Sun, Jun 24, 2018 - Page 6 News List

Out-of-context statistics mislead

By Emilio Venezian

The front page article “Cancer remains top cause as deaths hit record high” (Taipei Times, June 16) appears to have been designed to demonstrate the general incompetence of the Ministry of Health and Welfare and, in particular, of its Statistics Department. It cites, entirely out of context, a number of “statistics” that appear intended more to raise concern than to cast light on the issues.

As a first example, should we be concerned about the highlighted fact, emphasized in the headline, that there were 48,037 cancer deaths last year, representing an increase of 277 deaths from the year before?

Even someone with no background in statistics would understand that if nothing else changes, the number of deaths in any given year would not remain static, but would vary over some range.

A reasonably trained undergraduate could use a spreadsheet to show you that in four out of five years the actual number of cases would fall between 47,676 and 48,397 — so observing 47,760 would not be surprising.

Any competent statistician would have pointed out that if the population had not changed in any respect from 2016 to last year, the standard deviation of the number of cancer deaths would be approximately 220. A difference of 277 between the two years would be well within the expected variation by chance alone.

Unfortunately, pointing that out would not have made the report worthy of a headline, much less of placement on the front page.

Even worse, the Ministry of the Interior’s population projections for 2015 (medium variant) showed a total population of 23,546,946 for 2016 and of 23,595,450 for last year. If the mortality rate had remained the same in the two years, the change in population would have implied an expected increase of 99 cancer deaths from 2016 to last year.

It follows that more than one-third of the change in deaths between the two years might have resulted from change in the total population, reducing the expected difference from the reported 277 to about 179, less than the standard deviation. Should that rate any space in a newspaper?

That is not all. As has been written again and again over the years, Taiwan’s population is aging, and aging tends to increase average mortality, even when age-specific mortality is stable. That is why responsible statistical analysis of health statistics is commonly based on gender and age-standardized morbidity and mortality.

The article does not indicate that any standardization was considered and does not provide much information about age effects. Perhaps more was said at the conference at which the data were presented or might have been elicited by an interested spectator, but that is not available to me.

However, there is some information that might be worth considering.

The article points out that 123,543 of last year’s total deaths were of people older than 64. As that was 1,287 more than in 2016, the total deaths of people older than 64 in 2016 equals 122,256.

Using once again the Ministry of the Interior’s projections, the population total of those older than 64 was 3,108,154 for 2016 and 3,270,187 for last year. That implies average mortalities for the age group older than 64 of 3,933 in 2016 and 3,778 last year. That is a reduction in mortality of 3.95 percent. That might be newsworthy, but does not share the focus that the presented data seem to have.

Comments will be moderated. Keep comments relevant to the article. Remarks containing abusive and obscene language, personal attacks of any kind or promotion will be removed and the user banned. Final decision will be at the discretion of the Taipei Times.

TOP top