National survey sample and data
The data used in our analyses were drawn from the adolescent subsample of the 2000 National Household Survey on Drug Abuse (NHSDA), which is representative of all noninstitutionalized adolescents aged 12–17 in the United States. Suicide attempters, our study population of interest, were defined by a positive response to one item: "In the past 12 months, have you tried to kill yourself?" Of the 19,430 adolescents in the survey, 877 answered "yes," and they constituted the study sample. Sixty-six percent (N=579) of the 877 youths were white, 11% (N=96) were African American, 16% (N=140) were Hispanic, and 7% (N=64) were from other racial-ethnic groups. Most were female (70%, N=614), and most were between the ages of 14 and 17 (78%, N=684). Compared with youths in the survey who did not attempt suicide, a greater proportion of suicide attempters were not living with both parents (37%, N=324).
Data collection consisted of computer-assisted personal interviews conducted in respondents' homes. All information obtained was self-reported by adolescents. For questions about suicide attempt, substance use, and other sensitive issues, audio computer-assisted self-interviewing was used to increase the rate of honest reporting (
28 ). Further details on the sampling design and study procedures of the original study, including its consent procedures, are available elsewhere (
28 ). Analyses were carried out in full compliance with New York State Psychiatric Institute Institutional Review Board requirements.
Measures
Mental health service use. The mental health services used by adolescents in the past 12 months fell into three categories: inpatient, outpatient, and school-based services. Inpatient services included overnight stays in a hospital or residential treatment center to receive treatment or counseling for emotional or behavioral problems not caused by alcohol or drugs. Outpatient services included services received at a partial day hospital, day treatment program, or mental health clinic or center or treatment received from a private therapist, psychologist, social worker, counselor, pediatrician, or family doctor for these problems. School-based services included consultations with school counselors or psychologists or regularly scheduled meetings with teachers to deal with these problems.
The service use outcome variable is the type or types of service used, a Guttman scale measure (
29 ) that divides adolescents who had used any type of treatment services in the past 12 months into three mutually exclusive subgroups: inpatient service use (those who used inpatient mental health services, with or without any other types of services), outpatient service use (those who had used outpatient mental health services with or without school services but not inpatient services), and school-based service use only. Adolescents in the sample of suicide attempters who had not used any mental health services formed the reference group.
Demographic characteristics. Information on age, gender, ethnicity, and nativity was obtained in the survey. Three age groups were used—12–13, 14–15, and 16–17 years. Race-ethnicity was categorized as white, African American, Hispanic, or other. Adolescents who answered no to the question, "Were you born in the U.S.?" were considered foreign born.
Family characteristics. Family income was a four-category variable based on adolescent-reported annual household income. Health insurance was categorized as Medicare or Medicaid, other insurance, or no insurance. Residential instability is a measure of the frequency with which a respondent had moved or changed residences (coded yes for respondents who reported that they had changed residences three or more times during the past five years). A dichotomous variable derived from survey items about the parental figures residing in the respondent's household documented whether or not the child was living with both parents. A variable addressing the child's ability to talk with parents was included. Respondents who disagreed with the statement "There is no one I can talk to about serious problems" and reported that they could talk to their mother, father, or guardian about such problems were coded yes. Respondents were also asked to report the number of children under age 18 in their family.
Community characteristics. Neighborhood supportiveness is a scale-based variable derived from two survey items that asked about the degree to which people in the respondent's neighborhood help each other out and visit each other's homes. The internal consistency of this scale was found to be .52. Neighborhood quality is a scale-based variable created from four ordinally coded survey items addressing crime, drug selling, street fights, and the frequency with which residents moved in and out of the respondent's neighborhood. Internal consistency of this scale was .73. Population density is categorized as not a Metropolitan Statistical Area (MSA), an MSA with less than one million residents, or an MSA with one million or more residents (high population density).
Individual characteristics. Extracurricular activities was a dichotomous variable that was coded yes for youths reporting any participation in such activities in the year before the interview—for example, any school- or community-based clubs, after-school programs, community or volunteer groups, team sports, band or choir, or music programs. Criminal justice system involvement was a lifetime measure and was based on a positive response to one of two survey questions—ever being arrested or booked for breaking the law and having been on parole or probation status in the past year.
Individual spirituality was measured by a scale consisting of two items, each coded from "strongly agree" to "strongly disagree": "My religious beliefs are a very important part of my life" and "My religious beliefs influence my decisions." The internal consistency of this scale was .85. Self-perceived health was measured with the question: "Would you say your health in general is very good, good, fair, or poor?" Respondents reporting either fair or poor health were coded as having poor self-perceived health.
Measures of past-year psychopathology were adapted from the Diagnostic Interview Schedule for Children (DISC) Predictive Scales (DPS) (
30 ), a screening measure derived from the National Institute of Mental Health's DISC Version IV (
31 ). The DPS includes only the DISC items that are most predictive of
DSM-IV diagnoses (
32 ). Psychometric data on the DPS are reported elsewhere (
30 ). Information on 11 DPS symptom clusters, including seven anxiety symptom clusters, one depressive symptom cluster, and three disruptive behavior symptom clusters, was used in the analyses for this study. A symptom cluster is derived from
DSM-IV criteria for a particular disorder, not including criteria related to impairment or symptom duration. The cutoff points for the clusters were selected on the basis of previous methodological research (
33 ).
The anxiety symptom clusters assessed included the social phobia, separation anxiety disorder, agoraphobia, panic disorder, generalized anxiety disorder, specific phobia, and obsessive-compulsive disorder symptom clusters. The number of anxiety symptom clusters found for each individual was used as a measure of his or her overall level of anxiety problems. Probable depression was assessed using the depressive symptom cluster of the DPS. The disruptive behavior symptom clusters assessed included the oppositional defiant disorder, attention-deficit hyperactivity disorder, and conduct disorder symptom clusters. The count of disruptive behavior symptom clusters for each individual was used to measure the overall level of disruptive behavior problems. Dependence on and abuse of substances were measured according to
DSM-IV criteria (
32 ). A dichotomous variable was created to indicate whether the respondent either abused or was dependent on alcohol or an illicit drug or had nicotine dependence.