Sex differences in suicides among children and youth: the potential impact of misclassification.
Rhodes, Anne E. ; Khan, Saba ; Boyle, Michael H. 等
Suicide is the second leading cause of death among persons aged
15-24 in Canada. (1) Suicide is a rare event in children/youth; (2)
however, it contributes substantially to premature mortality worldwide
(3) and is a devastating life event for loved ones that may lead to
broader imitation effects. (4) Suicide rates typically increase during
the transition from childhood to adolescence and, in most regions of the
world, including Canada, are higher in males. (2,5,6) This may arise
from sex differences in psychopathology, suicidal intent, access to
methods, help-seeking or ascertainment of death. (7) Such explanations
have not been thoroughly examined in children/youth. A fuller
understanding of sex differences in youth suicide rates is needed to
ensure that suicide prevention strategies are built on accurate data.
This study addresses the "ascertainment" explanation; in
particular, the possible misclassification of suicides in boys and
girls. Suicides in general are thought to be under-reported as
accidental or undetermined deaths in many jurisdictions. Decedents or
their loved ones may not leave or provide evidence suggestive of suicide
for reasons of shame, stigma, or invalidation of life insurance. (8)
Furthermore, certain causes of death, such as alcohol/drug overdose,
pose greater uncertainty in suicide classification. (9)
Among youth, sex differences may arise from differential
underreporting of suicides in females compared to males. (7) However,
few studies have actively investigated misclassification of suicides in
children/youth. In the United States (US), potential changes in suicide
time trends (1979-1994) among those aged 10-24 were examined by
reclassifying suspect causes of death and undetermined deaths to
suicides. (8) While suicide rates were not altered appreciably, upward
corrections were most apparent in the 10-14 age group and in Black
males. In Norway, (10) suicides under age 15 were compared to suicides
in 15-19 year olds. Half of suicides under age 15 were reclassified as
equivocal or not suicides, compared to 4% of suicides in 15-19 year
olds. In a US national survey, 94 members of the National Association of
Medical Examiners were asked about factors related to making a suicide
determination in children. (11) All respondents identified a suicide
note and most identified a history of suicidal ideations or behaviours;
however, responses varied regarding the minimum age at which respondents
would classify a child death as a suicide.
Altogether, these studies imply that suicide deaths may be either
under- or over-reported among different age groups, but it remains
unclear if reclassification impacts observed sex differences in suicide
rates. The objective of this study was to examine sex differences in
suicide rates in children and youth in Ontario after adjusting for
potential misclassification of death.
METHODS
Study design
This is a retrospective study of suicide, accidental and
undetermined deaths in children and youth (aged 10-25 years) who had
been living in Ontario and died between January 1, 2000 and December 31,
2007. Data access was granted under a data sharing agreement between the
Office of the Chief Coroner for Ontario (OCC) and the Institute for
Clinical Evaluative Sciences (ICES), both in Toronto, Ontario. This
study was approved by the Research Ethics Boards of St. Michael's
Hospital and Sunnybrook Health Sciences Centre in Toronto.
Participants
All suicides (n=1,299), all undetermined deaths (n=256), and a
random sample of one third of accidental deaths (n=966) in the
aforementioned age and time ranges were selected. Of these 2,521 deaths,
9 had files that could not be located and 4 were excluded based on
eligibility criteria (i.e., they were outside of age range). Thus, a
total of 2,508 cases were analyzed (1,294 suicides, 254 undetermined
deaths, 961 accidental deaths).
Setting
All data were obtained from paper case files held at the OCC. In
Ontario, a coroner is a licensed physician appointed to investigate
medico-legal deaths within the province according to the Coroners Act,
R.S.O. 1990, c. C.37. (12) Most of these deaths have occurred suddenly
or unexpectedly. Individual case files contain the Coroner's
Investigation Statement (a narrative summary of the circumstances of
death) and additional documentation, including police reports.
