| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
POSTPONING DEATH |
From the Department of Psychiatry, Washington University School of Medicine, St. Louis, MO.
Address correspondence and reprint requests to Judith A. Skala, RN, PhD, Department of Psychiatry, Washington University School of Medicine, 4625 Lindell Blvd., Suite 420, St. Louis, MO 63108. E-mail: skalaj{at}bmc.wustl.edu
ABSTRACT
OBJECTIVE: To review the research literature on temporal variation in mortality rates around symbolically important occasions.
METHODS: Qualitative literature review.
RESULTS: Since the early 1970s, at least 18 studies have investigated whether death rates increase or decrease before, during, or after symbolically important occasions such as holidays and birthdays. Some studies but not others have found modest evidence of temporal effects. Methodological issues have raised questions about most of the positive findings. None of the studies provides any direct evidence that a psychophysiological mechanism enables people to postpone or hasten their own death.
CONCLUSION: Research over the past 3 decades has failed to provide convincing evidence that psychological phenomena such as "giving up" or "holding on" can influence the timing of death.
Key Words: data interpretation; • cause of death; • statistical; • death; • holidays; • mortality; • stress; • psychological.
INTRODUCTION
Many people, including some health care professionals, believe that we humans have an innate ability to hasten or delay our own deaths, that we can "give up" or "let go" or "hold on" as the end of life draws near. Many others believe that we have little or no control over the timing of death. This controversy arose in ancient times and continues to the present day. Both views are implied in the Old Testament: Deuteronomy 30:19 instructs us that "... I have set before you life and death, blessing and cursing: therefore choose life," but Ecclesiastes 3:1 to 2 tells us that "To every thing there is a season, and a time to every purpose under heaven: a time to be born, and a time to die... " The ancient Roman philosopher Seneca wrote, "Just as I shall select my ship when I am about to go on a voyage, or my house when I propose to take a residence, so I shall choose my death when I am about to depart from life." Nearly 2000 years later, Sigmund Freud proposed in Civilization and Its Discontents that we are endowed with Thanatos, a "death instinct" that drives us toward death. A few decades later, novelist and playwright Thornton Wilder wrote in The Eighth Day, "We do not choose the day of our birth nor may we choose the day of our death, yet choice is the sovereign faculty of the mind." Over the centuries, perhaps no one has portrayed either possibility more vividly than did Chief Dan George of the Coast Salish tribe of British Columbia in the 1970 film Little Big Man. After delivering his famous line, "It is a good day to die," the old chief goes up to the top of a hill, expecting to die forthwith, but he is merely doused by a passing rainstorm and decides with a shrug that perhaps it is not such a good day to die after all.
This issue was not informed by scientific research until the 1970s, when studies of temporal variability in mortality began to appear. Many of these studies have attempted to determine whether death rates dip or peak before, during, or after symbolically important occasions such as religious holidays or birthdays. This line of research has been controversial almost since its inception (1). Critics have asserted that some of the temporal patterns that have been reported were due to statistical artifacts, and they have questioned whether psychological processes explain them.
There are a number of different reasons why temporal variations in mortality rates might occur, and most of them have nothing to do with volitional control over the timing of one’s own death. For example, patients with certain kinds of acute, life-threatening conditions appear to be more likely to die in the hospital if admitted on a weekend than on a weekday, perhaps because hospital staffing levels are usually lower on weekends than on weekdays (2). Some of the reasons do involve psychological processes, but fairly mundane ones. For example, the mortality rate in the United States routinely increases during the first week of the month, a time when many paychecks are used to pay for substance abuse which, in turn, leads to an increase in deaths due to suicide, homicide, and accidents (3). Others show that psychological factors can indeed influence the timing of death, but not via volitional psychosomatic processes. For example, the 1994 Northridge earthquake triggered a sharp increase in sudden cardiac deaths, apparently due to severe emotional stress (4).
