How Is Machine Learning Different From Statistical Modelling Ncbi

doi: 10.1002/ehf2.13073. Epub 2020 Nov 17.

Machine learning vs. conventional statistical models for predicting centre failure readmission and mortality

Peter C Austin^one , Heather J Ross¹ , Husam Abdel-Qadir¹ , Cassandra Freitas¹ , George Tomlinson¹ , Davide Chicco^one , Meera Mahendiran¹ , Patrick R Lawler¹ , Filio Billia^one , Anthony Gramolini¹ , Slava Epelman¹ , Bo Wang¹ , Douglas S Lee¹

Affiliations

PMID: 33205591
PMCID: PMC7835549
DOI: x.1002/ehf2.13073

Gratis PMC commodity

Machine learning vs. conventional statistical models for predicting heart failure readmission and bloodshed

Sheojung Shin et al. ESC Heart Fail. 2021 Feb .

Free PMC article

Abstract

Aims: This study aimed to review the performance of automobile learning (ML) methods compared with conventional statistical models (CSMs) for predicting readmission and bloodshed in patients with heart failure (HF) and to nowadays an approach to formally evaluate the quality of studies using ML algorithms for prediction modelling.

Methods and results: Post-obit Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, we performed a systematic literature search using MEDLINE, EPUB, Cochrane Key, EMBASE, INSPEC, ACM Library, and Web of Science. Eligible studies included primary research manufactures published betwixt January 2000 and July 2020 comparing ML and CSMs in mortality and readmission prognosis of initially hospitalized HF patients. Data were extracted and analysed by two independent reviewers. A modified CHARMS checklist was adult in consultation with ML and biostatistics experts for quality cess and was utilized to evaluate studies for run a risk of bias. Of 4322 articles identified and screened by ii independent reviewers, 172 were deemed eligible for a full-text review. The final set comprised 20 manufactures and 686 842 patients. ML methods included random forests (n = 11), decision trees (north = five), regression trees (n = 3), support vector machines (n = 9), neural networks (n = 12), and Bayesian techniques (n = 3). CSMs included logistic regression (n = 16), Cox regression (n = iii), or Poisson regression (north = three). In 15 studies, readmission was examined at multiple time points ranging from 30 to 180 twenty-four hours readmission, with the majority of studies (due north = 12) presenting prediction models for 30 day readmission outcomes. Of a full of 21 time-point comparisons, ML-derived c-indices were higher than CSM-derived c-indices in xvi of the 21 comparisons. In 7 studies, mortality was examined at 9 time points ranging from in-hospital mortality to 1 year survival; of these 9, seven reported higher c-indices using ML. Ii of these seven studies reported survival analyses utilizing random survival forests in their ML prediction models. Both reported higher c-indices when using ML compared with CSMs. A limitation of studies using ML techniques was that the majority were non externally validated, and scale was rarely assessed. In the only written report that was externally validated in a separate dataset, ML was superior to CSMs (c-indices 0.913 vs. 0.835).

Conclusions: ML algorithms had better discrimination than CSMs in virtually studies aiming to predict risk of readmission and mortality in HF patients. Based on our review, there is a need for external validation of ML-based studies of prediction modelling. We propose that ML-based studies should also be evaluated using clinical quality standards for prognosis research. Registration: PROSPERO CRD42020134867.

Keywords: Expiry; Heart failure; Hospitalization; Machine learning; Mortality; Prognosis; Readmission; Statistical models.

Conflict of interest argument

The authors have no relevant disclosures.

Figures

**Figure ane**
Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) menstruum nautical chart.

**Figure 2**
Scatterplot of the highest reported c‐index for automobile learning and conventional statistical approaches for readmission studies. Circles of the same colour signal different time points in the aforementioned study publication. CSM, conventional statistical model; ML, machine learning.

