Citation
An analysis of suicide hotline caller data following suicide prevention public service announcements

Material Information

Title:
An analysis of suicide hotline caller data following suicide prevention public service announcements
Creator:
Layman, Deborah
Publication Date:
Language:
English
Physical Description:
xii, 109 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Hotlines (Counseling) ( lcsh )
Advertising, Public service ( lcsh )
Suicide -- Prevention ( lcsh )
Advertising, Public service ( fast )
Hotlines (Counseling) ( fast )
Suicide -- Prevention ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 103-109).
General Note:
Department of Psychology
Statement of Responsibility:
by Deborah Layman.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
62873083 ( OCLC )
ocm62873083
Classification:
LD1193.L645 2005m L39 ( lcc )

Full Text
AN ANALYSIS OF SUICIDE HOTLINE CALLER DATA FOLLOWING
SUICIDE PREVENTION PUBLIC SERVICE ANNOUNCEMENTS
by
Deborah Layman
B.A., University of Virginia, 2000
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts
Psychology
2005


This thesis for the Master of Arts
degree by
Deborah Layman
has been approved
by
i
i
m>313-00^
Date
i
i


Layman, Deborah Marie (M.A., Psychology)
An Analysis of Suicide Hotline Caller Data Following Suicide Prevention Public
Service Announcements
Thesis directed by Suzanne Kennedy Leahy, Senior Researcher, OMNI Research
and Training, Inc.
ABSTRACT
Colorado has ranked consistently among the top 12 states for high suicide rates. In
response to Colorados elevated suicide rate, individuals working with Colorados
Office of Suicide Prevention implemented a suicide prevention public service
announcement campaign. Intermittently from December 13, 2002 through July 2,
2003, Comcast, a television carrier, aired over 5,200 public service announcements
(PSAs) to educate viewers on the warning signs of suicide, urge viewers to seek
help for themselves or others, and promote utilization of the 1-800-SUICIDE hotline
telephone number. Relatively little is known about the impact of suicide prevention
public service messaging on the prevention of suicidal behaviors. The current study
analyzed whether there was a relationship between the local airing of these suicide
prevention messages and calls to the national hotline from within Colorado. A series
of linear regressions and One Way Analysis of Variance procedures were conducted
to examine the relationship between PSA and hotline caller data. Analysis resulted
in several findings: (1) increases in the number of PSAs predicted an increase in the
number of overall callers, male callers, and callers classified by telephone
counselors as depressed, suicidal, and in an emergency; (2) Colorados hotline crisis
center experienced a greater number of calls during the year in which PSAs aired,
when compared to the previous and subsequent years; (3) tentative evidence of a lag
in


effect for suicide prevention PSAs was found, in other words, individuals who view
suicide prevention messages may behaviorally respond to these messages a few
days after initially viewing an announcement; (4) there may be a minimum number
or threshold of PSA airings that must be reached in order to predict hotline behavior
and, (5) there may be a cumulative effect of PSAs on the number of females callers.
This preliminary study finds evidence supporting the utility of PSAs in the
prevention of suicide. Further research is needed to determine the most effective
production and delivery of suicide prevention messages and to evaluate whether
PSAs reduce the overall suicide rate.
This abstract accurately represents the content of the candidates thesis. I
recommend its publication.
Signed
SuzanneKennedy Leahy


DEDICATION
I dedicate this thesis to memory of Jeff Porter, Sr. and Jeff Porter, Jr. and to my
loving sister, Jennifer Porter. Jennifers journey in surviving the loss of her husband
to suicide shortly after losing her infant son has taught me what it is to be brave,
forgiving, and strong.


ACKNOWLEDGEMENT
The author wishes to express sincere appreciation to Suzanne Kennedy Leahy,
Ph.D. for volunteering so much time and effort to seeing me succeed in finishing
this project. In addition, I owe much gratitude to Andrej Biijulin, PhD for his time
devoted to assisting me in the data analysis as well as Leanne Candura, Katie
Rothery, and the staff at OMNI Research and Training, Inc. for their support and
understanding during this process. Thanks also to Eleanor Hamm at the Pueblo
Suicide Prevention Crisis Center; Karen Mason, PhD and Shannon Brieztman at
Colorados Office of Suicide Prevention; Cindy Hodge; Joel Bidman and Trisha
Jorgensen for their support and assistance in making this project possible. A special
thanks to my family in Virginia for encouraging my education and patiently waiting
for my return home and finally to my friends in Denver who have become my local
family.


CONTENTS
Figures.......................................................xi
Tables.......................................................xii
CHAPTER
1. INTRODUCTION......................................................1
National Suicide Trends........................................1
Preventing Suicide.............................................3
Suicide Prevention Crisis Telephone Hotline Centers............4
History of Crisis Telephone Hotline Centers.............4
Theoretical Basis for Suicide Prevention Crisis Telephone
Hotline Centers.........................................6
Evaluation of Suicide Prevention Crisis Hotline Centers.8
Suicide Prevention Hotline Callers: Gender of Callers and
Reason for Calling.....................................18
Public Service Announcements..................................21
History of Public Service Announcements (PSAs).........21
Evaluations of Public Service Announcements............22
Public Service Announcements to Prevent Suicide........28
Concerns Regarding PSAs With Suicide Prevention........32
Current Study.................................................34
Local Context: Colorado Data...........................34


Suicide Prevention Public Service Announcements in
Colorado....................................
36
Purpose of Study.........................................41
2. METHODS............................................................46
Participants...................................................46
Key Informant Interviews.................................46
Archival Data............................................49
Measures.......................................................49
Key Informant Interviews.................................49
Archival Data............................................51
Procedures.....................................................55
Key Informant Interviews.................................55
Archival Data............................................56
3. RESULTS............................................................61
Testing Hypotheses 1-3: Bivariate Regressions..................61
Testing Hypotheses 4-6: One Way Anovas.........................69
Testing Hypothesis 7...........................................75
4. DISCUSSION.........................................................79
Overall Discussion of Findings.................................80


Evidence for the Impact of PSAs
80
Spots as a Predictor of Caller Behavior: A Review of the
Independent Variable..............................81
Potential Lag Effects of Caller Behavior: Review of Results
Based on the Unit of Analysis.....................84
Potential Threshold Effects: Review of Time Periods A and B ....84
Potential Cumulative Effects For PSAs on Female Callers:
A Review of Caller Data by Quarters...............86
Alternative Explanations for the Increase in Calls......87
Limitations to the Current Study........................88
Future Research.........................................90
Conclusion..............................................91
APPENDIX
A. INTERVIEW QUESTIONS FOR ELEANOR HAMM, PUEBLO
SUICIDE PREVENTION CENTER............................94
B. INTERVIEW QUESTIONS FOR MICHAEL CAFASSO,
OFFICE OF SUICIDE PREVENTION ADVISORY BOARD
MEMBER (2001)........................................95
C. INTERVIEW QUESTIONS FOR JOEL BIDMAN, COMCAST.........96
D. INTERVIEW QUESTIONS FOR TRISHA JORGENSEN,
COMCAST..............................................97
E. CALLER LOG SHEET.....................................98


F. AFFIDAVIT OF PERFORMANCE .................99
G. SPSS VARIABLE VIEW OF CALLER DATA ........100
H. SPSS VARIABLE VIEW OF PSA DATA............101
I. APPENDIX I: SUMMARY OF CALLER DATA RECEIVED BY
THE PUEBLO SUICIDE PREVENTION CENTER.......102
REFERENCES.......................................103
x


FIGURES
Figure
1.1 Number of Suicide Prevention PSAs from December 2002 July 2003......38
1.2. Figure 1.2. Percent of PSAs per Time Segment for the First Block
(December 2002- January 2003).........................................39
1.3 Figure 1.3. Percent of PSAs per Time Segment for the Second Block
(April July 2003)....................................................40
2.1 Construction of Independent Variable (IV): Saturation of Public
Service Announcements..................................................60
3.1 Time Periods A and B in Regression Analysis of PSA and Caller Data....62
3.2 Mean Number of Overall and Distressed Suicide Hotline Callers Before,
During* and After Suicide Prevention PSA Project......................72
3.3 Mean Number of Calls per Week by Call Classification.................74
3.4 Mean Number of Male and Female Suicide Hotline Callers Before,
During, and After Suicide Prevention PSA Project.......................75
3.5 Mean Number of Overall, Male, Female and Distressed Callers
per Day per Quarter of the Year in 2003 ..............................77
3.6 Mean Number of Calls per Day per Quarter of the Year.................78


TABLES
Table
1.1 Expansion of Lesters (1997) Summary of Literature Regarding the
Impact of Suicide Prevention Centers on the Suicide Rate .............14
2.1 A Description of Interview Topics for Each Key Informant ..............50
3.1 Results of Bivariate Regressions for Saturation and DVs by Day........63
3.2 Results of Bivariate Regressions for Saturation and DVs by Week.......64
3.3 Results of Bivariate Regressions for Number of Spots and DVs by Day..65
3.4 Results of Bivariate Regressions for Number of Spots and DVs by Week ...66


CHAPTER 1
INTRODUCTION
A member of Colorados Office of Suicide Preventions advisory board collaborated
with a local production company and Comcast1 to develop and air suicide prevention
public service announcements (PSAs) throughout Colorado. From December 13,
2002 through January 9, 2003 and beginning again on April 17th until July 2, 2003,
Colorado residents viewed 5,287 suicide prevention messages which advertised the
national suicide prevention hotline number, 1-800-SUICIDE. Relatively little was
known about the impact of suicide prevention public service messaging on the
prevention of suicidal behaviors. The current study analyzes whether there was a
relationship between the local airing of these suicide prevention messages and calls
to the national hotline from within Colorado.
National Suicide Trends
In 2002, 31,655 Americans died by suicide; this number translates into 11 of every
100,000 Americans who die from a self-inflicted injury (McIntosh, 2004a).11 As the
1 PSAs were initially aired by AT&T before Comcast purchased AT&T.
112002 was the latest year that official national data were available.
1


11th leading cause of death for all ages and the 3rd leading cause of death for the
young, suicide is a national public health concern.
Suicide deaths disproportionately affect some subpopulations. Specifically, suicide
death rates vary according to gender, race, and age. With male to female rates in
2002 of 17.9 to 4.3 per 100,000 population, approximately 4 males to one female
die by suicide annually (McIntosh, 2004a). Suicide death rates also vary by age.
According to McIntosh (2004b), the elderly (ages 65 and above), made up 12.3% of
population in 2002 but comprised 17.5% of all suicides in the same year. In
contrast, young Americans (ages 15 24) represented 14.1% of 2002 population,
yet accounted for 12.7% of all suicide deaths in 2002. Overall, death rates indicate
that White males over the age of 75 have been the population most at risk for
suicide (Institute of Medicine, 2002).
The spectrum of suicidal behaviors ranges from thinking about ending ones life
(suicide ideation), threatening to attempt suicide, making a suicide attempt, and
dying by suicide. While death by suicide is the most permanent and serious result
from suicidal behaviors, individuals who consider, threaten, and/or attempt ending
their lives also exhibit suicidal behavior and, thus, are at risk of dying by suicide.
Individuals who have attempted suicide are 38% more likely to make another
attempt (American Psychiatric Association, 2003). Since individuals who attempt
2


suicide may not receive appropriate medical or mental health treatment, it is
difficult to accurately estimate the number of individuals who attempt per year.
However, research estimates 25 attempts per death by suicide. This estimate
translated into 790,000 annual attempts in the United States in 2002 (McIntosh,
2004a).
Like suicide death rates, rates of suicide attempts are experienced differently by
subpopulations. These patterns are different than trends of death rates. While it is
difficult to estimate the number of attempts per year, approximately 3 females to
one male attempt annually (McIntosh, 2004a). It also has been estimated that 100-
200 youth attempt suicide for every one death (McIntosh, 2004a; AAS, 2002).
Overall, research suggests that young females are the most at risk group for
attempting suicide. A review of death and hospitalization rates for self-inflicted
intentional injuries indicates that suicidal behaviors, more broadly than deaths alone,
are a national public health concern.
Preventing Suicide
Trends in intentional self-inflicted injuries serve to emphasize the importance of
intervention and prevention strategies. In conjunction with additional literature
regarding the etiology of suicide, precipitating factors, and warning signs, these
trends may serve to guide such strategies. Researchers using the psychological
3


autopsy method, which involves the systematic review of the life of suicide victims
and may include an analysis of medical records, mental health treatment records,
interviews with family and friends close to the deceased, and records from schools,
military, employment, etc., report that approximately 90% of suicide victims
suffered from a mental illness (Cavanagh, Carson, Sharpe, and Lawire, 2003). As a
means to reach individuals needing mental health treatment and prevent suicidal
behaviors, suicide prevention telephone hotlines emerged and flourished in the
1960s (Lester, 2002).
Suicide Prevention Crisis Telephone Hotline Centers
History of Crisis Telephone Hotline Centers
There are several thousand suicide prevention crisis hotline centers in the United
States (Seeley, 1996). Beginning in the late 1940s, counseling centers for suicide
prevention were first established in Vienna, England, and Germany (Wolfesrdorf,
Blattner, Grober, 1989). In the United States, the telephone was integrated into crisis
prevention with the Washington-Moscow hotline. This hotline was developed as a
warning system to handle the accidental release of nuclear weapons (Seeley 1996).
During the same period, the Los Angeles Suicide Prevention Center developed a
telephone hotline to serve clients after regular business hours (Litman, Farberow, &
Shneidman (1965). Both the Washington-Moscow and the Los Angeles Suicide
4