Research on psychosocial risk factors for mortality is one of the most active areas of investigation in psychosomatic medicine, and mechanistic studies are an important part of this effort. One of the barriers to mechanistic research is that in many studies, months, years, or even decades elapse between the measurement of the psychosocial factors of interest and the deaths of the affected members of the cohort (eg, 5). In contrast, most studies of temporal variation in mortality rates around symbolically important occasions seek effects that are apparent over much shorter intervals, ie, days or weeks. For example, an investigator who hypothesizes that people who are medically ill are able to briefly postpone death might look for a decrease in deaths during the day, week, or month preceding an important event. Psychosomatic processes that operate over such short time intervals might be more amenable to systematic investigation than ones that unfold over much longer periods. Thus, research on mortality rates around symbolically important occasions could potentially help to clarify how psychological processes might influence survival.
This paper will briefly review, in chronological order of publication, the research literature on temporal variation in mortality rates. It will also consider whether there is any evidence that medically ill humans have volitional control over the timing of their own death via psychosomatic processes. The studies are summarized in Table 1.
|
A 1975 study of 255,164 deaths of people aged 65 or over in England and Wales (8) found excess death rates during, and for the 3 months following, the birth month among people older than 75 within each of eight strata defined by gender and marital status. This statistically significant excess accounted for 1% of all deaths for the year. No effect was found for people 65 to 74 years of age. The author suggested that the excess deaths in the older sample might be due to psychological stress, but no stress-related data were available.
These two studies were critiqued by a number of authors. Some tried but failed to replicate the results, and some questioned the statistical methods used in the studies. Roger (9) reanalyzed Alderson’s (8) data with a correction for systematic trends due to increasing death rates and dwindling age-specified populations, and found "... little evidence for an increased rate in mortality after birthdate in the aged" (p. 65). Schulz and Bazerman (1) reanalyzed the data from the Phillips and Feldman study (6, 7) and also closely inspected the data from Alderson’s study (8). They concluded that the evidence provided in support of the death dip-rise phenomenon around important ceremonial occasions "... did not withstand close scrutiny and/or appropriate reanalysis" (1, p. 259).
Judith Tanur, the editor of the volume in which Phillip’s and Feldman’s work was first published (6), responded to Schulz and Bazerman’s critique. She encountered difficulties in reconstructing some of the data and also found that the pre-birthday death dip disappeared when a 31-day interval was used instead of calendar month. She concluded that "One wonders what flaws might be found if similar reanalyses were undertaken of less startling social science findings based on improbable distributions of events over categories chosen by the analyst" (10, p. 315). However, Wood and Cho (11, 12) replicated Alderson’s findings and concluded that the results were even stronger in the subgroup that died of heart disease. They also found differences among groups defined by gender and marital status and concluded that "Perhaps for older married people the birthday is an event to strive to survive (some who try don’t quite make it), and for middle-aged married females the birthday is an event to avoid" (12, p. 326).
During the 1980s, seven original studies of widely differing populations were published. Rago et al. (13) reported a flat mortality distribution among 463 institutionalized, profoundly mentally retarded (PMR) people. Their rationale for choosing this group was to remove psychological variables, such as anticipation of holidays, as potential explanations for variations in death rates. They also found that the variance of the PMR group did not differ from the famous Americans group, as reanalyzed by Schulz and Bazerman (1). However, the mean age at death for the PMR group was 43 years, well below the age at which Alderson (8) found significant effects. In a 1986 study of 212 suicides, 207 natural deaths, and 353 homicides in Philadelphia, Lester (14) found chance or close-to-chance rates of mortality in each of the three categories. Also in 1986, Zusne (15) examined birth and death dates of various groups differing with respect to age, gender, profession, eminence, and fame. The author concluded that whereas in men, the probability of death drops before the birthday and increases afterward, the opposite pattern occurs in women. By Zusne’s calculations (15), the cycle runs 36 days pre and post for men and 65 days pre and post for women. It is difficult to evaluate the validity of Zusne’s conclusions because the data are presented in graphical but not in tabular form, because the p values for statistical tests are presented without the statistical values on which they are based, and because the time windows for aggregating data points were not specified a priori.
