Scared to Death?
Gary Smith
Department of Economics
Pomona College
Claremont, California 91711
gsmith@pomona.edu
909.607.3135
Fax 909.621.8576
Phillips, et al, report a Hound of the Baskervilles
effect among Chinese- and Japanese-Americans who have abnormally high
cardiac mortality on the fourth day of the month because they think the number
4 is unlucky.1 Is it plausible that the number 4 is regarded with
the same terror that the fictional Charles Baskerville felt as he was pursued
in the dark by a savage dog with its flaming jaws and blazing eyes?
We encounter the number 4 every day. Could one more 4, the fourth day of a month,
be fatal?
Methods and Results
A natural test is a comparison of deaths on the third, fourth,
and fifth days of each month. A longer horizon is flawed because a systematic
cycle that causes deaths to peak near the fourth may show statistical significance
that reflects a bulge near the fourth rather than any special significance of
the fourth day itself.
Californias computerized mortality records include the racial
status recorded on death certificates. The first three data rows of the table
show California deaths caused by heart disease during the years 1989-1998 used
by Phillips. Of the Chinese- and Japanese-American deaths on days 3, 4, and
5, 33.9% occurred on day 4; the two-sided p value is 0.652. Comparing Chinese/Japanese
deaths with white deaths, the two-sided p value is 0.664.
Phillips argues that recorded death dates are more reliable for
inpatient data, which have lower p values. The 1989-1998 data use the ninth
revision of the International Classification of Diseases (ICD). In his studies
of Jewish deaths near Passover2 and Chinese deaths near the Harvest
Moon Festival,3 Phillips uses all heart-disease categories (390-398,
402, 404-429). In a study of Chinese-Americans with unlucky birth years,4
Phillips divides heart diseases into these three groups: (a) 410; (b) 412, 414;
and (c) 390-398, 402-409, 411, 413, 415-429. In the Baskervilles paper, Phillips
only considers 410-414, which he calls chronic heart diseases. These
categories have more deaths on day 4 than on days 3 or 5; the categories he
omits do not.
One way to check data dredging is to look at fresh data. We now
have data for 1999-2001; we also have 1969-1988 data that Phillips has used
in his other studies. (Pre-1969 data use ICD-7 classifications that are not
directly comparable to the later revisions.) The table shows no statistically
persuasive peaks on day 4 for total coronary deaths, inpatients, or Phillips
subset of heart diseases. In the 1969-1988 data, there are more deaths on day
5 than on day 4; in the 1999-2001 data, there are more deaths on day 3.
Newcombe notes that if day 4 is sufficiently stressful to affect
cardiac mortality, it should also affect suicides and accidental deaths.5
Phillips response (We were not interested in testing the unsurprising
hypothesis that psychological stress might trigger suicide.) indicates
that he agrees. The table shows that Chinese- and Japanese-Americans actually
had relatively few suicides on day 4, which casts doubt on the theory that day
4 is fatally stressful.
If the number 4 affected mortality, we should find similar effects
with other lucky and unlucky numbers. In Chinese culture, the number 7 can signify
death, while white Americans generally consider 7 to be a lucky number; white
Americans consider 13 to be unlucky; the luckiest Chinese number is 8. Fishers
test was applied to cardiac mortality on all days using data for the entire
period 1969-2001. None of the suggested days (4, 7, 8, or 13) are statistically
persuasive.
Comment
It seems implausible that Chinese- and Japanese-Americans are scared to death by the number 4. California death records do not show statistically persuasive patterns unless several types of cardiac mortality are omitted. The elevated mortality in this pruned subset of the 1989-1998 data is not present in 1969-1988 or 1999-2001 data. Nor are there statistically persuasive patterns in suicides or accidental deaths or on other lucky or unlucky days.
References
1. Phillips DP, Liu GC, Kwok K, Jarvinen JR, Zhang W, Abramson IS. The Hound of the Baskervilles effect: natural experiment on the influence of psychological stress on timing of death. British Medical Journal 2001; 323: 1443-1446.
2. Phillips DP, King EW. Death takes a holiday: Mortality surrounding major social occasions. The Lancet 1988; 2: 728732.
3. Phillips DP, Smith DG. Postponement of death until symbolically meaningful occasions. JAMA 1990; 263: 194751.
4. Phillips DP, Ruth TE, Wagner LM. Psychology and survival, The Lancet 1993; 342: 1142-1145.
5. Newcombe RG. Bad4U? British Medical Journal letters 2002; 324: 1098.
Table Day 4 Deaths as a Percentage of Deaths on Days 3, 4, and 5. The number of deaths on days 3, 4, and 5 and are shown in parentheses.
Day 4 Deaths (n)
|
two-sided p values
|
|||
White
|
Chinese/Japanese
|
Binomial*
|
Fisher†
|
|
1989-1998 | ||||
Total coronary deaths |
33.3 (60,839)
|
33.9 (1,391)
|
.652
|
.664
|
Inpatients |
33.8 (20,658)
|
36.5 (540)
|
.133
|
.211
|
Inpatients, chronic§ |
33.8 (15,084)
|
38.9 (404)
|
.022
|
.040
|
1999-2001 | ||||
Total coronary deaths |
33.2 (18,598)
|
30.9 (505)
|
.264
|
.293
|
Inpatients |
32.6 (6,061)
|
31.8 (195)
|
.710
|
.873
|
Inpatients, chronic§ |
32.2 (4,273)
|
34.6 (133)
|
.822
|
.626
|
1969-1988 | ||||
Total coronary deaths |
33.4 (120,301)
|
34.9 (992)
|
.317
|
.353
|
Chronic§ |
33.5 (98,257)
|
35.0 (734)
|
.353
|
.399
|
1969-2001 | ||||
Suicides |
33.5 (11,150)
|
27.8 (198)
|
.110
|
.100
|
Accidents |
33.7 (30,508)
|
36.2 (390)
|
.259
|
.326
|
* Binomial test of the null hypothesis that for each Chinese/Japanese death
on the third, fourth, or fifth day of a month, there is a 1/3 probability it
occurs on the fourth day.
† Fishers exact test with the data separated by race and by whether the death occurred on the fourth day or on the adjacent days.
§ The heart diseases selected by Phillips are ICD-8 410-414; ICD-9 410-414; ICD-10 I20-I24, I25.1-I25.9