ORIGINAL ARTICLE

Suicide Risk in Primary Care Patients

With Major Physical Diseases
A Case-Control Study
Roger T. Webb, PhD; Evangelos Kontopantelis, PhD; Tim Doran, MD;
Ping Qin, PhD; Francis Creed, MD; Nav Kapur, MD

Context: Most previous studies have examined suicide

risk in relation to a single physical disease.

Objectives: To estimate relative risk across a range of
physical diseases, to assess the confounding effect of clinical depression and effect modification by sex and age, and
to examine physical illness multimorbidity.
Design: Nested case-control study.
Setting: Family practices (n = 593) registered with the
General Practice Research Database from January 1, 2001,
through December 31, 2008. The case-control data were
drawn from approximately 10.6 million complete patient records, pertaining to approximately 8% of the total
population of the United Kingdom, with complete linkage to national mortality records.

Results: Among all patients, coronary heart disease,

stroke, chronic obstructive pulmonary disease, and osteoporosis were linked with elevated suicide risk, and,
with the exception of osteoporosis, the increase was explained by clinical depression. The only significantly elevated risk in men was with osteoporosis. Female effect
sizes were greater, with 2- or 3-fold higher risk found
among women diagnosed as having cancer, coronary heart
disease, stroke, chronic obstructive pulmonary disease,
and osteoporosis. In women with cancer and coronary
heart disease, a significant elevation persisted after adjustment for depression. Overall, heightened risk was confined to physically ill women younger than 50 years and
to older women with multiple physical diseases.

17 460 living controls matched on age and sex were studied. The reference group for relative risk estimation consisted of people without any of the specific physical illnesses examined.

Conclusions: Our findings indicate that clinical depression is a strong confounder of increased suicide risk among
physically ill people. They also demonstrate an independent elevation in risk linked with certain diagnoses, particularly among women. Health care professionals working across all medical specialties should be vigilant for
signs of undetected psychological symptoms.

Main Outcome Measures: Suicide and open verdicts.

Arch Gen Psychiatry. 2012;69(3):256-264

Participants: A total of 873 adult suicide cases and

Author Affiliations: Centre for

Suicide Prevention, Centre for
Mental Health and Risk
(Drs Webb and Kapur), Health
Sciences Primary Care Research
Group (Drs Kontopantelis and
Doran), and Mental Health and
Neurodegeneration Research
Group (Dr Creed), The
University of Manchester,
Manchester, England; and
National Centre for
Register-based Research,
University of Aarhus, Aarhus,
Denmark (Dr Qin).

HYSICALLY ILL PEOPLE ARE

more likely to have poor

mental health and vice versa,1
and suicide occurs more frequently with the coexistence of psychiatric and physical illnesses. 2 The relevant literature is
dominated by investigations of a single
physical illness, with higher risk indicated with asthma,3 cancer,4,5 coronary
heart disease,6 diabetes mellitus,7 epilepsy,8,9 and stroke.10 Few studies11-13 have
compared effects among different physical illnesses in the same cohort. The range
of conditions that confer higher risk may
be broader than the current literature suggests, and the underlying reasons for this

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

256

association are not fully understood. In

particular, a greater understanding of the
confounding role of clinical depression is
needed. The association between physical illness and suicide is a major clinical
and public health issue for several reasons. First, these patients usually have
regular health service contact, which offers more opportunities for monitoring and
intervention. Second, clinical depression
is the key risk factor for suicide14 and is
considerably more common in physically ill people. It can be treated effectively in this population15 and thereby reduce risk. 16 Third, the proportion of
suicides known to be associated with
physical illness is considerable for indi-

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

viduals 50 years or older,17 which implies a significant

population impact.
Conducting such research requires large cohorts because of the rarity of suicide. We used the United Kingdoms General Practice Research Database (GPRD),18,19
which has been used to investigate mortality11,20 and depression.21,22 Its potential has been further enhanced by
recent implementation of complete linkage to national
mortality records. We addressed the following research
questions: (1) How does relative risk of suicide vary among
different types of major physical illness among all registered primary care patients? (2) To what degree does
clinical depression detected by health care services explain elevated risk? (3) Is relative risk modified by sex
and age? (4) Does risk increase with increasing levels of
physical illness multimorbidity?
We hypothesized higher risk across all 11 physical diseases examined. We expected depression to largely explain any elevated risks observed and that risk would be
greatest in those with multiple illnesses.13 We also assessed age- and sex-specific effects because most previous studies12,13,23-25 have investigated older people. Evidence is lacking for effects at younger age26 and by sex.
METHODS

DATA SETS AND

OUTCOME ASCERTAINMENT
The case-control study was constructed using the GPRD,18,19
the worlds largest population-based, longitudinal, primary care
database. Before our study data set was created, the Independent Scientific Advisory Committee of the GPRD granted ethics approval. Consent from individual patients is not required
for GPRD-based studies. Although participating practices were
not randomly sampled, the database provides a broadly representative sample of all people registered with a family practice, which applies to virtually every resident of the United Kingdom. The September 2010 version we analyzed contained
approximately 10.6 million complete patient records from 593
family practices, including approximately 8% of the total population. It routinely records all primary care consultations for
registered patients, with comprehensive and detailed clinical
coding for symptoms, diagnoses, treatment (including prescribed medication), and referral to other forms of National
Health Service care and to other health care providers. During
2008, complete prospective and historic linkage to national mortality registration systems was implemented via the Office for
National Statistics, England and Wales, and the General Register Offices for Scotland and for Northern Ireland.
We included adult suicides if the individual died from January 1, 2001, through December 31, 2008, and had at least 2
full years of up-to-standard GPRD clinical data. These quality criteria were also applied in selecting living controls. In the
United Kingdom, most unnatural deaths of undetermined cause
(or open verdicts) among adults are considered likely to be suicides.27 To reduce false-negative misclassification28 and in line
with standard practice for conducting epidemiologic studies of
suicide in the United Kingdom, our case definition included
these open verdicts. We used the following ICD-10 codes: X60-84
and Y10-3429 (excluding Y33.9; ie, deaths with adjourned inquests that are mostly deemed subsequently to be homicides).30
By this definition, the male suicide rate in the United Kingdom in 2009 was 17.5 per 100 000 population and the female
rate was 5.2 per 100 000 population.31