In 1987, Byers and Zeller (16) found a 20% post-Christmas death rise in natural deaths in Ohio for the years 1979 through 1981, but they did not detect a corresponding pre-Christmas dip in mortality. However, close examination of the reported data reveals that the rise in death rate actually began during the 5 days before Christmas and that the death rate on Christmas itself was higher than the average rate for the preceding and following weeks. They also examined subgroups defined by age, gender race/ethnicity, marital status, and cause of death. They found larger increases for particular subgroups in each category. Examination of a wide variety of covariates failed to explain these results. The authors did, however, determine that weather was not a factor. A 1987 study by Angermeyer et al. (17) examined numerous biographies of famous people from around the world and of well-known Swiss citizens. They did not find a death dip or rise around the time of the birthday despite having very large samples at their disposal. They speculated that they may have been unable to detect an effect because many of the people in their sample were scientists who might be less committed than other people to social conventions and ceremonies.
A 1988 study by Phillips and King (18) found that males with "Jewish surnames" were more likely to show a 4% death dip before Passover and a corresponding 4% rise afterward compared with Jewish women, Chinese-Americans, and Japanese-Americans in Los Angeles. The authors provided a number of questionable, post hoc sociological explanations for the findings, eg, "Women play only a peripheral role in the religious rituals associated with Passover" (18, p. 730). They mentioned "psychosomatic processes" without further elaboration. This publication prompted a number of letters to the editor. Some of them proposed physiological mechanisms such as dietary change to explain the findings; others disputed certain explanations for the effect on males and not females. Of note is a letter from Emmanuel Lewis, offering an interesting alternative explanation of Phillips and King’s findings: "The Angel of Death has a habit disorder and continues to pass over observing Jewish households. My hypothesis also explains the sparing effect on males and not females" (19, p. 1030).
In a 1989 study of 4800 psychiatric patients at a Houston area Veteran’s Administration medical center (20), Greiner and Pokorny showed that the accident rate rose slightly before birthdays and that there was a rise-dip pattern of suicides around Christmas. They did not find changes in rates of death due to natural causes.
In the 1990s, Phillips and various co-authors accounted for three of six major publications in this area of inquiry. In 1990, Phillips and Smith (21) reported a 35% dip-rise mortality pattern among elderly Chinese-Americans around the Harvest Moon Festival but no such pattern among young Chinese-Americans or among Jews in California. In attempting to explain the failure to find effects among younger Chinese-Americans and older Chinese men, the authors noted that the Harvest Moon Festival "... places great emphasis on the symbolic importance of old women... indeed, the festival is sometimes disparaged as ‘an old woman’s holiday’" (21, p. 1948). Also in 1990, Walker and Walker (22) attempted to extend some of Phillips’ earlier findings by conducting a study of Roman Catholic clergy born between 1830 and 1860. They found that the date of death was not statistically associated with the person’s birthday or anniversary of ordination, or with Christmas or Easter.
In 1992, Idler and Kasl (23) reported several findings pertaining to observance of religious holidays. Unlike most of the other studies, deaths on the holidays of interest were included in the pre-holiday rather than post-holiday numbers. Male and more-observant Jews showed a dip-rise phenomenon around Passover and Yom Kippur but not around Rosh Hashanah, and Christians showed dip-rise effects around Christmas but not Easter. Subgroup analysis showed that the latter effect was evident primarily among Catholics, especially the more observant. Although the numbers of deaths before and after each holiday were presented, the subgroup sample sizes were not. Consequently, the effects sizes cannot be determined from the article.
In another 1992 study, Phillips et al. (24) reported on Californians aged 18 and over who died of natural causes. Men showed only an unspecified pre-birthday spike. Women showed an unspecified pre-birthday dip with a 3% post-birthday spike. This contrasts with Zusne’s (15) finding (discussed above) of a pre-birthday rise and post-birthday dip among women. The following year, Phillips et al. (25) compared 28,169 Chinese-Americans in California to 412,632 "matched controls." They found that Chinese-Americans who had a disease associated with their birth year on the Chinese astrological calendar tended to die approximately 1 year younger than expected compared with controls with similar diseases. Whereas Idler et al. concluded that mortality dip-rise phenomena were due to some aspect of the religious experience, Phillips et al. proposed unspecified psychosomatic processes as being responsible for their findings.
In 1999, Neumann et al. (26) analyzed the deaths of 626 Canadians with cancer who were admitted to a hospital for palliative care. They found fewer deaths during the evening/night hours but no significant effect for birth date. They speculated that the day/night phenomenon might be due either to circadian effects or to fear of dying alone during the night.
