Identifying Acute Low Back Pain Episodes in Primary Care Practice From Clinical Notes: Observational Study.
clinical notes
electronic health records
low back pain
machine learning
natural language processing
Journal
JMIR medical informatics
ISSN: 2291-9694
Titre abrégé: JMIR Med Inform
Pays: Canada
ID NLM: 101645109
Informations de publication
Date de publication:
27 Feb 2020
27 Feb 2020
Historique:
received:
01
11
2019
accepted:
15
12
2019
entrez:
5
3
2020
pubmed:
5
3
2020
medline:
5
3
2020
Statut:
epublish
Résumé
Acute and chronic low back pain (LBP) are different conditions with different treatments. However, they are coded in electronic health records with the same International Classification of Diseases, 10th revision (ICD-10) code (M54.5) and can be differentiated only by retrospective chart reviews. This prevents an efficient definition of data-driven guidelines for billing and therapy recommendations, such as return-to-work options. The objective of this study was to evaluate the feasibility of automatically distinguishing acute LBP episodes by analyzing free-text clinical notes. We used a dataset of 17,409 clinical notes from different primary care practices; of these, 891 documents were manually annotated as acute LBP and 2973 were generally associated with LBP via the recorded ICD-10 code. We compared different supervised and unsupervised strategies for automated identification: keyword search, topic modeling, logistic regression with bag of n-grams and manual features, and deep learning (a convolutional neural network-based architecture [ConvNet]). We trained the supervised models using either manual annotations or ICD-10 codes as positive labels. ConvNet trained using manual annotations obtained the best results with an area under the receiver operating characteristic curve of 0.98 and an F score of 0.70. ConvNet's results were also robust to reduction of the number of manually annotated documents. In the absence of manual annotations, topic models performed better than methods trained using ICD-10 codes, which were unsatisfactory for identifying LBP acuity. This study uses clinical notes to delineate a potential path toward systematic learning of therapeutic strategies, billing guidelines, and management options for acute LBP at the point of care.
Sections du résumé
BACKGROUND
BACKGROUND
Acute and chronic low back pain (LBP) are different conditions with different treatments. However, they are coded in electronic health records with the same International Classification of Diseases, 10th revision (ICD-10) code (M54.5) and can be differentiated only by retrospective chart reviews. This prevents an efficient definition of data-driven guidelines for billing and therapy recommendations, such as return-to-work options.
OBJECTIVE
OBJECTIVE
The objective of this study was to evaluate the feasibility of automatically distinguishing acute LBP episodes by analyzing free-text clinical notes.
METHODS
METHODS
We used a dataset of 17,409 clinical notes from different primary care practices; of these, 891 documents were manually annotated as acute LBP and 2973 were generally associated with LBP via the recorded ICD-10 code. We compared different supervised and unsupervised strategies for automated identification: keyword search, topic modeling, logistic regression with bag of n-grams and manual features, and deep learning (a convolutional neural network-based architecture [ConvNet]). We trained the supervised models using either manual annotations or ICD-10 codes as positive labels.
RESULTS
RESULTS
ConvNet trained using manual annotations obtained the best results with an area under the receiver operating characteristic curve of 0.98 and an F score of 0.70. ConvNet's results were also robust to reduction of the number of manually annotated documents. In the absence of manual annotations, topic models performed better than methods trained using ICD-10 codes, which were unsatisfactory for identifying LBP acuity.
CONCLUSIONS
CONCLUSIONS
This study uses clinical notes to delineate a potential path toward systematic learning of therapeutic strategies, billing guidelines, and management options for acute LBP at the point of care.
Identifiants
pubmed: 32130159
pii: v8i2e16878
doi: 10.2196/16878
pmc: PMC7068466
doi:
Types de publication
Journal Article
Langues
eng
Pagination
e16878Subventions
Organisme : NIOSH CDC HHS
ID : T42 OH008422
Pays : United States
Informations de copyright
©Riccardo Miotto, Bethany L Percha, Benjamin S Glicksberg, Hao-Chih Lee, Lisanne Cruz, Joel T Dudley, Ismail Nabeel. Originally published in JMIR Medical Informatics (http://medinform.jmir.org), 27.02.2020.
