The relationship between text message sentiment and self-reported depression.
Depression
Digital phenotyping
Language sentiment analysis
Machine learning
Personal sensing
Journal
Journal of affective disorders
ISSN: 1573-2517
Titre abrégé: J Affect Disord
Pays: Netherlands
ID NLM: 7906073
Informations de publication
Date de publication:
01 04 2022
01 04 2022
Historique:
received:
22
06
2021
revised:
15
11
2021
accepted:
18
12
2021
pubmed:
30
12
2021
medline:
24
2
2022
entrez:
29
12
2021
Statut:
ppublish
Résumé
Personal sensing has shown promise for detecting behavioral correlates of depression, but there is little work examining personal sensing of cognitive and affective states. Digital language, particularly through personal text messages, is one source that can measure these markers. We correlated privacy-preserving sentiment analysis of text messages with self-reported depression symptom severity. We enrolled 219 U.S. adults in a 16 week longitudinal observational study. Participants installed a personal sensing app on their phones, which administered self-report PHQ-8 assessments of their depression severity, collected phone sensor data, and computed anonymized language sentiment scores from their text messages. We also trained machine learning models for predicting end-of-study self-reported depression status using on blocks of phone sensor and text features. In correlation analyses, we find that degrees of depression, emotional, and personal pronoun language categories correlate most strongly with self-reported depression, validating prior literature. Our classification models which predict binary depression status achieve a leave-one-out AUC of 0.72 when only considering text features and 0.76 when combining text with other networked smartphone sensors. Participants were recruited from a panel that over-represented women, caucasians, and individuals with self-reported depression at baseline. As language use differs across demographic factors, generalizability beyond this population may be limited. The study period also coincided with the initial COVID-19 outbreak in the United States, which may have affected smartphone sensor data quality. Effective depression prediction through text message sentiment, especially when combined with other personal sensors, could enable comprehensive mental health monitoring and intervention.
Sections du résumé
BACKGROUND
Personal sensing has shown promise for detecting behavioral correlates of depression, but there is little work examining personal sensing of cognitive and affective states. Digital language, particularly through personal text messages, is one source that can measure these markers.
METHODS
We correlated privacy-preserving sentiment analysis of text messages with self-reported depression symptom severity. We enrolled 219 U.S. adults in a 16 week longitudinal observational study. Participants installed a personal sensing app on their phones, which administered self-report PHQ-8 assessments of their depression severity, collected phone sensor data, and computed anonymized language sentiment scores from their text messages. We also trained machine learning models for predicting end-of-study self-reported depression status using on blocks of phone sensor and text features.
RESULTS
In correlation analyses, we find that degrees of depression, emotional, and personal pronoun language categories correlate most strongly with self-reported depression, validating prior literature. Our classification models which predict binary depression status achieve a leave-one-out AUC of 0.72 when only considering text features and 0.76 when combining text with other networked smartphone sensors.
LIMITATIONS
Participants were recruited from a panel that over-represented women, caucasians, and individuals with self-reported depression at baseline. As language use differs across demographic factors, generalizability beyond this population may be limited. The study period also coincided with the initial COVID-19 outbreak in the United States, which may have affected smartphone sensor data quality.
CONCLUSIONS
Effective depression prediction through text message sentiment, especially when combined with other personal sensors, could enable comprehensive mental health monitoring and intervention.
Identifiants
pubmed: 34963643
pii: S0165-0327(21)01359-8
doi: 10.1016/j.jad.2021.12.048
pmc: PMC8912980
mid: NIHMS1775906
pii:
doi:
Types de publication
Journal Article
Observational Study
Research Support, N.I.H., Extramural
Langues
eng
Sous-ensembles de citation
IM
Pagination
7-14Subventions
Organisme : NIMH NIH HHS
ID : R01 MH111610
Pays : United States
Organisme : NIMH NIH HHS
ID : T32 MH115882
Pays : United States
Organisme : NCATS NIH HHS
ID : UL1 TR001422
Pays : United States
Informations de copyright
Copyright © 2021. Published by Elsevier B.V.
Références
J Biomed Inform. 2009 Apr;42(2):377-81
pubmed: 18929686
J Pers. 2010 Feb;78(1):313-38
pubmed: 20433621
Nat Hum Behav. 2021 Apr;5(4):458-466
pubmed: 33574604
Lancet Psychiatry. 2020 Aug;7(8):692-702
pubmed: 32711710
NPJ Digit Med. 2019 Sep 6;2:88
pubmed: 31508498
Ann Behav Med. 2018 May 18;52(6):446-462
pubmed: 27663578
J Clin Psychiatry. 2008 Dec;69(12):1916-9
pubmed: 19192467
Psychol Bull. 2017 Aug;143(8):783-822
pubmed: 28447828
J Clin Psychiatry. 2015 Feb;76(2):155-62
pubmed: 25742202
J Med Internet Res. 2020 Oct 29;22(10):e20631
pubmed: 33118946
J Affect Disord. 2015 Feb 1;172:96-102
pubmed: 25451401
Neuropsychopharmacology. 2018 Jul;43(8):1637-1638
pubmed: 29703995
PeerJ. 2016 Sep 29;4:e2537
pubmed: 28344895
J Med Internet Res. 2018 Jul 20;20(7):e241
pubmed: 30030209
Neuropsychopharmacology. 2021 Jan;46(1):45-54
pubmed: 32679583
Nat Rev Dis Primers. 2016 Sep 15;2:16065
pubmed: 27629598
Bull World Health Organ. 2020 Apr 1;98(4):270-276
pubmed: 32284651
Proc ACM Interact Mob Wearable Ubiquitous Technol. 2020 Mar;4(1):
pubmed: 34527853
PLoS One. 2013 Sep 25;8(9):e73791
pubmed: 24086296
Annu Rev Psychol. 2020 Jan 4;71:471-497
pubmed: 31518525
Annu Rev Clin Psychol. 2017 May 8;13:23-47
pubmed: 28375728
Nat Commun. 2019 Jul 23;10(1):3069
pubmed: 31337762
Curr Psychiatry Rep. 2015 Aug;17(8):602
pubmed: 26073363
JAMA. 2017 Oct 3;318(13):1215-1216
pubmed: 28973224
Eur Arch Psychiatry Clin Neurosci. 2005 Aug;255(4):215-22
pubmed: 16133740
PLoS One. 2019 Jun 17;14(6):e0215476
pubmed: 31206534
JAMA Netw Open. 2019 Apr 5;2(4):e192542
pubmed: 31002321
J Affect Disord. 2017 Jan 15;208:191-197
pubmed: 27792962
Br J Psychiatry Suppl. 1996 Jun;(30):38-43
pubmed: 8864147
World Psychiatry. 2018 Oct;17(3):276-277
pubmed: 30192103
J Biomed Inform. 2019 Jul;95:103208
pubmed: 31078660
NPJ Digit Med. 2020 Mar 25;3:45
pubmed: 32219186
JMIR Mhealth Uhealth. 2019 Apr 05;7(4):e12578
pubmed: 30950799
J Affect Disord. 2009 Apr;114(1-3):163-73
pubmed: 18752852
J Pers Soc Psychol. 2019 May;116(5):817-834
pubmed: 29504797
Proc Natl Acad Sci U S A. 2018 Oct 30;115(44):11203-11208
pubmed: 30322910
Neuropsychopharmacology. 2016 Jun;41(7):1691-6
pubmed: 26818126
Biometrics. 1988 Sep;44(3):837-45
pubmed: 3203132