Development of machine learning models to predict gestational diabetes risk in the first half of pregnancy.
Data augmentation
GDM risk prediction
Gestational diabetes mellitus (GDM)
Machine learning models
Widely available variables
Journal
BMC pregnancy and childbirth
ISSN: 1471-2393
Titre abrégé: BMC Pregnancy Childbirth
Pays: England
ID NLM: 100967799
Informations de publication
Date de publication:
23 Jun 2023
23 Jun 2023
Historique:
received:
18
04
2023
accepted:
08
06
2023
medline:
26
6
2023
pubmed:
24
6
2023
entrez:
23
6
2023
Statut:
epublish
Résumé
Early prediction of Gestational Diabetes Mellitus (GDM) risk is of particular importance as it may enable more efficacious interventions and reduce cumulative injury to mother and fetus. The aim of this study is to develop machine learning (ML) models, for the early prediction of GDM using widely available variables, facilitating early intervention, and making possible to apply the prediction models in places where there is no access to more complex examinations. The dataset used in this study includes registries from 1,611 pregnancies. Twelve different ML models and their hyperparameters were optimized to achieve early and high prediction performance of GDM. A data augmentation method was used in training to improve prediction results. Three methods were used to select the most relevant variables for GDM prediction. After training, the models ranked with the highest Area under the Receiver Operating Characteristic Curve (AUCROC), were assessed on the validation set. Models with the best results were assessed in the test set as a measure of generalization performance. Our method allows identifying many possible models for various levels of sensitivity and specificity. Four models achieved a high sensitivity of 0.82, a specificity in the range 0.72-0.74, accuracy between 0.73-0.75, and AUCROC of 0.81. These models required between 7 and 12 input variables. Another possible choice could be a model with sensitivity of 0.89 that requires just 5 variables reaching an accuracy of 0.65, a specificity of 0.62, and AUCROC of 0.82. The principal findings of our study are: Early prediction of GDM within early stages of pregnancy using regular examinations/exams; the development and optimization of twelve different ML models and their hyperparameters to achieve the highest prediction performance; a novel data augmentation method is proposed to allow reaching excellent GDM prediction results with various models.
Sections du résumé
BACKGROUND
BACKGROUND
Early prediction of Gestational Diabetes Mellitus (GDM) risk is of particular importance as it may enable more efficacious interventions and reduce cumulative injury to mother and fetus. The aim of this study is to develop machine learning (ML) models, for the early prediction of GDM using widely available variables, facilitating early intervention, and making possible to apply the prediction models in places where there is no access to more complex examinations.
METHODS
METHODS
The dataset used in this study includes registries from 1,611 pregnancies. Twelve different ML models and their hyperparameters were optimized to achieve early and high prediction performance of GDM. A data augmentation method was used in training to improve prediction results. Three methods were used to select the most relevant variables for GDM prediction. After training, the models ranked with the highest Area under the Receiver Operating Characteristic Curve (AUCROC), were assessed on the validation set. Models with the best results were assessed in the test set as a measure of generalization performance.
RESULTS
RESULTS
Our method allows identifying many possible models for various levels of sensitivity and specificity. Four models achieved a high sensitivity of 0.82, a specificity in the range 0.72-0.74, accuracy between 0.73-0.75, and AUCROC of 0.81. These models required between 7 and 12 input variables. Another possible choice could be a model with sensitivity of 0.89 that requires just 5 variables reaching an accuracy of 0.65, a specificity of 0.62, and AUCROC of 0.82.
CONCLUSIONS
CONCLUSIONS
The principal findings of our study are: Early prediction of GDM within early stages of pregnancy using regular examinations/exams; the development and optimization of twelve different ML models and their hyperparameters to achieve the highest prediction performance; a novel data augmentation method is proposed to allow reaching excellent GDM prediction results with various models.
Identifiants
pubmed: 37353749
doi: 10.1186/s12884-023-05766-4
pii: 10.1186/s12884-023-05766-4
pmc: PMC10288662
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
469Subventions
Organisme : Agencia Nacional de Investigación y Desarrollo
ID : Basal funding for Scientific and Technological Center of Excellence, IMPACT, #FB210024, FONDECYT 1231675
Organisme : Agencia Nacional de Investigación y Desarrollo
ID : Basal funding for Scientific and Technological Center of Excellence, IMPACT, #FB210024, FONDECYT 1231675
Organisme : Agencia Nacional de Investigación y Desarrollo
ID : Basal funding for Scientific and Technological Center of Excellence, IMPACT, #FB210024, FONDECYT 1231675
Organisme : Agencia Nacional de Investigación y Desarrollo
ID : Basal funding for Scientific and Technological Center of Excellence, IMPACT, #FB210024, FONDECYT 1231675
Informations de copyright
© 2023. The Author(s).
Références
BMC Pregnancy Childbirth. 2022 Dec 1;22(1):889
pubmed: 36456970
Diabetes Care. 2010 Sep;33(9):2018-20
pubmed: 20519664
Diabetes Care. 2020 Jul;43(7):1438-1441
pubmed: 32350020
BMC Pregnancy Childbirth. 2018 Dec 14;18(1):494
pubmed: 30547769
Diabetes Care. 2012 Mar;35(3):574-80
pubmed: 22301123
Int J Mol Sci. 2018 Oct 26;19(11):
pubmed: 30373146
Diabetes Res Clin Pract. 2018 Jul;141:200-208
pubmed: 29772286
BMC Pregnancy Childbirth. 2022 Jul 4;22(1):540
pubmed: 35787265
Nat Med. 2020 Jan;26(1):71-76
pubmed: 31932807
N Engl J Med. 2005 Jun 16;352(24):2477-86
pubmed: 15951574
Artif Intell Med. 2022 Oct;132:102378
pubmed: 36207076
J Int Med Res. 2020 Apr;48(4):300060519889199
pubmed: 31875748
JAMA. 2021 Aug 10;326(6):539-562
pubmed: 34374717
Diabetes Res Clin Pract. 2021 Sep;179:109001
pubmed: 34390760
NIH Consens State Sci Statements. 2013 Mar 06;29(1):1-31
pubmed: 23748438
Clin Diabetes Endocrinol. 2016 Apr 5;2:8
pubmed: 28702243
Front Pharmacol. 2015 May 20;6:103
pubmed: 26042038
Diabetes Res Clin Pract. 2022 Mar;185:109237
pubmed: 35124096
J Clin Endocrinol Metab. 2023 Feb 01;:
pubmed: 36723990
Diabetes Metab Res Rev. 2021 Jul;37(5):e3397
pubmed: 32845061
J Med Internet Res. 2022 Mar 16;24(3):e26634
pubmed: 35294369
Diabetes Care. 2010 Mar;33(3):676-82
pubmed: 20190296
Diabetes Care. 2022 Jan 1;45(Suppl 1):S17-S38
pubmed: 34964875
Diabetes Res Clin Pract. 2018 Apr;138:271-281
pubmed: 29496507
NPJ Digit Med. 2022 Jun 6;5(1):68
pubmed: 35668134
Diabetes Technol Ther. 2014 Mar;16(3):166-71
pubmed: 24224751
Med Biol Eng Comput. 2023 Jul;61(7):1649-1660
pubmed: 36848010
EClinicalMedicine. 2022 Feb 12;45:101294
pubmed: 35198924
Endocr Rev. 2022 Sep 26;43(5):763-793
pubmed: 35041752
J Big Data. 2021;8(1):101
pubmed: 34306963
Diabetes Care. 2011 Jun;34(6):e61-99
pubmed: 21617108
BMC Pregnancy Childbirth. 2020 Jan 20;20(1):43
pubmed: 31959134
J Clin Endocrinol Metab. 2021 Mar 8;106(3):e1191-e1205
pubmed: 33351102
Am J Obstet Gynecol. 2010 Jun;202(6):654.e1-6
pubmed: 20510967
Biomed Pharmacother. 2021 Nov;143:112183
pubmed: 34560536
BMC Pregnancy Childbirth. 2012 Mar 31;12:23
pubmed: 22462760
Eur J Health Econ. 2019 Apr;20(3):407-417
pubmed: 30229375
Front Endocrinol (Lausanne). 2023 Feb 24;14:1087994
pubmed: 36909340
BMC Pregnancy Childbirth. 2021 Dec 8;21(1):814
pubmed: 34879850
Fetal Diagn Ther. 2022;49(5-6):256-264
pubmed: 35671743
Front Endocrinol (Lausanne). 2022 Jun 06;13:868094
pubmed: 35733768
Am J Obstet Gynecol. 1973 Aug 1;116(7):895-900
pubmed: 4718216
Eur J Endocrinol. 2013 Nov 22;170(1):87-93
pubmed: 24114434
BMC Pregnancy Childbirth. 2019 Jul 19;19(1):252
pubmed: 31324151