Objective. Suicide completion has increased in the United States among 15- to 24-year-olds from 4.5 per 100 000 (1950) to 13.2 per 100 000 (1990). Suicide attempts have been found to be associated with depression, substance use, overall number of life stressors, gender, and impulsive behaviors. We hypothesized that suicide attempt is associated with many other health risk behaviors.
Design. To evaluate this hypothesis, we analyzed the data from the 1993 Massachusetts Youth Risk Behavior Survey from a representative sample of students in grades 9 through 12 in public and private high schools. We examined the relationships between ever attempting suicide and engaging in other health risk or problem behaviors (eg, sexual activity, substance use, violence, and seat belt nonuse).
Results. The 3054 students were distributed among the 9th to 12th grades; 50.7% were female, 77.7% white, 6.7% black, 5.9% Hispanic, 4.6% Asian, .8% Native American, and 4.4% other; and the mean age was 16 ± 1.2 years. Based on logistic regression analysis, ever attempting suicide was associated with physical fights in the past 12 months (odds ratio [OR] = 1.3[95% confidence interval (CI), 1.22–1.44]), regular cigarette use in the past 30 days (2.0[1.41–2.97]), female gender (3.2[2.21–4.71]), lack of seat belt use (1.3[1.21–1.34]), gun carrying in the past 30 days (1.4[1.12–1.70]), substance use before last sexual activity (1.4[1.09–1.84]), Native American or other (2.4[1.21–4.71]), and lifetime use of other drugs (1.2[1.04–1.40]). The full model correctly assigned 92% of the youth. There were small differences in models for males, females, and sexually active youth. Same gender sexual experiences for sexually active youth did not significantly improve the model (partial r = .0, OR = .87 [95% CI, .33–2.30]) and fewer (88%) of the youth were correctly assigned by the model.
Conclusions. Ever attempting suicide is associated with other problem behaviors. This research proposes an empirical model of the relationship between suicide attempts and other health risk behaviors. Identifying which adolescents are at risk for suicide attempts from the presence of accompanying risk behaviors will help clinicians to evaluate suicide risk when other problem behaviors are identified, thus enabling the design of possible prevention strategies.
There has been an increase in completed suicides in the adolescent population in the United States. The incidence rates have increased from 4.5 per 100 000 in 1950 to 13.2 per 100 000 in 1990 in the 15- to 24-year-old age group.1 The 1993 national Centers for Disease Control and Prevention (CDC) Youth Risk Behavior Survey (YRBS) reveals that 19% of boys and 23% of girls considered suicide in the past 12 months.2 Five percent of boys and 13% of girls attempted suicide with 2.7% receiving medical treatment for any attempt.2 Some attempts may be part of a cluster of suicide attempts after an index case in a high school or the news media.3
Slap and colleagues4 found that 14.5% of youth presenting to a medical clinic reported a previous suicide attempt. Most youth who reported suicide attempts came to the clinic unaccompanied by a guardian, were female, and seeking care for office gynecology, sexually transmitted diseases, or mental health.4 Few models exist to aid health care professionals in predicting which clients are at higher risk for suicide attempts and completions and most evaluations require additional interviews or scales not routinely utilized in the primary care setting.4-6 Many authors have described the association of suicide attempts with suicidal ideation, depression, substance use,7,8 and other mental disorders.1,9,10 Life stressors or abuse,11female gender,7,9 impulsive behaviors,12 family dysfunction,7,9 hopelessness,13,14 past suicide attempts,14 and recent relationship breakups15,16 have also been identified as possible precursors for suicide attempts.
Previous studies indicate that young women demonstrate more suicidal gestures while young men have more suicide completions.7,9This is due in part to young men using more potent methods such as firearms rather than drug overdoses in their suicide attempts.7,9 Gay and bisexual adolescents have been reported to exhibit high rates of depression and have been reported to have rates of suicidal ideation and attempts three times higher than other adolescents.17-19 Adolescents may reveal risk behaviors more clearly than classical symptoms of depression or suicidal ideation.
Most authors group suicide with depression and other psychiatric indicators and not with adolescent risk behaviors1,3,7,10,15,20. Jessor21 has proposed that adolescent risk and problem behaviors co-vary and cluster to form a risk behavior syndrome because different risk behaviors may meet similar psychosocial developmental needs. Consequently, intervention and prevention programs are likely to be most effective if the interventions impact broadly on risk behaviors and enhance specific skills that are protective factors for these behaviors.21
We hypothesized that suicide attempt is associated with other risk and problem behaviors. To evaluate this hypothesis, we analyzed the data from the 1993 Massachusetts YRBS. Identifying which adolescents are at risk for suicidal attempts will help clinicians recognize patients in need of suicide prevention interventions.