Data measures
Case files were reviewed by professional chart abstractors who
entered data into an encrypted electronic database in a secure location
at the OCC. Data were electronically transmitted and housed at ICES in a
secure fashion. Agreement between each abstractor pair was assessed for
all study variables in a subset of 5% of case files (n=121). Kappa
values on study variables ranged from 0.58-1.0, indicating substantial
to almost perfect agreement. (13) The following variables were analyzed:
* Age, sex, living situation: Identified at time of death.
* Death factor: Standardized cause-of-death coding used by
investigating coroners. As in the literature, (2,5) the major causes of
unnatural deaths are: asphyxia, alcohol/drug toxicity, drowning,
shooting, fall/jump, and motor vehicle collision (Kappas [greater than
or equal to] 0.88). Specific information about each death factor was
abstracted where possible.
* Means of death: Standardized death classification coding (Kappas
[greater than or equal to] 0.95) defined according to OCC:14
--Accident: Death caused by an external factor, where death or harm
was not foreseen or expected.
--Suicide: Death resulting from an intentional act of a person
knowing the probable consequence of his/her actions. While the legal
test to be satisfied is a balance of probability, a determination of
suicide can only be made where there is clear and convincing evidence.
--Undetermined: Death where there is no evidence for any specific
classification; or there is equal evidence or a significant contest
among two or more classifications; or the death is a suicide that does
not meet the legal test requiring a high degree of probability or an
apparent suicide under age 10 (as young children tend not to appreciate
the consequences of a suicidal act (15)).
* Evidence of pre-suicidal behaviour: Evidence of any suicidal
communications in month prior to death, including verbal, written or
other communication (e.g., Internet chat) (Kappas [greater than or equal
to] 0.70).
* Community size, neighbourhood income quintile: These residential
variables were created at ICES for subjects identified within the
Ontario Registered Persons Database (RPDB) as having an active Ontario
health card and valid Ontario postal code at the time of their death
(94% of the study population). (16) Postal code information from the
RPDB and the Statistics Canada Postal Conversion File (17) were used to
assign each subject's residence to its dissemination area (the
smallest geographic unit for which census data are produced) (18) to
define these variables.
Reclassification of death
Means of death were reclassified using an approach similar to
Mohler and Earls. (8) Two different reclassification criteria were
tested:
Criterion A: Suspect accidental and undetermined deaths (see Table
1) were reclassified to suicides (Criterion [A.sub.1]).
"Suspect" indicators of suicide were identified from the
literature, across all death factors (e.g., recent suicidal
communications (10,11,19)) and for specific death factors (e.g.,
single-occupant vehicle deaths (8,19-21)). As it was possible some
deaths classified as suicides were actually undetermined, particularly
with regard to those under age 15, (10) we also reclassified certain
suicide deaths to be undetermined, e.g., deaths with evidence of
recreational activity (Criterion [A.sub.2]).
Criterion B: All undetermined deaths were reclassified to suicides.
Analysis
Analyses were conducted in SAS 9.2. Counts and proportions for all
demographic variables and death factors were calculated by sex for
suicide cases, with 95% confidence intervals (CIs). Actual and
reclassified suicide rates were calculated for all subjects linked to
the RPDB, by sex, age group, and year of death, with 95% CIs. (An
accidental death was weighted by three to represent the total number of
deaths). Non-overlapping 95% CIs indicated a known statistically
significant difference. (22) Annually-adjusted Census population data
from Statistics Canada were used for rate denominators.
RESULTS
Study population characteristics
Table 2 displays demographic characteristics of the suicide deaths
by sex. The majority (75%) occurred in males, particularly in those aged
16-25 years. Just over 50% of cases lived with their biological parents
at the time of their death. There were no obvious sex differences in
living arrangement or community population size/income level.
Death factors
Asphyxia was the most common cause of suicides in both males and
females (Table 2). Most asphyxia deaths (92%) were hangings across both
sexes. There were no significant sex differences in causes of suicide,
except for alcohol/drug toxicity and shooting. Suicides by shooting were
almost exclusive to males and 84% involved long guns (e.g., rifles,
shotguns). Conversely, suicides due to alcohol/drug toxicity were
significantly higher in females than males (19.5% vs. 4.0%,
p<0.0001). Among alcohol/drug suicides, the most commonly reported
substances associated with death were: opioids (36% and 20% in males and
females, respectively) and anti-depressants (26% and 33% in males and
females, respectively). Twenty-seven percent of suicides reported
multiple (2+) types of substances involved with death. There were no
significant sex differences in types of substances involved in death.
