CoChat
Sign Up
Cited by 643Mental Health via WritingEmotion and Mood RecognitionSentiment Analysis and Opinion Mining

The DOI matches its Crossref record; 45 of the 46 references it cites were found in a registry or an index.

Facebook language predicts depression in medical records

Open access
Eichstaedt, Johannes C., Smith, Robert J., Merchant, Raina M., Ungar, Lyle H., Crutchley, Patrick, Preoţiuc-Pietro, Daniel, Asch, David A., Schwartz, H. Andrew·2018·Proceedings of the National Academy of Sciences
DOI 10.1073/pnas.1802331115PMC PMC6217418
Open this paper in CoChat — free account

Abstract

Significance: Depression is disabling and treatable, but underdiagnosed. In this study, we show that the content shared by consenting users on Facebook can predict a future occurrence of depression in their medical records. Language predictive of depression includes references to typical symptoms, including sadness, loneliness, hostility, rumination, and increased self-reference. This study suggests that an analysis of social media data could be used to screen consenting individuals for depression. Further, social media content may point clinicians to specific symptoms of depression.

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 · 10.1073/pnas.1802331115

Cites

45 of 45 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1. Prediction performances of future diagnosis of depression in the EMR based on demographics and Facebook posting activity, reported as cross-validated out-of-sample AUCs.
Fig. 1. Prediction performances of future diagnosis of depression in the EMR based on demographics and Facebook posting activity, reported as cross-validated out-of-sample AUCs.CC BY-NC-ND · PubMed Central
Fig. 2. ROC curve for a Facebook activity-based prediction model (all predictors combined; blue), and points as combinations of true and false positive rates reported by Noyes et al. (17) for different combinations of depression surveys (a and b, 9-item Mini-International Neuropsychiatric Interview–Major Depressive Episode Module; c and d, 15-item Geriatric Depression Scale with a cutoff >6) and time windows in Medicare claims data (a and c, within 6 mo before and after survey; b and d, within 12 mo).
Fig. 2. ROC curve for a Facebook activity-based prediction model (all predictors combined; blue), and points as combinations of true and false positive rates reported by Noyes et al. (17) for different combinations of depression surveys (a and b, 9-item Mini-International Neuropsychiatric Interview–Major Depressive Episode Module; c and d, 15-item Geriatric Depression Scale with a cutoff >6) and time windows in Medicare claims data (a and c, within 6 mo before and after survey; b and d, within 12 mo).CC BY-NC-ND · PubMed Central
Fig. 3. AUC prediction accuracies of future depression status as a function of time before the documentation of depression in the medical record. Shown in blue are the 6-mo time windows of Facebook data used for the predictions; the blue dots indicate the AUCs obtained for these windows. Error bars indicate SEs (based on the 10 cross-validation folds). Logarithmic trendline is shown to guide the eye.
Fig. 3. AUC prediction accuracies of future depression status as a function of time before the documentation of depression in the medical record. Shown in blue are the 6-mo time windows of Facebook data used for the predictions; the blue dots indicate the AUCs obtained for these windows. Error bars indicate SEs (based on the 10 cross-validation folds). Logarithmic trendline is shown to guide the eye.CC BY-NC-ND · PubMed Central
Fig. 4. Ten language topics most positively associated with a future depression diagnosis controlling for demographics (*P < 0.05, **P < 0.01, and ***P < 0.001; BHP < 0.05 after Benjamini–Hochberg correction for multiple comparisons). Font size reflects relative prevalence of words within topics. Color shading is to aid readability and carries no meaning.
Fig. 4. Ten language topics most positively associated with a future depression diagnosis controlling for demographics (*P < 0.05, **P < 0.01, and ***P < 0.001; BHP < 0.05 after Benjamini–Hochberg correction for multiple comparisons). Font size reflects relative prevalence of words within topics. Color shading is to aid readability and carries no meaning.CC BY-NC-ND · 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

Cited by

Sorted by citations

See all 643

Related

See all

Metadata

Authors
Eichstaedt, J. C., Smith, R. J., Merchant, R. M., Ungar, L. H., Crutchley, P., Preoţiuc-Pietro, D., Asch, D. A., & Schwartz, H. A.
Affiliations
  • University of Pennsylvania (US)
  • University of Pennsylvania Health System (US)
  • Positive Psychology Center, University of Pennsylvania, Philadelphia, PA 19104;
  • Philadelphia VA Medical Center (US)
  • Stony Brook University (US)
Corresponding author
Eichstaedt, Johannes C.
ORCID
Venue
Proceedings of the National Academy of Sciences
Publisher
National Academy of Sciences
Published
2018-10-15
Volume
115
Issue
44
Pages
11203-11208
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
DOI
10.1073/pnas.1802331115
PMC
PMC6217418
OpenAlex
W2897583329
Cited by
643
References
46
Citations by year
2026: 64 · 2025: 68 · 2024: 77 · 2023: 87 · 2022: 107 · 2021: 111 · 2020: 84 · 2019: 41 · 2018: 3
FWCI
80.90
Type
journal article
Language
en
Open access
Licence
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
Funders
  • Robert Wood Johnson Foundation (Pioneer Award)
  • Templeton Religion Trust (TRT0048)
  • HHS | NIH | National Heart, Lung, and Blood Institute (HL122457)
Topics
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
  • Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
Keywords
depression, Facebook posts, social media language, depression prediction, electronic health records, depression screening, emergency department patients, rumination, loneliness, sadness, self-absorption, hostility
MeSH terms
Adult, Depression, Depressive Disorder, Female, Humans, Language, Male, Surveys and Questionnaires, Electronic Health Records, Social Media

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