PHYSICAL DISEASES
AND CLINICAL DEPRESSION
We examined major physical illnesses that either place a heavy
burden on primary care services in the United Kingdom32 or are
indicated in the literature as being likely correlates of suicidal
behavior or clinical depression. We delineated them using the
Read/OXMIS codes that capture the complete clinical record of
each GPRD patient.33 Read/OXMIS is a hierarchical coding system widely used in primary care in the United Kingdom. For major disease groups, the first digit denotes the chapter heading,
with subheadings giving progressively more detail (eg, G, circulatory system diseases; G3, ischemic heart disease; G30, acute
myocardial infarction; and G300, acute anterolateral myocardial infarction). Recent validation showed that ICD-9 codes for
identifying multiple physical diseases can be readily translated
for examination in the GPRD.34 Two of us (E.K. and T.D.) have
previously examined the effect of national schemes to incentivize general practitioners to improve their quality of care.35 We
conducted an extensive consensus-development process with experienced academic general practitioners to identify valid codes
denoting major illnesses in the GPRD and other national routinely collected data sets. We then applied their coding ranges.
The conditions that we examined were cancer, coronary heart
disease, hypertension, stroke, diabetes, asthma, chronic obstructive pulmonary disease (COPD), osteoarthritis, osteoporosis, back
pain, and epilepsy (Table 1 and Figure 1).
In addition, Doran and colleagues35 had identified coding
ranges for clinical depression. We used these to delineate previous or current episodes recorded in the patients clinical record, according to diagnoses made before suicide by general practitioners or specialist mental health services. To achieve
comprehensive adjustment for depression as a confounder, patients with a history of depression but who were not known to
be depressed at the time of their suicide were classified as having depression unless their record clearly indicated that their
condition had remitted. For controls, it was necessary for all
physical illnesses and clinical depression to have been coded
in the GPRD before the matched cases date of death. The detailed lists of Read/OXMIS codes used are available on request
from the corresponding author.

STUDY DESIGN AND STATISTICAL ANALYSIS

The analyses were conducted using Stata statistical software (StataCorp). The 873 suicides in our case-control study were matched
to 17 460 living controls by sex, age (in years), and date of death.
Twenty controls were selected from the age-, sex-, and timespecific risk set of each case to maximize power for precise effect
estimation with physical illnesses of low prevalence.36 We calculated prevalence among cases and controls separately and generated conditional logistic regression models to estimate relative
risks as exposure odds ratios (ORs) across both sexes. These models were adjusted inherently for sex and age in the matched design. We then fitted sex interaction terms and formally tested their
significance. The incidence density sampling in the nested casecontrol design meant that the ORs were interpretable as hazard
ratios (relative risks), as would be derived from survival analysis
of the entire cohort.37 To make our estimates comparable among
specific physical illnesses, we applied a generic reference category (ie, patients with none of the 11 assessed physical illnesses). This was a heterogeneous group. In the United Kingdom, virtually the entire population is registered with a family
practice, meaning that some of these people will have been completely well, whereas others will have had conditions that we did
not examine. Younger registered patients are more likely to have
been fit and well than older ones, but age-matching ensured that
this did not confound our estimates.

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

257

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

Table 2. Relative Risk of Suicide in Patients With Specific Physical Illnesses: Evidence of Effect Modification by Sex a
No. (%)
Physical Illnesses
Any of the illnesses
Men
Women
12 e
P value
Cancer
Men
Women
12 e
P value
Coronary heart disease
Men
Women
12 e
P value
COPD
Men
Women
12 e
P value

Cases
(n = 873)

Controls
(n = 17 460)

Odds Ratio
(95% CI)

Adjusted Odds
Ratio (95% CI) b

Median
Age, y c

227/658 (34.5)
111/215 (51.6)

4504/13 160 (34.2)

1954/4300 (45.4)

1.02 (0.84-1.23)
1.39 (1.01-1.90)
2.8
.10

0.82 (0.68-1.00) d
1.10 (0.80-1.52)
2.2
.14

61
64

12/658 (1.8)
18/215 (8.4)

360/13 160 (2.7)

201/4300 (4.7)

0.66 (0.36-1.21)
2.23 (1.28-3.86)
8.5
.004

0.56 (0.30-1.04)
1.85 (1.04-3.26)
7.6
.006

76.5
69.5

47/658 (7.1)
24/215 (11.2)

817/13 160 (6.2)

218/4300 (5.1)

1.19 (0.84-1.68)
3.10 (1.81-5.29)
8.7
.003

0.87 (0.61-1.24)
2.04 (1.18-3.55)
6.3
.01

74
74

17/658 (2.6)
10/215 (4.7)

249/13 160 (1.9)

80/4300 (1.9)

1.41 (0.83-2.39)
3.23 (1.57-6.66)
3.3
.07

1.09 (0.63-1.88)
1.90 (0.89-4.06)
1.3
.26

66
63

Abbreviation: COPD, chronic obstructive pulmonary disease.

a Effects with a statistically significant (P .05) or borderline significant (P .05, P .10) sex interaction; all Table 1 main effects were refitted with a sex
interaction term added to each conditional logistic regression model. All statistically significant sex-specific odds ratios are highlighted in bold.
b Multivariate adjustment for clinical depression.
c Median age of cases (and age-matched controls) when case died.
d This effect was statistically significant (ie, upper confidence limit = 0.995).
e Interaction test.

(53.3% vs 43.2%; 12 = 6.2; P = .01), whereas among female cases, no difference in prevalence was found between these 2 groups (66.7% with physical illness vs
67.3% without; 12 = 0.01; P = .92). Prevalence among all
living controls with physical illnesses was 19.9% vs
12.6% in those with no physical illnesses (12 = 164.8;
P .001). Figure 1 shows the variation in prevalence
among suicides across the specific physical illnesses examined. Prevalence was approximately two-thirds or
greater in cases with asthma (64.7%) and back pain
(74.2%) and was lower than half in those with osteoporosis (47.8%) and stroke (44.1%). Across the specific
illnesses, we tested for correlation between the OR
point estimate and prevalence of depression in each illness (Table 1). The Pearson coefficient was 0.42 but
was nonsignificant (P = .20).
SEX- AND AGE-SPECIFIC EFFECTS
The median age of all female suicide cases (54 years) was
significantly greater than that of male suicide cases (48
years) (Wilcoxon rank-sum test, P.001), and the prevalence of physical illness was greater in women (Table 2).
Evidence of effect modification by sex is also given in
Table 2. Women with any of the physical illnesses assessed had significantly elevated suicide risk, with no effect at all evident in men, although the sex difference did
not reach significance (P = .10). Almost without exception, relative risks were greater in women than men across
the specific illnesses. Highly significant sex differences