Data for the study were obtained from the 1993 Massachusetts YRBS conducted between February and May 1993. Forty-five of the 51 randomly selected high schools across the state agreed to administer the survey. Approximately 70 students from five randomly chosen classrooms within each school (grades 9 to 12) participated for a total of 3054 students.22 The demographic characteristics of the students are presented in Table 1 as described previously.23 The other category includes biracial and other races and ethnicities plus .4% nonrespondents to this question.
The questionnaire consisted of 89 multiple choice questions based on a standardized YRBS instrument designed by the CDC and administered by the Department of Education in Massachusetts schools. The YRBS questionnaire was administered in both English and Spanish. The instrument was self-administered, and all participants were assured that the survey was anonymous, confidential, and voluntary. Schools had the option of obtaining parental consent for participation; fewer than 10 students' parents denied permission for their children to complete the questionnaire. Because student participation was voluntary and anonymous, schools were not required to obtain parental consent. Fewer than five of the selected students who were in school on the days that the questionnaire was administered did not complete the survey.22,23
Questions used to evaluate suicidal behaviors asked (1) whether or not the respondent had seriously considered suicide in the past 12 months, (2) how many times they had actually attempted suicide in the past 12 months, and (3) whether or not he/she had attempted suicide in the past 12 months that resulted in an injury requiring medical attention. Violence behaviors were assessed with questions about how many times he/she had been in a physical fight in the past 12 months, how many times in the previous 12 months he/she had been in a fight that resulted in an injury requiring medical attention, how many times in the past 30 days the respondent had carried a weapon (defined as a gun, knife, or club), and how many times in the past 30 days the respondent had carried a gun.
Sexual risk behaviors were measured with questions asking the number of people with whom the respondent had sexual intercourse both in the last 3 months and in his/her life, whether a condom had been used during the last intercourse, and whether the respondent had ever been told by a health care professional that he/she had a sexually transmitted disease such as genital herpes, genital warts, chlamydia, syphilis, gonorrhea, acquired immunodeficiency syndrome, or human immunodeficiency virus infection. Same gender sexual experiences were assessed by combining the student's gender with a variable that asked the persons with whom you have had sexual contact with options of male(s), female(s), male(s) and female(s), and “I have not had sexual contact with anyone.” Those with same gender sexual experiences were compared with sexually active youth who did not report same gender sexual experiences.
Other risk behavior questions addressed the number of times the student had ridden a motorcycle in the past 12 months, the frequency of having worn a helmet when riding a motorcycle in the past 12 months, and the frequency of having worn a passenger seatbelt. The number of times in the past 30 days the respondent had driven a car after substance use and had ridden as a passenger after the driver had been drinking alcohol were also measured. Tobacco, alcohol, marijuana, crack cocaine, and other drug use was assessed by whether they were ever used, age of first use, and use in the past 30 days.
Weighting procedures were used to correct for the sampling scheme in the data collection. Westat, Inc (Rockville, MD) performed the data cleaning and the computation of the weight under contract by the CDC. Weighting compensated for nonresponse and reflected the likelihood of sampling each student. Oversampling was performed to coordinate the state YRBS with the Boston school district's YRBS and weighting was used to adjust for the oversampling of Boston students.22 The weight used for estimation is given by:
W = W1*W2*f1*f2*f3
W1 = inverse of the probability of school selection
W2 = inverse of the probability of classroom selection
f1 = a school-level nonresponse adjustment factor calculated by school size
f2 = a student-level nonresponse adjustment factor calculated by the class
f3 = a poststratification adjustment factor by gender and grade.22
Investigators from the CDC and Westat, Inc recently published the results from a test-retest reliability study of the 1992 YRBS. They administered the YRBS questionnaire twice, 14 days apart, to 1679 students from grades 7 to 12.24 The authors computed a Kappa statistic for the 53 self-report items and compared group prevalence estimates from two testing occasions. The Kappas for the entire test ranged from .145 to .911, with 71.7% of the items considered to have good to excellent reliability (Kappa = .61 to 1.00). Questions concerning suicide attempts in the past 12 months had high Kappa's of .602 to .838.24 The lowest Kappa for the questions used in this study (0.602) was for suicide attempt resulting in injury which had the lowest prevalence of response (1.4% and 2.0% at time 1 and time 2, respectively).24 The anonymous and confidential administration of this survey was designed to improve the validity of the youth's self-report.