Prevention hotlines were the first efforts in the U.S. to use the telephone to prevent
or respond to a crisis.
As hotlines flourished, additional innovations and adaptations were developed to
establish a network of trained individuals available to respond to a crisis or other
social/health issues. For example, volunteers became telephone counselors with the
advent of CALL BRUCE which provided suicide prevention activities to the gay
community (Seeley, 1996). Many hotlines established to address suicide prevention
and other issues, such as sex, marriage, alcoholism, and smoking cessation,
developed additional services such as drop-in counseling and follow-up services.
Today prevention and intervention efforts use the telephone to address a spectrum of
social issues and include varying levels of service delivery (Seeley, 1996).
As crisis hotline centers emerged in the United States, media attention and literature
regarding the use of telephone communication to resolve a crisis emerged.
According to Seeley (1995), television, magazines, and newspaper promoted the use
of hotlines mainly by reporting stories/case studies of lives saved through the use of
hotlines. To address the growing number of unregulated hotlines appearing across
the country, crisis prevention professionals began publishing manuals on managing
crisis hotline centers.
5


The American Association of Suicidology (AAS) tackled the issue of unregulated
suicide prevention hotlines in the United States. Since the early 1970s, AAS has
certified suicide prevention hotlines to ensure that they were maintained and
managed according to specific standards developed by suicidology experts. The
AAS certification process requires independent hotlines to meet various standards
regarding administration, crisis worker training, service delivery, response to life-
threatening crisis, ethics, community integration, and evaluation efforts (AAS,
2001). Certified centers receive a ranking along a rating scale from Level I to Level
IV. The rating level represents the degree to which a crisis hotline center
administers and delivers services and provides follow-up with clients. According to
Eleanor Hamm, the Executive Director of the Pueblo Suicide Prevention Center,
certification and rating levels not only validate service delivery efforts but also
provide benchmarks for program growth (personal communication, January 24,
2005). Certification also provides hotlines with additional credibility when applying
for grants and provides crisis workers with a network of professional support.
Theoretical Basis for Suicide Prevention Crisis Telephone
Hotline Centers
According to Litman et al. (1965), hotlines lead to fewer suicides by deferring an
imminent suicide and referring individuals in crisis to effective treatment. This
rationale assumes that individuals in danger of death by suicide recently
1
6


experienced a stressful event, were acting on an impulse, but maintain some desire
to live (Shaffer, Gould, Fisher, & Trautman, 1988 as cited by Center for Disease
Control, 1992). Hotlines provide disturbed callers with an immediate, anonymous,
and convenient method of support. The range of services provided by suicide
prevention centers often varies according to the intended population (adolescents,
sexual minorities, professionals, military, etc.), association with other institutions
(schools, mental health clinics, other community services), and staff characteristics
(professionals, volunteers, youth peers, etc.).
The rationale behind hotline programs was founded on several assumptions: (1)
distressed individuals were aware of the crisis hotline center/hotline, (2) distressed
individuals felt inclined to utilize these services, (3) distressed individuals initiated a
call, (4) trained counselors accurately assessed and responded to suicide risk with an
immediate intervention or referral to effective treatment (5) referrals were available
(financially, logistically, and quickly available), (6) treatment was effective, and (7)
there was concurrent availability and utilization of social support networks (Center
for Disease Control, 1992). In addition, this rationale assumed that individuals in
crisis believe they can be helped, believe their circumstance can change, and are
willing to seek help.
7


This rationale suggests that in order to be effective crisis hotline centers must (1)
advertise services or provide outreach, (2) be accessible to distressed individuals (24
hour availability, free, require only a telephone), and (3) adequately train counselors
(Center for Disease Control, 1992).
However, evaluations of suicide crisis hotline centers have been equivocal (Lester,
1997). While there is strong theoretical evidence for the development of suicide
crisis hotline centers, there has been little empirical research to consistently validate
the effectiveness of suicide hotlines (Gould and Kramer 2001).
Evaluation of Suicide Prevention Crisis Hotline Centers
Impact of Suicide Prevention Crisis Hotline Centers on the Suicide Rate.
Variations in suicide crisis interventions make it difficult to summarize the effects
for hotlines specifically. According to Lesters (1997) meta-analysis, the most
common methods to determine the impact of hotlines on suicide deaths were
ecological, time series, or correlational studies. (1) Ecological studies have
examined the suicide rates for a number of comparable geographic regions with and
without a suicide prevention center. (2) Time series studies have investigated one
region before and after the implementation of a suicide prevention center. (3) Other
8


correlational studies tend to compare the density of suicide prevention centers
within a geographic location with their respective suicide rates.
Evaluations of suicide prevention centers/hotlines emerged in the 1960s. In the
United Kingdom, Bagley (1968) conducted the first evaluation of suicide prevention
centers. From 1957 1966, he compared the suicide death rate of several English
towns before and after the introduction of a crisis hotline center with towns that had
never developed a hotline. Similar to suicide prevention hotlines today, hotline
volunteers referred emotionally disturbed callers to psychiatric help. Bagley
reported a 5.8% decrease in the suicide rate after the development of a suicide
prevention hotline (Bagley 1968). This decrease did not occur in the control cities.
However, several investigators have failed to duplicate Bagleys findings. In 1980,
Lester published a reanalysis of Bagleys data and found an impact of the hotline in
only one town (Lester, 1980). According to subsequent researchers, Bagley used
poorly matched controls in his 1968 evaluation (Gould and Kramer, 2001). This
may explain why other English researchers were not able to replicate Bagleys
findings (Jennings, Barraclough, and Moss 1978; Barraclough, Jennings, and Moss,
1977).
i
9


Using four samples of English towns, Jennings et al. (1978) and Barraclough et al.
(1977) matched towns with suicide prevention centers (including Bagleys original
fifteen towns) to comparable English boroughs with similar suicide death rates and
single-person households. Their analysis failed to find a statistically significant
decrease in the suicide rate in the boroughs with suicide prevention centers. Similar
results were reported in the United States. Bridge, Potlin, Zung, and Soldo (1977)
compared cities in North Carolina with and without suicide prevention centers from
1970-1971. This single year ecological study reported no preventive effect for crisis
hotline centers.
However, when suicide death rates have been disaggregated for gender, a preventive
effect for suicide prevention centers emerged. Similar to the techniques described
above, Miller, Coombs, Leeper, and Barton (1984) compared American towns with
suicide prevention centers to matched comparison counties before and after the
implementation of a suicide crisis hotline center. Miller et al. (1984) found a
significant reduction (33% 55%) in the suicide rate for White females under the
age of twenty-four. However, a positive impact on the general population did not
emerge (Miller et al. 1984).
Lester also reported a preventive effect for certain subpopulations. In his U.S. study,
Lester analyzed the density of suicide prevention centers and the overall suicide rate
10


and found a positive effect for the female suicide death rate and the suicide death
rate for individuals ages 15-25 and 55 64 (Lester, 1993). However, attempts to
replicate this study in Canada did not find a statistically significant relationship
(Leenaars and Lester, 1995).
Leo, Buono, and Dwyer (2002) also reported a preventive effect from suicide
prevention centers. They found a decline in the suicide rate among elderly females
who used a telephone support line, compared to the general population in northern
Italy. However, this telephone support varied from typical suicide prevention
hotlines; telephone counselors called elderly participants twice a week in addition to
responding to ad hoc calls.
Other researchers have analyzed suicide rates while controlling for social variables
such as unemployment, marriage, and religion. In a single year ecological study,
Medoff (1984) included other variables such as age and religion in his study of the
effectiveness of U.S. suicide prevention centers. According to estimates by Medoff,
crisis hotline centers prevented 2,821 White male suicides and 789 White female
suicides annually. When controlling for social variables such as marriage, divorce,
and birth rates, Huang and Lester (1995) reported a preventive effect after the
establishment of a suicide prevention center in Taiwan from 1965-1985. A
11


preventive effect for the male suicide rate emerged in Japan when researchers
controlled for birth, divorce, and employment rates (Lester, Saito, and Abe, 1996).
Many studies, however, reported mixed results or no effects. Other studies have
found positive effects but also results that were mixed for different populations
based on gender, age, and social characteristics. In addition, some evaluations have
found a change in the negative direction. By comparing the suicide rate in twenty-
five German cities fifteen years prior and seven years after the establishment of a
suicide prevention center, Riehl, Marchner, and Miller (1988) reported an increase
in the suicide rate in several of these German cities (Riehl et al. 1988 as cited by
Lester, 1997). When disaggregated by gender, the suicide rate increased for males in
seven cities and decreased in three other cities, while the suicide rate for females
increased in three cities and decreased in one. Using a small sample of California
cities, Weiner (1969) reported an increase in one city, Los Angeles, after the
development of a suicide prevention center. According to Lester (1997), suicide
prevention efforts may have increased awareness and reduced the stigma related to
suicide which led medical examiners or coroners to classify ambiguous deaths as the
result of suicide.
A description of evaluation studies provides an unclear picture of the effectiveness
of crisis hotline centers. In an effort to understand the impact of prevention centers,
12


Lester conducted a meta-analysis of fourteen unique studies including many of the
evaluations outlined above. Overall, he reported that of the fourteen studies, seven
reported a preventive effect for some population after the establishment of a suicide
crisis hotline center (Lester, 1997). The following table was a summary of
evaluation research described in the narrative above.
13


Table 1.1:
Expansion of Lesters (1997) Summary of Literature Regarding the Impact of
Suicide Prevention Centers on the Suicide Rate
Name
Type of
Evaluation
Time
period
Population Location
Found a positive effect for the overall suicide rate
Findings
Bagley (1968) Ecological studies 1957- 1966 Overall suicide rate England Positive impact on all treatment towns
Lester (1980)' Ecological studies. 1957- 1964 Overall suicide rate England Preventive effect for only town
Medoff (1984) Single year ecological study 1979 By gender and other social variables US For both males and females
such as
religion
Found a positive effect for the whole population or sub populations
Miller et al. (1984) Ecological study 1968- 1973 All residents and then by race and gender US White females under age of 25
Lester (1993) Correlational study 1970- 1980 Overall suicide rate, rate by gender and age. US Total suicide rate, female suicide rate, ages 15-24 and 55-64
Huang and Lester (1995) Time series 1965- 1985 Overall suicide rate controlling for marriage, birth, and divorce rates Taiwan Preventive effect only when variables of domestic integration were controlled.
Lester, Saito, and Abe (1996) Time series 1970- 1989 Total suicide rate with controls, & by sex Japan More consistent preventive effect for male suicide rate
14