were seen among patients diagnosed as having cancer and

coronary heart disease. For these 2 conditions, the elevated risk among women was independent of clinical
depression, with an approximate doubling of risk indicated in the multivariate models. There was a particularly strong sex interaction with cancer; a more than 2-fold
significant increase in risk was seen among these women,
but there was no evidence of an effect in men. The OR
in men diagnosed as having cancer was lower than unity
but with no significant evidence of a protective effect
(12 = 1.8; P =.18). For COPD, a highly significant 3-fold
elevation in risk was seen in women, although the sex
interaction term was marginally nonsignificant. These sexspecific analyses revealed considerably greater relative
risks in women (than men) for cancer, coronary heart
disease, and COPD (effects that were hidden in the main
effects presented in Table 1).
Two additional sex-specific effects, with nonsignificant interaction terms and hence not listed in Table 2,
are also noteworthy. First, for osteoporosis, a comparable increase in risk was indicated in both sexes (men:
8 cases; OR, 2.72; 95% CI, 1.27-5.82; women: 15 cases;
OR, 2.44; 95% CI, 1.32-4.51). Osteoporosis was the only
specific physical illness examined that was associated with
a significantly elevated risk in men. Second, for stroke,
although the sex interaction term was nonsignificant
(12 = 1.0; P = .32), a more than doubled risk was indicated in these women (10 cases; OR, 2.14; 95% CI, 1.034.45), whereas in men, the increase was nonsignificant
(24 cases; OR, 1.38; 95% CI, 0.87-2.19).

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

259

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

each specific illness, but we have shown a linear relationship of increasing risk by increasing multimorbidity among women. This factor could be viewed as a proxy
for severity, although not all severely ill patients will have
experienced more than 1 major physical disease.
One final limitation was the lack of sociodemographic
data held by the GPRD. Risk factors, such as marital status, social class, unemployment, and living alone, are not
recorded during routine patient consultations. We, therefore, could not adjust for these important confounders.
To our knowledge, this is the first large populationbased study to examine suicide risk in a broad range of
physical diseases with systematic and complete linkage
to national mortality data. Our findings show that health
care professionals working across all medical specialties
should be vigilant for signs of undetected psychological
symptoms,2,57 especially when treating women diagnosed as having cancer, coronary heart disease, COPD,
osteoporosis, or stroke. Previous or current clinical depression did not explain elevated risk seen in women with
cancer and coronary heart disease, and it only partially
explained higher risk seen in younger, physically ill
women and those with multiple diseases. For these groups
of women, our results challenge the notion that physical illness and associated disability lead to depression,
which leads to suicide. It could be that younger women
feel threatened by serious physical disease, particularly
those that are widely known to be the most common
causes of premature death. The specific mechanisms involved are likely to vary greatly according to the nature
of the physical illness being considered, and further research is therefore needed to understand the causal pathways. Training for a wider range of health care professionals in suicide risk management and reduction may
be beneficial. Addressing this need could greatly improve patients quality of life by tackling their psychological distress, as well as potentially reducing populationlevel suicide rates.
Submitted for Publication: July 15, 2011; accepted August 19, 2011.
Correspondence: Roger T. Webb, PhD, Centre for Suicide Prevention, Centre for Mental Health and Risk, University of Manchester, Jean McFarlane Building, Oxford
Road, Manchester, England M13 9PL (roger.webb
@manchester.ac.uk).
Financial Disclosure: None reported.
Funding/Support: The investigation was unfunded. The
study data set was obtained by the authors free of charge
via a Single Study Dataset Agreement between the GPRD
and the Medical Research Council under the aegis of the
Medical Research Council License Agreement.
Additional Contributions: Tarita Murray-Thomas, MSc,
GPRD, Medicines and Healthcare Products Regulatory
Agency, London, England, supplied the GPRD data set
in the form of a nested case-control study.
REFERENCES
1. Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. Depression, chronic
diseases, and decrements in health: results from the World Health Surveys. Lancet.
2007;370(9590):851-858.