Among undetermined deaths, the most common cause of death was
alcohol/drug toxicity, accounting for about one quarter of undetermined
deaths in both sexes. Among accidental deaths, motor vehicle collisions
were most common (67.4% of males and 76.1% of females).
Suicide rates and reclassification
The average annual suicide rate over the study period was 8.48 per
100,000 in males and 3.02 per 100,000 in females. When rates were
examined by age group, children/youth aged 16-25 had significantly
higher rates than those aged 10-15 years (Figure 1). This pattern
remained when suicides were reclassified using Criterion A or B. Sex
differences in both actual and reclassified rates presented in the 16-25
year olds. Among this age group, the application of Criterion A had a
greater impact on suicide rates in males than in females.
In each study year, actual suicide rates were significantly higher
in males than females (see Figures 2 and 3). Sex differences remained
stable upon reclassification of suicides using Criterion A or B. Of
note, among males, Criterion A reclassified rates were almost always
significantly higher than the actual suicide rates in each year.
DISCUSSION
Sex differences in actual and reclassified suicide rates emerged
between the 10-15 and 16-25 year age groups. In each year, both actual
and reclassified suicide rates were higher in males than in females.
Thus, it is unlikely that sex differences in suicide rates emerging in
adolescence are due to potential misclassification.
Limitations
There is no gold standard for suicide determination; therefore,
"true" rates remain unknown. Reclassifying deaths provides an
opportunity to test the robustness of sex differences in suicide rates
by age and over time. However, our criteria for reclassifying suicides
must be contextualized. For example, it would be inappropriate to apply
Criterion A in settings where there is limited access to cars and
trains. Furthermore, if suicides under age 10 were officially
recognized, undetermined deaths under age 10 would have been
reclassified as suicides under Criterion B. Still, the number of
undetermined deaths in 2000-2007 for children ages 5 to 9 was quite low
(<10 deaths), minimally impacting our findings.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
The selection of Criterion A was based on existing literature and
heavily influenced by motor vehicle collisions, particularly with regard
to single-occupant vehicle deaths. While abstractor agreement on recent
suicidal communications was substantial, there is no standard requiring
that it be assessed in these or other accidents. Thus, without further
investigation, suspicious motor vehicle collision deaths may be better
conceptualized as falling along a continuum of risky behaviours with
varying or undetermined intent. (23,24) We were unable to identify
specific criteria to reclassify other suspect "accidental"
causes of death as suicides, especially alcohol/drug deaths. While some
have suggested that multiple drug consumption prior to death may
indicate suicidal intent, (25) it is also possible that the decedent
misjudged the effect or tolerance of additional drugs, especially during
a relapse.
CONCLUSIONS
Sex differences in suicide rates emerging in adolescence are
unlikely to be the result of misclassification of suicides. However, for
the purpose of reporting absolute counts or rates of suicide,
underreports should not be discounted. In particular, differences
between actual and reclassified rates were most evident among males aged
16-25 years when Criterion A was applied. When testing associations of
risk/protective factors with suicide in children/youth, given a strict
definition of suicide, it seems unlikely that associations found would
be invalidated by suicide misclassification. Other proposed explanations
for the emergence of sex differences in suicide rates in adolescence
merit further investigation.
Acknowledgements: Funding for this project was provided by an
operating grant from the Canadian Institutes of Health Research, MMG
103246. This study was supported by the Institute for Clinical
Evaluative Sciences, which is funded by an annual grant from the Ontario
Ministry of Health and Long-Term Care, and the Office of the Chief
Coroner for Ontario. In addition, we acknowledge the support of The
Child Welfare League of Canada, the Ontario Association of
Children's Aid Societies, the Ontario Centre of Excellence for
Child and Youth Mental Health, and the Injury and Child Maltreatment
Section, Health Surveillance and Epidemiology Division, Public Health
Agency of Canada. A special thanks to June Lindsell, Anne Marie Mior,
Monique Kerr-Taylor and Nancy Cooper for their assistance in data
collection. Preliminary results of this study were presented at the
Canadian Association for Suicide Prevention conference in Vancouver,
Canada in October, 2011.
