Temporal Variation in Adolescent Suicide Attempts
The authors wish to acknowledge the Queen's Medical Center and Kapiolani Medical Center, Honolulu, HI, for their assistance with medical records. They also wish to thank the UH Department of Psychiatry faculty for their suggestions; Deborah Nakamura, RN, for her assistance with chart reviews and data coding; and Gary Belcher, MEd, for his helpful comments.
Abstract
The purpose of this study was to examine the association between adolescent suicide attempts and temporal cycles in a multiethnic population. Medical records at two hospitals in the state of Hawaii were reviewed for all adolescents (ages 12–18) admitted for suicide attempts during the years 1987–1991. Results showed 296 adolescents attempted suicide, and as hypothesized, temporal factors were associated with the attempts. A significant increase in attempts was found during the afternoon/evening and on Mondays and Tuesdays, transition points between home and school. The results are discussed and implications for delivery and temporal theory development considered.
Adolescent suicide continues to present a significant health problem in the United States and current research is directed toward understanding the multiple risk factors involved (Pfeffer et al., 1991). In the evaluation of attempted and completed suicides, precipitating emotional, social, and cognitive factors are often identified but the less salient temporal factors may be overlooked. The aim of this study is to investigate the relationship between temporal cycles and adolescent suicide attempts.
The majority of temporal variation studies have focused on trends such as seasonal fluctuations of suicide rates in the adult population (MacMahon, 1983; Wenz, 1977). Only a few studies have addressed adolescent suicide completions (Frank & Lester, 1988; Phillips & Wills, 1987), and even fewer have specifically focused on adolescent suicide attempts (Christoffel, Marcus, Sagerman, & Bennett, 1988). In general, the findings have suggested that suicide rates tend to be affected by temporal cycles in consistent ways.
For example, in a recent regional study in the United States, adolescents have demonstrated more symptoms of mood and behavioral problems during the winter, particularly in northern climates (Carskadon & Acebo, 1993). Christoffel et al. (1988) have found that adolescent suicidal behavior in the Chicago, Illinois, area occurred more often in the winter months. A similar study conducted in the Toronto, Canada, area found that 30% of the adolescents attempted suicide in the winter months of January, February, and March (Garfinkel, Froese, & Hood, 1982).
In addition to seasonal cycles in suicide, some studies have examined variation over days of the week. The general population in the United States has shown a peak in suicide completions on Mondays, with a decline throughout the week to a trough on Saturday or Sunday (MacMahon, 1983). Adolescents and young adults in the. United States have shown a peak in completed suicide on Sundays (Frank. & Lester, 1988).
Suicide rates and special times of the year such as holidays and birthdays have also been looked at, but the results are inconsistent. One study with adolescents has shown that suicidal behavior was associated with the holiday times of Thanksgiving and between December 18, and January 7 (Christoffel et al., 1988), whereas another has shown no increase or a decrease in suicide around holidays (Phillips & Wills, 1987). Also, birthdays have shown an association with the timing of suicides for adolescents and the elderly, but not for adults in between these ages (Barraclough & Shepherd, 1976; Shaffer, 1974). In summary, the incidence of suicide attempts by day of the week and time of the day for adolescents in the United States has not been examined adequately, even though the potential importance of this risk factor for prevention and service delivery distribution seems critical.
Given the rapid increase in adolescent suicide rates during the last three decades, it is important to differentiate adolescents from other age groups and to determine whether certain risk factors are consistently associated with suicidal behavior in different subgroups of adolescent populations (Pfeffer, Plutchik, Mizruchi, & Lipkins, 1986). This study examines whether temporal cycles such as seasons, days of the week, times of the day, holidays, and birthdays are associated with adolescent suicide attempts in a large metropolitan area, Honolulu, Hawaii. The authors hypothesized that: (1) Mondays would be the peak day of the week for suicide attempts. (2) Seasonality would not be related to suicide attempts, particularly since there are no seasonal weather changes in the study area and the climate is temperate all year round. (3) Birthdays would be associated with suicide attempts.