2. Copsey Spring TR, Yanni LM, Levenson JL. A shot in the dark: failing to recognize the link between physical and mental illness. J Gen Intern Med. 2007;22
(5):677-680.
3. Kuo CJ, Chen VC, Lee WC, Chen WJ, Ferri CP, Stewart R, Lai TJ, Chen CC, Wang
TN, Ko YC. Asthma and suicide mortality in young people: a 12-year follow-up
study. Am J Psychiatry. 2010;167(9):1092-1099.
4. Louhivuori KA, Hakama M. Risk of suicide among cancer patients. Am J Epidemiol.
1979;109(1):59-65.
5. Allebeck P, Bolund C, Ringback G. Increased suicide rate in cancer patients: a
cohort study based on the Swedish Cancer-Environment Register. J Clin Epidemiol.
1989;42(7):611-616.
6. Larsen KK, Agerbo E, Christensen B, Sndergaard J, Vestergaard M. Myocardial
infarction and risk of suicide: a population-based case-control study. Circulation.
2010;122(23):2388-2393.
7. Tseng CH. Mortality and causes of death in a national sample of diabetic patients in Taiwan. Diabetes Care. 2004;27(7):1605-1609.
8. Nilsson L, Tomson T, Farahmand BY, Diwan V, Persson PG. Cause-specific mortality in epilepsy: a cohort study of more than 9,000 patients once hospitalized
for epilepsy. Epilepsia. 1997;38(10):1062-1068.
9. Christensen J, Vestergaard M, Mortensen PB, Sidenius P, Agerbo E. Epilepsy
and risk of suicide: a population-based case-control study. Lancet Neurol. 2007;
6(8):693-698.
10. Teasdale TW, Engberg AW. Suicide after a stroke: a population study. J Epidemiol Community Health. 2001;55(12):863-866.
11. Haste F, Charlton J, Jenkins R. Potential for suicide prevention in primary care?
an analysis of factors associated with suicide. Br J Gen Pract. 1998;48(436):
1759-1763.
12. Waern M, Rubenowitz E, Runeson B, Skoog I, Wilhelmson K, Allebeck P. Burden of illness and suicide in elderly people: case-control study. BMJ. 2002;
324(7350):1355-1357.
13. Juurlink DN, Herrmann N, Szalai JP, Kopp A, Redelmeier DA. Medical illness and
the risk of suicide in the elderly. Arch Intern Med. 2004;164(11):1179-1184.
14. Isometsa ET, Henriksson MM, Aro HM, Heikkinen ME, Kuoppasalmi KI, Lonnqvist
JK. Suicide in major depression. Am J Psychiatry. 1994;151(4):530-536.
15. National Institute for Clinical Excellence (NICE). Depression in Adults With a Chronic
Physical Health Problem: Treatment & Management. London, England: National
Collaborating Centre for Mental Health; 2009. National Health Service (NHS) NICE
clinical guideline 91.
16. Rihmer Z. Can better recognition and treatment of depression reduce suicide rates?
a brief review. Eur Psychiatry. 2001;16(7):406-409.
17. Hendin H. Suicide, assisted suicide, and medical illness. J Clin Psychiatry. 1999;
60(suppl 2):46-52.
18. Walley T, Mantgani A. The UK General Practice Research Database. Lancet. 1997;
350(9084):1097-1099.
19. Khan NF, Harrison SE, Rose PW. Validity of diagnostic coding within the General Practice Research Database: a systematic review. Br J Gen Pract. 2010;
60(572):e128-e136. doi:10.3399/bjgp10X483562.
20. Osborn DP, Levy G, Nazareth I, Petersen I, Islam A, King MB. Relative risk of
cardiovascular and cancer mortality in people with severe mental illness from
the United Kingdoms General Practice Research Database. Arch Gen Psychiatry.
2007;64(2):242-249.
21. Watson L, Baird J, Hosel V, Peveler R. The effect of concurrent pain on the management of patients with depression: an analysis of NHS healthcare resource utilisation using the GPRD database. Int J Clin Pract. 2009;63(5):698-706.
22. Byford S, Barrett B, Despiegel N, Wade A. Impact of treatment success on health
service use and cost in depression: longitudinal database analysis.
Pharmacoeconomics. 2011;29(2):157-170.
23. Grabbe L, Demi A, Camann MA, Potter L. The health status of elderly persons in
the last year of life: a comparison of deaths by suicide, injury, and natural causes.
Am J Public Health. 1997;87(3):434-437.
24. Conwell Y, Duberstein PR, Caine ED. Risk factors for suicide in later life. Biol
Psychiatry. 2002;52(3):193-204.
25. Quan H, Arboleda-Florez J, Fick GH, Stuart HL, Love EJ. Association between
physical illness and suicide among the elderly. Soc Psychiatry Psychiatr Epidemiol.
2002;37(4):190-197.
26. Viilo KM, Timonen MJ, Hakko HH, Sarkioja T, Meyer-Rochow VB, Rasanen PK.
Lifetime prevalences of physical diseases and mental disorders in young suicide victims. Psychosom Med. 2005;67(2):241-245.
27. Linsley KR, Schapira K, Kelly TP. Open verdict v. suicideimportance to research.
Br J Psychiatry. 2001;178:465-468.
28. Neeleman J, Wessely S. Changes in classification of suicide in England and Wales:
time trends and associations with coroners professional backgrounds. Psychol
Med. 1997;27(2):467-472.
29. World Health Organization. Manual of the International Classification of Diseases,
10th Revision (ICD-10). Geneva, Switzerland: World Health Organization; 1992.

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

263

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

Table 3. Sensitivity Analysis Comparing Significant

Sex-Specific Odds Ratios With and Without Open Verdicts
Included in the Suicide Case Definition a
Odds Ratio (95% CI)
Physical Illnesses
Cancer in women
Coronary heart
disease in women
Stroke in women
COPD in women
Osteoporosis
Men
Women

Suicide and
Open Verdicts

Suicide
Verdicts Only

2.23 (1.28-3.86)
3.10 (1.81-5.29)

2.15 (1.15-4.03)
2.09 (1.05-4.16)

2.14 (1.03-4.45)
3.23 (1.57-6.66)

1.77 (0.70-4.46)
2.78 (1.11-6.96)

2.72 (1.27-5.82)
2.44 (1.32-4.51)

2.48 (1.04-5.95)
1.90 (0.88-4.11)

Abbreviation: COPD, chronic obstructive pulmonary disease.

statistically significant sex-specific odds ratios are highlighted in bold.

a All

3.5

Odds Ratio (95% CI)

3.0
2.5
2.0
1.5
1.0
0.5
0.0

Men < 50 y
n = 58

Women < 50 y
n = 29

Men 50 y
n = 169

Women 50 y
n = 82

Figure 2. Sex-specific relative risk of suicide in patients diagnosed as having

any of the physical illnesses stratified by age (12 = 6.5, P =.01, for sex
interaction at younger than 50 years).

For the 5 female and 1 male disease-specific ORs that

were significantly elevated, we performed a sensitivity
analysis assessing the impact of including open verdicts
in the suicide case definition (Table 3). With only definite suicide cases examined, the effect sizes were attenuated somewhat, and 2 of the female-specific effects (osteoporosis and stroke) became nonsignificant, perhaps
partly due to reduced power. The findings overall, however, were essentially unchanged.
We assessed the sex-specific effect for having any of
the 11 physical illnesses in cases 50 years or older and
those younger than 50 years (Figure 2). At 50 years or
older, there was no evidence of a difference in risk by sex
(interaction test: 12 = 0.0; P = .96). However, in those
younger than 50, there was an almost doubled risk in
women and no increased risk for men (interaction: 12 =
6.5; P=.01). The effect in women younger than 50 was
partially explained by clinical depression (unadjusted OR,
1.85; 95% CI, 1.16-2.92; adjusted OR, 1.38; 95% CI, 0.862.22). There was a suggestion of higher prevalence of depression among the younger female suicide cases (23 of
29 [79.3%]) compared with those 50 years or older (51
of 82 [62.2%]), although this difference was nonsignificant (12 = 2.8; P = .09).