Disclaimer: The opinions, results and conclusions reported in this
paper are those of the authors and do not necessarily reflect the
official policy or position of the affiliated or acknowledged
organizations. No endorsement by these organizations is intended or
should be inferred.
Conflict of Interest: None to declare.
Received: January 4, 2012
Accepted: March 31, 2012
REFERENCES
(1.) Statistics Canada. Leading Causes of Death in Canada, 2008.
Available at: http://www.statcan.gc.ca/pub/84-215-x/
2011001/tbl/t003-eng.pdf (Accessed December 5, 2011).
(2.) Statistics Canada. Suicides and suicide rate, by sex and by
age group, 2008. Available at: http://www40.statcan.gc.ca/
101/cst01/hlth66d-eng.htm (Accessed December 5, 2011).
(3.) World Health Organization. Suicide Prevention. Available at:
http://www.who.int/mental_health/prevention/suicide/suicideprevent/en/
index.html (Accessed December 5, 2011).
(4.) Insel B, Gould M. Impact of modeling on adolescent suicidal
behavior. Psychiatr Clin North Am 2008;31(2):293-316.
(5.) Varnik A, Kolves K, Allik J, Arensman E, Aromaa E, van
Audenhove C, et al. Gender issues in suicide rates, trends and methods
among youths aged 15-24 in 15 European countries. J Affect Disord
2009;113(3):216-26.
(6.) Bursztein C, Apter A. Adolescent suicide. Curr OpinPsychiatry
2009;22(1):1-6.
(7.) Beautrais AL. Gender issues in youth suicidal behaviour. Emerg
Med 2002;14(1):35-42.
(8.) Mohler B, Earls F. Trends in adolescent suicide:
Misclassification bias? Am J Public Health 2001;91(1):150-53.
(9.) Parais J, Kreiger N, Tomlinson G, Adlaf E. The validity of the
certification of manner of death by Ontario Coroners. Ann Epidemiol
2006;16(11):805-11.
(10.) Groholt B, Ekeberg O. Suicide in young people under 15 years:
Problems of classification. Nord J Psychiatry 2003;57(6):411-17.
(11.) Crepeau-Hobson F. The psychological autopsy and determination
of child suicides: A survey of medical examiners. Arch Suicide Res
2010;14(1):24-34.
(12.) Government of Ontario. Coroners Act, R.S.O. 1990, Chapter C.
37. Available at: http://www.e-laws.gov.on.ca/html/
statutes/english/elaws_statutes_90c37_e.htm (Accessed December 5, 2011).
(13.) Landis JR, Koch GG. The measurement of observer agreement for
categorical data. Biometrics 1977;33(1):159-74.
(14.) Paediatric Death Review Committee. Report of the Paediatric
Death Review Committee: Annual Report 2009. Toronto, ON: Office of the
Chief Coroner, Province of Ontario, 2009.
(15.) Mishara BL. Conceptions of death and suicide in children aged
6 to 12 and their implications for suicide prevention. Suicide Life
Threat Behav 1999;29(2):105-18.
(16.) Iron K, Zagorski B, Sykora K, Manuel D. Living and dying in
Ontario: An opportunity for improved health information. ICES
Investigative Report. Toronto, ON: Institute for Clinical Evaluative
Sciences, 2008.
(17.) Wilkins R. PCCF + Version 5E User's Guide
(Geocodes/PCCF): Automated Geographic Coding based on the Statistics
Canada Postal Code Conversion Files. Ottawa, ON: Statistics Canada,
2009.
(18.) Statistics Canada. Geographic Units: Dissemination Area (DA).
Available at: http://geodepot.statcan.ca/2006/Reference/COGG/
LongDescription_e.jsp?GE O_LEVEL=35&REFCODE=10&TYPE=L (Accessed
June 20, 2011).
