Professional Documents
Culture Documents
December 2010, Vol 100, No. 12 | American Journal of Public Health Knox et al. | Peer Reviewed | Research and Practice | 2457
RESEARCH AND PRACTICE
2458 | Research and Practice | Peer Reviewed | Knox et al. American Journal of Public Health | December 2010, Vol 100, No. 12
RESEARCH AND PRACTICE
development of integrated delivery system and (P < .01). The estimated correlation coefficient results suggest an overall higher level of AFSPP
community action information boards, (6) en- between consecutive quarterly observations is implementation in 2006 than in 2004. Median
hancement of community mental health services, 0.431 (P < .001), and the estimated variance implementations for the 7 prevention do-
and (7) instituting policies. of the disturbance term is 0.513. During the mains in 2004 were all below 90%, and the
The items in the domains are best described postintervention period, in the third quarter of lowest was 56% (Figure 2). In comparison,
as ‘‘operational measurements,’’ after the work 2004, there was a significant upward spike in during 2006 the overall implementation
of Hand5; this kind of measure is also called an suicide rates (P < .001); subsequently, suicide values for 2 of the 7 prevention domains (con-
‘‘indicator measure’’ by Fayers and Hand.6 As rates fell and have remained within the tinuous professional military training and en-
Fayers and Hand6 state, in contrast to psycho- expected range of the low rates seen soon after hancement of community mental health
metric measures, the goal of using indicator initial implementation of the program. services) were 100% for 95% of all bases. For
measures is to construct an index that consists of We observed an inverse relationship be- prevention activities in the 5 other prevention
the combined values of the measured variables. tween population size and suicide rates in the domains in 2006, at least half of the bases
We then used these indexes to calculate an preintervention period: a smaller population were found to have high levels of implemen-
implementation score for each of the 7 pre- size tended to be associated with a higher tation (Figure 3).
vention domains. (To preserve the most infor- suicide rate than that observed for a larger
mation for analyses of implementation levels, population size. All population sizes above DISCUSSION
data were analyzed at the installation level.) This 500 000 occurred during the period from 1981
permitted us to measure operationally whether through 1990, with the population declining In recent years there has been a marked
an air force installation carried out the activities during the remainder of the preintervention increase in research on translating the findings
described in the box on the previous page, which period. The regression model for the pre- and of efficacy and effectiveness studies into actual
are the direct result of complying with the 11 postintervention periods included an indicator health practices.7,8 The AFSPP provides an
Initiatives. Each question was assigned a score of of implementation of the AFSPP in 1997, in opportunity to study the implementation of
1if a respondent answered ‘‘yes’’ and 0 otherwise. addition to population size. It showed a linear public health practices intended to reduce deaths
The data were weighted according to the nonsignificant relationship between the rate from suicide. This opportunity is unique in 3
number of implementation indicators from the and the population size, with a negative slope of ways. First, the AFSPP was developed well
11 Initiatives that were grouped within each of –1.38 (P > .05). Because the population size before implementation science was acknowl-
the 7 prevention domains. Levels of imple- and the intervention period are confounded, edged as a field of study. Second, the current
mentation were then determined across all the regression intervention effect (–4.9; operational structure of the AFSPP evolved over
installations, and the scores were represented P < .05) is larger than in the bivariate analysis time, even though its principal initiatives were
as the percentage of the maximum possible (–2.9; P < .05) unadjusted for population size. described at the outset; thus, attention to sus-
score for each of the 7 domains. Air force A similar trend was found when the relation- tained implementation of its core components
leaders and installation commanders com- ship between the change in population size emerged iteratively. Third, this public health
pleted the 11 Initiatives survey in 2004 and the and suicide rate was investigated. prevention program was not originally devel-
11 Initiatives checklist in 2006, using adminis- To give air force leadership tools for early oped on a theoretical basis, which is now rec-
trative records of activities monitored at each detection of future increases in suicide rates, we ommended as the best way to strengthen the
installation. Data were reported anonymously developed risk indicators on the basis of the credibility of measured outcomes.7 In spite of this
to minimize the potential for reporting bias forecasted suicide rate for 2008 (9.3 per latter shortcoming, the subsequent identifica-
and were compiled both at the base level and at 100 000). Rates less than or equal to 1 standard tion of theoretical prevention domains for the
the level of the 9 major commands, which are deviation from the forecast rate (<12.1 per AFSPP proved to be relatively straightforward.
