VOLUME 17 NUMBER 10 | OctOBER 2009 | www.obesityjournal.org
nature publishing group
Suicide is a public health problem of enormous consequence,
accounting for over 32,600 deaths in 2005 and representing the
11th leading cause of death in the United States (1). Although
several risk factors for completed suicide are established,
suicide rates have declined only slightly in recent decades (2),
underscoring the need to identify new, potentially remediable
suicide risk factors.
The epidemic of obesity in the United States has spurred
increasing interest in its relationship with mental health
disorders. If obesity is positively associated with depression
and anxiety symptoms, as is commonly found (3,4), it might
be hypothesized that overweight and obese individuals would
have a higher risk of suicide. Surprisingly, a large series of
prospective studies have now found the opposite (5–12), with
relatively consistent lower risks of completed suicide with
greater BMI. However, the possibility of confounding of BMI
at the individual-level, particularly by changes in weight related
to depression, has not been fully resolved. To date, only one
nationally representative study had addressed this association
(12), leaving its generalizability to the broader population of
US adults uncertain. Further, no previous study has examined
BMI, suicide attempts, completed suicides, and method-
specific case–fatality ratios simultaneously, an approach that
allows us to explore possible mechanisms empirically.
To determine the association of BMI with risk of completed
suicide across the United States, we conducted an ecological
analysis of statewide rates of obesity, suicide attempts, and
cause-specific completed suicides, based upon representative
surveys conducted throughout the United States in recent years.
Because statewide rates of obesity are not subject to the same
confounding factors as individual BMI, these analyses offer the
opportunity to re-examine the BMI–suicide relationship with
a complementary approach.
Methods And Procedures
We compiled state-level information on risk factors and rates of
attempted and completed suicides from a variety of contemporaneous
sources. In virtually all cases, the rank ordering of states for these factors
is nearly identical year-to-year during the time period under study.
Suicide mortality data for each state were obtained through the Centers
for Disease Control and Prevention’s Web-based Injury Statistics Query
and Reporting System (13). Suicide data were further grouped by firearm
(ICD-10 E-codes X72–X74) and nonfirearm methods (E-codes X60–X71,
X75–X84, Y87.0, and U03). Reported mortality data were aggregated
during the 2-year period of 2004–2005.
BMI and Rates of Suicide in the United States:
An Ecological Analysis
Kenneth J. Mukamal1, Christina C. Wee1 and Matthew Miller2
BMI has been inversely associated with risk of completed suicide in several cohort studies, but putative mechanisms
for this association and its generalizability throughout the United States are uncertain. We ascertained recent population-
based, state-level data on rates of obesity, completed suicide (by method), firearm ownership, smoking, major
depression, income, education, white race, and nonmetropolitan residence, compiled from federal agencies and surveys,
and determined the adjusted population-weighted correlations of statewide obesity rates with measures of completed
and attempted suicide. Statewide prevalence of obesity was strongly inversely correlated with age-adjusted suicide
rate (multivariable-adjusted r = −0.66; P < 0.001). The correlation was somewhat stronger for rates of nonfirearm-related
(r = −0.75; P < 0.001) than firearm-related suicides (r = −0.53; P < 0.001), and was of similar magnitude as the positive
correlations of firearm prevalence with suicide rate (r = 0.75; P < 0.001) or of obesity with prevalence of diabetes (r = 0.41;
P = 0.006). In analyses of fatal and nonfatal suicidal acts, obesity rates were inversely correlated with rates of suicidal
acts using firearms (r = −0.53; P = 0.02) and suffocation (r = −0.76; P < 0.001) but not other methods. Obesity rates were
also inversely correlated with the case–fatality ratios of acts using poisoning (r = −0.51; P = 0.01). Thus, statewide rates
of obesity are strongly inversely correlated with rates of completed suicide in multivariable analyses, a finding that
appears to relate to fewer attempts by suffocation and a lower case–fatality ratio for poisonings, although the mechanism
for the inverse correlation with firearm-related suicides requires further elucidation.