All analyses were performed on weighted data. Pairwise associations for suicidal behaviors with the other high-risk behaviors were initially examined using Pearson's χ2 analyses. The variable, “During the past 12 months, how many times did you actually attempt suicide?” had the strongest association by Phi or contingency coefficient with other risk behaviors, and a sufficient number of respondents to be analyzed further. This variable was dichotomized due to sparse cells regarding multiple attempts; the new variable grouped no attempted suicides versus one or more attempted suicides in the last 12 months. The statistical software that we employed assumes that the data were drawn from simple random samples. Because the YRBS has a complex survey design there is the possibility that the error terms that were computed could be artificially low. A P < .01 was used as the level of significance instead of P≤ .05 to reduce the possibilities of a type I error. All but one of the independent variables had probability levels of ≤.00001 (genderP = .0037). From each area of risk behavior, the one variable which was most strongly associated with suicide attempts and had a phi or contingency coefficient ≥11% was entered into a forward stepwise multiple logistic regression model using the Wald procedure. Age and gender were included in the variables in the final model as important potential confounders regardless of their lower contingency coefficients. Ordinal variables with sparse cells were recoded as dichotomous variables as having not engaged in the behavior versus one or more times. Ethnicity was dichotomized into two groups: whites, blacks, Hispanics, Asians (frequency of suicide attempts 9.2%, 10.1%, 10.8%, 9.2%, respectively) versus Native Americans and others (including mixed racial youth) (frequency 30.0%, 24.5% of suicide attempts). For variables measured on ranked scales the odds ratio (OR) reflects the change in risk of a suicide attempt associated with an increase in each rank of the behavior.
The model was performed by gender subgroups and for sexually active youth only. The impact of same gender sexual experiences was forced into the final regression model for sexually active youth as a potential confounder despite the fact that the relationship was not selected in the stepwise logistic regression model. Statistical analyses were performed using SPSS for Windows.25
In this sample of Massachusetts students, 9.4% (288) reported ever attempting suicide. Those youth who had attempted suicide one or more times in the past 12 months were more likely to participate in other risk behaviors. Table 2 shows that there were significant relationships between suicide attempt and female gender, poor class standing, lack of seat belt and helmet use, and multiple substance use, including use before driving and use before sexual activity. Suicide attempts were also associated with histories of fighting, gun use, sexually transmitted diseases, and same gender sexual experiences. The strongest relationships included cigarette, alcohol, marijuana, crack and other drug use, physical fighting, substance use before sexual activity, and gun carrying.
Forward stepwise logistic regression indicated that physical fights in the past 12 months, regular cigarette use in the past 30 days, female sex, lack of seat belt use, gun use in the past 30 days, alcohol or drug use before last sexual activity, Native American and other race/ethnic group categories, and lifetime use of other drugs were significant predictors of ever attempting suicide (Table3). In this model, 92.1% of the youth were correctly assigned. The stepwise logistic regression analyses were repeated separately for male and female genders (Table 3). The significant predictors for males were age, lack of seat belt use, carrying a gun in the past 30 days, physical fights in the past 12 months, cigarette use in the past 30 days, and injection drug use with 94.2% correctly assigned by the model. For females Native American or other race, lack of seat belt use, physical fights in the past 12 months, cigarette and smokeless tobacco use in the past 30 days, and substance use before last sexual activity were significant predictors with 90.43% correctly assigned by the model. The models for both genders included variables concerning physical fights, lack of seat belt use and substance use including tobacco products; but race, substance use before last sexual activity and smokeless tobacco use were significant predictors for females and age and injection drug use were significant predictors for males.
When we limited the study sample to sexually active youth, fewer predictor variables were significant in the stepwise logistic regression model (Table 4) and slightly fewer (88.4%) of the youth were correctly assigned by the model. For sexually active students, ethnicity, and substance use before last sexual intercourse were no longer significant predictors once the other variables were entered into the model.
