As the data turns (amazingly interesting talk)

This clip is long -- about 20 minutes -- but fascinating. I was never bored. Not for a moment. It's a talk given at the 2006 TED Conference in Monterey by Professor Hans Rosling of Sweden's Karolinska Institute. It's about . . . well you decide what this is about. It starts out being about international health and global development and winds up being about data, data access and data presentation. A tour de force:

More like this

"data" is plural, so they turn.

Rick: As the typical pedantic epidemiologist I almost always use data as a plural, too, although the dictionary now says it can be used either way. In this case I deliberately chose the singular to make it match with "As the World Turns," the title of a famous soaper and an appropriate one for the subject of the talk. So you see, some hidden thought went into it.

Revere:
I have a question about comparing infant mortality rates across countries. The numerator is dead infants; that is not a problem and the case definition is very reliable in different nations. The tricky part is the denominator, namely, live births. What counts in one place as a live birth may be different, if I am not greatly misinformed. A very premature delivery that leads to a neonate who is ventilated immediately and later dies might not have been ventilated in a less developed setting and might count as a still birth. Do you happen to know where this is discussed in a clear way? Thanks

By Ed Whitney (not verified) on 22 Feb 2010 #permalink

Rosling is awesome and I'm only 5 minutes in...

By antipodean (not verified) on 22 Feb 2010 #permalink

WHO summarizes succinctly the problem. The key quote is at the end:
The reliability of the neonatal mortality estimates depends on accuracy and completeness of reporting and recording of births and deaths. Underreporting and misclassification are common, especially for deaths occurring early on in life.
Perinatal mortality, defined as number of stillbirths and deaths in the first week of life per 1,000 live births, is a useful additional indicator, and work is ongoing to improve estimates of stillbirth rates, a major component of perinatal mortality.

The underreporting and misclassification of live births is the issue I was wondering about. WHO has infant mortality tables by nation at http://apps.who.int/whosis/database/mort/table2.cfm. You have to download one country at a time, one year at a time. But when I tried a few countries, some more developed and some less developed, and looked at the percentage of infant deaths that were reported as occurring at less than one day, there was considerable variation. Using a calculator, and using the most recent year with data, I found these percentages of all infant deaths reported as occurring at less than one day of age:
Sweden 2005 22.6%
Cuba 200510.0%
USA 200539.9%
Canada 200041.5%
Haiti 200325.8%
Ecuador 200519.5%
Slovakia 200514.3%
Australia 200340.9%
Mexico 200520.6%
Brazil 200424.4%

Such large variations make me think that some of the comparisons of infant mortality may be due to variations in classification of live births; it seems likely that an event that would be classified in Cuba as a stillbirth could be classified in the USA or Canada as a live birth.

But I was having a tough time finding any systematic reviews of the problem of misclassification (or variations in classification) of live birth; the definition of whether the neonate âbreathes or shows any other evidence of life - e.g. beating of the heart, pulsation of the umbilical cord or definite movement of voluntary musclesâ could depend on how the criteria are applied.

When I searched PubMed using âlive birth/epidemiology[MAJR]â as the search term, I found very little to shed light on this question. A few studies from Brazil looked at some methodological issues with the Brazilian Information System on Live Births . But most of the articles dealt with specific diseases in specific populations.

I do not want to distract you from keeping your grant proposal afloat, but I figured that there are some pretty sophisticated epidemiologists who read this blog, and that one of them has some insights into this issue.

By Ed Whitney (not verified) on 23 Feb 2010 #permalink

Hans Rosling is the man! I loved this, and showed it to the studnets in the intro biology course I teach at the local community college. It hits the right note on a bunch of levels.

Ed

Rosling is looking at this information over many years though.

As the years progress the bias is probably going to be toward a greater recording of live births over still births as the technology for keeping neonates alive improves (NICU emergency ceasarean techniques improve etc).

So if, in reality, infant mortality really isn't improving then the trend in the data would be for countries to tend to have worse apparent infant mortality.

But, as we see in the video the opposite is happening.

In the strict sense you are absolutely correct. There are always issues in comparing routinely collected data across countries with very different systems. But in this presentation the bias is probably working against the clear trend and the trend is in many many countries over 40 years.

It's very cool stuff with a very wide focus approach. If you got more focussed the problems would start to matter.

By antipodean (not verified) on 23 Feb 2010 #permalink

Thanks for the information, antipodean. I agree that the trend toward improved infant mortality will be masked if more live births are recorded, when the same events would have been previously masked as stillbirths. It was the comparison across countries that I was wondering about, though. If the infant mortalities of Cuba and the US are compared in 2005, the infant mortality of Cuba could be underestimated, creating a bias in that comparison. Rosling's looking at the trend is likely to give a conservative estimate of the improvements over time in infant survival when this one issue is taken into account. But the comparisons between national data could still be biased in a way that makes Cuban or Slovokian infant survival appear better than it might otherwise.

By Ed Whitney (not verified) on 24 Feb 2010 #permalink

Ed

I completely agree. A one-to-one comparison at a single timepoint would be sensitive to the issues you wisely list.

One of the ways you could get around this (a little) is to look at the income gradients within cuba and the USA and then compare the slopes liek Rosling does. These health gradients are why the USA has such an appalling health record compared to its wealth. The poor people are not cared for and thus the gradient is very steep compared with other developed nations and this is why it ranks behind Cuba in many measures.

By antipodean (not verified) on 24 Feb 2010 #permalink