Obesity (2009) 17, 1946–1950. doi:10.1038/oby.2009.122
1Department of Medicine, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA; 2Department of Health
Policy and Management, Harvard School of Public Heath, Boston, Massachusetts, USA. Correspondence: Kenneth J. Mukamal (firstname.lastname@example.org)
Received 26 January 2009; accepted 22 March 2009; published online 23 April 2009. doi:10.1038/oby.2009.122
obesity | VOLUME 17 NUMBER 10 | OctOBER 2009 1947
Hospital admissions for deliberate self-harm were used to
calculate rates of deliberate self-harm and method-specific case–
fatality ratios. These data were obtained from the Healthcare
Cost and Utilization Project (HCUP), which is maintained by the
Agency for Healthcare Research and Quality (14). HCUP data
are derived from hospital discharge summaries and abstracts. In
total, 26 states provided data to HCUP in 2004 and 2005. The
HCUP State Inpatient Database contains all of the community
hospital inpatient discharge records from each participating state.
Deliberate self-harm admissions included any hospitalization
with a secondary diagnosis of ICD-9 E-codes 950.0–959.9.
The method used in attempted or completed suicide was
grouped into four categories: (i) poisoning with solid or liquid
substances (E950.0–E950.9; X60–X66, X68, X69); (ii) hanging,
strangulation, and suffocation (E953.0; E953.1; E953.8; E953.9;
X70); (iii) firearms (E955.0–E955.4; X72–X74); (iv) and other
and unspecified means (all other E-codes).
In these analyses, we estimate the number and case–fatality
ratio for suicidal acts (i.e., completed and attempted suicides),
rather than individuals. The case–fatality ratio for a particular
method is defined as the state-specific number of completed
suicides with that method divided by the number of completed
suicides and hospitalizations (estimated by HCUP) with that
method. As a result, the case–fatality ratios calculated here are
slight underestimates because the small number of hospitaliza-
tions that result in death are double-counted (15).
State-level data on the percentage of individuals living in
households with firearms (gun prevalence) were obtained from
the 2004 Behavioral Risk Factor Surveillance System (BRFSS)
(16). The BRFSS, the world’s largest telephone survey (303,822
respondents in 2004), is an ongoing data collection program
sponsored by the Centers for Disease Control and Prevention,
with all 50 states participating. Data collected are representative
at the state and national level after appropriate use of the complex
sampling and weighting scheme. BRFSS questionnaires and data
are publicly available at www.cdc.gov/brfss. The BRFSS was also
used to obtain state-level proportions of respondents with obesity
(BMI derived from self-reported height and weight ≥30 kg/m2)
and current smoking in 2005 and self-reported, physician-
diagnosed diabetes in 2004 (other than Hawaii, for which the
mean of 2003 and 2005 estimates was used).
State-level measures of alcohol and illicit substance abuse and
dependence and serious mental illness were obtained from the
2004–2005 National Survey on Drug Use and Health (17). The
survey-weighted hierarchical Bayes methodology used to arrive
at state estimates is described in detail elsewhere (18). Serious
mental illness is defined among respondents ≥age 18 as having a
diagnosable mental, behavioral, or emotional disorder that met the
criteria found in the 4th edition of the Diagnostic and Statistical
Manual of Mental Disorders and resulted in functional impairment
that substantially interfered with or limited one or more major
life activities. Rates of alcohol or illicit substance dependence or
abuse are based on definitions found in the 4th edition of the
Diagnostic and Statistical Manual of Mental Disorders. Estimates
of illicit substance abuse and dependence were the percentage
of respondents reporting having used any illicit drug other than
marijuana in the year before the survey and includes cocaine,
crack, heroin, hallucinogens, inhalants, or any prescription-type
psychotherapeutic agent used nonmedically. Estimates of alcohol
abuse and dependence pertain to the percentage of respondents
reporting abuse or dependence in the year before the survey.
