Suicide Misclassification: Evolution of a Skeptic

1
Suicide Misclassification Evolution of a Skeptic

• Ian R.H. Rockett, PhD, MPH
• Department of Community Medicine
• and the
• Injury Control Research Center
• West Virginia University School of Medicine
• Public Health Grand Rounds
• February 15, 2006

2
Objectives

• To develop a healthy skepticism towards official
  health statistics
• To comprehend sensitivity and specificity as
  concepts with application to suicide
  certification
• To consider the value of multiple cause of death
  information in assessing suicide data quality

3
Evolution of a Skeptic
4
Evolution of a Skeptic Reinterpreted

5
Trip Down Memory Lane
6
Approach

• Evolution of interest in suicide and suicide
  misclassification
• From Description to Analysis

7
History

• Morselli and Durkheim - 19th century
• Earliest suicide data skeptics? Zilboorg and
  Rumsey

8
Emile Durkheim

• Hypothesized that followers of religions or
  religious denominations which foster a high
  degree of social integration are less
  suicide-prone than their opposites
  whose religious affiliation encourages or is more
  permissive towards individualism and the
  pursuit of free inquiry
  (Catholics compared with Protestants)

9
Emile Durkheim

• Alternative explanation differential
  suicide case ascertainment related to variation
  in social condemnation of suicide

10
Personal History

• 5-country study US, United Kingdom, West
  Germany, France, Japan
• 3-country study United Kingdom, Australia,
  New Zealand
• Hypothesis formulation Japanese study

11
Data

• Age-sex-cause-specific mortality data
  published by the World Health Organization

12
Rockett Smith. Homicide, Suicide, Motor Vehicle
Crash, and Fall Mortality United States
Experience in Comparative Perspective. American
Journal of Public Health. 1989, 79(10),
1396-1400.
13
(No Transcript)
14
Rockett Smith. Homicide, Suicide, Motor Vehicle
Crash, And Fall Mortality United States
Experience in Comparative Perspective. American
Journal of Public Health. 1989,
79(10), 1396-1400.
15
Rockett Smith. Homicide, Suicide, Motor Vehicle
Crash, And Fall Mortality United States
Experience in Comparative Perspective. American
Journal of Public Health. 1989,
79(10), 1396-1400.
16

• Covert Suicide among Elderly Japanese Females
  Rockett, IRH Smith, GS,
  Social
  Science and Medicine,
  36 (11) 1993 1467-1472.

17
Data

• Age-sex-cause-specific mortality data from
  the World Health Organization
• Cause-specific suicide data for the Japanese
  suicide study came from the Japanese Ministry of
  Health and Welfare
• Annualized three years of data to stabilize
  the rates

18
Annual Average Suicide Rates by Age and Sex
Japan, 1979-1981
Rockett Smith. Covert Suicide Among Elderly
Japanese Females. Social Science and Medicine.
36(11),1467-1472,1993.
19
Big Bang rather than Evolution
20
Return of the Ape
21
Annual Average Unintentional Drowning Rates by
Age and Sex Japan, 1979-1981
Rockett Smith. Covert Suicide Among Elderly
Japanese Females. Social Science and Medicine.
36(11),1467-1472,1993.
22
Rockett Smith. Covert Suicide Among Elderly
Japanese Females. Social Science and Medicine.
36(11),1467-1472,1993.
23
Annual Average Drowning Suicide Rates by Age and
Sex Japan, 1979-1981
Rockett Smith. Covert Suicide Among Elderly
Japanese Females. Social Science and Medicine.
36(11),1467-1472,1993.
24
Ratio of Drowning Suicides to Unintentional
Drownings by Age and Sex Japan 1979-1981
Rockett Smith. Covert Suicide Among Elderly
Japanese Females. Social Science and Medicine.
36(11),1467-1472,1993.
25
The Rest of the Title

• Covert Suicide among Elderly Japanese Females
  Questioning Unintentional Drownings
• Rockett, IRH Smith, GS, Social Science and
  Medicine, 36 (11) 1993 1467-1472.