Références
JMLR Workshop Conf Proc. 2016 Aug;56:301-318
pubmed: 28286600
J Biomed Inform. 2001 Oct;34(5):301-10
pubmed: 12123149
Nature. 2015 May 28;521(7553):436-44
pubmed: 26017442
Am Fam Physician. 2012 Feb 15;85(4):343-50
pubmed: 22335313
Arch Intern Med. 2009 Feb 9;169(3):251-8
pubmed: 19204216
J Am Med Inform Assoc. 2015 Apr;22(e1):e141-50
pubmed: 25769682
AMIA Jt Summits Transl Sci Proc. 2016 Jul 20;2016:41-50
pubmed: 27570647
JAMA. 2008 Feb 13;299(6):656-64
pubmed: 18270354
Spine (Phila Pa 1976). 2011 Oct 1;36(21 Suppl):S19-42
pubmed: 21952188
Nat Rev Genet. 2012 May 02;13(6):395-405
pubmed: 22549152
Spine (Phila Pa 1976). 2007 Jul 15;32(16):1754-60
pubmed: 17632396
Ann Intern Med. 2019 May 14;:
pubmed: 31083729
Spine (Phila Pa 1976). 2000 Jan 15;25(2):251-8; discussion 258-9
pubmed: 10685491
Spine (Phila Pa 1976). 2006 Dec 15;31(26):3052-60
pubmed: 17173003
Ann Intern Med. 2001 Aug 21;135(4):262-8
pubmed: 11511141
N Engl J Med. 1995 Feb 9;332(6):351-5
pubmed: 7823996
BMC Med Inform Decis Mak. 2016 Nov 3;16(1):138
pubmed: 27809908
Anesth Analg. 2017 Nov;125(5):1769-1778
pubmed: 29049121
BMC Musculoskelet Disord. 2017 May 12;18(1):188
pubmed: 28499364
PLoS One. 2014 Feb 13;9(2):e87555
pubmed: 24551060
Hum Mol Genet. 2018 May 1;27(R1):R56-R62
pubmed: 29659828
Sci Rep. 2016 May 17;6:26094
pubmed: 27185194
J Occup Rehabil. 2016 Sep;26(3):286-318
pubmed: 26667939
J Occup Environ Med. 2019 Jun;61(6):445-452
pubmed: 31167221
J Family Med Prim Care. 2018 Nov-Dec;7(6):1185-1192
pubmed: 30613495
Dialogues Clin Neurosci. 2012 Mar;14(1):77-89
pubmed: 22577307
MMWR Morb Mortal Wkly Rep. 2009 May 1;58(16):421-6
pubmed: 19407734
J Am Med Inform Assoc. 2018 Oct 1;25(10):1419-1428
pubmed: 29893864
Spine (Phila Pa 1976). 2004 Apr 15;29(8):884-90; discussion 891
pubmed: 15082989
Spine J. 2004 Jan-Feb;4(1):56-63
pubmed: 14749194
J Biomed Inform. 2018 Nov;87:12-20
pubmed: 30217670
JMIR Med Inform. 2019 Apr 27;7(2):e12239
pubmed: 31066697
CMAJ. 1997 Aug 15;157(4):408-16
pubmed: 9275952
N Engl J Med. 2004 Feb 12;350(7):722-6
pubmed: 14960750
Brief Bioinform. 2018 Nov 27;19(6):1236-1246
pubmed: 28481991
Pain Med. 2006 Mar-Apr;7(2):143-50
pubmed: 16634727
Pac Symp Biocomput. 2018;23:145-156
pubmed: 29218877
Clin Orthop Relat Res. 2006 Feb;443:139-46
pubmed: 16462438
IEEE J Biomed Health Inform. 2018 Jan;22(1):244-251
pubmed: 28475069
NPJ Digit Med. 2018 May 8;1:18
pubmed: 31304302
BMC Med Inform Decis Mak. 2017 Aug 22;17(1):126
pubmed: 28830409