State-level data on median income, education, and urbaniza-
tion derive from Census 2000. These data are publicly available
Because we have previously found BMI and firearm ownership
to be positively associated on an individual basis in the BRFSS
(19), all results were adjusted for firearm ownership. As
described previously (20), we performed population-weighted
Pearson correlations of state-specific rates of obesity, age-
adjusted suicide, and related factors, including all 50 states.
These analyses estimate the cross-sectional correlation across
states of two continuously distributed characteristics—the
prevalence of obesity and the age-adjusted rate of suicide—
while adjusting for covariates. Population weighting gives
greater influence to larger states, where more suicides occur,
but we also tested our results without such weighting.
Both obesity and suicide were approximately normally
distributed; joint tests of skewness and kurtosis did not reject
normality in either case (P = 0.77 and P = 0.10). Initial models
derived partial correlation coefficients with adjustment for
firearm prevalence; multivariable models further adjusted for
statewide prevalence of white non-Hispanic race/ethnicity,
college education, current smoking, and nonmetropolitan area
residence, and state-specific median household income and
composite depression index score. To examine the magnitude
of change in state-specific suicide rates with increments of
rates of obesity, we performed linear regression models, using
the same variables as in multivariable correlation models, with
robust standard errors.
We first examined the associations of obesity with total
suicide mortality, then examined its associations with suicide
related to firearms and suffocation (methods with generally
high case–fatality ratios), and poisonings. Analyses of case–
fatality were restricted to the 26 states in the HCUP, although
the association of BMI and suicide was similar in these states as
in the entire United States. None of the associations of BMI and
suicide were significantly changed if the prevalence of normal
weight, rather than obesity, was used (observed correlations
were in the inverse direction, as expected).
We attempted to provide context for these associations in
two ways. First, we examined the correlation of obesity rates
with rates of diabetes, a disorder with perhaps the best-known
association with obesity. Second, we recalculated the correla-
tion of suicide rate with firearm prevalence, which we have
previously demonstrated to be very strongly associated with
firearm-related suicide (16,20–24).
All analyses were performed with Stata Intercooled 10.1
(StataCorp, College Station, TX). No external funding was
used for these analyses, and all analyses were conducted on
publicly available, deidentified data.
VOLUME 17 NUMBER 10 | OctOBER 2009 | www.obesityjournal.org
The median proportion of adult residents who were obese was
24.4%, ranging from 17.8% in Colorado to 30.9% in Mississippi.
As seen in Table 1, states with higher rates of obesity tended to
have higher rates of firearm ownership and smoking and lower
rates of college education and median household income.
The median age-adjusted suicide rate was 11.7 per 100,000,
with extremes of 6.0 in the New York and 23.1 in Alaska.
Table 2 shows the correlation of statewide prevalence of obesity
with suicide. There was a strong inverse correlation of obesity
(r = −0.59) with completed suicide rates that was strengthened
(r = −0.66) by multivariable adjustment. In comparison, the
correlation coefficient in these models for firearm prevalence
was 0.75 (P < 0.001); the only other variable that was statistically
significant was the proportion of population that was white
non-Hispanic (r = −0.36; P = 0.02). The correlation coefficient
of obesity with diabetes in similar models was 0.41 (P = 0.006).
In linear regression models, an increase in the statewide
prevalence of obesity of 3% (the standard deviation in obesity
rates) was associated with a decrease in the age-adjusted
suicide rate of −2.6 per 100,000 (95% confidence interval, −1.5
to −3.7). This corresponds to roughly the difference between
the 50th and 75th percentiles of state-specific suicide rates.