26
Cause-of-Death Categories hiding Suicides

• Injury of Undetermined Intent
• Unintentional Drownings
• Unintentional Poisonings
• Ill-defined and Unknown Causes (formerly
  Symptoms, Signs and Ill-defined
  Conditions)

27
Personal History

• International Collaborative Effort
  on Injury Statistics (ICE) meeting
  at NIH
• Suicide Misclassification in an
  
  International Context

28
(No Transcript)
29
Ratio of Combined Deaths from Accidental
Drowning, Accidental Poisoning, and Other
Violence to Suicides by Country, 1990
Rockett Smith. Suicide Misclassification in an
International Context. Proceedings of the
International Collaborative Effort on Injury
Statistics. 1995.
30
Ratio of Combined Deaths from Accidental
Drowning, Accidental Poisoning, Other Violence,
and Symptoms, Signs, and Ill-defined Conditions
to Suicides by Country, 1990
Rockett Smith. Suicide Misclassification in an
International Context. Proceedings of the
International Collaborative Effort on Injury
Statistics. 1995.
31
Personal History

• Reliability and Validity
• Sensitivity and Specificity
• (two sides of validity inclusions and
  exclusions)

32
What is Sensitivity of Suicide Certification?

• Measures the degree to which true suicides
  are classified as suicides ()

33
What is Sensitivity of Suicide Certification?

• Conceptual definition true positives
• Operational definition
• -- numerator official suicide count
• -- denominator count of suicides deaths
  categorized as unintentional drownings,
  unintentional poisonings, and undetermined
  injury intent (other violence)

34
What is Specificity of Suicide Certification?

• Measures the degree to which true nonsuicides
  (true negatives) are officially classified as
  nonsuicide deaths NASH
• Assumption in my research that specificity of
  suicide certification is 100

35
Rationale for 100 Specificity Assumption

• Medicolegal authorities are impeded from
  ruling even probable suicides as suicide without
  strong corroborative evidence, most notably from
  reliable witnesses, psychiatric records, or a
  letter of suicidal intent left by the decedent.

36
Rationale for 100 Specificity Assumption

• Further constraints against ruling a probable
  suicide a suicide in higher income democratic
  countries include concerns about
• precipitation of familial stigma
• awareness of the reluctance or unwillingness of
  insurance companies to redeem
  life insurance policies on suicide decedents
• political pressure

37

• Reliability and Sensitivity of Suicide
  Certification in Higher Income Countries Rockett,
  IRH Thomas, BM,
  Suicide and Life Threatening Behavior,
  1999 29 (2) 141-149.

38
Ratio of Combined Death Rate for Unintentional
Poisoning, Unintentional Drowning,
and Other Violence to Suicide Rate Over All
Ages by Country Females, 1988-1990
39
Ratio of Combined Death Rate for Unintentional
Poisoning, Unintentional Drowning, and Other
Violence to Suicide Rate Over All Ages by
Country Males, 1988-1990
40
Rockett Thomas. Reliability and Sensitivity of
Suicide Certification in Higher Income Countries.
Suicide and Life-Threatening Behavior. 29(2),
141-149, 1999.
41
Sensitivity of Suicide Certification by Age and
Country Females, 1988-1990
Sensitivity percentage estimates are calculated
by dividing the suicide rate by the combined
death rate of suicide, unintentional poisoning,
unintentional drowning, and other violence, and
then multiplying by 100.
42
Sensitivity of Suicide Certification by Age and
Country Males, 1988-1990
Sensitivity percentage estimates are calculated
by dividing the suicide rate by the combined
death rate of suicide, unintentional poisoning,
unintentional drowning, and other violence, and
then multiplying by 100.
43
Rockett Thomas, 1999

• Disaggregating age/sex/nation offering a
  surveillance tool
• Israeli study (Barel et al., 1996) --
  Jewish males ages 18-21 in 1987-89 --
  amended suicide rate twice the official rate