Table 1.1 (Cont.)
Name Type of Evaluation Time period Population Location Findings
Preventive effect with different research design
Glatt et al. Case Study 1984- Phone used 10 times
(1986) (emergency 1985 in one year. Five
phone at a individuals died by
high risk site). suicide.
Lester Meta-analysis 1968- Preventive effect
(1997) 1996 found in 7 of 14 studies
Results overall indicate an increase in suicide rate
Weiner Time series 1955- Total US (LA Suicide rate increased
(1969) 1967 suicide rate and San in 1 of 4 cities. The
Fran) remaining 3 had no effect.
Riehl et al. Time Series 1945+ Suicide rate 25 # of cities
(1988) total & by German +
sex Cities Males 1 3 Females I 2
Found no differences
Jennings et Ecological 1957- Overall England
al. (1978)' studies 1973 suicide rate
Barrac lough Ecological 1957- Overall England
et al. (1977)'" studies. 1973 suicide rate
Leenaars Single year 1985- Suicide rate Canada
and Lester ecological 1991 total, by
(1995)'v study sex & age.
Bridge et al. Single year 1970- Total US (NC)
(1977) ecological 1971 suicide rate
This investigation was a reanalysis or attempted replication of the Bagley (1968) study.
IV This study was a replication of Lesters (1991) study in Canada.
15


Overall, there was some evidence for a preventive effect for suicide prevention
hotlines especially for some subpopulations. However, the research that has
attempted to determine the effect of suicide prevention crisis hotline centers on the
suicide rate has continued to report inconsistent results. There were several possible
explanations for these inconsistencies such as the varying levels of service delivery
and the limitations associated with conducting research on suicide.
Limitations of Suicide Research and Suicide Rates. Although the research
on hotlines, such as those studies described above, has yet to validate or invalidate
the preventive effect of suicide prevention, assuming these efforts were ineffective
was problematic since most studies suffer from important limitations. Limitations
include problems associated with computing and monitoring suicide rates. The
suicide rate was computed using the number of deaths marked as suicide by the
medical examiner or coroner per population. Classifying a death as self-injurious
requires having enough evidence that the victim intended to harm him/herself.
Communities, which were cognizant of suicide, may be more vigilant in accurately
recording suicide deaths. Lester (1997) notes that the development of crisis hotline
centers may raise awareness of suicide and lead to a more accurate classification of
death and nonfatal attempts of suicide. For this reason, changes in local suicide rates
16


may be attributable to changing trends in classifying death as suicide. In summary,
researchers cannot assume suicide prevention centers were ineffective or effective
due to the limitations on data quality.
National Trends in the Suicide Rate. In addition, researchers also must note
the national trends of suicide rates. Changes in local trends may be due to changes
in classification and suicide prevention efforts underway, but it also may be
influenced by the economy, unemployment rate and national catastrophic events
such as the tragedy of September 11, 2001. According to Miller et al. (1984), the
suicide rate in the United States was increasing during the time that communities
were first developing suicide prevention hotlines. This indicates that additional
variables influence the suicide rate and prevent researchers from making definitive
conclusions regarding suicide prevention hotline studies.
Temporal Fluctuations in Utilization of Suicide Hotlines. Another
consideration in reviewing data regarding suicide hotlines was the temporal
fluctuations in the utilization of suicide hotlines. According to previous studies,
suicide calls vary according to day of the week, day of the month, and season
(Noble, 1996). According to Noble (1996), suicide related calls to a crisis center
were the highest on Wednesday and in May. The lowest month for suicide related
j
17


calls to this crisis center was August. This differs from the temporal variations
related to suicide deaths (i.e., deaths were most likely to occur from April to May,
on Mondays and least likely to occur in December). Voracek and Sonneck (1999)
did not replicate some of Nobles findings. Specifically they reported a peak of calls
on Mondays/
Suicide Prevention Hotline Callers: Gender of Callers
and Reason for Calling
The impact of the availability of suicide prevention hotlines on the general
population and the population most at risk for suicidal behaviors was unclear.
However, there was some preliminary research on the reasons why individuals call
suicide prevention hotlines.
Reasons for Calling a Suicide Prevention Hotline. All hotlines establish their
own criteria for determining a individuals reason for calling. Lethality for suicide
death was usually based on the presence of a suicide plan, lethality of plan (access
and plans to use a firearm was the most dangerous), and imminence for dying (caller
plans to imminently end her life). According to most published reports, danger of
suicide was infrequently the most common reason for calling a hotline. Since each
v However Voracek and Sonneck (1999) analyzed hotline data from a center in Vienna that was
closed on the weekends.
18


crisis hotline center collects and operationalizes caller information differently, it
was difficult to compare the reason for calling or the event precipitating a callers
suicidal ideation across sites. Some hotline crisis centers may not even collect this
data. However, most published reports of caller data include information on suicidal
callers based on crisis center records. These reports consistently suggest that the
majority of callers of suicide prevention hotlines were troubled but were rarely at
high risk for an imminent suicide (Anda and Smith, 1993). According to Anda and
Smith (1993), only 3% of callers to the Los Angeles County suicide help line were
classified by the telephone counselors as being at high risk for death by suicide. The
reason for considering suicide and calling a suicide prevention hotline varies from
center to center, but many call centers report that depression was the most frequent
reason for considering suicide (Anda and Smith, 1993). The differences in the type
of calls may depend on the telephone hotlines publicity, institution affiliations, the
suicide prevention centers mission, the needs of the community, and how
counselors were instructed and trained in recording the content of each call.
Gender of Suicide Hotline Callers. As previously discussed, White males
over the age of 75 are at the greatest risk of dying by suicide, whereas young White
females are at the greatest risk for attempting suicide (Institute of Medicine, 2002).
Several theories have been developed to explain this gender difference. Specifically,
19


males typically choose more lethal means than females; seek mental healthcare less
than females and suffer from gender stereotypes which dictate that males should be
strong enough to handle problems independently (Institute of Medicine, 2002;
Pilgrim and Rogers, 1993).
Despite the higher suicide death rates for males, most suicide prevention centers
report a higher volume of young female callers (Adna and Smith, 1993; Boehm and
Campbell, 1995; Leo, Buono, and Dwyer, 2002; Wolferdor, Blattner, Grober, 1989,
Fakhoury, 2000; Voracek and Sonneck, 1999) and as described above, some
evaluations have found a preventive effect for this population (Miller et al., 1984).
According to Gould and Kramer (2001), the evidence for a preventive effect in the
female population, a population which utilizes suicide prevention hotlines, indicates
that the limited impact of suicide prevention crisis hotline centers may be due to
their low utilization rate by those populations at greatest risk. This implies that
additional efforts should be directed to increasing help-seeking behaviors among
males, decreasing the stigma of seeking help for suicidal feelings, and advertising
the hotline.
One such advertising effort, public service announcements (PSAs), has been widely
used in promoting prosocial behaviors and substance abuse prevention. While PSAs
1
i
20


have been a powerful tool in educating the public, the field of suicidology has
focused efforts on educating the media about responsible reporting of suicide cases
(Gould, Jamieson, and Romer, 2003; Jamieson, Jamieson, and Romer, 2003).
Research on suicide contagion following media portrayals of suicide death (Bollen
and Phillips, 1982; Stack, 2002) creates specific concern regarding the impact of
PSAs. Concern that PSAs may cause a negative effect appears to be unique to
suicide and has not been investigated by researchers to date. Due to concerns
regarding the deleterious effects of media stories on suicide (Bollen and Phillips,
1982), suicide prevention messages in public service announcements have been
approached with caution. Using PSAs to prevent suicide had not been explored until
the American Foundation of Suicide Prevention (AFSP) released the Suicide
Shouldnt Be a Secret national teen suicide prevention campaign in 2000
(American Foundation for Suicide Prevention, n.d.). Overall, PSAs as a tool for
suicide prevention, has yet to be evaluated.
Public Service Announcements
History of Public Service Announcements (PSAs')
According to the Federal Communications Commission (FCC), a PSA is any
announcement which promotes program, activities, or services of federal state, or
21


local governments, or the programs, activities or services of a non-profit
organization and other announcements regarding community interests excluding
time signals, routine weather announcements, and promotional announcements
(Dessart, n.d.). PSAs have a long history of use by different groups to bring about
broad changes in the awareness, attitudes and behaviors of the general public. Well-
known PSAs include the long-running Smokey the Bear campaign and anti-drug use
campaigns, like the ad showing eggs in a frying pan and the message that this is
your brain on drugs. Before television viewers were exposed to prosocial
characters like Smokey the Bear or tag lines like this was your brain on drugs, the
first PSA encouraged Americans during World War II to purchase war bonds
(Dessart, n.d.). Raising awareness was believed to change attitudes and, to a lesser
extent, change behaviors.
Evaluations of Public Service Announcements
Several quasi-experimental or correlational evaluation studies have been conducted
to assess the effectiveness of public service announcements. Different from
commercials, media time is donated and therefore PSA projects cannot control the
level of exposure (through controlling the timing, frequency, and channel) of
messages. Many evaluation studies of PSAs investigate relationships between
knowledge, attitudes, and behaviors regarding the target message and the
22


implementation of the PSA project. Typically, evaluation studies involve repeated
survey measures that sample the target audience, before, during, and after the
messages were aired; some have included comparison group designs. Other methods
include universal telephone surveys and analysis of official data records such as
child abuse rates, death rates, etc.
Evaluations of public service announcements have found some contradictory results.
While several large community trials have reported no overall or minimal effects
(COMMIT, 1995; Farquhar et al., 1990; Luepker, 1994), there have been other
studies which have found evidence that public health education has led to significant
changes in health behaviors (Lorch, Pugzles, 2002; Palmgreen, Donohew, Pierce,
Macaskill, and Hill, 1990, Worden and Flynn, 2002). The most widely referenced
public health study, the Stanford Three Community Study conducted from 1972 -
1975 reported some decreases in cardiovascular risk factors. Later, the Stanford
Five City Project conducted from 1980-1986 promoted messages urging
cardiovascular risk reduction including weight loss, cigarette smoking cessation,
proper nutrition, physical activity, and blood pressure control. The results indicated
a positive gain for treatment cities in only two of the six risk factors (Homik, 2002).
According to Homik, the treatment cities that were exposed to the risk reduction
messages displayed only minimal effects. Later programs similar to the Stanford
i
i
23
i


Five City Project and the Stanford Three Community Study demonstrated some
behavior change in treatment cities for some of the intended outcomes. However,
these later programs, which used a more rigorous evaluation design, reported less
success than the Stanford projects.
Several evaluations, which did not use a comparison group in the analysis, found
significant positive effects. For example, the National High Blood Pressure
Education Program (NHBPEP) evaluated prevention efforts using a time series
analysis and found a significant decrease in age-adjusted stroke mortality rate, not
accounted for by improved treatment and medications (Homik, 2002). An anti-
smoking campaign, which ran from 1967 1970, led to a significant decline (10%)
in cigarette consumption. Similar effects were found for later anti-smoking
campaigns. Self-reported increase in condom use following the introduction of a
public service announcements campaign promoting safe sex practices in European
countries mirrored a sales increase in condom sales (Coffman, 2003). Declines in
Sudden Infant Death Syndrome were reported in European countries and in the U.S.
after the introduction of public education campaigns. Similar declines were seen for
campaigns regarding seatbelt use, Reyes syndrome, and anti-sunburn (Coffman,
2003).
24