(19.) Wyatt J, Squires T, Collis S, Broadley R. Road traffic
suicides. J Forensic Leg Med 2009;16(4):212-14.
(20.) Conolly J, Cullen A, McTigue O. Single road traffic
deaths--accident or suicide? Crisis 1995;16(2):85-89.
(21.) Murray D, de Leo D. Suicidal behavior by motor vehicle
collision. Traffic Inj Prev 2007;8(3):244-47.
(22.) Schenker N, Gentleman JF. On judging the significance of
differences by examining the overlap between confidence intervals. Am
Stat2001;55(3):182-86.
(23.) Rockett I, Hobbs G, De Leo D, Stack S, Frost J, Ducatman A,
et al. Suicide and unintentional poisoning mortality trends in the
United States, 1987-2006: Two unrelated phenomena? BMC Public Health
2010;10:705.
(24.) Steinberg L. Adolescents' risky driving in context. J
Adolesc Health 2011;49(6):557-58.
(25.) Stanistreet D, Gabbay M, Jeffrey V, Taylor S. Are deaths due
to drug use among young men underestimated in official statistics?
Drugs: Education, Prevention and Policy 2004;11(3):229-42.
Anne E. Rhodes, PhD, [1-4] Saba Khan, MPH, [4] Michael H. Boyle,
PhD, [5] Christine Wekerle, PhD, [6] Deborah Goodman, PhD, [7,8] Lil
Tonmyr, PhD, [9] Jennifer Bethell, MSc, [1,3] Bruce Leslie, MSW, [8,10]
Ian Manion, PhD [11]
Author Affiliations
[1.] The Suicide Studies Research Unit and the Keenan Research
Centre at the Li Ka Shing Knowledge Institute of St. Michael's
Hospital, Toronto, ON
[2.] Department of Psychiatry, Faculty of Medicine, University of
Toronto, Toronto, ON
[3.] Dalla Lana School of Public Health, Faculty of Medicine,
University of Toronto, Toronto, ON
[4.] The Institute for Clinical Evaluative Sciences, Toronto, ON
[5.] Department of Psychiatry and Behavioural Neurosciences and
Offord Center for Child Studies, McMaster University, Hamilton, ON
[6.] Department of Pediatrics, McMaster University, Hamilton, ON
[7.] The Children's Aid Society of Toronto, Child Welfare
Institute, Toronto, ON
[8.] Factor-Inwentash Faculty of Social Work, University of
Toronto, Toronto, ON
[9.] The Injury and Child Maltreatment Section, Health Surveillance
and Epidemiology Division, Public Health Agency of Canada, Ottawa, ON
[10.] The Catholic Children's Aid Society of Toronto, Toronto,
ON
[11.] The Ontario Centre of Excellence for Child and Youth Mental
Health, Ottawa, ON
Correspondence: Dr. Anne E. Rhodes, 2 Shuter Wing, Suite 2010f, St.
Michael's Hospital, 30 Bond Street, Toronto, ON M5B 1W8, Tel:
416-864-6060, ext. 2693, Fax: 416-864-5996, E-mail: RhodesA@smh.ca
Table 1. Reclassification Criteria for Suicides, Accidental and
Undetermined Deaths
Criterion A
A1: A2:
Suspect ACC or UD Suspect suicides
[??] [??]
Suicides UD
N=186 Reclassified N<6* Reclassified
N (%) N (%)
Indicators Across Reported recent
Death Factors suicidal communication
in month prior to death
16 (8.6%)
Indicators Specific
to Death Factor
Asphyxia History of choking game
N<6*
Evidence of autoerotic
activity
N<6*
Shooting Evidence of
recreational activity
N<6*
Motor vehicle Single-occupant driver,
collision no evidence of:
* Hazardous weather
* Vehicle malfunction
146 (78.5%)
Pedestrian hit by rail
train or subway, no
evidence of:
* Occupational activity
24 (12.9%)
Drowning
Alcohol/drug
toxicity
Fall/jump
Other
Criterion B
All UD
[??]