When states were categorized in quartiles of obesity rates,
there was a stepwise decrease in age-adjusted suicide rates
across quartiles, with decreases of −0.2 (95% confidence
interval, −2.8 to 2.4), −2.3 (95% confidence interval, −5.6
to 1.1), and −3.9 (95% confidence interval, −7.8 to 0.01) in
We conducted several sensitivity analyses to test the
robustness of our findings. Additional adjustment for alcohol
and illicit drug use and use of four mental health measures
(rather than the composite depression index) did not greatly
alter the correlation of obesity and completed suicide
(r = −0.65; P < 0.001). The association tended to be similar
in areas of lower and higher prevalence of firearm ownership,
with correlation coefficients in ascending quartiles of firearm
ownership of −0.34, −0.86, −0.52, and −0.50 (adjusted for
actual firearm prevalence). Without population weighting, the
multivariable-adjusted correlation of obesity and completed
suicide was similar but numerically higher (r = −0.68; P <
0.001). The multivariable-adjusted correlation of obesity rate
with the square root of state-specific suicide rate (the best
fitting model form) was −0.62 (P < 0.001).
Table 2 also shows the association of obesity with suicide
of various methods. The multivariable-adjusted correlation
of obesity with nonfirearm-related suicide (r = −0.75; P <
0.001) was higher than that with firearm-related suicide. For
the latter, multivariable adjustment substantially strengthened
the correlation, predominately due to addition of smoking
and race/ethnicity as covariates. Among nonfirearm-related
methods, obesity appeared to be similarly associated with both
poisonings and suffocation.
Table 3 presents the correlation of obesity rates with
statewide rates of suicidal acts (i.e., attempted and completed
suicides) and statewide case–fatality ratios for specific methods
in 26 US states. Obesity rates were not strongly associated with
suicidal acts overall. However, they were strongly correlated
table 1 Pairwise population-weighted Pearson correlation coefficients among 50 us states in prevalence of selected
risk factors for suicide
Normal weight−0.92 1.00
Firearm ownership 0.68−0.591.00
White non-Hispanic 0.16−0.050.50 1.00
Household income −0.660.64 −0.66 −0.161.00
College degree −0.740.71−0.72−0.26 0.891.00
Current smoking0.62−0.51 0.590.51−0.67−0.71 1.00
Depression index 0.09−0.060.430.49 −0.35 −0.390.35 1.00
0.55 −0.46 0.840.56−0.57−0.62 0.57 0.38 1.00
table 2 Partial population-weighted Pearson correlation coefficients among 50 us states between prevalence
of obesity and completed suicide
Adjusted for firearm
−0.59 (P < 0.001)−0.35 (P = 0.01)−0.68 (P < 0.001)−0.69 (P < 0.001)−0.47 (P = 0.001)
Multivariable-adjusted−0.66 (P < 0.001)−0.53 (P < 0.001) −0.67 (P < 0.001)−0.65 (P < 0.001) −0.51 (P < 0.001)
Multivariable model included statewide prevalence of firearm ownership, white non-Hispanic race/ethnicity, college education, current smoking, and nonmetropolitan area
residence, and median household income and composite depression index.
obesity | VOLUME 17 NUMBER 10 | OctOBER 2009 1949
with lower rates of acts using firearms and suffocation/hanging,
two highly lethal methods, and almost completely uncorrelated
with rates of acts using poisonings and other methods, which
tend to be least lethal. In addition, obesity rates were correlated
with lower state-specific case–fatality ratio for poisonings but
not for other methods.
In this study of statewide rates of obesity and suicide, higher rates
of obesity were strongly correlated with lower rates of completed
suicide. There were lower rates of firearm-related and nonfirearm-
related suicides, related to fewer firearm and suffocation-related
acts and a lower case–fatality ratio for poisonings.
An emerging body of literature supports an inverse
relationship of BMI with risk of suicide death. In Paffenbarger
and Asnes’ original study of male college students, the
ponderal index (height divided by the cube root of weight—a
marker of leanness) was significantly higher among men who
later committed suicide than among controls (5). Subsequent
studies in recent years have also found an inverse relationship
of BMI and risk of suicide, particularly among Scandinavian
populations (6–8,10) (who have among the highest rates of
suicide worldwide); other, generally smaller studies have also
suggested an inverse relationship (9,11,12). An ecological
study in the Caribbean Islands was null, but did not account
for firearm availability (25). In the Norwegian HUNT Study,
BMI was inversely associated with risk of suicide despite a
generally positive association with depression (7).