More studies appeared in 2000 and 2001, with interesting and confusing findings. Lee and Smith (27) studied a sample of Jews who had received funeral services at Jewish mortuaries. In contrast to Phillips and King’s (18) findings, Lee and Smith (27) found that males were more likely to die before Passover (11% rise before, 11% dip after). Males were not more likely to die before other holidays, and in the sample as a whole, there were spikes in mortality within a few weeks on both sides of the birthday. The authors speculated that this might be due to the effects of stress but also acknowledged that the findings might represent Type I errors because the report was based on a large number of statistical analyses. Anson and Anson (28) again studied this phenomenon among Jews, this time in Israel, and found a dip-rise pattern of approximately one death per day, on average, around the Sabbath among men of all ages, but not among women less than 75 years of age. Interestingly, they did not find a dip-rise pattern around the major holy days or similar patterns among Jewish children or non-Jewish Israelis. These authors posited more mechanisms than most, but all of them were nonspecific. They suggested that altruistic activity might be protective for women before Sabbath, yet elsewhere they speculate that a Wednesday rise in deaths among younger Israeli women might be due to the burden of preparing for Sabbath. They further suggested that death dips may be due to augmentation of coping resources by social activity.
Finally, in 2001, Phillips et al. (29) reported another culturally determined mortality phenomenon among Chinese and Japanese Americans. They found a 7% increase in cardiac deaths in these groups on the 4th of every month, compared with the 3 days before and after the 4th day of the month. They suggested that this is due to the stress associated with the number "4," which is considered to be so unlucky in these cultures that many Chinese-Americans and Japanese-Americans in California take steps to avoid having it in their addresses and telephone numbers. The study did not address whether death peaks are associated with any other unlucky numbers (eg, the number 7) or whether death dips are associated with any lucky numbers (eg, the number 8).
In conclusion, this review spans a 30-year history of research on temporal variability in mortality rates around symbolically important occasions. This research has spawned debates about larger questions such as the relationship between mind and mortality and the elusive nature of psychosomatic processes. The studies published to date have not convincingly established that death can be postponed through force of will or hastened by the loss of the desire to live. Modest effects have been found in some subgroups but not others, around some occasions but not others, and by some investigators but not others, and critics have raised methodological concerns about many of the studies. This research has also produced very little information about any specific biopsychosocial mechanisms that might account for temporal variations in mortality around symbolically important occasions. When speculating about possible mechanisms, the authors of these studies have tended to cite research on stress reactions and on the protections afforded by social support. However, they have not provided any data on these factors.
It seems unlikely that these questions will ever be resolved by retrospective analyses of large population datasets, especially ones that contain little or no information about psychosocial characteristics. It is possible that they might be resolved by prospectively studying groups of patients with similar terminal illnesses. In such studies, the personal meanings attached to symbolic occasions such as birthdays and religious holidays, and the strength of the patient’s desire to survive and experience these events, could be ascertained. It would be essential to specify a priori hypotheses rather than relying on post hoc speculations to explain any phenomena that might be found. It would also be very desirable to collect data pertaining to any hypothesized psychosomatic mechanisms.
Perhaps future studies will yield more convincing results and will suggest mechanisms that are both plausible and amenable to systematic investigation. Until then, speculation will undoubtedly continue about the differences between people who seem to lose steam near the end of a race and those who appear to find one last, impossible spurt of strength within themselves; between those who are more physiologically reactive to internal beliefs and those who are more reactive to external conditions; and between those who struggle to live while others seem to give up on life and drift into death.
Received for publication July 29, 2003.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  J. Rodriquez Attributions of Agency and the Construction of Moral Order: Dementia, Death, and Dignity in Nursing-home Care Social Psychology Quarterly, June 1, 2009; 72(2): 165 - 179. [Abstract] [PDF]
|
|
|
|
 |
|
 D. C. Young and E. M. Hade Holidays, Birthdays, and Postponement of Cancer Death JAMA, December 22, 2004; 292(24): 3012 - 3016. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. D. Lane and R. D. Lane LETTERS TO THE EDITOR: POSTPONING DEATH: ANOTHER FAILURE TO REPLICATE Psychosom Med, November 1, 2004; 66(6): 973 - 974. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