One public service campaign designed to encourage positive attitudes regarding
depression was conducted in the early 90s in the United Kingdom (Paykel, Hart,
and Priest, 1998). According to Paykel, Hart, and Priest (1998), the Defeat
Depression Campaign led to increased positive attitudes toward depression and
antidepressants. This evaluation found overall acceptance of counseling as a method
for dealing with depression. However, attitudes toward treatment through general
practitioners was inconsistent, and despite gains in positive attitudes toward
antidepressants only 24% of survey participants agreed that people suffering from
depression should be offered this medication while 74% maintained antidepressants
were addictive. Since this campaign targeted a change in attitude, it was unclear
whether a change in attitudes toward depression would lead to changes in help-
seeking behaviors. In other words, would distressed individuals feel people with
depression should seek help, and were they able to recognize the symptoms and
need for help themselves?
In addition to these published evaluations, several technical evaluations have
reported a positive impact of public service announcements on prosocial behaviors.
One such evaluation conducted by the Ad Council reported small differences in gun
storage behavior after the implementation of a public service announcement
campaign designed to promote gun safety (Coffman, 2003). Results of this
25


evaluation indicated a 6% increase in gun owners who reported moving their guns
within the home to safer storage 6 months into the campaign with a total increase of
12% at the end of the campaign.
Similar to the gun storage campaign, another public service campaign sought to
promote a helpline through the promotion of help seeking and socially responsible
behavior (Coffman, 2003). Evaluation of the helpline included descriptive analysis
of hotline callers, random telephone surveys of Vermont residents, interviews with
child abuse prevention key decision-makers, surveys of clinicians, and telephone
interviews with the county prosecutors office. Evaluators found an increase in
knowledge regarding sexual abuse, knowledge about available child abuse
intervention resources; increases in positive responses from key decision-makers
and stakeholders; and an increase in individuals who voluntarily sought treatment
and legal help.vl Like many telephone hotlines, the majority of callers were female.
Over half of the callers became aware of the hotline through the media.
Another public service campaign regarding ground level ozone found positive
changes in awareness and behavior (Coffman, 2003). This public information
campaign sought to raise awareness and decrease the number of miles driven and
w From the evaluation report, it was unclear if the number of individuals seeking help for child abuse
perpetration was higher after the implementation of the public service announcement campaign.
26


the number of driving trips. Evaluators at the Applied Research Center at Georgia
State University conducted rolling sample surveys (daily tracking surveys) which
assessed driving behavior, recent mode of transportation, attitudes toward personal
health risks, self efficacy in impacting ground level ozone, and awareness of ozone
alerts. Survey data indicated that ozone alerts raised awareness about ground level
ozone and decreased the number of miles driven during ozone alert days.
According to his review, Homik (2002) attributed the discrepancy in effectiveness
between the earlier rigorously evaluated programs and the later less rigorous
evaluations to research design. The controlled designs, he suggested, may not be a
successful method to detect change from public service announcements since
information and prevention activities may leak to the control cities.
Another problem may be the amount of exposure. The Stanford project presented 25
hours of messages over 5 years regarding several heart disease factors. There may
not have been enough information for each health behavior independently to affect
changes in behaviors related to heart disease. According to Homik (2002), public
service announcements that rely solely on donated time and therefore air
infrequently, air during low hours of viewership, and air on less popular channels
will fail even when messages were well-designed.
27


Despite some inconsistencies, evaluations of PSAs have indicated that public
service campaigns are effective in increasing positive changes in knowledge,
attitudes, and behaviors. However, the general population must be exposed to these
messages with adequate frequency to cause an effect. In addition, these messages
may spread beyond the intervention area, making it difficult for evaluation designs
with controls to detect positive change.
Public Service Announcements to Prevent Suicide
Potential Benefits of using PSAs in Suicide Prevention: Encouraging Help
Seeking Behaviors and Reducing the Stigma of Mental Illness. Given that suicide
was related to untreated mental health illness, especially mood disorders (Cavanagh,
Carson, Sharpe, and Lawire, 2003), prevention should include efforts to encourage
distressed individuals to receive professional mental health treatment. Researchers
such as Horowitz (1977) and Kadushin (1958) have investigated the pathways or
stages taken by individuals who receive professional psychiatric care. According to
Kadushin (1958), there are four stages to becoming a psychiatric patient: individuals
1) recognize they have a psychiatric problem, 2) decide to discuss problems with
friends or family, 3) decide whether to seek professional help, and finally 4) decide
28


whether or not to receive psychiatric treatment. Horowitz (1977) contended that
individuals do not recognize the presence of a psychiatric problem due to shame or a
lack of understanding of their problems. Instead, an individual would be prompted
to seek help in order to resolve general distress such as interpersonal problems,
feelings of depression, anxiety, etc. (Horowitz, 1977). An individuals pathway to
professional treatment initially begins with seeking help from their social support
network, members of which encourage troubled individuals to seek professional
psychiatric treatment.
Later research has investigated demographic, social, and psychiatric differences
between individuals who seek out and receive mental health treatment. Studies of
psychiatric utilization have indicated several differences between individuals who
seek help and those who do not. One major discriminating difference between help-
seekers and non-help-seeking individuals was gender. Research on help-seeking
behaviors has indicated that males seek help less than females, have more negative
attitudes toward help-seeking than females, and use more maladaptive coping skills
than females (Bland, Newman, and Om, 1997; Leong, and Zachar, 1999; Schonert-
Reichl and Muller, 1996; and Timlin-Scalera, Ponterotto & Blumberg, 2003). In
their interviews with male adolescents, Timlin-Scalera, Ponterotto & Blumberg
(2003), found that males reported a lack of awareness and understanding of
29


professional mental health services, increased concern regarding confidentiality, and
believed individuals who sought help were considered weak and troubled. Males
who did seek formal mental health services were more likely to be from homes that
encouraged openness, had positive experiences with a helping professional, had
problems that were considered serious, and were encouraged by someone to seek
professional help (Timlin-Scalera, Ponterotto & Blumberg, 2003).
Overall, these findings suggest that individuals, especially males, may be barred
from Kadushins (1958) first step in the pathway to psychiatric treatment due to
feelings of shame and a lack of understanding of mental disorders. Without a
willingness to independently seek help, reticent individuals will only enter the
psychiatric treatment through the urgings of the social network (Horowtiz, 1997).
This secondary pathway may decrease the chance that a male may be exposed to
professional resources. Since 44% of sampled outpatient and short term inpatient
males, compared to 14% of respective females, report having no friends, the need
for others to prompt distressed individuals to seek treatment may account for some
of the gender differences in service utilization. Lack of social support coupled with
barriers caused by stigma and misunderstanding of mental illness may explain why
gender differences did not emerge after a need for help was identified (Saunder,
Resnick, Hoberman, and Blum, 1994).
30


Since individuals with mental illness may be socially rejected, feared, distrusted,
and disliked, feelings of shame are reasonable concerns (Hinshaw and Cicchetti,
2000). Efforts to reduce stigma regarding mental illness would not only improve the
quality of life of people who suffer from severe mental illness (Corrigan and Penn,
1999), but may also reduce social barriers to accessing mental health treatment. As
described earlier, the negative effects of stigma may disproportionately affect males,
which, in turn, may lead to gender differences among psychiatric service utilization.
According to Timlin-Scalera, Ponterotto, & Blumbergs (2003) grounded theory,
males experience a high amount of stress related to high expectations concerning
success, wealth, and acceptance. According to Timlin-Scalera, Ponterotto, &
Blumbergs (2003), help-seeking behaviors do not align with such expectations,
therefore, a gender-specific stigma of recognizing psychiatric problems and seeking
has emerged and has barred males from entering treatment. In other words, males
who admit to struggling with a mental health issues were admitting to being weak
and troubled. To make services more available to males, Timlin-Scalera, Ponterotto,
& Blumberg (2003) suggested that normalizing and exposing males to resources
and to the process of help-seeking was needed.
31


Increased public awareness about the warning signs of suicide and the suicide
prevention hotline as an immediate, convenient, and confidential resource may help
address stigma associated with seeking help for mental health issues. Educating
viewers on suicide and providing them with an anonymous resource may encourage
males, who traditionally have not accessed suicide prevention hotlines, to seek help.
These campaigns may also encourage any viewer who recognizes warning signs in
herself or someone else to seek help.
Concerns Regarding PSAs with Suicide Prevention
While the PSAs may be a powerful tool to promote prosocial behavior, this method
may not be appropriate to prevent suicide due to legitimate concerns that
irresponsible media coverage of suicide can cause harm (Bollen and Phillips, 1982).
Describing suicide to mass audiences in ways which romanticize or idealize the life
of the victim may lead to suicide contagion (Center for Disease Control, 2001).
Suicide contagion refers to behavior imitation of suicides portrayed in the media
(Gould, Jamieson, and Romer, 2003). Based on social learning theory, researchers
view suicide contagion and clusters of suicide deaths as the result of observational
learning or modeling. In other words, vulnerable viewers learn from the media that
suicide is an option to end despair and individuals who die by suicide are idealized.
I
32


While it is important to increase awareness, it is vital to ensure that suicide
prevention messages encourage individuals in crisis to seek help while not
communicating suicide as a common solution to despair. According to the National
Institute of Mental Health in collaboration with the Center for Disease Control,
World Health Association, New Zealand Youth Suicide Prevention Strategy, and the
National Swedish Center for Suicide Research, extensive research on suicide
contagion or suicide clusters has found that suicide contagion was more likely to
occur with increased overall and prime time coverage of one individuals suicide,
increased details of the method of suicide, and when sensational headlines were
used (Center for Disease Control, 2001).
Thus, suicide prevention messages should avoid inadvertently normalizing or
romanticizing self-inflicted death. Because of the potential for suicide contagion, the
National Institute of Mental Health and other experts in the field caution that
messages delivered to mass audiences regarding suicide should be carefully crafted
as they may have deleterious effects (Center for Disease Control, 2001; Gould et al.,
2003). Media guidelines serve to educate journalists and recommend specific
language and story angles. Since these guidelines refer to coverage of individual
suicides in the media, they do not provide specific guidance in how to craft public
i
33


service announcements. To date, information on the negative or positive effects of
suicide prevention messaging is sparse.
Current Study
Local Context: Colorado Data
The need for suicide prevention activities such as suicide prevention hotlines and
suicide prevention public service announcement campaigns is especially relevant in
Colorado. Suicide death rates in the United States are the highest in the west
(McIntosh, 2004a). With 727 suicides in Colorado and over 16 out of every 100,000
Coloradans dying by suicide, compared to the rest of the nation, Colorado ranked 7th
in the nation in 2002. (McIntosh, 2004a, McIntosh, 2004b).vn Since 1990, Colorado
has ranked consistently among the top 12 states for high suicide rates.V1" Suicide
remains the states eighth leading cause of death for all ages and the second leading
cause of death for individuals ages 10-34 (National Center for Injury Prevention and
Control, 2002).
Similar to national trends, death and hospitalization rates vary by age and gender. In
2003, Colorado males (25.4 per 100,000) had an age adjusted suicide rate that was
2002 was the most recent year that national rankings were available.
VU1 With the exception of 1998, Colorado has consistently ranked in the top ten.
34


four times higher than the rate for females (6.7 per 100,000) (Colorado Health
Information Dataset, 2003). In addition to suicide deaths, suicide attempts were a
national and local public health concern.
Thehospitalization rate for suicide attempts from 2000-2002 was significantly
higher for females, 67.5 per 100,000, compared to males, 42.7 per 100,000
(Colorado Health and Hospital Association as cited by the CDPHE, in press).
Several anonymous surveys have found that an average of 10% of high school
students report attempting suicide nationally (Kalafat and Elias, 1992). According to
the 1995 Youth Risk Behavioral Surveillancelx system, 22.2% of Colorado students
seriously considered attempting suicide; 16.9% made a plan about how they would
attempt suicide, and 8.2% attempted suicide in the past year (Center for Disease
Control, 1995). According to the AAS, suicidal teens report seeking help from peers
before talking to an adult. These data indicated that youth experienced, and were
exposed to peers that were experiencing, a spectrum of suicidal behaviors.
In 1998, in response to the consistently high rate of suicide among Coloradans, the
Governors Suicide Prevention Advisory Commission was formed (CDPHE, n.d.).
The Commissions report led to state legislation establishing the Office of Suicide
>* 1995 YRBS data rather than data from subsequent years, because 1995 was the last year for which
weighted state estimates from the Colorado YRBS data were available.
35