Our results extend these previous findings in several
potentially important ways. First, studies of BMI and health
outcomes are prone to individual-level confounding (26) by
health conditions and behaviors (like smoking (27)) that lead
to both lower weight and higher risk of completed suicide.
Depression has been hypothesized to be one such confounder,
although it is possible that adiposity itself influences mood,
implicating depression as a potential intermediate rather than
a confounder alone. Regardless, statewide levels of obesity
are highly unlikely to be influenced by weight loss related to
medical conditions such as depression, a supposition supported
by the weak correlation of obesity and depression at the state
level. As a result, our results strongly imply that the lower risk
of completed suicide associated with heavier BMI is not merely
the result of confounding by such factors.
Second, ours is the first study to our knowledge to relate obesity
to method-specific rates of suicide attempts, completed suicides,
and case–fatality ratio. Our results preliminarily implicate
specific mechanisms that may play a role in the observed BMI–
suicide association, although these will require confirmation in
prospective studies. For example, obesity rates were associated
with lower case–fatality ratios from poisoning, a result that, while
intuitive, has not been previously established. Also, obesity rates
were associated with lower rates of acts by suffocation (which
primarily includes hanging), a method that tends to have a high
case–fatality ratio (28,29). Although the rationale for this lower
rate is necessarily speculative, the steps involved in hanging may
be burdensome, uncomfortable, or otherwise aversive for heavier
individuals. The correlation of obesity rates with lower rates of
completed suicides and suicidal acts using firearms requires
further investigation. However, firearm suicides often appear
impulsive (30,31), and hence this observation is consistent with
the hypothesis that high circulating insulin levels influence
tryptophan metabolism and reduce impulsivity (6,32,33).
Third, our results provide the most current and generalizable
information available to date about BMI and suicide in the
US population. Because of the relative rarity of suicide in
cohort studies, they necessarily require very large, typically
unrepresentative samples followed for long periods of time to
accrue sufficient numbers of completed suicides. As a result,
even recent studies of this topic rely upon measurements of BMI
collected one to two decades in the past, and do not necessarily
reflect the current experience of the broader US population.
Our findings, synthesized within the last few years from high-
quality, nationally representative surveys administered by
highly experienced federal agencies, extend the results of cohort
studies to a larger and more contemporaneous audience.
Even if the relationship observed in this study were causal,
the risks of obesity far outweigh any potential benefit on
suicide prevention. Indeed, using this same approach, we
clearly identified the expected strong correlation between
rates of obesity and diabetes. However, the magnitude of
correlation with obesity was nearly identical for diabetes and
completed suicide, despite the far greater body of evidence
linking adiposity with diabetes. Given this, we urge that more
attention and resources be devoted to the mechanisms by
which obesity may influence suicide risk, given how strongly
these factors relate.
table 3 Partial population-weighted Pearson correlation coefficients among 26 us states between prevalence of obesity
and suicidal acts or method-specific case–fatality ratio
Suicidal acts (completed + attempted)Case–fatality ratio (completed/total)
TotalFirearms Suffocation Poisonings
(P = 0.34)
(P = 0.12)
(P < 0.001)
(P = 0.54)
(P = 0.38)
(P = 0.08)
(P = 0.67)
(P = 0.01)
(P = 0.64)
(P = 0.58)
(P = 0.02)
(P < 0.001)
(P = 0.75)
(P = 0.98)
(P = 0.07)
(P = 0.44)
(P = 0.01)
(P = 0.29)
Suicidal acts represent the sum of fatal suicidal acts reported on death certificate and suicide attempts admitted to medical hospitals. Multivariable models included
statewide prevalence of firearm ownership, white non-Hispanic race/ethnicity, college education, current smoking, and nonmetropolitan area residence, and median
household income and composite depression index.