the operational units of the service. 100 000) were identified as indicators of con- The effects of the AFSPP are inevitably
cern. Rates greater than 1 standard deviation confounded with the activation of the air force
RESULTS from the forecast rate (12.1–14.8 per 100 000) for warfare, beginning immediately after the
were defined as indicators of warning, and attacks of September 11, 2001, and accelerating
Figure 1 depicts the observed quarterly rates greater than 2 standard deviations from into the wars in Afghanistan and Iraq. The
suicide rate from 1991 through 2008. The the forecast rate (>14.8 per 100 000) were possibility that these conflicts had an effect on
horizontal lines represent the mean pre- and identified as critical indicators of a change in suicide rates, regardless of any changes in
postintervention quarterly suicide rates, and the pattern of suicide rates. program content or implementation, cannot be
the deviations from this mean are depicted as Installations reported variation in the extent ruled out. Military morale is expected to be
outliers across the decades. The estimated to which they implemented the AFSPP’s suicide higher at the start of a conflict, particularly after
mean suicide rate per quarter during the in- prevention activities as specified by the 11 a domestic attack, but as these 2 conflicts
tervention period was 2.387 per 100 000, Initiatives. Figures 2 and 3 show the levels of continued, morale may have suffered from
compared with 3.033 per 100 000 for the implementation across the 7 prevention do- a variety of factors, including stop-loss mea-
preintervention mean, for a change of 0.646 mains in 2004 and 2006, respectively. These sures that barred service personnel from
December 2010, Vol 100, No. 12 | American Journal of Public Health Knox et al. | Peer Reviewed | Research and Practice | 2459
RESEARCH AND PRACTICE
Note. The US Air Force Suicide Prevention Program was implemented in 1997.
FIGURE 1—Quarterly suicide rates: US Air Force, 1981–2008.
leaving the military at the end of their enlist- the program’s presumed effectiveness. Our program. In 2006, levels of implementation
ments early in the war, the absence of any new data did not allow us to estimate exactly when were again measured. When suicide levels in
attacks in the United States, and the cumulative implementation efforts diminished; rather, they 2004 and 2006 were compared with levels
effects of repeated deployments on military only gave us a snapshot at the end of 2004, of implementation in 2004 and 2006, it
personnel and family members in areas such as a time when implementation likely had been appeared that diminished implementation of the
relationships and finances. It is beyond the diminishing for several years in the face of AFSPP may have played a role in the reversal of
scope of this study to elucidate any contribu- heavy demands from both the Afghanistan and the program’s apparent effectiveness.
tions from these factors. Iraq wars. Regardless, air force leadership felt it The air force now measures compliance with
The upward spike in suicide rates observed was imperative to address the possibility that established AFSPP procedures on an annual
during 2004 (Figure 1) raised important diminished implementation of the program basis. Organizational capacity for monitoring
questions about whether the 2 ongoing wars, played a role in the increase in suicide rates and compliance with the program is now coupled
a decreasing force size, or diminished imple- initiated actions to ensure community-wide with development of population risk indi-
mentation of the AFSPP had taken its toll upon compliance with all of the components of the cators that are used to monitor suicide rates
2460 | Research and Practice | Peer Reviewed | Knox et al. American Journal of Public Health | December 2010, Vol 100, No. 12
RESEARCH AND PRACTICE
Note. Horizontal lines represent the 5th, 25th, 50th, 75th, and 95th percentiles. Dots represent points beyond the 95th percentile.