We fitted age interaction terms (50 vs 50 years)

against each specific physical illness for both sexes combined because of small cell values for some of the sexand age-specific ORs. With COPD, there was a much
stronger effect at younger age (50 years: 4 cases; OR,
16.11; 95% CI, 4.32-60.03; 50 years: 23 cases; OR, 1.55;
95% CI, 0.97-2.47; age interaction: 12 = 10.8; P=.001),
although this result should be viewed cautiously because of the small number of younger suicide cases. There
was no evidence of an age difference in the effect linked
with osteoporosis (P=.17).
PHYSICAL ILLNESS MULTIMORBIDITY
Overall, there was no evidence of increasing risk linked
with increasing multimorbidity, and this was also true
for men only (Table 4). However, risk in women with
3 or more physical illnesses was more than twice as high
as in those without any of the examined conditions, with
strong evidence of a linear trend. These female cases generally died at an older age (median, 77 years), and the
higher risk was only partially explained by clinical depression (adjusted OR, 1.53; 95% CI, 0.90-2.59). There
was a significant sex interaction among patients with 3
or more physical diseases (12 = 6.5; P=.01). Comparing
these results with those shown in Figure 2, we expected
to find a more marked effect of multimorbidity in younger
women. However, this was hard to assess because there
were just 4 female cases younger than 50 years with 2
physical illnesses and 3 younger cases with 3 or more conditions. Nonetheless, the point estimate was large for
younger women with 3 or more illnesses (OR, 9.51; 95%
CI, 2.32-39.01). This effect was highly significant (12 =
9.8; P =.002) but should be viewed cautiously because
of its imprecision. In men, there was no evidence of higher
risk with multimorbidity at younger than 50 years or 50
years or older.
DETAILED INVESTIGATION OF ELEVATED RISK
IN WOMEN WITH CANCER
We examined the records to identify reasons for the strong
sex interaction. The 18 female suicide cases died at a
younger age than did the 12 male cases (median, 69.5 vs
76.5 in men) but not significantly so (Wilcoxon ranksum test, P =.21). Also, a longer time elapsed between
first diagnosis of cancer and suicide in the female cases
(median, 5.99 years in women vs 1.52 years in men), but
again the difference was not significant (P =.15). Half of
the female suicide cases with cancer (9 of 18) had experienced malignant neoplasm of the breast. Suicide risk
was approximately doubled in this subgroup (OR, 2.06;
95% CI, 0.99-4.29) as it was in all women with cancer.
COMMENT

SUMMARY OF FINDINGS
Among all primary care patients, we found significantly
higher suicide risk with coronary heart disease, stroke,
COPD, and osteoporosis. Female effect sizes were gen-

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

260

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

Table 4. Relative Risk of Suicide by Increasing Level of Physical Illness Multimorbidity a

No. (%)
No. of Physical Illnesses
All people
0
1
2
3
12 c
P value
Men
0
1
2
3
12 c
P value
Women
0
1
2
3
12 c
P value

Cases
(n = 873)

Controls
(n = 17 460)

Odds Ratio
(95% CI)

Median Age, y b

535 (61.3)
202 (23.1)
80 (9.2)
56 (6.4)

11 002 (63.0)
3916 (22.4)
1592 (9.1)
950 (5.4)

1.00 [Reference]
1.09 (0.91-1.29)
1.08 (0.83-1.41)
1.29 (0.94-1.78)
2.2
.14

43
55
70.5
77

431/658 (65.5)
144/658 (21.9)
53/658 (8.1)
30/658 (4.6)

8656/13 160 (65.8)

2787/13 160 (21.2)
1087/13 160 (8.3)
630/13 160 (4.8)

1.00 [Reference]
1.04 (0.84-1.27)
0.97 (0.71-1.34)
0.95 (0.62-1.44)
0
.89

43
55
70
77.5

104/215 (48.4)
58/215 (27.0)
27/215 (12.6)
26/215 (12.1)

2346/4300 (54.6)
1129/4300 (26.3)
505/4300 (11.7)
320/4300 (7.4)

1.00 [Reference]
1.27 (0.89-1.79)
1.42 (0.88-2.30)
2.27 (1.35-3.84)
8.6
.003

46
53.5
73
77

a Multimorbidity is measured as the number of physical illnesses from the list of 11 illnesses examined in this study, not the total number of illnesses recorded
in the General Practice Research Database.
b Median age of cases (and age-matched controls) when case died.
c Linear trend test.

erally stronger than their equivalent male effects, and the

only significantly elevated risk found in men was with
osteoporosis. Significantly increased risk was seen in
women who had any of the physical illnesses examined,
and a 2- or 3-fold higher risk was indicated in women
with cancer, coronary heart disease, stroke, COPD, and
osteoporosis. Elevated risk in women with cancer or coronary heart disease was independent of clinical depression. Risk was greater in younger, physically ill women
per se and in older women with multimorbidity.
COMPARISON WITH EXISTING EVIDENCE
AND INTERPRETATION
A few case-control studies have previously been reported. For example, a smaller study,11 conducted from
1991 through 1993 using the GPRD, found no evidence
of higher risk with a range of physical diseases. It lacked
systematic linkage to national mortality data and thus may
have been prone to ascertainment bias and low power.
A second small case-control study,12 conducted among
elderly people in Gothenburg, Sweden, reported that serious physical illness and a greater total burden of illness predicted higher risk, with these effects restricted
to men. With just 85 cases it also lacked power, and less
than two-thirds of those eligible for sampling as controls consented to participate. By contrast, we found
greater female effect sizes.
Our failure to observe elevated risk in physically ill
men requires explanation, especially given that rates are
high for elderly men in many Western countries.38 A
study39 of the US National Health Interview Survey linked

with the National Death Index reported that chronic physical illness did not predict suicide independent of functional limitation. Thus, associated disability rather than
illness per se could be the key determinant in men. Our
findings suggest that increased risk may be restricted to
certain groups; for example, younger, physically ill
women, older women with multiple illnesses, and those
with only some of the 11 specific diseases examined. Insufficient statistical power might explain why we failed
to find more significant effects across the full range of
illnesses examined, in both sexes, and across the adult
age range. However, this seems unlikely because our study
was much better powered to detect male effects than female ones (there were 3 times more male suicides) and
because some of the illnesses indicating no association
were in fact adequately powered to detect modest effect
sizes. This was so for asthma, diabetes, hypertension, and
osteoarthritis, for example. A more likely explanation lies
with the intrinsic nature of this primary care population. Stronger effects across a broader range of illnesses
and in all demographic subgroups might be seen by investigating a cohort treated for these illnesses in secondary care.
Scandinavian registry studies have shown higher risk
in patients diagnosed as having myocardial infarction,6
stroke,10 and epilepsy.8,9 We did not find an association
with epilepsy, perhaps due to low power, because we observed only 10 suicide cases with this diagnosis. National registry studies conducted in Taiwan showed higher
risk with asthma3 and diabetes,7 but these findings may
not generalize to Western nations, and the asthma study
was restricted to young people. We found a 2- to 3-fold