Prevention (OSP) within the Colorado Department of Public Health and
Environment in 2000. The OSP addresses suicide and suicidal behavior among
Coloradans by coordinating statewide suicide prevention efforts. One of the longest
enduring suicide prevention efforts in Colorado has been through suicide prevention
crisis centers, namely suicide prevention crisis hotlines.
Currently, Colorado hosts only one suicide prevention hotline certified by the AAS,
which takes calls from the national suicide prevention hotline 1-800-SUICIDE. An
individual within Colorado who calls the national number will be directed to the
Pueblo Suicide Prevention Center. According to the Executive Director, the Pueblo
Suicide Prevention Center has been maintaining a suicide prevention hotline since
1968 and became the smallest center certified by the AAS in 1985 (personal
communication, January 28, 2005).
Suicide Prevention Public Service Announcements
in Colorado
During employment at the Colorado Office of Suicide Prevention, the primary
investigator became familiar with suicide prevention activities across the state. After
learning of the efforts of the suicide prevention hotline and a suicide prevention
public service announcement project from the staff of the Office of Suicide
Prevention, the primary investigator conducted several key informant interviews.
36


For this project, the key informant interviews allowed the investigator to understand
the history surrounding the suicide prevention hotline and the suicide prevention
public service announcement campaign from individuals who were the most
intimately involved with the production and airing of the PSAs and who managed
the Pueblo Suicide Prevention Center.
The OSP launched the suicide prevention campaign in 2003. Comprised of 5,287
televised public service announcements across Colorado, this campaign was one of
the first suicide prevention PSA campaigns in the nation. This special project was
spearheaded by a member of the Office of Suicide Preventions advisory board who
approached AT&T (later Comcast) and requested airtime to spread a message of
suicide prevention. Donated efforts by Capture It Productions, a local production
company, led to the development of three suicide prevention messages. These
announcements displayed suicide statistics and included narration from an adult
male, adult female and an adolescent female. As illustrated in Figure 1.1, PSAs
aired intermittently between December 13, 2002 through January 9, 2003 and again
from April 17th until July 2, 2003. PSAs aired in the most uninterrupted block of
time from April 17th until July 2nd. However, even during this period there were
three weeks (May 1st May 7th, May 22nd May 28th, and June 12th to June 18th) in
which PSAs were not aired. There were also three days without PSAs during the
37


December 2002 through January 2003 blocks. Figure 1.1 illustrates the time period
and quantity of PSAs, which sporadically aired from December 2002 July 2003.
As demonstrated by Figure 1.1, the number of PSAs varied dramatically between
the initial block of donated time (airing from December 2002 -January 2003) and
the later block of donated airtime (April July 2003), whereas few PSAs aired in
the first block (PSAs aired from 0-18 per day) compared to the second block (0-139
per day).
Figure 1.1. Number of Suicide Prevention PSAs from December 2002 July
2003
Number of PSAs
The first and second block of donated time also varied in the percentage of spots per
different day segment. Figures 1.2 and 1.3 illustrate the percentages of PSAs per
38


time of day. Numbers within each pie slice represent the popularity of each time slot
on a scale from 1-5, with 5 being the most popular time, based on Comcast ratings.
According to these percentages, 29.3% of the 297 PSAs in the first block of donated
time (December 2002 January 2003) were aired in prime time (from 6:00pm to
10:59pm). In the second block of donated time, only 12.9% of 4990 PSAs were
aired during prime time hours.
Figure 1.2. Percent of PSAs per Time Segment for the
First Block (December 2002- January 2003)
1:00 am 5:59 am B 6:00 am 8:59 am
9:00 am 2:59 pm B 3:00 pm -5:59 pm
B6:00 pm 10:59 pm (Best Time)____________ 11:00 pm 12:59 am
Note: Numbers within each pie slice corresponds to the popularity or value of each time segment, with 5 being
the most popular and one being the least popular.
39


Figure 1.3. Percent of PSAs per Time Segment for the
Second Block (April July 2003)
1:00 am 5:59 am B 6:00 am 8:59 am
9:00 am-2:59 pm S 3:00 pm-5:59 pm
6:00 pm 10:59 pm (Best Time)____________E3 11:00 pm 12:59 am
Note: Numbers within each pie slice corresponds to the popularity or value of each time segment, with 5 being
the most popular and one being the least popular.
As illustrated by Figure 1.3, during the second block of the PSA campaign, a
smaller percentage of PSAs aired during the best rated time spots (percentages
dropped from 29.3% to 12.9%) and more PSAs were aired in the worst time
segment (percentages increased from 0% to 20.6%). However, due to the overall
increase in the number of PSAs in the second block of time (4990) compared to the
first block (297), there was a larger number of PSAs aired in the best time slot.
40


Purpose of Study
According to Litman et al.s (1965) theory of how a hotline works, individuals who
have experienced a stressful event must be willing to seek help and be aware of the
hotline. By promoting help-seeking behaviors, that is, utilization of the hotline,
more callers may use the hotline and be referred to treatment. The more individuals,
especially individuals at risk, who receive treatment, the more effective the hotline.
According to Gould and Kramer, (2001) and based on the findings of researchers
such as Miller et al. (1984), a preventive effect for populations who use the hotline
has been demonstrated. Therefore, an increase in the number of calls to the hotline
may ultimately reduce suicide attempts and deaths. The purpose of this study was to
determine whether the suicide prevention PSA campaign in Colorado led to an
increase in total hotline calls and an increase in calls from vulnerable populations.
Such evidence would indicate that PSAs are a viable means to increase the
utilization of hotlines. Given how little research has been conducted in this area, this
was an exploratory study of PSAs as a tool in suicide prevention activities.
Anecdotal evidence from the Executive Director of Pueblo Suicide Prevention
Center suggested that Colorados suicide prevention center received an increased
number of calls in response to the 2003 PSAs. It was unclear whether there was a
statistically significant increase of callers and, specifically, male callers who are at
41


greater risk for death by suicide than female callers and depressed, suicidal or
callers experiencing an emergency following the airing of the suicide prevention
messages. With permission from Pueblo Suicide Prevention Center and the Office
of Suicide Prevention, the primary investigator analyzed the suicide prevention
centers caller data and the airing of the public service announcements to explore
whether there was a relationship between the public service announcements and
hotline calls.
Since evaluations of hotlines report that hotline callers were not typically at high
risk for suicide and were more likely to be female, the primary investigator also
explored whether the implementation of a public service campaign targeting a broad
audience would lead to an increase in callers from populations who traditionally do
not utilize hotlines. The primary investigator predicted that PSAs would facilitate
more calls to the hotlines by educating viewers on the warning signs of suicidal
behaviors and encouraging viewers to seek help by calling the suicide prevention
hotline, 1-800 SUICIDE. Since this was a global campaign (PSAs targeted adults,
seniors, and adolescents of both genders), the primary investigator predicted an
increase in overall calls as well as an increase in calls from vulnerable populations
(males and distressed individuals) who may not have otherwise known about
42


available help or who were inhibited due to the stigma surrounding help-seeking and
mental illness.
Since Homiks (2002) review suggested that the duration of public service
campaigns influences whether these messages impact behavior, the investigator
further questioned whether the hotline received more calls toward the end of the
suicide prevention public service announcement project. Specifically, the primary
investigator explored whether there was a higher number of calls in 2003 (the year
the public service campaign was implemented) compared to the year prior to the
initiation of the campaign. The year subsequent to 2004 was included in the analysis
as well; however, during this year, an unknown proportion of calls were misdirected
to crisis prevention centers outside of Colorado. Due to this limitation, 2004 was
included in selected analytic procedures but was not included in the primary
investigators hypotheses.
In addition, differences between the first and final quarters of 2003 were examined
in order to examine any potential cumulative effects of exposure to PSAs on the
number of calls. To answer these research questions, the current study utilized two
different and unique anonymous existing data sets compiled by the Pueblo Suicide
43


Prevention Center and Comcast. In summary, the current study tested the following
hypotheses:
Hypothesis (1) The Pueblo Suicide Prevention Center hotline experienced a
statistically significant higher number of callers during days/weeks with a higher
saturation of public service announcements.
Hypothesis (2) The Pueblo Suicide Prevention Center hotline experienced a
statistically significant higher number of male callers during days/weeks with higher
saturation of public service announcements.
Hypothesis: (3) The Pueblo Suicide Prevention Center hotline experienced a
significantly higher number of distressed callers (Depression, Suicide, Emergency)
during days/weeks with a higher saturation of public service announcements.
Hypothesis (4) The Pueblo Suicide Prevention Center hotline experienced a
significant higher number of callers in 2003 (intervention year) versus 2002.
Hypothesis (5) The Pueblo Suicide Prevention Center hotline experienced a
significant higher number of male callers in 2003 (intervention year) versus 2002.
Hypothesis (6) The Pueblo Suicide Prevention Center hotline experienced a
significant higher number of distressed callers (Depression, Suicide, Emergency) in
2003 (intervention year) versus 2002.
44


Hypothesis (7) The Pueblo Suicide Prevention Center hotline experienced a
significantly higher number of callers during in the third quarter of 2003 (July
September) versus earlier quarters in 2003.
45


CHAPTER 2
METHODS
This study primarily examined archival data on suicide prevention hotline calls and
records of public service announcements. In addition to gaining access to this
secondary data, the primary investigator conducted key informant interviews to gain
a historical understanding of the public service announcement project and the
development and operation of the Pueblo Suicide Prevention Center. This
information guided how archival data was organized and analyzed.
Participants
Key Informant Interviews
To understand suicide prevention hotline efforts in Colorado and the process of data
collection for the 1-800-SUICIDE calls from Colorado, the primary investigator
conducted a series of interviews with four key informants, that is, individuals who
had different yet important roles in the suicide prevention PSA campaign and the
Pueblo Suicide Prevention Center in Colorado. The primary investigator conducted
several phone interviews with Eleanor Hamm, Executive Director of the Pueblo
Suicide Prevention Center in order to learn how telephone counselors were trained
46


in crisis intervention and data collection, the criteria for classifying the reason for
calling, the history of the Pueblo Suicide Prevention Crisis Center including the
AAS certification process, and the daily operation of the Pueblo Suicide Prevention
Center. See Appendix A for a list of interview questions used with Eleanor Hamm
to inform this study. Eleanor Hamm was selected as a key informant due to her
position at the Pueblo Suicide Prevention Center and her longevity within the field
of suicide prevention.
In addition, the primary investigator conducted key informant interviews with
individuals who were integral in the development and airing of the suicide
prevention PSAs. Key informants included Michael Cafasso, Joel Bidman, and
Trisha Jorgensen. Michael Cafasso, an Office of Suicide Prevention advisory board
member in 2001, initiated the PSA project and was granted free airtime for the
airing of the public service announcements. The primary investigator conducted the
interview in order to learn about the history of the suicide prevention PSA
campaign, how the messages were developed and produced, and how the project
received donated airtime. See Appendix B for a list of the interview questions that
were asked of Michael Cafasso. Through a key informant interview with Joel
Bidman, an Account Executive at Comcast, the primary investigator gained insight
into the scheduling of airings of the suicide prevention PSAs. Joel Bidman also
47


provided information on the value of donated time and on how to interpret Comcast
records, referred to as the Affidavit of Performance. See Appendix C for a list of
interview questions used with Joel Bidman. An interview with Trisha Jorgensen, the
Local Sales Manager, also was conducted to obtain additional information regarding
the PSA data. Trisha Jorgensen was selected as a key informant because of her
position at Comcast and because she had information regarding later stages of the
project that Joel Bidman, who left Comcast, did not have. See Appendix D for a list
of interview questions used with Trisha Jorgensen.
During interviews with Comcast staff, Joel Bidman and Trisha Jorgensen provided a
ranking of each channel from one to five, based on industry standards of the
channels popularity. Trisha Jorgensen also reported a 5 point ranking of each time
spot based on the industry cost of airtime based on time of day. In addition, Trisha
Jorgensen recommended equally weighting channel popularity and time of airing,
based on industry standards. Information gathered through key informant interviews
was not available from the archival data and was helpful in guiding appropriate
entry, coding and analysis of the data.
48