In the full sample, 111 (3.6%) reported same gender sexual experiences. Because youth with same gender sexual experiences have been previously reported to be at a higher risk for suicide,12,14,15 this variable was assessed further even though it was not associated with suicide after adjusting for the other variables in the main stepwise logistic regression model. When same gender sexual experiences was forced into the model and all variables significant at this bivariate level were entered in a stepwise fashion, same gender sexual activity did not contribute to the model (partial r = .0, OR = .87 [95% confidence interval (CI), .33–2.30]) and fewer (87.8%) of the youth were correctly assigned by the model. When age, gender, and same gender sexual experiences were all driven as potential confounders into the model first, age (partial r = .0, OR = 1.0 [95% CI, .89–1.18]) and same gender sexual experiences (partialr = .0, OR = .7 [95% CI, .20–2.28]) did not contribute significantly and did not increase the number correctly assigned by the model. When the model was limited to males only, same gender sexual experience still did not contribute significantly to the model (partial r = .0, OR = 1.2 [95% CI, .17–6.90]). Same gender sexual experiences were significantly related to gun carrying (contingency coefficient = .17, P< .00001), physical fights (contingency coefficient = .11,P < .00001), lifetime other drug use (Phi = .19,P < .00001), and alcohol use in the past 30 days (contingency coefficient = .15, P < .00001) through bivariate analyses.
Ever attempting suicide in the past 12 months was associated with other risk and problem behaviors. The 9.4% ever attempting suicide in this sample is similar to the national YRBS survey.2This research proposes an empirical model of the relationship between suicide attempts and other health risk and problem behaviors. Our study showed a significant relationship between suicide attempts and other major risk and problem behaviors including sexual behavior, substance use, lack of injury protection, violence, and school standing. The main stepwise regression model (Table 3) correctly classified 92% of the study population. There were small differences between models for males and females separately: smokeless tobacco and substance abuse before last sexual activity were significant predictors for females and injection drug use for males. Restricting the population to only sexually active youth did not contribute significantly to the model. Forcing same gender sexual experiences into the model did not offer any additional contribution, in spite of the fact that adolescents who reported same gender sexual experiences were likely to report a suicide attempt in the bivariate analyses.
Jessor defines risk behavior as any behavior that can interfere with successful adolescent psychosocial development. Problem behaviors are risk behaviors that elicit either formal or informal social sanctions when performed by adolescents. Problem behaviors are often age-graded; they may be considered a problem behavior for younger adolescents, but only a risk behavior for older adolescents.21 For example, buying cigarettes is often illegal before 17 or 18 years of age, but remains a health risk behavior even after the legal age to purchase cigarettes. The delineation of when sexual activity becomes a problem behavior is unclear and may be arbitrary. However, carrying a gun, physical fights, lifetime use of drugs and the use of substances before sexual activity more clearly have negative repercussions. According to the data, many adolescents report one or more risk or problem behaviors, but the combination of these risk or problem behaviors were predictive of suicide attempts in this population. The development of a theoretical model which separates problem behaviors from adolescent experimental or risk-taking behaviors might be useful for future evaluations.
This study supports the concept that risk-taking and problem behaviors are related to suicide attempts. A school-based survey in Minnesota demonstrated the clustering of health-compromising behaviors, including suicidal attempts with unhealthy weight loss methods.26Clark and colleagues27 found that suicide attempts were not related to all reckless behaviors but were most closely related to smoking, drug use, and choosing bad company. Irwin28proposed a theoretical concept of how volitional behaviors interact to produce possible negative health outcomes. In the area of substance use, Kandel29 and colleagues recognize a sequence of progression of substance use: alcohol, cigarettes, marijuana, and other illicit substances. Cigarettes, alcohol, and marijuana use can predict cocaine use in subsequent years.30 Our study shows that cigarette and substance use are associated with suicide attempts. These models should be expanded to evaluate the relationship of risk behaviors with suicide completions in larger populations. More prospective studies are needed to evaluate the sequence of initiation of health impairing behaviors in adolescents.
There are several potential limitations to this study. The risk and problem behaviors are self-reported, and may represent a conservative estimate.31 Youth who are not available to complete the questionnaire due to truancy, absence, or dropout are likely to be at higher risk for suicide attempts and other risk behaviors. This would lessen the strength of the associations measured in our study. The weighting procedures use two nonresponse factors to correct for sampling biases and error due to nonresponse to minimize these effects. Same gender sexual experiences may be a precursor to and be collinear with other risk and problem behaviors making this a difficult variable to fully evaluate in this format. The observation that same gender sexual experiences are also significantly related to other risk behaviors (such as gun carrying, physical fights, substance, and alcohol use) supports the possibility of collinearity in this sample. The previously recognized increased risk of suicide attempts among gay and bisexual youth17,32 may not be well-represented by reports of same gender sexual experiences or suicide attempts in this study. In-school youth reporting same gender sexual experience may not accurately represent the subgroup of self-identified gay/lesbian, bisexual youth, or even youth with homosexual/bisexual preferences who are not sexually experienced. Future surveys should include questions about sexual preferences and identity to address this issue better. Although not the focus of this study, youth who report suicide attempts may not fully represent youth who complete suicides and are no longer available to answer the questionnaire. Further studies are needed to follow cohorts of youth who report suicide attempts to assess predictors of suicide completion.
