Simpson’s Paradox and U.S. Infant Mortality

by Clive Riddle, December 3, 2009

There’s a little tingle you feel when you encounter a term for a phenomenon you’ve previously observed. Maybe not Spiderman’s “spider-sense” tingle, but a tingle none the less. So, a tingling I did detect when reading yesterday’s  Wall Street Journal article, “When Combined Data Reveal the Flaw of Averages.” There I learned about Simpson’s Paradox. Surely I came across it countless times in statistics classes and conferences, but my eyes must have been glazed over.

In non-technical terms, as the WSJ article puts it, “Simpson's Paradox reveals that aggregated data can appear to reverse important trends in the numbers being combined….’It's the magic of weighted averages,"’ says Princeton University economics professor Henry Farber.”

The WSJ article cites Simpson’s Paradox impacting current unemployment figures, baseball comparisons between Derek Jeter and David Justice, UC Berkeley admission rates and a study of Kidney Stone treatments.

For me, it brought déjà vu regarding a report CDC's National Center for Health Statistics released last month: "Behind international rankings of infant mortality: How the United States compares with Europe."

Consistent with numerous studies indicating the U.S. compares unfavorably with Europe regarding various key health outcomes, this report indicates the U.S. infant mortality rate lags behind our European counterparts, and in fact ranks 30^th in the world. As indicated in the table below compiled from the study, the U.S. incurs 6.9 infant deaths per 1,000 live births, worse than all but one country cited below and far off the mark from a rate of 2.4 in Sweden.

So what’s really going on behind the numbers? If one digs one level deeper, one would attribute the higher overall U.S. numbers to a much higher preterm birth rate, given that pre-term births globally experience a higher mortality rate. Stopping there, the conclusion would be to focus solely on reducing the preterm birth rate (an admirable goal.)

Fair enough, the U.S overall infant mortality rate would decline if the preterm birth rate declined. But dig another level deeper, into morality rates by gestation, and one will see that the U.S. mortality rate is highest for full term births, but compares much more favorably with premature births. Thus if the U.S. lowers its preterm birth rate, it won’t improve in overall comparisons, because of unfavorable experience with full term births, which comprise close to 88% of all U.S. births. Thus focus needs to be given on the mortality for our full term cases.

	Infant Mortality Rates (Deaths/1,000 Live Births)
Country	%Preterm	Overall	22 wks+	24-27 wks	28-31 wks	32-36 wks	37 wks+
U.S.	12.4%	6.9	5.8	236.9	45.0	8.6	2.4
Austria	11.4%	4.2	4.1	319.6	43.8	5.8	1.5
Czech Republic	7.0%	3.4	3.7
Denmark	6.9%	4.4	4.4	301.2	42.2	10.3	2.3
England	7.5%	5.0	4.9	298.2	52.2	10.6	1.8
Finland	5.6%	3.0	3.4	315.8	58.5	9.7	1.4
France	6.3%	3.6	3.9
Germany	8.9%	3.9	4.1
Greece	6.0%	3.8	4.0
Hungary	8.6%	6.2	6.6
Ireland	5.5%	4.0	4.6
Italy	6.8%	4.7	4.0
N. Ireland	6.6%	6.3	4.0	268.3	54.5	13.1	1.6
Netherland	7.4%	4.9	4.6
Norway	7.1%	3.1	3.0	220.2	56.4	7.2	1.5
Poland	6.8%	6.4	6.8	530.6	147.7	23.1	2.3
Portugal	6.8%	3.5	3.9
Scotland	7.6%	5.2	4.9	377.0	60.8	8.8	1.7
Slovakia	6.3%	7.2	6.7
Spain	8.0%	4.1	4.0
Sweden	6.3%	2.4	3.0	197.7	41.3	12.8	1.5

Comments

[h1n1]

Such a usefule blog…wow !!!!