Archival Data
Secondary data examined in this study consisted of two datasets, hotline calls and
Comcast records on the airing of public service announcements. Hotline caller
information was collected by telephone counselors from the Pueblo Suicide
Prevention Center and recorded on Caller Log Sheets. See Appendix E for a sample
Caller Log Sheet used by the telephone counselors from the crisis center. Caller
information reflected anonymous data collection from individuals who called 1-800
SUICIDE from Colorado, collected for the purpose of describing the volume, reason
for calls, and type of callers. The Executive Director of the Pueblo Suicide
Prevention Center regularly collects and reports this information to grant funders
such as the Colorado Office of Suicide Prevention to describe hotline activities. The
Executive Director indicated that, each year, the Pueblo Suicide Prevention Center
receives approximately 2,000 calls through the 1-800-SUICIDE number.
Measures
Key Informant Interviews
The key informant telephone interviews with Eleanor Hamm, Mike Cafasso, Joel
Bidman, and Trisha Jorgensen occurred over several points in time in the
development of the current study. The investigator prepared several questions
49


regarding the history of the hotline or the public service announcement project and
the interpretation of secondary data. A description of interview topics per key
informant is provided in Table 2.1.
Table 2.1
A Description of Interview Topics for each Key Informant
Interview Topics Eleanor Hamm Mike Cafasso Joel Bidman Trisha Jorgensen
History of Pueblo Suicide Prevention Crisis Center V
Interpreting Caller Log Sheet V
PSA Project History V V
Creation of PSAs V
Interpreting Affidavit of Performance V
Value of Donated Time V V
Assistance in Creating Independent Variable V V
Popularity Ratings of Channels V V
Weighting of Day Segments V
50


For a list of specific interview questions, see Appendices A D. Initial interviews
ranged from 45 90 minutes. Approximately four follow-up interviews occurred
with Eleanor Hamm ranging from 30-60 minutes each. The primary investigator
recorded interviews through hand written notes. This information was then used to
summarize information about the project in the background section of this document
and to inform the preparation and analysis of data.
Archival Data
Caller Data. Information for every suicide hotline caller was recorded on the
Caller Log Sheet developed by the Pueblo Suicide Prevention Center and enhanced
during the 1985 certification process with the American Association of Suicidology
(AAS). See Appendix E for a sample Caller Log Sheet Form. The AAS certification
process requires that certain information is collected by suicide prevention centers,
but it does not require a specific data collection tool. For this reason, Caller Log
Sheets vary from state to state. Attempts to replicate this study in other states will
vary according to the unique caller data collection procedures and variables in other
states.
Telephone counselors working for the Pueblo Suicide Prevention Center recorded
demographic information related to the caller as well as information related to the
51


telephone counseling session. Counselors recorded the date of the call, gender, age
(adult or teenx), city and county of caller, specific reason precipitating the hotline
call (depression, suicidality, divorce, unemployment, etc.), and the target of the call
(for self, family, or friend). In addition, when recording call information, the
telephone counselors categorized the main reason or focus of the call into one of
five categories referred to as Depression, Suicide, Emergency, Information, and
General Mental Health. Categorization of calls was based on whether calls primarily
addressed symptoms of depression, suicidal ideation (thoughts of suicide), general
requests for information, questions or concerns about general mental health issues
linked to specific precipitating events (e.g., unemployment, sexual orientation,
recent stressful events, housing), or an immediate emergency (e.g., imminent
suicide or homicide). Although callers categorized as suicidal also might be
suffering from depression or depressed mood, counselors were trained to categorize
calls as Depression only when the caller reported depressed mood without suicidal
thoughts/ideation. In contrast, if the caller reported suicidal thoughts and a plan for
ending his/her life, this call was classified in either the suicide category or the
emergency category, depending on the callers current level of risk. According to
the Executive Director of the Pueblo Suicide Prevention Center, the calls labeled
emergency were calls which required an immediate response by the telephone
counselor due to a life or death situation (personal communication, January 19,
x The Pueblo Suicide Prevention Center does not maintain records on the exact age of the caller.
52


2005). Typically, emergency calls required an immediate admission into the
hospital and/or a police intervention.
To analyze the subset of callers who were at risk for suicide at the time of their call,
the primary investigator combined the depression, suicide, and emergency
categories into a new variable referred to as Distressed. This decision was
informed by psychological autopsy studies which indicate that mood disorders,
particularly depression, and feelings of hopelessness were associated with death by
suicide (Cavanagh, Carson, Sharpe, and Lawire, 2003). These three categories
represented the varying levels of lethality of risk (depression being the lowest and
emergency being the highest). Combined, these categories represented callers who
were at heightened risk for suicide at the time of their call compared to other hotline
callers. Moreover, the primary investigator combined these categories due to the
relatively small percentage of the calls classified as Emergency and Suicide.
Applicants who passed the initial interview process at the Pueblo Suicide Prevention
Center received 40 hours of intensive training in crisis intervention and data
collection. The training and the specification of data collection standards should
increase inter-rater reliability among telephone counselors recording and classifying
calls. The majority of telephone counselors were graduate students in a counseling
53


program at local universities. Counselors were required to volunteer for one six-
hour shift per week. Any additional hours over the expected volunteer hours were
financially compensated ($25 per additional six hour shift). According to the
Executive Director, efforts to compensate counselors have decreased attrition
among telephone hotline counselors. Maintaining a well-trained staff increases
efficiency and reliable data collection. Relative stability among the individuals who
performed the role of telephone counselors made the primary investigator
reasonably confident in the inter-rater reliability of the caller data.
PSA Data. The Colorado Office of Suicide Prevention gave the primary
investigator hard copies of the Affidavit of Performance, containing information
regarding the airing of the public service announcements. Comcast recorded the
channel, date, day of the week, time, zone (geographic area), title of the
announcement, and duration of the public service announcements on the Affidavit
of Performance. A sample of the Affidavit of Performance is attached in Appendix
F. Based on these records, three suicide prevention messages were aired. Each
message was 30 seconds long and was aired at all hours, on several different
channels, across Colorado.xl
X1 As illustrated by Figure 1.2, PSAs did not air from 1:00 am 5:59 am during the first block of
donated time.
54


Procedures
Key Informant Interviews
After learning of the PSA project through Colorados Office of Suicide Prevention,
the primary investigator initially contacted Eleanor Hamm, the Executive Director
of the Pueblo Suicide Prevention Crisis Center. The purpose of this call was to
introduce the study and request Eleanor Hamm to participate in the study through
telephone interviews and the sharing of anonymous caller data. Concurrently, the
primary investigator contacted Joel Bidman, the Office of Suicide Preventions
contact at Comcast. Interviews with Joel led the primary investigator to Michael
Cafasso, who was the advisory board member who spearheaded the project. Michael
Cafasso also provided the primary investigator with a copy of the video-taped public
service announcements. During the course of the study, additional information
regarding the PSAs was provided by Trisha Jorgensen, who filled in information
about the project that occurred after Joel Bidman left Comcast. Interview questions
were generated during developmental stages of the study to inform the study design.
55


Archival Data
Caller Data. The primary investigator received the 2002, 2003, and 2004
caller data electronically from the Executive Director of the Pueblo Suicide
Prevention Center. The primary investigator did not have access to the Caller Log
Sheets. Instead the primary investigator received a summary of calls organized by
each call within Microsoft Excel. See Appendix I for a sample of caller data
received by the primary investigator. After importing the information into SPSS, the
primary investigator organized caller information by date. New variables created or
recoded for analysis included date, number of calls received on this date, number of
male callers, and the number of callers categorized as Depression, Suicide,
Emergency, Information, and General Mental Health. Information gained from key
informant interviews regarding the criteria for classifying the focus of each call
guided the primary investigator in deciding which categories to be included in an
analysis of distressed callers. For a summary of how variables were organized in
SPSS, see Appendix G.
PSA Data. The primary investigator received hard copies of this data (the
Affidavit of Performance) from the Office of Suicide Prevention. The primary
investigator entered the following information regarding aired PSAs into an Excel
56


spreadsheet for data preparation purposes: the date, day of the week, average
popularity of channels which aired PSAs per day, and the number of spots, defined
as the number of PSAs per day. Then, data were imported into SPSS. For a
summary of how PSA variables were organized in SPSS, see Appendix H.
Since many messages aired every day, the primary investigator collapsed PSA data
by day and later by week.
In order to examine the relationship between viewers exposure to PSAs and the
number of calls made to the hotline, a variable, the saturation of PSAs, was created
based on available information about PSAs. To create the saturation of PSAs
variable, the primary investigator weighted two key variables, time slot and channel,
to capture their relative influence on overall exposure. Rankings for the time spots
and channel popularity were provided (on a 1-5 ordinal scale) by professionals at
Comcast during key informant interviews. In order to collapse the data by day and
compute the independent variable, the primary investigator computed the number of
messages aired per time spot per day and the average popularity rating of channels
on which PSAs were aired each day. These procedures were repeated for another
unit of analysis, weekly calls.
57


Since experts at Comcast stated that the channel and timing were equally important
in understanding the overall viewership for the airing of an individual PSA, these
variables were equally weighted. The primary investigator multiplied the total
number of PSA spots that aired per day by the ranking for the time segment,
calculated the sum of these scores and multiplied this product by the average
popularity of all channels that aired that day. See Figure 2.1 for a summary of the
computation of the independent variable.
The saturation of PSAs may also be influenced by other variables such as the
intended audience for channels, the zones (cities and counties) which viewed the
PSA, and the program which the PSAs interrupted. For the purpose of this study,
several variables were not included in the development of this investigations
saturation variable because information was either not available or did not appear to
add incremental validity to the independent variable. Data concerning differences in
exposure according to which program the PSA interrupted was not available,
however. In addition, data concerning the intended audience was not included in
analyses, because programs spanned a broad range of ages that did not significantly
vary across channels. Finally, zone was not included since PSAs were evenly
distributed across geographic zones. The primary investigator created a variable that
represented the most vital elements of airtime that would influence the probability
58


of a viewers exposure to a PSA. Since no other research on suicide prevention
messages and caller behaviors has been conducted previously, the primary
investigator anticipates that future studies may incorporate more sophisticated
means of capturing the saturation of PSAs.
59


Figure 2.1. Construction of Independent Variable (IV): Saturation of Public Service
Announcements
Number of PSAs (per weighted time spot)
Time Spot #of PSAs' X Ranking"
6:00 am 8:59 am Varied from 0-139 spots per day 4
9:00 am 2:59 pm 2
3:00 pm 5:59 pm 3
6:00 pm 10:59 pm 5
11:00 pm 12:59 am 3
1:00 am 5:59 am 1
Average popularity of channels
Channel Ratingxiv
CMDY 3
CNBC 3
E! 2
ESPN2 3
FAM 4
Hallmark 1
HGTV 5
HIST 1
LIFE 4
MTV 5
OLN 1
OXYG 2
SCIFI 1
SPEN 1
TLC 1
TNN 3
TTC 5
TVLD 1
USA 5
WTBS 5
't_______________________________________________^
Equal weighting
Independent
Variable
<
£
o
§
1
2
X1' See F igure 1.1. for a distribution of the PSAs aired during the suicide prevention campaign.
X1" Scale ranged from 1-5 with 5 being the most popular time spot.
v Scale ranged from 1-5 with 5 being the most popular channel.
60


CHAPTER 3
RESULTS
Testing Hypotheses 1-3: Bivariate Regressions
Since the primary investigator explored whether one independent continuous
variable (saturation of PSAs) was related to a continuous dependent variable
(number of calls, number of male callers, number of distressed callers), bivariate
regressions were conducted to test the first three hypotheses. The standard alpha
level of .05 was used to determine significance in order to minimize the chance of
falsely rejecting the null hypothesis (Type I error). The investigator conducted all
%
regressions by day and then by week. Running the regression by day provided a
maximum number of cases for analysis to ensure a large enough effect size and
minimize the chance of failing to reject the null hypothesis (Type II error). However
since there was no reason to believe that a behavioral response to a PSA may occur
in the same day as viewing the public service announcement, an analysis of the data
by week was also conducted. These weekly analyses provided a smaller number of
cases but took into account any potential lag effects (time needed between viewing a
PSA and acting on the message). In addition, regressions were run for different
periods of time. Time period A was comprised of data for both blocks of donated
time plus two weeks before and after first each time period airing of PSA, and time
61


period B was comprised of the second block of donated time only, including two
weeks before and after the PSA campaign during that time (April to July of 2003).
See Figure 3.1 for an illustration of time periods A and B. The purpose of
conducting regressions with different time periods was to explore the caller data
before and after PSAs aired and to see whether the number of PSAs aired during a
given period of time affected the strength of predictive relationship between
independent and dependent variables.
Figure 3.1. Time Periods A and B in Regression Analysis of PSA and Caller
Data
V
Ok'





