CoChat
Sign Up
Cited by 506Mental Health via WritingSuicide and Self-Harm StudiesMachine Learning in Healthcare

The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.

Natural Language Processing of Social Media as Screening for Suicide Risk

Open access
Coppersmith, Glen, Leary, Ryan, Crutchley, Patrick, Fine, Alex·2018·Biomedical Informatics Insights
DOI 10.1177/1178222618792860PMC PMC6111391
Open this paper in CoChat — free account

Abstract

Suicide is among the 10 most common causes of death, as assessed by the World Health Organization. For every death by suicide, an estimated 138 people’s lives are meaningfully affected, and almost any other statistic around suicide deaths is equally alarming. The pervasiveness of social media—and the near-ubiquity of mobile devices used to access social media networks—offers new types of data for understanding the behavior of those who (attempt to) take their own lives and suggests new possibilities for preventive intervention. We demonstrate the feasibility of using social media data to detect those at risk for suicide. Specifically, we use natural language processing and machine learning (specifically deep learning) techniques to detect quantifiable signals around suicide attempts, and describe designs for an automated system for estimating suicide risk, usable by those without specialized mental health training (eg, a primary care doctor). We also discuss the ethical use of such technology and examine privacy implications. Currently, this technology is only used for intervention for individuals who have “opted in” for the analysis and intervention, but the technology enables scalable screening for suicide risk, potentially identifying many people who are at risk preventively and prior to any engagement with a health care system. This raises a significant cultural question about the trade-off between privacy and prevention—we have potentially life-saving technology that is currently reaching only a fraction of the possible people at risk because of respect for their privacy. Is the current trade-off between privacy and prevention the right one?

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1177/1178222618792860

Cites

35 of 35 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. One example: person’s interaction with the health care system (in red hashes) and with a social media platform (in blue hashes) over a period of 4 years (x-axis, left is earlier in time, right is later in time). Social media provides information in the “clinical whitespace”15 between interactions with the health care system. Summarized from health record data presented in Padrez et al.16
Figure 1. One example: person’s interaction with the health care system (in red hashes) and with a social media platform (in blue hashes) over a period of 4 years (x-axis, left is earlier in time, right is later in time). Social media provides information in the “clinical whitespace”15 between interactions with the health care system. Summarized from health record data presented in Padrez et al.16CC BY-NC · PubMed Central
Figure 2. Fictitious example: posts of genuine statements of a suicide attempt (top), genuine statements of a suicide attempt indicating a time for the attempt (middle), and disingenuous statements of a suicide attempt (bottom). Only data from the middle are include in the analysis here (because we can ascertain a date and have a chance to obtain data prior to that date). In line with Benton et al,41 we do not reproduce text directly from social media data and instead paraphrase to protect the user’s identity.
Figure 2. Fictitious example: posts of genuine statements of a suicide attempt (top), genuine statements of a suicide attempt indicating a time for the attempt (middle), and disingenuous statements of a suicide attempt (bottom). Only data from the middle are include in the analysis here (because we can ascertain a date and have a chance to obtain data prior to that date). In line with Benton et al,41 we do not reproduce text directly from social media data and instead paraphrase to protect the user’s identity.CC BY-NC · PubMed Central
Figure 3. Histogram of age of users, separated by gender. Females in red, males in blue, non-binary in gray. The mean age of each gender is indicated by a vertical line of the same color.
Figure 3. Histogram of age of users, separated by gender. Females in red, males in blue, non-binary in gray. The mean age of each gender is indicated by a vertical line of the same color.CC BY-NC · PubMed Central
Figure 4. Fictitious rephrasing of posts scored highly by the suicide risk model. The bottom panel indicates what the attention mechanism within the neural architecture is highlighting as an important signal for a person at risk. Larger words (eg, “cutting” and “again”) have more power in the current prediction than smaller words (eg, “be” or “been”).
Figure 4. Fictitious rephrasing of posts scored highly by the suicide risk model. The bottom panel indicates what the attention mechanism within the neural architecture is highlighting as an important signal for a person at risk. Larger words (eg, “cutting” and “again”) have more power in the current prediction than smaller words (eg, “be” or “been”).CC BY-NC · PubMed Central
Figure 5. Scores for each user’s 200 posts preceding the suicide attempt (and matched controls) are plotted with a moving window average; y-axis magnitude corresponds to risk for suicide (likelihood of messages belonging to a user from the positive class). Lines in blue are users who will go on to attempt suicide, whereas users in green are their matched controls.
Figure 5. Scores for each user’s 200 posts preceding the suicide attempt (and matched controls) are plotted with a moving window average; y-axis magnitude corresponds to risk for suicide (likelihood of messages belonging to a user from the positive class). Lines in blue are users who will go on to attempt suicide, whereas users in green are their matched controls.CC BY-NC · PubMed Central
Figure 6. ROC curves for models separating users prior to a suicide attempt from their matched controls. The green line only uses data for the month prior to the suicide attempt to make the classification (30 to 0 days prior), the blue line uses data from 2 months prior (60 to 0 days prior), and so on. The black line indicates performance using all of the data available for that user prior to their attempt. ROC indicates receiver operating characteristic.
Figure 6. ROC curves for models separating users prior to a suicide attempt from their matched controls. The green line only uses data for the month prior to the suicide attempt to make the classification (30 to 0 days prior), the blue line uses data from 2 months prior (60 to 0 days prior), and so on. The black line indicates performance using all of the data available for that user prior to their attempt. ROC indicates receiver operating characteristic.CC BY-NC · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 32 references