Suicides
N=253 Reclassified
N (%)
Indicators Across
Death Factors
Indicators Specific
to Death Factor
Asphyxia 29 (11.5%)
Shooting 6 (2.4%)
Motor vehicle 27 (10.7%)
collision
Drowning 17 (6.7%)
Alcohol/drug 65 (25.7%)
toxicity
Fall/jump 11 (4.3%)
Other 98 (38.7%)
ACC: Accidental deaths, UD: Undetermined deaths.
* Small cells (n<6) and corresponding proportions suppressed to
protect confidentiality.
Table 2. Characteristics of Suicide Cases, by Sex (n=1294)
Males
N %* (95% CI)
966 75
Age (years)
10-15 75 7.8 (6.1-9.4)
16-25 891 92.2 (90.5-93.9)
Living situation
Lived with biological parents 526 54.4 (51.4-57.6)
Lived alone 68 7.0 (5.4-8.7)
Lived with friends/roommates 67 6.9 (5.3-8.5)
Lived with spouse/partner 66 6.8 (5.2-8.4)
Lived with other family members 38 3.9 (2.7-5.2)
Other 75 7.8 (6.1-9.5)
Unknown 126 13.0 (10.8-15.1)
Community size (population)
1,500,000+ 244 25.3 (22.5-28.0)
500,000-1,499,999 100 10.4 (8.4-12.3)
100,000-499,999 265 27.4 (24.6-30.2)
10,000-99,999 90 9.3 (7.5-11.2)
<10,000 211 21.8 (19.2-24.4)
Unknown 56 5.8 (4.3-7.3)
Neighbourhood income quintile
1 (lowest) 234 24.2 (21.5-26.9)
2 138 14.3 (12.1-16.5)
3 185 19.2 (16.7-21.6)
4 169 17.5 (15.1-19.9)
5 (highest) 154 15.9 (13.6-18.2)
Unknown 86 8.9 (7.1-10.7)
Death factor
Asphyxia 601 62.2 (59.2-65.3)
Fall/jump 105 10.9 (8.9-12.8)
Shooting 119 12.3 (10.2-14.4)
Alcohol/drug toxicity 39 4.0 (2.8-5.3)
Motor vehicle collision 67 6.9 (5.3-8.5)
Drowning 19 2.0 (1.1-2.8)
Other 16 1.7 (0.8-2.5)
Females
N % (95% CI)
328 25
Age (years)
10-15 64 19.5 (15.2-23.9)
16-25 264 80.5 (76.2-84.8)
Living situation
Lived with biological parents 172 52.4 (47.0-57.8)
Lived alone 27 8.2 (5.2-11.2)
Lived with friends/roommates 25 7.6 (4.7-10.5)
Lived with spouse/partner 21 6.4 (3.7-9.0)
Lived with other family members 19 5.8 (3.3-8.3)
Other 21 6.4 (3.7-9.0)
Unknown 43 13.1 (9.4-16.8)
Community size (population)
1,500,000+ 93 28.4 (23.5-33.2)
500,000-1,499,999 30 9.2 (6.0-12.3)
100,000-499,999 79 24.1 (19.4-28.7)
10,000-99,999 21 6.4 (3.7-9.0)
<10,000 87 26.5 (21.7-31.3)
Unknown 18 5.5 (3.0-8.0)
Neighbourhood income quintile
1 (lowest) 75 22.9 (18.3-27.4)
2 46 14.0 (10.3-17.8)
3 52 15.8 (11.9-19.8)
4 54 16.5 (12.4-20.5)
5 (highest) 55 16.8 (12.7-20.8)
Unknown 46 14.0 (10.3-17.8)
Death factor
Asphyxia 195 59.4 (54.1-64.8)
Fall/jump 34 10.4 (7.1-13.7)
Shooting <6t -- --
Alcohol/drug toxicity 64 19.5 (15.2-23.8)
Motor vehicle collision 20 6.1 (3.5-8.7)
Drowning <6t -- --
Other 6 1.8 (0.4-3.3)
CI: Confidence Interval.
* Percentages may not add to 100% due to rounding.
([dagger]) Small cells (<6) and corresponding proportions and
confidence intervals are suppressed to protect confidentiality.