44
New Focus

• United States Black and White
  Suicide Rates
  

45
Background

• Suicide was the 10th leading cause of death
  for US whites in 2002
• Crude suicide rate for the period 1999-2002
  for whites was 12 per 100,000 -- gt double the
  black suicide rate (5 per 100,000) -- same with
  age- adjusted suicide rates

46
Underlying Cause-of-Death Data

• age/sex/underlying cause-specific mortality
  and corresponding population data for blacks and
  whites from the Compressed Mortality File (CDC
  WONDER) for 1999-2002
• 15 years and older population

47
Suicide Rates by Age and Race United
States,1999-2002
48
Research Question

• Are Black Suicide Rates
  Really Much Lower than
  White Rates?

49
Racial Gap a Paradox

• Warshauer Monk, 1978 Peck, 1983 Phillips
  et al., 1993 Mohler Earls, 2001
• Suicide Risk Factors
• (1) Alcohol/other drugs for 1999-2001 the
  age- adjusted rate for alcohol-induced
  deaths and other drug-induced deaths both
  29 higher for blacks than whites
• (2) HIV/Cancer 1999-2002 the black
  age- adjusted HIV death rate 7.7 times
  higher for blacks and the corresponding
  cancer rate 26 higher

50
More Suicide Risk Factors

• (3) Lower education -- blacks
  overrepresented among the less educated
• (4) Unemployment -- unemployed at twice
  the risk of white collar workers e.g.
  unemployment rates in 2003 black males (12) vs
  white males (6) and black females (10) vs
  white females (5)
• - figures ignore discouraged workers who are
  likely more common among blacks than
  whites

51
Sensitivity of Suicide Certification by Age, Sex,
and Race United States, 1999-2002
52
DEATH CERTIFICATION A PROBLEMATIC PROCESS
53
Death Certification

• Transcending their legal importance, death
  certificates form the centerpiece for conducting
  epidemiologic surveillance, planning and
  prioritizing public health expenditures and
  services
• Mortality data from the national vital
  statistics system are more universal,
  standardized, and timely than data from other
  major health databases. (Rosenberg, 1999)

54
Death Certification

• Critics draw on compelling empirical evidence
  from validation studies to charge that death
  certificates are seriously deficient (e.g.
  Kircher Anderson, 1987 Lloyd-Jones et al.,
  1998 Smith et al., 2001)
• The current state of death certificates in the
  United States would constitute a national
  embarrassment if it were not for the fact that
  the rest of the world seems to be similarly
  afflicted (Hill Anderson, 1990).

55
Death Certification

• A survey of over 700 medical examiners and
  coroners, the official responsible for
  medicolegal postmortems and autopsies, showed
  that few of them received any formal training in
  death certification in their medical school
  training or residencies (Goodin Hanzlick,
  1997).
• Death certification still receives scant
  attention in medical school and hospital training
  (Lakkireddy et al., 2004 Myers Farquhar, 2005
  Swain et al., 2005), and the autopsy rate is low
  and becoming lower (Rosenbaum et al., 2000 Welsh
  Kaplan, 1998).

56
United States Autopsy Rates
57
Implications

• The training deficit and the low autopsy rate
  have important implications not only for the
  quality of death certification in general, but
  for the quality of suicide certification in
  particular

58
Enter Social Constructionism

• Social constructionists critique the process of
  how humans create knowledge, including knowledge
  founded on quantification and measurement

59
Social Constructionism and
Suicide Data Quality

• Facts strongly suggest that the relationship
  between statistical organizations and the
  suicide rates they produce is subject to the
  following principle other things being equal,
  suicide rates vary directly with the degree
  of professional medical training of the
  categorizers, the average rate of manhours
  devoted to cause of death categorization, and
  the independence of the categorizers from
  interested parties (Douglas,
  1967)

60
Social Constructionism and
Suicide Data Quality

• Proper diagnosis of suicidal death rests upon
  adequate personnel, appropriate legislation, and
  financial and community support (Davis
  Spelman, 1968).
• -- proper diagnosis not only connotes accurate
  suicide case ascertainment, but harbors positive
  implications for the existence and nature of
  disease comorbidities listed on the death
  certificate of a suicide or possible suicide

61
Social Constructionism and
Suicide Data Quality

• Social constructionism is particularly salient to
  research on racial disparities in the quality of
  suicide data, especially data comparing blacks
  and whites when rate differences might at least,
  in part, be an artifact of variable practice by
  medicolegal authorities

62
Assumption

• Disease diagnoses listed on the death certificate
  of suicides and possible suicides precede most
  suicides (slow suicide a rare phenomenon but also
  likely to be highly prone to misclassification)
• -- may also be causal
• National data for 2001-2002 showed 99
  concordance between a suicidal mention on the
  death certificate and its assignment as the
  underlying cause of death (Redelings et
  al., 2006)

63
INSIDE THE BLACK-WHITE SUICIDE PARADOX
Comorbidity and
MisclassificationCo-Investigators Dr. Jeff
Coben Dr.
Erdogan Gunel
Dr. June Lunney
64
Multiple Cause-of Death DataNational Center
for Health Statistics

65

66

67
Central Hypothesis

• Documentation of black suicides is less complete
  in content and coverage than that of white suicide

68
Working Hypotheses

• Black suicides manifest a lower ratio of disease
  diagnoses to underlying suicide on their death
  certificates than white suicides
• Black decedents, whose underlying cause of death
  has been variously classified as suicide,
  unintentional drowning, unintentional poisoning,
  injury of undetermined intent, or ill-defined and
  unknown causes, collectively have a smaller
  likelihood than white counterparts of being
  classified as suicide

69
Rationale for Hypothesis 1

• Surveys show that blacks have higher rates of
  morbidity, disability, and mortality than whites
  (Arias et al., 2003)
• Blacks have less access to healthcare services
  than whites (Mayberry et al., 2000 Weinick et
  al., 2000), and when they access care they are
  less likely to receive routine medical procedures
  and more likely to experience lower quality
  services than whites (Smedley et al., 2003)
  spawning less detailed medical histories
• Blacks had a smaller number of diagnoses listed
  on their death certificate than whites (Havlik
  Rosenberg, 1992)
• Preliminary data

70
Preliminary
Data

71
Ratio of Diagnostic Disease Mentions to Suicides
by Race and Age United States, 1999-2002
72
Rationale for Hypothesis 2

• Evidence presented from the literature regarding
  Hypothesis 1
• Preliminary data

73
Ratio of Unintentional Poisoning Deaths to
Suicides by Age and Race United States, 1999-2002
74
Ratio of Deaths of Undetermined Injury Intent to
Suicides by Age and Race United States, 1999-2002
75
Ratio of Unintentional Drownings to Suicides by
Age and Race United States,
1999-2002
76
Ratio of Deaths of Ill-Defined and Unknown Causes
to Suicides by Age and Race
United States, 1999-2002
77
MultivariateModel

78
Multivariate Model

• Age
• Sex
• Race (black/white)
• Marital Status
• Educational Attainment
• Rurality
• Foreign Born
• Place of Death
• Type of State Medicolegal System

79
The scientific purist, who will wait for medical
statistics until they are nosologically exact, is
no wiser than Horaces rustic waiting for the
river to flow away Major
Greenwood, 1948
80
Guiding Principle

• High quality suicide data are essential for
  risk group delineation, risk factor
  identification, and program planning and
  evaluation

81
Long Term Goal

• Long term goal of this research is to provide
  information necessary for improving the quality
  of death certification
• Accuracy is an imperative for measuring and
  monitoring mortality, morbidity, and disability,
  formulating sound health policy, and instituting
  evidence-based prevention programs

82
Thats All Folks for now. . .

83

• Please complete the survey for the presentation
  at
• http//tinyurl.com/8663x
• and then follow the link to obtain your free
  continuing education credit.