This study suggests that suicide attempts are associated with other problem behaviors. However, the results do not provide information about a causal relationship. Potential antecedents to problem behaviors such as previous history of sexual abuse, life stressors,11impulsive behaviors,12 lack of optimism,33hopelessness,13,14 and depression9 should be studied further. Identifying, treating, and preventing precursors of problem behaviors may reduce the important sequelae of suicide attempts for adolescents and young adults.
Recognizing that recent suicide attempts cluster with other risk and problem behaviors underscores the importance of developing interventions that address a broad range of risk and problem behaviors and providing the specific skills needed for each area of risk. Intervention programs for at risk youth should evaluate suicide prevention components.1 Clinicians should recognize the increased potential for risk of suicide attempt among patients who report lack of seat belt use, substance use, physical fighting and gun carrying, smokeless tobacco use, drug use, and unprotected sexual intercourse after substance use. Intervening before an attempted suicide is important to reduce morbidity and mortality in this age group because suicide attempts are associated with suicide completions. Identifying which adolescents are at risk for suicide attempts from the presence of associated risk behaviors can advance the design and evaluations of office-based prevention strategies.
Supported, in part, by Project #MCJ-MA 259195 from the Maternal and Child Health Bureau (Title 5, Social Security Act), Health Resources and Services Administration, Department of Health and Human Services; and Special Projects of National Significance Project #BRH 970155–03–0, Health Resources and Services Administration, Department of Health and Human Services.
We would like to thank Annie Faulkner, Kevin Cranston, and the Massachusetts Department of Education for access to the data set for this study. We appreciate S. Jean Emans, MD, for editing and reading the manuscript; Elizabeth Goodman, MD, for her helpful comments; Robin Guilfoy for preparation of the manuscript; and Rissa Abriam for her literature searches.
- Received May 30, 1996.
- Accepted November 18, 1996.
Reprint requests to (E.R.W.) Division of Adolescent/Young Adult Medicine, Children's Hospital, 300 Longwood Ave, Boston, MA 02115.
Presented in part to the Society for Adolescent Medicine, March 23, 1996, Washington, DC.
The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the Maternal and Child Health Bureau or Special Projects of National Significance.
- CDC =
- Centers for Disease Control and Prevention •
- YRBS =
- Youth Risk Behavior Survey •
- OR =
- odds ratio •
- CI =
- confidence interval
- ↵Centers for Disease Control. Youth Risk Behavior Surveillance—United States, 1993. MMWR. 1995;44(SS-1):1–56
- Centers for Disease Control
- ↵Brent DA. Depression and suicide in children and adolescents. In: Green M, Haggarty RJ, eds. Ambulatory Pediatrics (IV). Philadelphia, PA: WB Saunders; 1990:522–529
- Brent DA,
- Perper JA,
- Moritz G,
- et al.
- Swedo SE,
- Rettew DPC,
- Kuppenheimer M,
- Lum D,
- Dolan S,
- Goldberger E
- Remafedi G,
- Farrow JA,
- Deisher RW
- Remafedi G
- ↵Boston Department of Education. Massachusetts 1993 Youth Risk Behavior Survey Results. Boston, MA: Department of Education; 1994. Report No. 17576–70–5000–6/94-DOE
- Middleman AB,
- Faulkner AH,
- Woods ER,
- Emans SJ,
- DuRant RH
- Brener ND,
- Collins JH,
- Kann L,
- Warren CW,
- Williams BI
- ↵Norusis MJ. SPSS Advanced Statistics 6.1. Chicago, IL: SPSS, Inc; 1994
- ↵Irwin CE. The theoretical concept of at-risk adolescents. Adolesc Med: State of the Art Rev. 1990;1:1–14
- ↵MacMahon B, Pugh TF. Epidemiology. Principles and Methods. Boston, MA; Little, Brown and Company; 1970
- ↵Remafedi G, French S, Story M, Resnick MD, Blum R. The relationship between suicide risk and sexual orientation: results of a population-based study. Abstract presented at the Society for Adolescent Medicine; March 23, 1996; Washington, DC
- Goodman E
- Copyright © 1997 American Academy of Pediatrics