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

261

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

elevated risk with COPD in women and osteoporosis in

both sexes. To our knowledge, these associations have
not been reported previously, although recent reviews
have indicated higher rates of depression in these 2
groups.40,41 The higher risk in men and women with osteoporosis, a novel finding, may be linked with fracture
and trauma, although we could find no evidence of such
an association in the existing literature. As with most of
the other conditions we examined, elevated risk in patients with osteoporosis may be linked with unhealthy
lifestyles, such as smoking42 and heavy alcohol use,43
which are also established risk factors for suicide.44,45
Previous population-based studies have shown higher
risk after diagnosis with cancer, with conflicting evidence of a sex difference. For example, relative risk was
higher for women than men during the 1950s and 1960s
in the Finnish Cancer Registry4 and vice versa in the Swedish Cancer-Environment Register during the 1960s and
1970s.5 Recent reports from Sweden46 and Italy47 indicate decreasing risk associated with cancer, perhaps due
to improved treatment outcome and better communication of diagnosis.47 Suicide is especially likely soon after
diagnosis,5,48 although long-term follow-up of almost
724 000 women in the United States and Scandinavia
showed that elevated risk persists for up to 2 to 3 decades after breast cancer diagnosis,49,50 the most common cancer in women. Increased levels of psychological distress among breast cancer survivors may be due
to the adverse effects of mastectomy.51,52 This procedure
had occurred in only 3 of the female suicide cases we observed, and so we could not estimate relative risk specifically in these women. Overall, we did not see a higher
risk specific to neoplasm of the breast vs other cancers
in women.
STRENGTHS AND LIMITATIONS
The GPRD uniquely provides a large, detailed, and nationally representative computerized cohort of primary
caretreated mental and physical illness, with complete
linkage and follow-up of cause-specific mortality. Investigation of rare exposures and events is possible, and biases that commonly flaw epidemiologic studies are minimized. For example, information bias is reduced because
the data are collected prospectively, irrespective of subsequent outcome. National registers in Denmark and Sweden have collected details of all hospital admissions during several decades, but we could investigate milder forms
of physical illness among people not hospitalized. We
could also make a more comprehensive assessment of the
confounding role of clinical depression, as detected by
all national public health care services in both primary
and secondary care settings.
Selection into primary care was an unlikely source of
major bias because almost all residents of the United Kingdom are registered with a general practitioner soon after their birth. When people move and register with a new
practice, their complete primary health care record is
transferred automatically. Thus, the GPRD is a representative community sample. Female patients, especially
those who are mentally ill, may see their general practitioner more often than their male counterparts, giving

physicians more opportunities to diagnose physical illnesses. This bias may pertain to conditions such as back
pain, osteoporosis, and hypertension because they may
not require urgent medical attention. However, it is unlikely to have affected our assessments of the other 8 physical illnesses. These are conditions that need active clinical care, which in the United Kingdom is mostly provided
through primary care. Patients general practitioners are
formally notified of diagnoses made by physicians in public or private hospitals, and this information is thereby
captured in the GPRD. Thus, for most of the illnesses we
examined, biased selection into primary care is an unlikely explanation for the excess risk we found in female patients.
The main limitation of our study was potential misclassification of exposure status, which occurred for 4 main
reasons. First, our generic reference group consisted of patients without any of the assessed physical illnesses. Patients in this group may have had other medical conditions that we did not examine. However, we adopted a
pragmatic approach that enabled direct comparison of effects across the specific illnesses. Delineation of GPRD patients with no major physical illnesses of any sort would
be an infeasible task with such a large data set containing
so much detailed and complex coding. Second, our adjustments for clinical depression were not fully comprehensive because we could not assess depression among
people who did not seek treatment or that was not detected by health care services. We attempted additional adjustments for all Axis I disorders, including anxiety disorders and substance dependence. However, we were not
completely confident of identifying all episodes of these
disorders in the Read/OXMIS coding, and so we adjusted
for clinical depression only. Third, because of logistical
constraints in examining multiple diseases, we did not use
GPRD medication data. In defining clinical depression, we
opted not to use antidepressant medication data because
of historic concerns about subtherapeutic dosage and prescription for indications other than depression in primary care in the United Kingdom.53,54 We anticipate that
not including medication in defining the assessed conditions may have attenuated some effect estimates and would
therefore not have generated false-positive results.55 Finally, although the clinical records of each GPRD practice can be assumed to be complete from when the practice was computerized, for the precomputerization era,
historic coding of physical diseases and depression was
somewhat incomplete.
Examining a primary health care data set that includes milder illness forms might be viewed as a strength,
but it could also hinder interpretation. Standardized measures of illness severity are not recorded in the GPRD,
and the Read/OXMIS systems do not classify this in a structured manner. Attempts have been made to measure severity by proxy using medication dosage and duration
and by number and length of hospitalizations, but we
could not feasibly conduct such detailed examinations
across multiple illnesses. Validation of these proxies is
also unusual. Thus, medication-based severity stratification has been validated for COPD,56 for example, but
not for most of the other physical illnesses we examined. We, therefore, did not attempt to assess severity for

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

262

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

each specific illness, but we have shown a linear relationship of increasing risk by increasing multimorbidity among women. This factor could be viewed as a proxy
for severity, although not all severely ill patients will have
experienced more than 1 major physical disease.
One final limitation was the lack of sociodemographic
data held by the GPRD. Risk factors, such as marital status, social class, unemployment, and living alone, are not
recorded during routine patient consultations. We, therefore, could not adjust for these important confounders.
To our knowledge, this is the first large populationbased study to examine suicide risk in a broad range of
physical diseases with systematic and complete linkage
to national mortality data. Our findings show that health
care professionals working across all medical specialties
should be vigilant for signs of undetected psychological
symptoms,2,57 especially when treating women diagnosed as having cancer, coronary heart disease, COPD,
osteoporosis, or stroke. Previous or current clinical depression did not explain elevated risk seen in women with
cancer and coronary heart disease, and it only partially
explained higher risk seen in younger, physically ill
women and those with multiple diseases. For these groups
of women, our results challenge the notion that physical illness and associated disability lead to depression,
which leads to suicide. It could be that younger women
feel threatened by serious physical disease, particularly
those that are widely known to be the most common
causes of premature death. The specific mechanisms involved are likely to vary greatly according to the nature
of the physical illness being considered, and further research is therefore needed to understand the causal pathways. Training for a wider range of health care professionals in suicide risk management and reduction may
be beneficial. Addressing this need could greatly improve patients quality of life by tackling their psychological distress, as well as potentially reducing populationlevel suicide rates.
Submitted for Publication: July 15, 2011; accepted August 19, 2011.
Correspondence: Roger T. Webb, PhD, Centre for Suicide Prevention, Centre for Mental Health and Risk, University of Manchester, Jean McFarlane Building, Oxford
Road, Manchester, England M13 9PL (roger.webb
@manchester.ac.uk).
Financial Disclosure: None reported.
Funding/Support: The investigation was unfunded. The
study data set was obtained by the authors free of charge
via a Single Study Dataset Agreement between the GPRD
and the Medical Research Council under the aegis of the
Medical Research Council License Agreement.
Additional Contributions: Tarita Murray-Thomas, MSc,
GPRD, Medicines and Healthcare Products Regulatory
Agency, London, England, supplied the GPRD data set
in the form of a nested case-control study.
REFERENCES
1. Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. Depression, chronic
diseases, and decrements in health: results from the World Health Surveys. Lancet.
2007;370(9590):851-858.

