.

Aside from 2018, this year does not even look like much of an outlier: https://mobile.twitter.com/TLennhamn/status/1295269505984344... . 2012 seems to have been deadlier, for Sweden, than 2020, which was 8 years ago.

What about other countries?

Other coutries used lockdowns.

Sweden basically did a lockdown - it just wasn't state-mandated.

If you look at the mobility data [1], you'll see that Sweden followed a very similar trend to its neighbors, and actually maintained lower mobility after other state-mandated lockdowns let up.

[1] https://www.teliacompany.com/sv/om-foretaget/uppdatering/mob...

The arbitrary choice of the reference point makes that bad data.

Either way, using Stockholm as the geographical reference point the chosen point correlates against dark cold Januari/Februari with average temperatures of -1°C and 7-8 hours of twilight. Considering that keeping the relative number below at the height of summer is quite impressive.

To get a better idea of why zooming out 20 years isn't reasonable, take a look at the past 120 years in the US (Sweden's numbers likely follow the same trajectory): https://www.cdc.gov/nchs/data-visualization/mortality-trends...

Directly comparing 2000 to 2020 makes little sense given the consistent downward trend and magnitude of difference between 2001-2005 and 2015-2019.

Given the downward trend in adjusted deaths per million over time, a more reasonable approach is to average the last 5 years and compare it to 2020. Doing so, you will find that 2020 has ~750 more deaths per million in Sweden than prior periods. (Keep in mind this approach likely understates the affect of COVID on mortality rate due to the aforementioned trend).

I am afraid it is more complicated than that.

For example, before we decide that the last 5 years are a good average we should look at what the trend of the last 5 years looked like. If for example there was an upwards trend (which it was for many countries around the world) then even without COVID you'd expect that 2020 would have continued on that trend for whatever reason it was going that way. In that case it would be fairly unexciting to say that 2020 was above the 5 year average in the context of COVID.

In my opinion the fairest comparison would be to check if 2020 continued the trend of the last 5 years, or if it had a non linear jump upwards from that trend, but looking at figures such as "above average" doesn't mean anything if every year in the last 5 years was "above average" because it started to trend upwards again.

Swedish life expectancy has been trending up and death rate trending down (the latter from 10.5 to 8.6 per thousand) since 2000.


I think if you look at "excess mortality" rates from other countries and average them out, the 14% increase of deaths in 2020 might make more sense [1]. A +14% increase in mortality rates is statistically significant when compared to the UN projection of +0.44% made before covid-19 [2].

Looking at one county's stats and ignoring all others is making a fragile counter-argument on worldwide death rates.

[1] https://ourworldindata.org/excess-mortality-covid [2] https://www.macrotrends.net/countries/WLD/world/death-rate

> If you zoom out 200 years, this year's mortality looks awesome!

If you compare it with 1666, this is paradise.

Sweden had pretty good numbers over the summer, but sadly have a spike of cases over the last month. Essentially half of the total covid cases in Sweden have occurred in time. So with some lag, I'd expect another jump in deaths soon, too.

The twitter thread linked above is from August, so definitely does not account for these rapidly changing numbers.

(Recent data is available all over the place--here's one source: https://www.worldometers.info/coronavirus/country/sweden/ )

What does "Cumulative Excess Deaths" mean and why does it spend a lot of time trending down?


And what is "Age Adjusted + Population Adjusted Deaths"? (Yes, I get "population adjusted".)

[As of 2019, Sweden's population has increased 16% since 2000, 13% in the 0-17 yrs and 35% in 65+ yrs groups. The population of "Foreign Born" is up 97% (!) vs 11% growth in Swedish citizens. In that time, life expectancy has gone from 77.4 to 81.3 for men, 82.0 to 84.7 for women. "Crude death rate" has gone from 10.5 to 8.6 (per 1000). Source: https://www.scb.se/en/finding-statistics/statistics-by-subje... As a bottom line, I'd suggest Sweden's demographics changes in the last 20 years have been crazy pants.]

Conceptually, age adjustment is something like population adjustment with consideration for age or age bins. As a simple example, suppose you find that the death rate for people 0-64 is X% and 65+ is Y% in 2019.

These rates change in 2020, but the proportion of the population in these age groups changes, too. So you set a baseline proportion (maybe equal to 2019), compute the per-age-group death rates in 2020, and rescale the total death rate so that the age proportions are the same as the 2019 baseline.

The ultimate goal is to correct for the effects of changing age demographics in computing the death rate. You'd use it if want to look at changes that are corrected for, for example, the population as a whole aging.

So many people looking at Sweden which is strange since it’s so middle of the pack.

Looking at the outliers Norway, Denmark on one side and Peru and Ecuador is so much more interesting: https://www.economist.com/graphic-detail/2020/07/15/tracking...

There are underlying trends in that chart that aren't evident on first glance. The latest data on that chart is from July 2020, several months ago when Sweden had their numbers under decent control.

Even at that take a look at the last few columns, more particularly the slope from month to month in the last couple years. Note how from July 2019 through March 2020 the levels were at or below the previous years, however in the last few months the slope accelerates to an increasing upward monthly trend. I would be quite curious to see this chart with August, Sept, and Oct 2020 data included to see if that acceleration continued. Consider this is what Sweden's case count has looked like: https://i.imgur.com/ofUhXvK.png

In essence this chart is really showing only about 4 months of pandemic influence, and at a time when Sweden was doing pretty well. But in those last 4 months there is a clear accelerating trend in relation to the previous years. I won't say this was cherry picked to be misleading, but at the least it is an incomplete representation of what is happening now.

Looks like 10 normal months and a large spike in April and May to me.

You realize the year ain't over yet.

The data is linearly adjusted to fill out the 2020 data. You are right though, that we don't have the full year's data.

Does it take into account that more people tend to die in December, January, and February than other months? In other words, did they adjust December properly?


I agree that we can't look at the whole year until it's over, due to the irregular level of spread throughout the year.

However, it's interesting to note that during the first ten weeks of 2020, 8% fewer people died in Sweden compared to the average of 2015-2019, and 2020 had in fact the lowest number of deaths of the last six years[1] during the first ten weeks of each year.

[1] https://scb.se/hitta-statistik/statistik-efter-amne/befolkni... (Excel file, Tabell 5, fairly easy to translate)

Well, that does pose a bit of a problem comparing across years, doesn't it? I'd like to see this sort of statistical analysis done after the calendar flips over to 2021, taking into account the trend of fewer deaths per year, so we could see if there is a "COVID spike" in the data, and what the magnitude of it is.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact