Journal of Affective Disorders 122 (2010) 277–279

Contents lists available at ScienceDirect

Journal of Affective Disorders

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a d

Brief report

Internet monitoring of suicide risk in the population

Michael J. McCarthy ⁎
University of California, San Diego, Department of Psychiatry, 9500 Gilman Dr Mail Code 0603, La Jolla, CA 92093, United States of America

a r t i c l e i n f o a b s t r a c t

Article history: Background: Public health statistics are often released too late to affect reversible societal
Received 19 June 2009 factors affecting suicide. Increasingly, internet search volume is used in epidemiology, but this
Received in revised form 21 August 2009 method has not yet been applied to suicide.
Accepted 21 August 2009 Methods: Google internet search engine activity for suicide-related terms from the years 2004–
Available online 12 September 2009
2009 was measured and correlated to available suicide and intentional self-injury data from the
Centers of Disease Control (CDC).
Keywords: Results: Google search volumes correlated to CDC statistics for both suicide and self-injury, but
Suicide
in patterns that differed by age. Whereas internet search activity was negatively correlated to
Internet
the suicide rate in the general population, it was positively correlated to both intentional self-
Youth
Season injury and completed suicides among youth.
Risk Conclusions: Monitoring changes in search volumes on the internet may provide an early
Epidemiology indicator of suicide risk within the population. Furthermore, youth may utilize the internet in
ways that differ from the general population with respect to suicide.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction particular pathogen such as “Salmonella”, “diarrhea” and “food

poisoning” (Brownstein et al., 2009). The present study
Efforts to track suicide in populations are hampered by investigated the feasibility of monitoring the volume of
difficulties in data collection. At present, lag times for suicide-related internet searches as a tool to more rapidly
reporting are three years for data in the United States through identify trends that could influence suicide risk on a popula-
the Centers of Disease Control (CDC, 2009) and five years or tion-wide level.
more for the World Health Organization (WHO, 2009). In
contrast, social or economic conditions affecting the suicide
rate may shift rapidly. Critically, some of these protective or 2. Methods
predisposing factors may also be modifiable. However, using
current methods, by the time a trend is associated with The terms “suicide”, “teen suicide”, “depression”, “divorce”
elevated suicide risk, it may have already peaked and waned, and “unemployment” were entered into Google Trends (Google
thereby closing the window for intervention. Trends, 2009), an online log of internet search volumes. Results
In the field of infectious disease, large-scale monitoring were limited to the United States. Search terms were chosen for
rapidly identifies new patterns in the spread of illnesses. As a their widespread use in the lay public and hypothesized
result, when an outbreak occurs resources can be rapidly relevance to suicide risk. Less commonly used terms like
mobilized to the affected area to contain further spread. “anhedonia” fail to generate sufficient search volumes for
Epidemiologists now monitor the use of internet search engines analysis and were not used. Data for each term were
to track epidemics, inferring their movements through changes normalized to the search volume for the word during the first
in the search volumes for phrases that are closely linked with a week of January, 2004 (earliest data available). As only Google
Trends information from the United States was considered, it
could be directly compared to official CDC statistics by category
⁎ Tel.: +1 858 534 4040; fax: +1 858 534 7653. for the years 2004–2007. Data from the CDC for completed
E-mail address: mmccarthy@ucsd.edu. suicides (up to 2006) and intentional self-injury (up to 2007)

0165-0327/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jad.2009.08.015
278 M.J. McCarthy / Journal of Affective Disorders 122 (2010) 277–279