2. Copsey Spring TR, Yanni LM, Levenson JL. A shot in the dark: failing to recognize the link between physical and mental illness. J Gen Intern Med. 2007;22
(5):677-680.
3. Kuo CJ, Chen VC, Lee WC, Chen WJ, Ferri CP, Stewart R, Lai TJ, Chen CC, Wang
TN, Ko YC. Asthma and suicide mortality in young people: a 12-year follow-up
study. Am J Psychiatry. 2010;167(9):1092-1099.
4. Louhivuori KA, Hakama M. Risk of suicide among cancer patients. Am J Epidemiol.
1979;109(1):59-65.
5. Allebeck P, Bolund C, Ringback G. Increased suicide rate in cancer patients: a
cohort study based on the Swedish Cancer-Environment Register. J Clin Epidemiol.
1989;42(7):611-616.
6. Larsen KK, Agerbo E, Christensen B, Sndergaard J, Vestergaard M. Myocardial
infarction and risk of suicide: a population-based case-control study. Circulation.
2010;122(23):2388-2393.
7. Tseng CH. Mortality and causes of death in a national sample of diabetic patients in Taiwan. Diabetes Care. 2004;27(7):1605-1609.
8. Nilsson L, Tomson T, Farahmand BY, Diwan V, Persson PG. Cause-specific mortality in epilepsy: a cohort study of more than 9,000 patients once hospitalized
for epilepsy. Epilepsia. 1997;38(10):1062-1068.
9. Christensen J, Vestergaard M, Mortensen PB, Sidenius P, Agerbo E. Epilepsy
and risk of suicide: a population-based case-control study. Lancet Neurol. 2007;
6(8):693-698.
10. Teasdale TW, Engberg AW. Suicide after a stroke: a population study. J Epidemiol Community Health. 2001;55(12):863-866.
11. Haste F, Charlton J, Jenkins R. Potential for suicide prevention in primary care?
an analysis of factors associated with suicide. Br J Gen Pract. 1998;48(436):
1759-1763.
12. Waern M, Rubenowitz E, Runeson B, Skoog I, Wilhelmson K, Allebeck P. Burden of illness and suicide in elderly people: case-control study. BMJ. 2002;
324(7350):1355-1357.
13. Juurlink DN, Herrmann N, Szalai JP, Kopp A, Redelmeier DA. Medical illness and
the risk of suicide in the elderly. Arch Intern Med. 2004;164(11):1179-1184.
14. Isometsa ET, Henriksson MM, Aro HM, Heikkinen ME, Kuoppasalmi KI, Lonnqvist
JK. Suicide in major depression. Am J Psychiatry. 1994;151(4):530-536.
15. National Institute for Clinical Excellence (NICE). Depression in Adults With a Chronic
Physical Health Problem: Treatment & Management. London, England: National
Collaborating Centre for Mental Health; 2009. National Health Service (NHS) NICE
clinical guideline 91.
16. Rihmer Z. Can better recognition and treatment of depression reduce suicide rates?
a brief review. Eur Psychiatry. 2001;16(7):406-409.
17. Hendin H. Suicide, assisted suicide, and medical illness. J Clin Psychiatry. 1999;
60(suppl 2):46-52.
18. Walley T, Mantgani A. The UK General Practice Research Database. Lancet. 1997;
350(9084):1097-1099.
19. Khan NF, Harrison SE, Rose PW. Validity of diagnostic coding within the General Practice Research Database: a systematic review. Br J Gen Pract. 2010;
60(572):e128-e136. doi:10.3399/bjgp10X483562.
20. Osborn DP, Levy G, Nazareth I, Petersen I, Islam A, King MB. Relative risk of
cardiovascular and cancer mortality in people with severe mental illness from
the United Kingdoms General Practice Research Database. Arch Gen Psychiatry.
2007;64(2):242-249.
21. Watson L, Baird J, Hosel V, Peveler R. The effect of concurrent pain on the management of patients with depression: an analysis of NHS healthcare resource utilisation using the GPRD database. Int J Clin Pract. 2009;63(5):698-706.
22. Byford S, Barrett B, Despiegel N, Wade A. Impact of treatment success on health
service use and cost in depression: longitudinal database analysis.
Pharmacoeconomics. 2011;29(2):157-170.
23. Grabbe L, Demi A, Camann MA, Potter L. The health status of elderly persons in
the last year of life: a comparison of deaths by suicide, injury, and natural causes.
Am J Public Health. 1997;87(3):434-437.
24. Conwell Y, Duberstein PR, Caine ED. Risk factors for suicide in later life. Biol
Psychiatry. 2002;52(3):193-204.
25. Quan H, Arboleda-Florez J, Fick GH, Stuart HL, Love EJ. Association between
physical illness and suicide among the elderly. Soc Psychiatry Psychiatr Epidemiol.
2002;37(4):190-197.
26. Viilo KM, Timonen MJ, Hakko HH, Sarkioja T, Meyer-Rochow VB, Rasanen PK.
Lifetime prevalences of physical diseases and mental disorders in young suicide victims. Psychosom Med. 2005;67(2):241-245.
27. Linsley KR, Schapira K, Kelly TP. Open verdict v. suicideimportance to research.
Br J Psychiatry. 2001;178:465-468.
28. Neeleman J, Wessely S. Changes in classification of suicide in England and Wales:
time trends and associations with coroners professional backgrounds. Psychol
Med. 1997;27(2):467-472.
29. World Health Organization. Manual of the International Classification of Diseases,
10th Revision (ICD-10). Geneva, Switzerland: World Health Organization; 1992.