METHOD
Subjects and Procedure
After obtaining institutional permission, all medical records from two major hospitals in Honolulu, Hawaii, were examined for adolescent patients (ages 12–18) admitted for a suicide attempt to either a pediatric unit, a psychiatric unit, or a medical unit during the years 1987–1991. The two hospitals represent the largest general and child and adolescent civilian medical centers in the state, and the major emergency rooms to which suicide attempts present for admission.
A total of 225 adolescents were admitted to the children's hospital and 71 adolescents to the general hospital for suicide attempts over the 5-year study period, providing a total sample of N = 296. Chart review revealed that 17 of these adolescents had been admitted to the hospital more than once for a suicide attempt during the years 1987–1991. For these 17 individuals, only information about their first attempt was included in this study so that the statistical assumption of independent observations would not be violated. Furthermore, chart review did not indicate any suicide “pacts” in this sample. Chi-square analyses, with Bonferonni correction, were performed on the data.
The majority of attempts were made by females (77%), and 23% were made by males (χ2 = 86.48, df = 1, p < .05). The mean age of all adolescents was 15.4 years, and males were slightly older than females (male M = 15.8, SD = 1.7; female M = 15.3, SD = 1.5; t = 2.35, p < .05). The largest percentage of attempters were of Hawaiian or part-Hawaiian ethnic background (20%), followed by Japanese (16%) and Filipino (15%) groups. The remaining ethnic distribution of the sample was: mixed cosmopolitan – 12%, Eurasian – 9%, Caucasian – 8%, Pacific Islander – 4%, Asian/Asian mix (e.g., Chinese-Korean) – 4%, Chinese – 2%, Korean – 2%, other – 4%, and unknown – 5%. The 1990 census data for ethnic distribution on the island of Oahu (where this study was conducted) identifies Caucasians as approximately 25% of the population, Japanese as 21%, Hawaiian/part-Hawaiian as 16%, and Filipino as 11% (Hawaii Department of Business, 1993).
Though socioeconomic status could not be computed from information in the records about education and income, a rough distribution could be determined and was representative of the population at large. Most of the adolescents or their parents had medical insurance (68%), 17% had Medicaid insurance, and 11% had no insurance or were “self-paying.” Insurance status was unavailable for 4% of the adolescents. More than half (56%) of the adolescents attended public school, 14% attended private school, 4% were in college, 2% went to an alternative school, and 10% were not in school. For 15% of the sample the type of school could not be determined. Although the clinical variables were not a focus in this study, a summary of the primary clinical characteristics of the sample is provided in Table 1.
| Females (n = 228) | Males (n = 68) | Total (N = 296) | ||||
|---|---|---|---|---|---|---|
| Characteristic | % | (n) | % | (n) | % | (N) |
| Method of attempt | ||||||
| Drug overdose | 87 | (198) | 78 | (53) | 85 | (251) |
| Laceration | 4 | (8) | 3 | (2) | 3 | (10) |
| Jump from height | <1 | (1) | 9 | (6) | 2 | (7) |
| Combination/other | 9 | (21) | 10 | (7) | 9 | (28) |
| Discharge diagnosis | ||||||
| Adjustment disorder | 40 | (92) | 35 | (24) | 39 | (116) |
| Affective disorder | 11 | (25) | 9 | (6) | 10 | (31) |
| Disruptive disorder | 1 | (3) | 7 | (5) | 3 | (8) |
| Thought disorder | 1 | (2) | 9 | (6) | 3 | (8) |
| Combination/other | 23 | (51) | 21 | (14) | 22 | (65) |
| None | 24 | (55) | 19 | (13) | 23 | (68) |
| Self-reported drug use | ||||||
| Alcohol | 11 | (24) | 10 | (7) | 10 | (31) |
| Marijuana | 6 | (13) | 3 | (2) | 5 | (15) |
| Amphetamines | 3 | (7) | 2 | (1) | 3 | (8) |
| Other | 5 | (12) | 13 | (9) | 7 | (21) |
| Polydrug use | 13 | (30) | 18 | (12) | 14 | (42) |
RESULTS
There was no significant difference in the frequency of attempts per year during the 5 years of the study (χ2 = 7.75, df = 4, p < .10). In combining the data for the years 1987–1991, we see that one third of the attempts occurred during the three months of January (12%), February (8%), and March (10%). January was the overall highest month for attempts and July was the second highest month for attempts. The number of attempts declined throughout the year, with fewest number of attempts occurring in December. There was no statistically significant difference in number of attempts between the months (χ2 = 16.55, df= 11, p < .12).