were not strongly correlated (r = −0.0145), nor were the

terms “suicide” and “divorce” (r = 0.064).
Weekly search volumes for “suicide” and “teen suicide”
were summed for each year (52 wk) period. These summed
volumes were then compared to the CDC data for the
corresponding years for both completed suicide and intentional
self-injury (Table 1). Among the general population there was
no correlation between search volume for the term “suicide”
and purposeful self-injury (r = −0.0436). In contrast, there
was a strong negative correlation between the internet search
term “suicide” and suicide deaths. The data for youth differed
markedly from those of the general population. Search volume
for “suicide” was positively correlated with both intentional
self-injury (r = 0.498) and completed suicides (r = 0.699).
Fig. 1. Google internet search volume for the term “suicide” averaged across
the year. Data reflect the averages of 2004 to 2009. Week 1 corresponds to
the first week of January for each year. Data are normalized to the volume of 4. Discussion
searches performed during the first week of 2004. Error bars +/− SEM.
Depression is a risk factor for suicide (reviewed by Rihmer,
2007). The time course, relationship to official suicide/self-injury
are publicly available at the CDC Web-based Injury Statistics statistics and positive correlation between the terms “suicide”
Query and Reporting System. The data were evaluated for both and “depression” suggest that searches for these phrases reflect
sexes combined, but separated into two age groups: the general interest in overlapping psychiatric constructs, rather than
population, defined as 1–85+ years old and youth, defined as homonyms that are non-psychiatric in nature (e.g. “depression”
15–25 years old. Averaged data were analyzed by one-way as economic collapse or “suicide” related to terrorism). Therefore,
ANOVA with Tukey–Kramer post-hoc analysis. Significance for internet searches for the terms “suicide” and “teen suicide” are
all statistical tests was defined as p < 0.05. likely relevant to psychiatric epidemiology.
There is an inverse correlation between internet searches
and suicides in the general population. These findings suggest
3. Results that the internet is used by many to seek help or otherwise
reduce suicide risk. In support of this interpretation, Recupero
The terms “depression”, “suicide” and “teen suicide” et al. (2008) suggested that internet searches more frequently
showed a cyclical pattern of internet search volume. When result in anti-suicide messages compared to suicide promotion,
years were averaged to evaluate recurring changes in volume, thereby making it probable that an internet search would
search activity for the term “suicide” peaked in the first quarter identify multiple options for support before leading one to
of each year, troughed during the middle half, and rose again for information encouraging suicide.
the final quarter. Search volume for the term “suicide” was Speaking against a non-specific cyclicity to internet search
significantly decreased in the third quarter of each year (weeks volumes, there was no seasonal variation in search volume for
27–39, early-July to late September) relative to the other three the terms “divorce” and “unemployment”. Instead, volume for
(p < 0.001 vs quarters 1 and 4, p < 0.01 vs quarter 2, see Fig. 1). “unemployment” grew rapidly and specifically from 2008–
Search volumes for the terms “divorce” and “unemployment” 2009, a period of severe economic difficulty, when the
did not differ on a regular cycle. unemployment rate increased sharply, thereby providing a
Search volumes for the terms “suicide” and “depression” positive control of sorts for the present study.
were highly correlated with each other (r = 0.848), as were Internet search volume for the term “suicide” is lowest
“teen suicide” and “depression” (r = 0.900). Search volume for during the summer, the time where most studies report a peak
the term “unemployment” greatly increased over the year in suicide deaths (Bjorksten et al., 2009; Reutfors et al., 2009;
2008–2009, however the terms “suicide” and “unemployment” Sebestyen et al., 2009). The interpretation above explains this

Table 1
Combined yearly Google search volumes for the term “suicide” compared to yearly CDC statistics for completed suicides and intentional self-injurious behavior
(SIB).

Year 2004 2005 2006 2007 Correlation to Google

Total population
Google volume “suicide” 53.74 55.48 48.07 45.04 N/A
CDC-completed suicides 32,428 32,629 33,292 Not available − 0.900178
CDC-intentional SIB 425,603 372,722 395,146 395,175 − 0.043599

Youth 15–25
Google volume “teen suicide” 47.92 37.99 27.62 22.38 N/A
CDC-completed suicide 4,820 4,734 4,757 Not available 0.698595
CDC-intentional SIB 148,041 134,134 138,862 139,821 0.498077

Data are shown for general population (above) and for youth (below).
 M.J. McCarthy / Journal of Affective Disorders 122 (2010) 277–279 279

discrepancy between internet search activity and the season- limited, these findings nonetheless provide testable hypotheses
ality of suicide. If internet search activity is related to suicide in a that will be supported, refuted or refined in coming years as
reciprocal manner, the negative correlation of internet activity more epidemiological data accumulate. Like systems utilized to
to suicide for the general population fit the existing model, and monitor the spread of infectious pathogens, internet search
suggests a corollary that people are less likely to seek help on volumes may be a useful indicator for tracking suicide trends.
the internet during the summer months. Given the risky climate of economic instability, large numbers
of returning war veterans and widespread disconnection from
4.1. Age differences in suicide-related internet searches medical care, efforts to track large-scale mental health crises in
real time could lead to focused interventions in sensitive
In contrast to the general population, internet searches are populations.
positively correlated with self-injury and suicide among
youth, suggesting this group uses the internet to facilitate Role of funding source
self-injury. The literature on youth suicide and the internet is The author receives support from the NIMH-funded UCSD research
resident training track (R25MH074508).
small, and originates primarily from outside of the United
States. Nonetheless, internet use has been associated with
Conflict of interest
greater suicide risk, suicide pacts and suicide epidemics
The author has no financial conflict of interests.
among youth (Katsumata et al., 2008; Hagihara et al., 2007;
Nicolis, 2007). It has also been shown that the quality of
internet suicide support resources varies considerably, Acknowledgments
possibly requiring considerable effort and motivation to
navigate (Recupero and Rainey, 2006). These traits may be Thanks to Sidney Zisook M.D. and Barry Lebowitz M.D. who
underdeveloped in some youth, causing them to choose less added helpful comments to the manuscript while in preparation.
selectively among the available options or to be drawn to the
most salient choices. References