Research works printed but not linked: 31

  1. World Health Organization, 2013, Mental health action plan 2013-2020
  2. Curtin SC, 2016, NCHS Data Brief, 241, 1
  3. 10.1037/0022-006X.47.2.343 · DOI · Paper page
  4. Coppersmith G, Paper presented at: 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), 393
  5. Chung C, Frontiers of Social Psychology: Social Communication, 2007, 343
  6. De Choudhury M, Paper presented at: Proceedings of the 7th International AAAI Conference on Weblogs and Social Media (ICWSM)
  7. Coppersmith G, Paper presented at: Proceedings of the 8th International AAAI Conference on Weblogs and Social Media (ICWSM
  8. Cohan A, Paper presented at: Proceedings of the 3rd Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality, 16
  9. Coppersmith G, Paper presented at: Proceedings of the Shared Task for the NAACL Workshop on Computational Linguistics and Clinical Psychology
  10. Resnik P, Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  11. Preotiuc-Pietro D, Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  12. Pedersen T., Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  13. Mitchell M, Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  14. Coppersmith G, Paper presented at: Proceedings of the ACL Workshop on Computational Linguistics and Clinical Psychology
  15. Coppersmith G, Paper presented at: Joint Statistics Meetings Proceedings, Statistical Computing Section (JSM)
  16. Kumar M, Paper presented at: Proceedings of the 26th ACM conference on Hypertext & Social Media
  17. Wood A, Paper presented at: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
  18. Kiciman E, Paper presented at: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
  19. Coppersmith G, Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  20. Dinakar K, Paper presented at: Proceedings of the 20th International Conference on Intelligent User Interfaces, 417
  21. Coppersmith G, Paper presented at: Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality (North American Chapter of the Association for Computational Linguistics)
  22. Loveys K, Paper presented at: Proceedings of the Fourth Workshop on Computational Linguistics and Clinical Psychology—From Linguistic Signal to Clinical Reality, 85
  23. Benton A, Paper presented at: Proceedings of the First Workshop on Ethics in Natural Language Processing, 94
  24. Sap M, Paper presented at: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), 1146
  25. Yang Z, Paper presented at: HLT-NAACL
  26. Pennington J, Paper presented at: Empirical Methods in Natural Language Processing (EMNLP), 1532
  27. Xu K, Paper presented at: International Conference on Machine Learning, 2048
  28. Kann L, 2014, MMWR Suppl, 63, 1
  29. De Choudhury M, Paper presented at: Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing (CSCW ’17), 353
  30. Facebook. Getting our community help in real time (press release), 2017. https://newsroom.fb.com/news/2017/11/getting-our-community-help-in-real-time/.
  31. Parker J, Paper presented at: Proceedings of the 2013 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM ’13), 556

Cited by

Sorted by citations

Items 501–505 of 505

Related

See all

Metadata

Authors
Coppersmith, G., Leary, R., Crutchley, P., & Fine, A.
Affiliations
Qntfy, Boston, MA, USA
Corresponding author
Coppersmith, Glen
ORCID
Venue
Biomedical Informatics Insights
Publisher
SAGE Publications
Published
2018-01
Volume
10
Pages
1178222618792860-1178222618792860
ISSN
1178-2226 (ISSN-L)
DOI
10.1177/1178222618792860
PMC
PMC6111391
OpenAlex
W2889391310
Cited by
506
References
32
Citations by year
2026: 45 · 2025: 72 · 2024: 56 · 2023: 61 · 2022: 73 · 2021: 116 · 2020: 62 · 2019: 18
FWCI
7.44
Type
journal article
Language
en
Open access
Licence
Locations
  • Biomedical Informatics Insights (publishedVersion, cc-by-nc)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Topics
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
  • Social Sciences › Psychology › Clinical Psychology › Suicide and Self-Harm Studies
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
suicide risk screening, suicide attempts, social media, natural language processing, primary care, privacy, ethics, mental health, preventive interventions

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review