PSA and Caller Data Time Period A Time Period B
62


The purpose of these regressions was to determine whether an increase in the
saturation of PSAs predicted an increase in the number of callers, the number of
male callers and the number of distressed callers (hypotheses 1-3). Tables 3.1 and
3.2 present the regression data that resulted from the testing of the investigators
hypotheses. Table 3.1 displays the regression results when PSA and caller data was
analyzed by day; Table 3.2 displays the results when PSA and caller data was
analyzed by week.
Table 3.1
Results of Bivariate Regressions for Saturation and DVs by Day
DV df R2 F P slope
Time Period A
No. of callers 157 .08 14.00 <.001 0.0
No. of male callers 157 .04 6.12 .01 0.0
No. of distressed callers 157 .08 13.59 <.001 0.0
Time Period B
No. of callers 101 .14 16.18 <.001 0.0
No. of male callers 101 .07 7.91 <.01 0.0
No. of distressed callers 101 .20 25.31 <.001 0.0
Note. Bolded data indicate that the relationship between the independent variable and dependent variables
was found to be statistically significant. Time Period A was the duration of the PSA project plus two
weeks before and after each block of donated time. Time Period B was the April July block of donated
time plus two weeks before and after this block of donated time. A portion of the gap period during which
no PSAs aired was excluded from both sets of calculations (see Figure 3.1).
63


Table 3.2
Results of Bivariate Regressions for Saturation and DVs by Week
DV df R2 F P slope
Time Period A
No. of callers 24 .17 4.78 .04 0.0
No. of male callers 24 .14 3.96 .06 0.0
No. of distressed callers 24 .17 4.79 .04 0.0
Time Period B
No. of callers 14 .44 10.91 <.01 0.0
No. of male callers 14 .24 4.53 .05 0.0
No. of distressed callers 14 .52 15.32 <.01 0.0
Note. Bolded data indicate that the relationship between the independent variable and dependent variables
was found to be statistically significant. Time Period A was the duration of the PSA project plus two
weeks before and after each block of donated time. Time Period B was the April July block of donated
time plus two weeks before and after this block of donated time. A portion of the gap period during which
no PSAs aired was excluded from both sets of calculations (see Figure 3.1).
After testing the first three hypotheses, the primary investigator conducted
additional regressions to test whether other variables, such as the number of spots
alone (without influence of channel or timing) influenced the number of calls. The
purpose of testing the impact of the number of spots was to determine whether
channel popularity and time of day were important factors in predicting changes in
64


caller behavior. Twelve regressions were performed identical to the regressions
described above to test for this relationship. The results of these regressions were
described in Table 3.3 and 3.4.
Table 3.3
Results of Bivariate Regressions for Spots and DVs by Day
DV df R2 F P slope
Time Period A
No. of callers 157 .09 16.05 <.001 .03
No. of male callers 157 .05 7.43 <.01 .01
No. of distressed callers 157 .08 13.84 <.001 .02
Time Period B
No. of callers 101 .158 18.96 <.001 .04
No. of male callers 101 .087 9.59 <.01 .01
No. of distressed callers 101 .208 26.53 <.001 .02
Note. Bolded data indicate that the relationship between the independent variable and dependent variables
was found to be statistically significant. Time Period A was the duration of the PSA project plus two
weeks before and after each block of donated time. Time Period B was the April July block of donated
time plus two weeks before and after this block of donated time. A portion of the gap period during which
no PSAs aired was excluded from both sets of calculations (see Figure 3.1).
65


Table 3.4
Results of Bivariate Regressions for Spots and DVs by Week
DV df R2 F P slope
Time Period A
No. of callers 24 .18 5.27 .03 .03
No. of male callers 24 .15 4.25 .05 .01
No. of distressed callers 24 .17 4.87 .04 .02
Time Period B
No. of callers 14 .47 12.25 <.01 .03
No. of male callers 14 .25 4.58 .05 .01
No. of distressed caller 14 .55 16.91 <.001 .03
Note. Bolded data indicate that the relationship between the independent variable and dependent variables was
found to be statistically significant. Time Period A was the duration of the PSA project plus two weeks before
and after each block of donated time. Time Period B was the April July block of donated time plus two weeks
before and after this block of donated time. A portion of the gap period during which no PSAs aired was
excluded from both sets of calculations (see Figure 3.1).
Analysis of the saturation variable revealed that saturation accounted for 8% (Time
Period A by day) and 44 % (Time Period B by week) of the variance in the overall
number of callers when analyzed by day and by week respectively. The number of
spots, however, accounted for 9% (Time Period A by day) and 48% (Time Period B
by week) of the variance in the caller data.
66


The primary investigator entered the saturation and spots variables by day for Time
Period B into a multiple regression model with forward selection and found that the
spots variable was statistically different from the saturation variable Flange (1,100)
= 3.82, p=.05. Multiple regression models did not show a statistical significant
difference between the saturation and the spots variable for other units of analysis
and time periods. However, the slope remained 0.0 for the saturation, while it
ranged from .01 .04 for the spots variable. This indicates that for every 100 PSAs
aired in Colorado, the hotline received an additional one to four calls. Assuming that
these individuals received adequate counseling and referral services, an increase of
one to four distressed individuals received help during a crisis. Due to the greater
amount of variance accounted for by the spots variable and the higher slope, the
primary investigator determined that the spots variable was a better predictor of
caller behavior.
Regression results also indicated that the relationship between PSAs and the number
of calls to the hotline was stronger when regressions were conducted on the second
block of donated time (Time Period B). When PSAs aired for a continual period and
aired between 46- 139 times per day (except for brief interruptions in which no daily
PSAs aired), the data indicate that PSAs accounted for close to half (47%) of the
variance or R in caller data. However, the variance dropped to 18% when the
67


December/ January block of time was included in the analysis (when PSAs aired
between 0-18 times per day). This may suggest that a minimum number of PSAs
must be aired to influence an increase in calls to the hotline. In other words, there
may a minimum threshold of PSAs needed to increase calls.
In addition, weekly analysis of PSA airings resulted in greater variance (18% 47%)
accounted for in the caller data versus a daily (9% -16%) unit of analysis. The
observed differences between R2 may have indicated that individuals do not
behaviorally respond to suicide prevention messages the same day they view PSAs.
Thus, there may have been a lag effect in response to PSAs which may explain the
difference in results according to the daily/ weekly unit of analysis.
Two different SPSS files were created to facilitate analysis by different levels of
analysis. For example, regressions were run from one file which organized the PSA
and caller data by day and for Time Period A. For the purposes of this study, data
files could not be merged for the purposes of conducting multiple regressions
because variables were measured at four different levels of analysis. In other words,
the data were stored in a way that precluded significance testing between the R2 of
Time Period A and B and of the daily versus weekly unit of analysis. Differences in
68


2 t
R between the units of analysis were noteworthy, however, given the exploratory
nature of this study.
Testing Hypotheses 4-6: One Wav ANOVAs
Since time of year, day of the week and month of the year were categorical variables
and the dependent variable was continuous (number of callers, number of male
callers, and the number of distressed callers), an ANOVA was the best statistical
procedure to answer the second set of research questions. The purpose of
conducting these ANOVAs was to determine whether there was a significant
difference between 2002 (year before PSAs aired) and 2003 (year PSAs aired). 2004
data (the year after PSAs aired) was also examined. However, as mentioned earlier,
an unknown proportion of calls from within Colorado were routed out of state;
therefore, results must be interpreted with caution.
Additional ANOVAs were performed to determine whether there was a significant
difference between the first and final quarters of 2003, in order to examine any
potential cumulative effects of exposure to PSAs on the number of calls. These
results are described under the heading Testing Hypothesis 7
69


One Way ANOVAs revealed a significant difference between 2002, 2003, and 2004
and the total number of daily calls, F(2, 1082) = 75.425, £<.001, r|2=.122; number
of male callers F(2, 1082) = 19.408, £<.001, r|2=.035; and distressed callers F(2,
1082) = 48.623,£<-001, n2=082.
Post hoc tests were run to explore the significance of the differences between the
numbers of callers per year. The Tukey post hoc test is typically conducted after
ANOVA results indicate a significant F value; Tukey compares each pair of means
to determine statistical significance. The Tukey post hoc test indicated a statistically
significant greater number of callers overall in 2003 (M= 9.72) versus 2002
(M=5.67) and 2004 (M=8.85). This supported hypothesis number four that the
Pueblo Suicide Prevention Center experienced a greater number of callers overall
during the year PSAs aired versus the years before and after. It was noteworthy that
in 2004, the year after the intervention, the mean number of calls (M=8.85) was
significantly higher than 2002 (M=5.67), the year before the PSA project was
implemented. This may have suggested that the increase in calls in 2003 was
somewhat maintained in 2004 but not at the same intensity as the intervention year.
Since an unknown number of calls to the hotline in 2004 were not included in this
analysis, the difference between 2004 and 2002 may be much larger. Further, while
post hoc tests did not detect a difference between 2003 and 2004, limitations to the
70


2004 data did not allow the primary investigator to conclude that the difference was
not statistically significant.
In addition, the Tukey post hoc test indicated a greater number of distressed callers
in 2003, the intervention year (M=4.70) versus years that did not air PSAs, 2002
(M=2.90) and 2004 (M=4.18). These data supported hypothesis number six that the
Pueblo Suicide Prevention Center experienced a greater number of distressed callers
overall during the year PSAs aired versus the years before and after. Figure 3.2
illustrates the mean values for the total number of overall and distressed calls for
2002, 2003 and 2004 per day. The increase in distressed callers mirrored the overall
rise in total callers.
71


Figure 3.2. Mean Number of Overall and Distressed Suicide Hotline Callers
Before, During, and After Suicide Prevention PSA Project
As described in the Methods section of this paper, the primary investigator
combined calls categorized as depression, suicide, and emergency to determine the
number of distressed callers.
72


Figure 3.3 describes the three categories that comprise the total number of distressed
callers and displays the means of each category by week.xv In addition, mean values
for the remaining categories, general mental health and information are displayed in
this figure. One Way Analysis of Variance was performed to test whether there was
a significant difference between 2002, 2003, and 2004 for mean values per week of
depression, suicide, and emergency. Analyses revealed a statistically significant
difference between years and the number of calls categorized for depression F(2,
156) = 14.532, £<.001, r|2=. 157 and emergencyJ:(2,156) = 28.508,£<.001,
r| =.268. No statistically significant differences between years and the number of
suicide calls were detected F(2, 156) = 2.329, £=.101, r|2=.029.
Similar to the relationship shown in Figure 3.1, there was an increase in depressed,
suicidal, and emergency calls in 2003, which appeared to be maintained in 2004.
However, conclusions regarding differences between 2003 and 2004 were limited.
As illustrated by Figure 3.3, the most common distressed-related call was by callers
reporting depression. The most common reason overall was classified as a general
mental health call.
xv Mean values were aggregated in Figure 3.2 per week due to the low frequency of Emergency and
Suicide related calls.
73


Figure 3.3. Mean Number of Calls per Week by Call Classification
Depression HSuicide -- Emergency
| XInformation_________X General Mental Health
Also, Tukey post hoc tests indicated a statistically significant higher number of male
callers to the Colorado hotline in 2003 (M=3.41) versus 2002 (M=2.28) and 2004
(M=2.84). This supported hypothesis five that a greater number of males utilized the
suicide prevention hotline during the year in which the suicide prevention public
service announcements aired. It was unclear whether the difference between 2003
and 2004 reflected an actual decrease in male callers in 2004. As illustrated by the
Figure 3.4, there were consistently more female callers than males but an increased
number of males utilized the hotline during the intervention year compared to other
years.
74