FIGURE 2—Levels of US Air Force Suicide Prevention Program (AFSPP) implementation, distributed across installations, in 2004.
for identification of early shifts in patterns of point.6 This is a limitation of the current study relationship between population size and sui-
suicide rates. There are limitations to this when viewed from the perspective of prevention cide rate when left uncorrected for the in-
approach, but it reflects the importance of close science or therapeutic trials. This limitation is tervention effect, rendering this inference
tracking of programmatic activities for reducing addressed to some degree through the develop- somewhat tentative. We also recognize that
deaths from suicide and the critical need to ment of theoretical prevention domains that a reduction in the number of service members
move beyond descriptive, epidemiologic stud- remained constant over time under which the with mental health problems could limit any
ies of suicide risk. measures were grouped and compared. To date, conclusion regarding a sustainable program-
The AFSPP has been continuously and in- however, there has been no external validation matic impact over time. However, air force–
crementally improved since its launch, includ- of these domains and measures. These data are specific data from the Department of Defense
ing the adoption of formalized prevention being used as early-generation studies of imple- Survey of Health Related Behaviors found that,
domains in 2004. The measures of imple- mentation of a multicomponent suicide preven- in 1998, 9.5% of air force personnel received
mentation that were introduced in 2004 and tion program to inform the next generation of mental health care; in 2002, 13.5% received
refined for 2006 represent ongoing continuous implementation studies, which should include mental health care; and in 2005, 13.3% re-
quality improvement efforts. Given this drive such external validation measures. ceived mental health care, suggesting that the
to enhance effectiveness as rapidly as possible, If the relationship between suicide rates and air force is not decreasing its population of
the 2004 and 2006 measures evaluated the population size in the air force were linear, as it personnel with mental health problems.9–11 It
same key implementation prevention domains appears to be when corrected for the inter- also is worth noting in this regard that the air
but were obtained somewhat differently. Al- vention effect of the AFSPP, a higher suicide force encourages early help-seeking behavior for
though studies have demonstrated that this rate would have been expected during the a mental health problem, and 97% of air force
approach is appropriate when effecting changes postintervention years (1997–2008), when the personnel who seek mental health care do not
in large organizations, it results in incrementally air force population was declining. However, experience any negative consequences to their
different measures being taken at each time we are mindful of the nonsignificant inverse military careers as a result.12
December 2010, Vol 100, No. 12 | American Journal of Public Health Knox et al. | Peer Reviewed | Research and Practice | 2461
RESEARCH AND PRACTICE
Note. Horizontal lines represent the 5th, 25th, 50th, 75th, and 95th percentiles. Dots represent points beyond the 95th percentile.
FIGURE 3—Levels of the US Air Force Suicide Prevention Program (AFSPP) implementation, distributed across installations, in 2006.
Knowledge about risk factors for suicide feasible and effective. The AFSPP is the first long- seek help because of stigma may become
includes evidence that aggression, impulsivity, term sustained effort of its kind to serve as an particularly significant in light of a recent
risk-taking, acute and chronic stresses (often example of what communities can accomplish in report based on a prospective study of a US
including interpersonal or occupational related reducing morbidity and mortality attributable to military cohort of 77 047 military active duty,
stress13–18), and alcohol or substance use19,20 are suicidal behaviors if there is ongoing commit- reserve, and National Guard personnel by
powerfully associated with suicidal behaviors, ment to do so. Smith et al.21 These investigators found that
which are moderated in their expression by Because of the wars in Afghanistan and deployed individuals who experienced combat
gender and age. Despite these compelling data, Iraq, we can expect a large population of exposure had a 3-fold increase in new onset of
there has been no reduction in overall suicide combat veterans to experience mental health self-reported posttraumatic stress disorder. In
rates in the US civilian population since the disorders, and many of these individuals may a study by Boscarino,22 veterans with posttrau-
1940s, when national rates fell after the Great not seek care. Stigma attached to mental matic stress disorder continued to be at height-
Depression and during the nation’s involvement health issues is a pervasive cultural phenom- ened risk for suicide 30 years after separation
in World War II. Thus, many policymakers and enon in the general US population, and it is from the service.
clinicians remain uncertain whether systematic even more pronounced in the military. The These findings highlight the importance of
approaches to reducing deaths from suicide are potential reluctance of military personnel to the role of specific combat exposures. It will be
2462 | Research and Practice | Peer Reviewed | Knox et al. American Journal of Public Health | December 2010, Vol 100, No. 12
RESEARCH AND PRACTICE
critical for clinicians who encounter returning W. Talcott is with the 59th Mental Health Squadron, 8. Wang S, Moss JR, Hiller JE. Applicability and trans-
Lackland Air Force Base, San Antonio, TX. Rick L. ferability of interventions in evidence-based public
military personnel to be trained to recognize
Campise is with the 1st Medical Operations Squadron, health. Health Promot Int. 2005;21(1):76–83.