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

263

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013

30. Hill C, Cook L. Narrative verdicts and their impact on mortality statistics in England and Wales. Health Stat Q. 2011;49(49):81-100.
31. Office for National Statistics. Suicide Rates in the United Kingdom, 2000-2009.
Statistical Bulletin. London, England: Office for National Statistics; 2011.
32. McCormick A, Fleming D, Charlton J. Morbidity Statistics From General Practice: Fourth National Study 1991-1992. London, England: HMSO; 1995.
33. Hammad TA, McAdams MA, Feight A, Iyasu S, Dal Pan GJ. Determining the predictive value of Read/OXMIS codes to identify incident acute myocardial infarction in the General Practice Research Database. Pharmacoepidemiol Drug Saf.
2008;17(12):1197-1201.
34. Khan NF, Perera R, Harper S, Rose PW. Adaptation and validation of the Charlson Index for Read/OXMIS coded databases. BMC Fam Pract. 2010;11:1. doi:
10.1186/1471-2296-11-1.
35. Doran T, Kontopantelis E, Valderas JM, Campbell S, Roland M, Salisbury C, Reeves
D. Effect of financial incentives on incentivized and non-incentivized clinical activities: longitudinal analysis of data from the UK Quality and Outcomes Framework.
BMJ. 2011;342:d3590. doi:10.1136/bmj.d3590.
36. Hennessy S, Bilker WB, Berlin JA, Strom BL. Factors influencing the optimal controlto-case ratio in matched case-control studies. Am J Epidemiol. 1999;149(2):
195-197.
37. Clayton D, Hills M. Statistical Models in Epidemiology. Oxford, England: Oxford
University Press; 1993.
38. Shah A, Bhat R, McKenzie S, Koen C. Elderly suicide rates: cross-national comparisons and association with sex and elderly age-bands. Med Sci Law. 2007;
47(3):244-252.
39. Kaplan MS, McFarland BH, Huguet N, Newsom JT. Physical illness, functional
limitations, and suicide risk: a population-based study. Am J Orthopsychiatry.
2007;77(1):56-60.
40. Mikkelsen RL, Middelboe T, Pisinger C, Stage KB. Anxiety and depression in patients with chronic obstructive pulmonary disease (COPD): a review. Nord J
Psychiatry. 2004;58(1):65-70.
41. Mezuk B, Eaton WW, Golden SH. Depression and osteoporosis: epidemiology
and potential mediating pathways. Osteoporos Int. 2008;19(1):1-12.
42. Law MR, Hackshaw AK. A meta-analysis of cigarette smoking, bone mineral density and risk of hip fracture: recognition of a major effect. BMJ. 1997;315(7112):
841-846.
43. Felson DT, Zhang Y, Hannan MT, Kannel WB, Kiel DP. Alcohol intake and bone
mineral density in elderly men and women: the Framingham Study. Am J Epidemiol.
1995;142(5):485-492.

44. Schneider B, Wetterling T, Georgi K, Bartusch B, Schnabel A, Blettner M. Smoking differently modifies suicide risk of affective disorders, substance use disorders, and social factors. J Affect Disord. 2009;112(1-3):165-173.
45. Flensborg-Madsen T, Knop J, Mortensen EL, Becker U, Sher L, Grnbaek M.
Alcohol use disorders increase the risk of completed suicideirrespective of other
psychiatric disorders: a longitudinal cohort study. Psychiatry Res. 2009;167
(1-2):123-130.
46. Bjorkenstam C, Edberg A, Ayoubi S, Rosen M. Are cancer patients at higher suicide risk than the general population? a nationwide register study in Sweden from
1965 to 1999. Scand J Public Health. 2005;33(3):208-214.
47. Miccinesi G, Crocetti E, Benvenuti A, Paci E. Suicide mortality is decreasing among
cancer patients in Central Italy. Eur J Cancer. 2004;40(7):1053-1057.
48. Crocetti E, Arniani S, Acciai S, Barchielli A, Buiatti E. High suicide mortality soon
after diagnosis among cancer patients in central Italy. Br J Cancer. 1998;77
(7):1194-1196.
49. Schairer C, Brown LM, Chen BE, Howard R, Lynch CF, Hall P, Storm H, Pukkala
E, Anderson A, Kaijser M, Andersson M, Joensuu H, Fossa SD, Ganz PA, Travis
LB. Suicide after breast cancer: an international population-based study of 723,810
women. J Natl Cancer Inst. 2006;98(19):1416-1419.
50. Twombly R. Decades after cancer, suicide risk remains high [editorial]. J Natl
Cancer Inst. 2006;98(19):1356-1358.
51. Maguire GP, Lee EG, Bevington DJ, Kuchemann CS, Crabtree RJ, Cornell CE.
Psychiatric problems in the first year after mastectomy. Br Med J. 1978;1(6118):
963-965.
52. Al-Ghazal SK, Fallowfield L, Blamey RW. Comparison of psychological aspects
and patient satisfaction following breast conserving surgery, simple mastectomy and breast reconstruction. Eur J Cancer. 2000;36(15):1938-1943.
53. MacDonald TM, McMahon AD, Reid IC, Fenton GW, McDevitt DG. Antidepressant drug use in primary care: a record linkage study in Tayside, Scotland. BMJ.
1996;313(7061):860-861.
54. Donoghue J, Hylan TR. Antidepressant use in clinical practice: efficacy v.
effectiveness. Br J Psychiatry Suppl. 2001;42:S9-S17.
55. Copeland KT, Checkoway H, McMichael AJ, Holbrook RH. Bias due to misclassification in the estimation of relative risk. Am J Epidemiol. 1977;105(5):488-495.
56. Soriano JB, Maier WC, Visick G, Pride NB. Validation of general practitionerdiagnosed COPD in the UK General Practice Research Database. Eur J Epidemiol.
2001;17(12):1075-1080.
57. Rifkin A. Depression in physically ill patients: dont dismiss it as understandable.
Postgrad Med. 1992;92(3):147-154.

ARCH GEN PSYCHIATRY/ VOL 69 (NO. 3), MAR 2012

264

WWW.ARCHGENPSYCHIATRY.COM

2012 American Medical Association. All rights reserved.

Downloaded From: http://archpsyc.jamanetwork.com/ on 07/05/2013