Also, there was no significant seasonal variation in suicide attempts over the 5-year period (χ2 = 1.62, df = 3, p < .65), or for each individual year (χ2 = 8.48, df = 12, p < .75). Attempts occurring in the spring (March 20-June 20) accounted for 26% of the sample, 23% occurred during the summer (June 21-September 21), 24% in autumn (September 22-December 20), and 27% in winter (December 21-March 19). There was no significant difference between males and females for any of the temporal variables.
Of the 296 adolescents, 53 (18%) attempted suicide on Monday and 62 (21%) on Tuesday, both of which were higher than chance expectations (χ2 = 17.82, df = 6, p < .05; see Fig. 1). An accurate assessment for the time of day that suicide attempts occurred was available for 87% of the sample (158/296). There was a significant variation in the time of day, with the majority of attempts occurring during the afternoon/evening rather than the morning (χ2 = 89.67, df = 23, p < .05). Peak times for attempts were 4 p.m. and 8 p.m. (see Fig. 2). After 10 p.m. there was a decline in attempts, with the fewest occurring at 4 a.m. Figure 3 shows a 168-hour plot of the suicide attempts (i.e., every hour of each day of the week) for the years 1987–1991 combined (N = 258).
Percentage of adolescent suicide attempts by day of week (n = 295).
Percentage of adolescent suicide attempts by time of day (n = 258).
Incidence of adolescent suicide atttempts: a 168-hour plot for the years 1987–1991 (n = 258).
The day before, the day of, and the day after major holidays (New Year's, Valentine's Day, Easter, Memorial Day, Labor Day, Thanksgiving, and Christmas) were each examined separately and revealed no deviation in suicide attempts from the approximately one attempt per day that would be expected by chance. Easter and Christmas showed somewhat less than expected rates, though not significant, with no attempts occurring in the three-day period surrounding these holidays.
An increase in attempts occurred right after the New Year's holiday though. The week prior to Christmas (December 21–24), the week of Christmas (December 25–31), and the week of New Year's (January 1–7) were examined and revealed only one attempt occurring the week before Christmas, one attempt the week of Christmas, and 12 attempts during the first week of the New Year's (χ2 = 17.28, df = 2, p < .05).
The day of suicide attempt in relation to the adolescents' birthdays showed that only one individual attempted suicide on his birthday. Ten percent of the sample attempted suicide within 30 days prior to their birthday (9% expected by chance). Four percent attempted within 30 days following their birthday, slightly less than expected by chance alone.
DISCUSSION
The main purpose of this study was to examine whether temporal cycles are a relevant factor associated with adolescent suicide attempts in a large multiethnic population. Results showed that suicide attempt rates did not significantly vary by year, by month, by season, or by gender. Also, there was no association with holidays or birthdays. There was a significant day of the week variation, with an increase in attempts on both Monday and Tuesday, offering partial support for our first hypothesis that Monday would be the peak day for attempts. We found that the suicide rate begins to rise on Monday, peaks on Tuesday, and then declines throughout the week. Furthermore, the significant time of day findings showed that the rate climbs throughout the day and peaks after school (4 p.m. and 8 p.m.), suggesting a relationship to home and free time, or postschool-related stress, at least on the surface.