VOLUME 17 NUMBER 10 | OctOBER 2009 | www.obesityjournal.org
Our analyses have strengths but also limitations. All
ecological studies are limited by an inability to infer that the
observed relationships apply to individuals, rather than to the
groups (in this case, the states) under study. That is, the BMI of
individuals who commit suicide may not, in fact, differ across
states regardless of overall state rates of obesity. However, our
previous findings from cohort studies provide substantial
reassurance that BMI and suicide are indeed linked at both the
individual and state levels. We were also limited to 50 units
of observation (i.e., states) for analyses of completed suicide
and to 26 states for analyses that included attempted suicide,
restricting the power of our analyses, and to rates of obesity
and suicide measured at a single cross-section in time. Further
studies that evaluate changes in obesity and suicide rates
longitudinally may provide additional insight into these issues
(24); on a national level, the 1990s saw increases in obesity rates
but declines in suicide rates that support our findings (2).
In conclusion, states with higher rates of obesity tend to have
lower adjusted rates of completed suicide, with a magnitude
of correlation similar to other well-established correlates of
obesity and suicide. Obesity rates also tend to be associated
specifically with lower rates of suicidal acts using firearms
and suffocation, two of the methods most likely to prove fatal,
and with a lower case–fatality ratio for poisonings. Given the
strength of the observed relationships, further research to
identify its mechanisms could lead to important insights into
prevention of this important cause of death.
All of the authors participated in the conception and design of the study,
analysis and/or interpretation of data, critical review and revision of the
manuscript, and provision of statistical or content expertise. Dr Mukamal
had full access to all the data in the study and takes responsibility for the
integrity of the data and the accuracy of the data analysis. No sources of
support were used in the conduct of these analyses.
the authors declared no conflict of interest.
© 2009 The Obesity Society
1. Kung HC, Hoyert DL, Xu J, Murphy SL. Deaths: final data for 2005. Natl Vital
Stat Rep 2008;56:1–120.
2. Kessler RC, Berglund P, Borges G, Nock M, Wang PS. Trends in suicide
ideation, plans, gestures, and attempts in the United States, 1990–1992 to
2001–2003. JAMA 2005;293:2487–2495.
3. Scott KM, Bruffaerts R, Simon GE et al. Obesity and mental disorders in the
general population: results from the world mental health surveys. Int J Obes
4. Petry NM, Barry D, Pietrzak RH, Wagner JA. Overweight and obesity
are associated with psychiatric disorders: results from the National
Epidemiologic Survey on Alcohol and Related Conditions. Psychosom
5. Paffenbarger RS, Asnes DP. Chronic disease in former college students. 3.
Precursors of suicide in early and middle life. Am J Public Health Nations
6. Golomb BA, Tenkanen L, Alikoski T et al. Insulin sensitivity markers:
predictors of accidents and suicides in Helsinki Heart Study screenees.
J Clin Epidemiol 2002;55:767–773.
7. Bjerkeset O, Romundstad P, Evans J, Gunnell D. Association of adult body
mass index and height with anxiety, depression, and suicide in the general
population: the HUNT study. Am J Epidemiol 2008;167:193–202.
8. Osler M, Nybo Andersen AM, Nordentoft M. Impaired childhood
development and suicidal behaviour in a cohort of Danish men born in
1953. J Epidemiol Community Health 2008;62:23–28.
9. Gasse C, Derby LE, Vasilakis C, Jick H. Risk of suicide among users of
calcium channel blockers: population based, nested case-control study.
10. Magnusson PK, Rasmussen F, Lawlor DA, Tynelius P, Gunnell D. Association
of body mass index with suicide mortality: a prospective cohort study of
more than one million men. Am J Epidemiol 2006;163:1–8.
11. Mukamal KJ, Kawachi I, Miller M, Rimm EB. Body mass index and risk of
suicide among men. Arch Intern Med 2007;167:468–475.
12. Kaplan MS, McFarland BH, Huguet N. The relationship of body weight to
suicide risk among men and women: results from the US National Health
Interview Survey Linked Mortality File. J Nerv Ment Dis 2007;195:948–951.