Figure 3.4. Mean Number of Male and Female Suicide Hotline Callers Before,
During, and After Suicide Prevention PSA Project
U
CL
cd
U
Ct-I
O
3
Z
C
2
Testing Hypothesis 7
The primary investigator hypothesized that a greater number of individuals would
call the suicide prevention hotline toward the end of the suicide prevention
campaign due to a potential cumulative effect of PSAs from the beginning of the
campaign. Since the last messages occurred in the first week of July 2003, the
campaign ended in the third quarter of the year. To test whether there was a greater
number of callers in the third quarter (the end of the PSA project versus the first
quarter), three One Way ANOVAs were conducted on total number of callers
£(3,361)= 4.664, £=.003, r|2=.037; number of male callers; £(3,361)= .553 £=.646,
75


r|2= 005 ; the number of distressed callers £(3,361)= 1.761, £=.154, r|2=.014 and the
number of females callers F(3,361)=7.224,£<.001, r|2=.057. Significant F values
were found between first and third quarters for the total number of callers and the
number of female callers only. Tukey post hoc tests conducted on 2003 caller data
overall indicated a significant difference between the 3rd (M=l 1.43) and the 2nd
quarter (M=8.48) of the year for all calls. In addition, Tukey post hoc tests for the
number of female callers resulted in statistically significantly differences between
the 3rd quarter (M=7.78) and the 1st (M=5.96) as well as the 2nd (M=5.12) quarter.
This suggested a potential cumulative effect for the PSA campaign for the overall
number of calls driven by an increase in the 3rd quarter by female callers. However,
evidence indicated that any potential cumulative effect did not occur for males or
distressed callers, the populations most vulnerable to suicide.
76


Figure 3.5. Mean Number of Overall, Male, Female and Distressed Callers per
Day per Quarter of the Year in 2003
To further explore differences between the number of calls by quarter of the year,
ANOVAs and post hoc tests were conducted on years 2002 £(3,352)= 3.114,
£=.026, r|2=.026 and 2004 £(3,360)= 16.002, £<.001, rj2=.l 18.
As illustrated in Figure 3.6, the increase in calls in the 3rd quarter (M=l 1.4) of 2003
was not replicated in either 2002 (M=5.1) or 2004 (M=6.0). However, since there
were an unknown number of missing calls in 2004, the decrease in the 3 rd quarter of
2004 must be interpreted with caution. Further, due to the low base rate of calls and
77


Mean number of calls per day
the variability in quarters, several years of caller data would be required to
determine trends in caller data by quarter.
Figure 3.6. Mean Number of Calls per Day per Quarter of the
Year
78


CHAPTER 4
DISCUSSION
The primary investigator explored whether suicide prevention PSAs, which
educated viewers on the warning signs of suicide, urged viewers to seek help for
themselves or others, and promoted the 1-800-SUICIDE hotline telephone number,
would lead the Pueblo Suicide Prevention Crisis Center to experience a greater
number of calls. Of particular interest to the primary investigator was whether the
crisis center would experience a greater number of calls from individuals considered
vulnerable to suicide. This research question led the primary investigator to develop
several hypotheses regarding the relationship between PSAs and caller data from the
Pueblo Suicide Prevention Crisis Center. Analysis of the data suggested that there
was a predictive relationship between the number and saturation of PSAs and the
daily and weekly number of calls, including calls from those individuals most
vulnerable to suicide.
79


Overall Discussion of Findings
Evidence for the Impact of PSAs
Results from the bivariate regressions indicated that increases in the number of
PSAs predicted an increase in the number of overall callers, male callers, and callers
classified by telephone counselors as depressed, suicidal, and in an emergency. This
supported hypotheses 1-3 and indicated that PSAs may be a useful tool in promoting
the use of a suicide prevention hotline crisis number. Moreover, it suggested that
PSAs may promote help-seeking behaviors and enhance suicide prevention efforts
in general. The Pueblo Suicide Prevention Center experienced a greater number of
calls during the year in which PSAs aired, when compared to the previous and
subsequent years. However, while the differences in caller behavior between years
2002 and 2003 suggested that PSAs influenced caller behavior, there may be other
alternative explanations which might account for the increase in calls. For example,
PSAs may have unintentionally increased suicidality in callers which in turn may
have led to the increase in calls during the intervention year. However, the ANOVA
analysis along with the results of the regression indicated that PSAs were a viable
explanation for the change in caller behavior in 2003. Overall data analysis from
both independent variables (saturation and spots) predicted an increase in overall
calls and subpopulations considered at risk for suicide.
80


Spots as a Predictor of Caller Behavior: A Review of the
Independent Variable
Incorporating the popularity and timings of channels did not predict the number of
calls as strongly as the number of spots or number of times the PSAs aired alone.
This indicated either that the amount of donated time alone influenced the amount
of exposure and, ultimately, the number of callers, or that exposure may not have
been accurately measured by the saturation variable. The saturation variable was
computed by rating the average popularity of channels and the timing on a scale
from one to five. Comcast professionals determined this rating, but the development
of the saturation variable did not use the actual logarithms of viewership used by
companies such as Comcast to calculate the cost of time. The actual gross rating
points (a media term which refers to the rating used to determine the saturation
level of an advertising campaign (PSA Research Center, n.d.)) were not available
for this PSA project, as this is proprietary information. The PSA Research Centers
online reports state that the calculated total number of expected television viewers
for a particular advertising schedule, known as the gross impressions, should be
computed by multiplying the reach and the number of times a commercial was run.
The reach, or number of individuals who were exposed to a message at least once in
an average week, was computed based on the segment of the day (early morning,
daytime, early fringe, prime time, late evening, and late night) and segment of the
81


week (week day or weekend day). While the segments of the day were incorporated
into the saturation variable, the segments of the week were not. Incorporating
elements such as segments of the week into the computation of the saturation
variable may provide future researchers with a more precise measure of PSA
exposure. In addition, other variables, which may influence the measurement of
PSAs, such as the program in which the PSA aired, intended audience, geographic
zone, were not included in this analysis.
The purpose of this preliminary investigation was exploratory and sought to
determine whether there was a relationship between suicide prevention public
service announcements and calls to a suicide prevention hotline. The primary
investigator predicts that future analyses with more sophisticated means of
determining the saturation of a public service announcement campaign may find
these variables important for predicting caller behavior. To achieve a more accurate
operationalization of PSA saturation, suicide prevention researchers must establish a
greater partnership with media evaluators. An additional means to improve the
current study beyond strengthening the independent variable would be to analyze
the intended audience or popularity of individual programs with the type of PSA.
For example, future research may analyze which programs/channels were more
often viewed by elderly males and determine whether PSAs targeting elderly males
82


i
l
i
!


leads to an increase in calls from this demographic population, the population most
at risk for dying by suicide. Despite the limitations of the independent variable,
results indicated a relationship between PSAs and caller behavior based on the
frequency PSAs were aired and the saturation variable created by the primary
investigator. Further studies may find an even stronger relationship between
increases in PSAs and hotline caller data.
However, future research may also find that media estimates of recording the
viewership public service announcements may not be a good measure of the number
of individuals vulnerable to suicide that were reached. For example, prime time
from 6:00 p.m. -11:00 p.m. was considered the time day which has the greatest
amount of television viewers. However, this may not be the time that individuals
who are at risk for suicidal behaviors are most likely to be watching television;
therefore, media formulas for determining saturation of PSAs may not apply to this
population. For example, individuals who were preoccupied with a stressful event or
experiencing a major depression episode may instead watch television on unpopular
channels in the middle of the night while the typical television viewers were asleep.
More research is needed to understand the times that are best to air suicide
prevention PSAs in order to reach those that were most vulnerable to suicide.
83


Potential Lag Effect of PSA Airings on Caller Behavior:
Review of Results Based on the Unit of Analysis
Analysis of PSA and caller data by week indicated a stronger relationship between
PSAs and caller data than analyzing the data by day. Since there was a greater
number of cases in the daily analysis (102 or 158 depending on the time period)
than the weekly analysis (15 or 25), one might expect a stronger relationship due to
greater statistical power. The results, however, suggest a lag effect in that
individuals exposed to suicide prevention messages did not appear to have
responded to the message the same day they viewed the PSA. Future analyses may
further investigate the interval of time for potential lag effects. For example, was the
greatest effect truly one week or a shorter interval of 3 days?
Potential Threshold Effect of PSA Airings: Review of
Time Periods A and B
Performing the regressions on the second block of donated time (April July block
of donated time plus two weeks before and after this block of donated time)
indicated a stronger relationship between PSAs and caller data. As described earlier
in this report, PSAs were intermittently delivered in two blocks of time: between
December 13, 2002 through January 5, 2003 and again from April 17th until July 2,
2003. The PSAs aired in the most uninterrupted block of time from April 17th until
July 2nd. The number of spots and length of time PSAs aired between the earlier
84


and later blocks of time varied, whereas fewer PSAs aired in the December/January
block (PSAs aired from 0-18 per day for a total of 297 spots) and the later block of
airings ranged from (0-139 per day for a total of 4990 spots). Differences in the
strength of the PSA campaign in the first and second time blocks emerged despite
the fact that Comcast aired a higher percentage of PSAs during the best time slot
from December 2002 January 2003 (See Figure 1.2). These results suggest that
PSAs delivered at prime time may not affect behavior unless a sufficient number of
PSAs are aired overall. The differences found in the results for both blocks of time
and the second block alone may indicate a threshold effect, as predicted by Homik
(2002), and/or a confounding difference between December/ January and four
months later. When PSAs aired for a continual period and aired between 46-139
times per day (except for brief interruptions in which no PSAs aired), the data
indicates that PSAs accounted for close to half (46.7%) of the variance or R2 in
caller data. However, the variance dropped to 18% when the December/ January
block of time was included in the analysis.
There may also have been external events in December 2002 and January 2003 or
April July 2003 that caused a change in the variance. According to Noble (1996),
suicide attempts, deaths, and the number of calls to suicide prevention hotlines vary
according to the season, with more deaths and calls in the spring (usually the month
85


of May) and August being the month with the fewest calls to crisis centers. The
current data did not replicate these fluctuations. Results of the One Way ANOVAs
showed a peak of total calls and number of calls and female calls in the 3rd quarter
of 2003 and a statistically significant difference between this quarter and quarter 3
in 2002 and 2004, when PSAs were not aired. This peak in the 3rd quarter was not
present in the caller data in either 2002 or 2004. The additional analysis suggested
that the increase in calls during the weeks of April July 2003 was related to the
increase of PSAs as opposed to temporal fluctuations. However, as mentioned in the
results section of this document, several years of data would be required to
understand the pattern of calls in non-intervention years. These data were not
available.
Potential Cumulative Effects for PSAs on Female Callers:
A Review of Caller Data by Quarters
ANOVA and Tukey post hoc tests found a statistically significant increase in the 3rd
quarter of 2003 for female callers. This increase was similar to results for overall
callers, however, evidence for a cumulative effect for overall callers was weaker
since post hoc tests indicated a difference only between the 2nd and 3rd quarters of
the year. Results do not support a cumulative effect for male or distressed callers.
These findings suggest that lengthening a PSA campaign will not necessarily lead to
86


an increase in calls from populations of greatest interest (males and distressed
individuals).
Alternative Explanations for the Increase in Calls
According to the theory of telephone crisis hotlines first outlined by Litman et al.
(1965), individuals call a suicide prevention telephone hotline in response to a
stressful event, an impulse, a willingness to seek help, and knowledge of the hotline.
Based on the current study, PSAs work within this theory by increasing the number
of individuals who were willing to seek help and were aware of the hotline.
Messages that encourage help-seeking for suicidal thoughts and provide a resource
were directed toward these viewers who may feel inhibited to seek treatment and
who were unaware of local resources.
However, other explanations for the increase in calls were possible. PSAs may have
had deleterious effect on caller behavior. The PSAs, especially the messages that
informed viewers that 55 Coloradoans die every month, may have normalized
suicide as a common option for ending despair. Thus, the PSAs may have invoked
suicidality within viewers and as a result led to an increase in hotline calls.
However, suicide death rates in Colorado (16.0 and 15.7 for 2002 and 2003
respectively) were not significantly different from 2002 to 2003. This suggests that
87