the early risk factors and warning signs of Langley Air Force Base, Hampton, VA. Alina Bajorska is 9. Bray RM, Sanchez RP, Ornstein ML, et al. Department
suicidal behaviors, and specifically those asso- with the Department of Community and Preventive Medi- of Defense Survey of Health Related Behaviors Among
cine, University of Rochester Medical Center, Rochester. Xin
ciated with combat exposures. Although an Active Duty Military Personnel: 1998. Washington, DC:
Tu is with the Department of Biostatistics and Computa- US Dept of Defense; 1999.
earlier study23 carried out among combat- tional Biology, University of Rochester Medical Center,
Rochester. Eric D. Caine is with the Department of 10. Bray RM, Hourani LL, Rae KL, et al. Department
exposed Vietnam veterans found a significant
Psychiatry, University of Rochester Medical Center, of Defense Survey of Health Related Behaviors Among
dose–response effect related to being Active Duty Military Personnel: 2002. Washington, DC:
Rochester.
wounded, it is unlikely that many physicians Correspondence should be sent to Kerry L. Knox, VA US Dept of Defense; 2003.
and clinicians are aware of the importance of Center of Excellence at Canandaigua, Canandaigua VA 11. Bray RM, Hourani LL, Olmsted KL, et al. Department
Medical Center, 400 Fort Hill Ave, Canandaigua, NY of Defense Survey of Health Related Behaviors Among
assessing trauma exposure among those who
(e-mail: kerry.knox@va.gov). Reprints can be ordered at Active Duty Military Personnel: 2005. Washington, DC:
have served. The larger challenge for com- http://www.ajph.org by clicking the ‘‘Reprints/Eprints’’ link. US Dept of Defense; 2006.
munities worldwide is whether the pervasive This article was accepted October 28, 2009. 12. Campise RL, Rowan Anderson B. A multisite study
stigma associated with mental health disor- of Air Force outpatient behavioral health treatment-
Contributors seeking patterns and career impact. Mil Med. 2006;
ders and psychosocial problems will be over-
K. L. Knox conceptualized and directed the study. K. L. 171(11):1123–1127.
come as a result of acceptance that these
Knox, S. Pflanz, G. W. Talcott, R. L. Campise, and E. D. 13. Conner KR, Cox C, Duberstein PR, Tian L, Nisbet
significant, adverse mental health outcomes Caine established the formal prevention domains that PA, Conwell Y. Violence, alcohol, and completed suicide:
are a normal human response to the expo- guided the development of implementation measures. a case-control study. Am J Psychiatry. 2001;158(10):
K. L. Knox, J. E. Lavigne, and X. Tu supervised the data 1701–1705.
sures associated with serving in the military.
analysis. A. Bajorska carried out the data analysis. All 14. Beautrais AL, Joyce PR, Mulder RT. Precipitating
In conclusion, the US Air Force showed, authors participated in interpreting the results and factors and life events in serious suicide attempts among
through its efforts to reduce deaths from writing and editing the article. youths aged 13 through 24 years. J Am Acad Child
suicide, that (1) it is possible to reduce the rate Adolesc Psychiatry. 1997;36(11):1543–1551.
of suicide across a period of years using Acknowledgments 15. Duberstein PR, Conwell Y, Caine ED. Interpersonal
This project was supported by National Institute of stressors, substance abuse, and suicide. J Nerv Ment Dis.
a multifaceted, overlapping, community-
Mental Health (grants K01 MH055317, R01 1993;181(2):80–85.
based approach, and (2) reductions in suicide MH075017-01A1, and P20 MH071897). 16. Mahon MJ, Tobin JP, Cusack DA, et al. Suicide
rates cannot be simply maintained by virtue Note. The funding agency was not involved in the among regular-duty military personnel: a retrospective
design and conduct of the study; collection, management,
of a program’s inherent momentum. Pro- case-control study of occupation-specific risk factors
analysis, and interpretation of the data; and preparation, for workplace suicide. Am J Psychiatry. 2005;162(9):
grammatic efforts must be continuously sup- review, or approval of the article. 1688–1696.