13. Centers for Disease Control and Prevention NCfIPaC. Web-based Injury
Statistics Query and Reporting System (WISQARS). <www.cdc.gov/ncipc/
wisqars> (2005). Accessed 27 November 2008.
14. Steiner C, Elixhauser A, Schnaier J. The healthcare cost and utilization
project: an overview. Eff Clin Pract 2002;5:143–151.
15. Coben JH, Steiner CA. Hospitalization for firearm-related injuries in the
United States, 1997. Am J Prev Med 2003;24:1–8.
16. Miller M, Lippmann SJ, Azrael D, Hemenway D. Household firearm
ownership and rates of suicide across the 50 United States. J Trauma
2007;62:1029–34; discussion 1034.
17. Wright D. State Estimates of Substance Use from the 2002 National Survey
on Drug Use and Health. Report No.: DHHS Publication No. SMA 04-3907,
NSDUH Series H-23. Substance Abuse and Mental Health Services
Administration, Office of Applied Studies: Rockville, MD, 2004.
18. Folsom RE, Shah B, Vaish A. Substance abuse in states: a methodological
report on model based estimates from the 1994–1996 National Household
Surveys on Drug Abuse. In: Proceedings of the American Statistical
Association, Survey Research Methods Section. American Statistical
Association: Washington, DC, 1999.
19. Mukamal KJ, Miller M. BMI and risk factors for suicide: why is BMI inversely
related to suicide? Obesity (Silver Spring) 2009;17:532–538.
20. Miller M, Hemenway D, Azrael D. Firearms and suicide in the northeast.
J Trauma 2004;57:626–632.
21. Miller M, Azrael D, Hemenway D. Household firearm ownership and suicide
rates in the United States. Epidemiology 2002;13:517–524.
22. Miller M, Azrael D, Hemenway D. Firearm availability and suicide,
homicide, and unintentional firearm deaths among women. J Urban Health
23. Miller M, Azrael D, Hemenway D. Firearm availability and unintentional
firearm deaths, suicide, and homicide among 5–14 year olds. J Trauma
2002;52:267–274; discussion 274.
24. Miller M, Azrael D, Hepburn L, Hemenway D, Lippmann SJ. The association
between changes in household firearm ownership and rates of suicide in the
United States, 1981–2002. Inj Prev 2006;12:178–182.
25. Bridges FS, Tran XV. Body mass index, suicide, and homicide among
inhabitants of the Caribbean Islands. Percept Mot Skills 2008;106:650–652.
26. Manson JE, Bassuk SS, Hu FB et al. Estimating the number of deaths due
to obesity: can the divergent findings be reconciled? J Womens Health
27. Miller M, Hemenway D, Bell NS, Yore MM, Amoroso PJ. Cigarette smoking
and suicide: a prospective study of 300,000 male active-duty Army soldiers.
Am J Epidemiol 2000;151:1060–1063.
28. Miller M, Azrael D, Hemenway D. The epidemiology of case fatality rates for
suicide in the northeast. Ann Emerg Med 2004;43:723–730.
29. Spicer RS, Miller TR. Suicide acts in 8 states: incidence and case fatality rates
by demographics and method. Am J Public Health 2000;90:1885–1891.
30. Azrael D, Hemenway D, Miller M, Barber CW, Schackner R. Youth suicide:
insights from 5 years of Arizona Child Fatality Review Team data. Suicide Life
Threat Behav 2004;34:36–43.
31. de Moore GM, Plew JD, Bray KM, Snars JN. Survivors of self-
inflicted firearm injury. A liaison psychiatry perspective. Med J Aust
32. Virkkunen M, Goldman D, Nielsen DA, Linnoila M. Low brain serotonin
turnover rate (low CSF 5-HIAA) and impulsive violence. J Psychiatry
33. Kwok RP, Juorio AV. Facilitating effect of insulin on brain 5-hydroxytryptamine
metabolism. Neuroendocrinology 1987;45:267–273.