Bjorksten, K.S., Kripke, D.F., Bjerregaard, P., 2009. Accentuation of suicides but
4.2. Demographic characteristics of suicide and internet use not homicides with rising latitudes of Greenland in the sunny months. BMC
Psychiatry 9, 20.
Google search volumes reflect a large proportion of all Brownstein, J.S., Freifeld, C.C., Madoff, L.C., 2009. Digital disease detection —
harnessing the web for public health surveillance. New Engl. J. Med. 360
available internet search data on the chosen terms, thereby
(21), 2153–2157.
capturing a vast amount of information. While the dataset is Centers for Disease Control Web-based Injury Statistics Query and Reporting
large, it is also missing key demographic information. Age System (WISQARSTM), http://www.cdc.gov/injury/wisqars/index.html
Accessed June, 2009.
bias in particular could jeopardize the validity of comparing
Google Trends, 2009 (http://www.google.com/trends). Accessed May, 2009.
internet search statistics to suicide rates. The CDC suicide Hagihara, A., Tarumi, K., Abe, T., 2007. Media suicide-reports, internet use
statistics used in this study were re-evaluated against the age and the occurrence of suicides between 1987 and 2005 in Japan. BMC
ranges used by Jones and Fox (2009) to measure internet use. Public Health 7 (321).
Jones, S., Fox, S., 2009. Generations Online in 2009. Pew internet and American
In accord with previous reports (Vastag, 2001), the youngest life project. http://www.pewinternet.org/Reports/2009/Generations-
group (18–32 years old) is over represented on the internet Online-in-2009.aspx2009Accessed June, 2009.
relative to their suicide rate, whereas the oldest age group Katsumata, Y., Matsumoto, T., Kitani, M., Takeshima, T., 2008. Electronic
media use and suicidal ideation in Japanese adolescents. Psychiatry Clin.
(age 74 +) is over represented among suicide victims relative Neurosci. 62 (6), 744–746.
to internet use. However, among those aged 33–73 years old, Nicolis, H., 2007. Nonlinear dynamics and probabilistic behavior of adolescent
the proportional contributions to both total internet use and suicidal outbreaks. Nonlinear Dyn. Psychol. Life Sci. 11 (4), 451–472.
Recupero, P.R., Rainey, S.E., 2006. Characteristics of e-therapy web sites. J. Clin.
total suicides were similar across the entire range. This Psychiatry 67 (9), 1435–1440.
intermediate age cohort represents ~ 65% of the totals of both Recupero, P.R., Harms, S.E., Noble, J.M., 2008. Googling suicide: surfing for
internet users and suicide deaths. As such, these findings suicide information on the internet. J. Clin. Psychiatry 69 (6), 878–888.
Reutfors, J., et al., 2009. Seasonality of suicide in Sweden: Relationship with
suggest that while appropriate in general, caution should be psychiatric disorder. J. Affect. Disord. 119, 59–65.
used in interpreting suicide–internet correlations as they may Rihmer, Z., 2007. Suicide risk in mood disorders. Curr. Opin. Psychiatry 20,
not generalize to specific populations. 17–22.
Sebestyen, B., Rihmer, Z., Lajos, B., Szokontor, N., Gonda, X., Gyarmati, B., Bodecs, T.,
Sandor, J. 2009. Gender Differences in Antidepressant Use-Related Seasonality
5. Conclusion Change in Suicide Mortality in Hungary, 1998–2006. (in review).
Vastag, B., 2001. Suicide prevention plan calls for physicians' help. JAMA 285,
A connection between suicide and internet search activity 2701–2703.
World Health Organization (WHO), 2009. Suicide statistics by country for the
has been supported, suggesting internet searches for suicide- year 2003. http://www.who.int/mental_health/prevention/suicide/suici-
related terms may predict actual self-injury and death. While derates/en/2009Accessed May.