ported and monitored to ensure sustained
17. Helmkamp JC. Occupation and suicide among males
effects. This may mean that many communi- Human Participant Protection in the US Armed Forces. Ann Epidemiol. 1996;6(1):
ties and organizations will not easily be able This study received approval from the institutional re- 83–88.
view board at the University of Rochester and the
to launch large-scale suicide prevention ef- 18. Marzuk PM, Nock MK, Leon AC, et al. Suicide among
Wilford Hall institutional review board for the US Air New York City police officers, 1977–1996. Am J Psy-
forts on a scale comparable to the AFSPP, Force. chiatry. 2002;159(12):2069–2071.
especially in developing countries.24 We sus-
19. Conner KR, Duberstein PR, Conwell Y. Age-related
pect that there may be real limitations on the References patterns of factors associated with completed suicide in
feasibility of an exact replication of the AFSPP to 1. Knox KL, Litts DA, Talcott GW, Feig JC, Caine ED. men with alcohol dependence. Am J Addict. 1999;
Risk of suicide and related adverse outcomes after expo- 8(4):312–318.
other settings. Nevertheless, the enduring public
sure to a suicide prevention program in the US Air Force: 20. Ilgen MA, Harris A, Moos RH, Tiet QQ. Predictors of
health message from 12 years of this program is cohort study. BMJ. 2003;327(7428):1376–1380. a suicide attempt one year after entry into substance use
that suicide rates can be reduced, and that 2. Box GEP, Jenkins GM, Reinsel GC. Time Series disorder treatment. Alcohol Clin Exp Res. 2007;31(4):
program success requires interventions to be Analysis: Forecasting and Control. Hoboken, NJ: John 635–642.
consistently supported, maintained, and moni- Wiley & Sons; 2008. 21. Smith TC, Wingard DL, Ryan MA, Kritz-Silverstein
D, Slymen DJ, Sallis JF; Millennium Cohort Study Team.
tored for compliance. This is a message that all 3. SAS Institute. SAS/STAT 9.1 User’s Guide. Cary,
NC: SAS Institute; 2004. Prior assault and posttraumatic stress disorder after
communities and organizations worldwide can combat deployment. Epidemiology. 2008;19(3):505–
embrace while considering how to appropriately 4. Knox KL, Conwell Y, Caine ED. If suicide is a public 512.
health problem, what are we doing to prevent it? Am J
structure programs and interventions at a local Public Health. 2004;94(1):37–45.
22. Boscarino JA. Postraumatic stress disorder and
mortality among US Army veterans 30 years after
level. j 5. Hand DJ. Statistics and the theory of measurement. military service. Ann Epidemiol. 2006;16(4):248–256.
J R Stat Soc Ser A Stat Soc. 1996;159(3):445–492.
23. Bullman TA, Kang HK. The risk of suicide among
6. Fayers PM, Hand DJ. Causal variables, indicator wounded Vietnam veterans. Am J Public Health. 1996;
About the Authors 86(5):662–667.
variables and measurement scales: an example from
Kerry L. Knox and Jill E. Lavigne are with the Canandaigua
quality of life. J R Stat Soc Ser A Stat Soc. 2002;165(2): 24. Hawton K, van Heeringen K. Suicide. Lancet. 2009;
VA Medical Center, US Department of Veterans Affairs,
233–261. 373(9672):1372–1381.
Canandaigua, NY, and the Department of Psychiatry,
University of Rochester Medical Center, Rochester, NY. 7. Rychetnik L, Frommer M, Hawe P, Shiell A. Criteria
Steven Pflanz is with the 579th Medical Operations for evaluating evidence on public health interventions.
Squadron, Bolling Air Force Base, Washington, DC. Gerald J Epidemiol Community Health. 2002;56(2):119–127.
December 2010, Vol 100, No. 12 | American Journal of Public Health Knox et al. | Peer Reviewed | Research and Practice | 2463