The second hypothesis, that there would be no seasonal variation in attempts, was also partially supported. Peak times for suicide attempts occurred at the beginning of the year, during the “winter” months of January, February, and March (30% of attempts), even though the site of the present study does not experience a typical “winter” season. However, the traditional four seasons, as defined by calendar divisions, and the 12 months of the year were not significantly different from chance expectations. Thus, our findings seem to parallel the ambiguous seasonal variation in light experienced in Hawaii, with mild but not significant variations in daylight hours during the winter months. This lack of clear seasonal findings in Hawaii supports the association of suicide attempts and light/dark variations found elsewhere.
Other findings in this study were a Christmas dip and a subsequent New Year's increase in attempts, but no association with other major holidays or birthdays. The nonsignificant association between the adolescents' birthdays and suicide attempts was somewhat surprising since previous literature indicates this may be a vulnerable time period for suicidal behavior in adolescents. The trends for holidays, however, have also been found in studies of completed suicides in the general population (Lester, 1985; Phillips & Wills, 1987) and in a clinical sample of psychiatric emergency room visits (Hillard, Holland, & Ramm, 1981). It may be that adolescents, even more than adults, are vulnerable during post-Christmas time, feeling let down and experiencing stress from new expectations at the beginning of the year.
Limitations of this study are mainly due to methodological weaknesses inherent to a retrospective approach. We cannot be certain about the reliability of information extracted from the records, because recording habits of individual physicians may vary substantially. The temporal variables, however, are less subject to individual interpretation and were available for the majority of the cases.
Despite these limitations, the results have several implications and may further our understanding of the relationship between suicide and temporal cycles. This study can be considered a valuable addition to the rather scarce empirical literature on temporal patterns of adolescent suicidal behavior in the United States. A strength of the study is the unique multiethnic distribution of the sample. Studies of adolescents hospitalized for suicidal behavior have typically focused on youths from middle-class Caucasian backgrounds, with few minority patients represented (Trautman, Rotheram-Borus, Dopkins, & Lewin, 1991).
In this study, adolescents of Hawaiian/part-Hawaiian and Filipino ethnic extractions were among the highest represented groups. They have also been shown to have the highest rates for adolescent suicide completions in the state of Hawaii (Tseng, Hsu, Omori, & McLaughlin, 1992). Comparing our findings with the ethnic distribution for the island of Oahu suggests that adolescents from the Hawaiian/part-Hawaiian and Filipino ethnic groups were overrepresented, whereas the Caucasian and Japanese groups were underrepresented.
Though the sample is culturally diverse, and the geographic location of the study is unique, our findings are very consistent with other temporal patterns found in the mainland United States. That is, the day of the week and time of day findings, as well as the increase in attempts at the beginning of the year, have been shown in other studies of suicidal behavior (Christoffel et al., 1988; Garfinkel et al., 1982; Maldonado & Kraus, 1991).
Also, these findings add support for the preliminary temporal theory of fluctuations in suicide, proposed by Gabennesch (1988), which suggests that suicide rates are higher at the beginnings of cycles as opposed to the ends of cycles. To elaborate, an underlying assumption of the theory is that a person's affective state can be negatively influenced by circumstances that tend to encourage the expectation for feeling better. As a new period approaches — for example, the beginning of a new year — it can raise hope or expectations for a “fresh start” in someone who is currently contemplating suicide. However, when the new period approaches and the individual experiences the disappointment of unmet expectations, it can create a sense of relatively more suffering, thus impelling the suicidal person to take action. This idea also helps to explain why the ends of cycles are characterized by a decline in attempts.
Finally, the findings may have implications for preventive and service delivery systems. Knowledge of the temporal associations with adolescent suicide attempts may be useful information for adequately staffing crisis hotlines, hospitals, and mental health services for intervention. Suicide prevention programs in the schools may also utilize information about these patterns in their educational efforts.
In conclusion, the present study furthers our understanding of temporal factors and their association with suicide attempts in adolescents, but it also reminds us that much research is still needed to understand the complexity of this major health and social concern.