**Effigy 3**
Cluster‐bar plot of the departure in the highest reported c‐index for automobile learning and conventional statistical approaches for readmission studies. Confined of the same color indicate different time points in the same study publication. Bars on the right side of the zero line point that c‐indices were higher with ML; bars on the left of the goose egg line bespeak c‐indices were higher with CSMs. Outcomes: Ben‐Assuli 2019 (i) = 90 day readmission; Ben‐Assuli 2019 (ii) = 60 day readmission; Ben‐Assuli 2019 (three) = 30 day readmission; Mortazavi 2016 (1) = 30 twenty-four hour period all‐cause readmission, Mortazavi 2016 (2) = 30 twenty-four hour period HF readmission, Mortazavi 2016 (iii) = 180 twenty-four hour period all‐crusade readmission, Mortazavi 2016 (4) = 180 day HF readmission, Sohrabi 2019 (1) = 1 month HF readmission, Sohrabi 2019 (ii) = 3 month readmission. CSM, conventional statistical model; ML, machine learning.

Similar manufactures

Machine Learning Compared With Conventional Statistical Models for Predicting Myocardial Infarction Readmission and Mortality: A Systematic Review.

Cho SM, Austin PC, Ross HJ, Abdel-Qadir H, Chicco D, Tomlinson M, Taheri C, Foroutan F, Lawler PR, Billia F, Gramolini A, Epelman Due south, Wang B, Lee DS. Cho SM, et al. Tin can J Cardiol. 2021 Aug;37(viii):1207-1214. doi: 10.1016/j.cjca.2021.02.020. Epub 2021 Mar 5. Can J Cardiol. 2021. PMID: 33677098
Application of motorcar learning in predicting hospital readmissions: a scoping review of the literature.

Huang Y, Talwar A, Chatterjee S, Aparasu RR. Huang Y, et al. BMC Med Res Methodol. 2021 May 6;21(ane):96. doi: x.1186/s12874-021-01284-z. BMC Med Res Methodol. 2021. PMID: 33952192 Gratis PMC article. Review.
Analysis of Machine Learning Techniques for Heart Failure Readmissions.

Mortazavi BJ, Downing NS, Bucholz EM, Dharmarajan K, Manhapra A, Li SX, Negahban SN, Krumholz HM. Mortazavi BJ, et al. Circ Cardiovasc Qual Outcomes. 2016 Nov;9(6):629-640. doi: 10.1161/CIRCOUTCOMES.116.003039. Epub 2016 November 8. Circ Cardiovasc Qual Outcomes. 2016. PMID: 28263938 Gratuitous PMC article.
Machine learning-based prediction of heart failure readmission or expiry: implications of choosing the correct model and the right metrics.

Awan SE, Bennamoun Grand, Sohel F, Sanfilippo FM, Dwivedi Yard. Awan SE, et al. ESC Centre Fail. 2019 April;6(2):428-435. doi: 10.1002/ehf2.12419. Epub 2019 Feb 27. ESC Heart Fail. 2019. PMID: 30810291 Gratis PMC article.

Cited by 10 articles

Predictive modeling for COVID-19 readmission chance using machine learning algorithms.

Shanbehzadeh M, Yazdani A, Shafiee M, Kazemi-Arpanahi H. Shanbehzadeh M, et al. BMC Med Inform Decis Mak. 2022 May twenty;22(1):139. doi: 10.1186/s12911-022-01880-z. BMC Med Inform Decis Mak. 2022. PMID: 35596167 Free PMC article.
Predicting infirmary readmission risk in patients with COVID-xix: A machine learning approach.

Afrash MR, Kazemi-Arpanahi H, Shanbehzadeh Grand, Nopour R, Mirbagheri E. Afrash MR, et al. Inform Med Unlocked. 2022;xxx:100908. doi: 10.1016/j.imu.2022.100908. Epub 2022 Mar eight. Inform Med Unlocked. 2022. PMID: 35280933 Free PMC article.
Graphical calibration curves and the integrated calibration alphabetize (ICI) for competing risk models.

Austin PC, Putter H, Giardiello D, van Klaveren D. Austin PC, et al. Diagn Progn Res. 2022 Jan 17;half-dozen(1):2. doi: x.1186/s41512-021-00114-6. Diagn Progn Res. 2022. PMID: 35039069 Free PMC commodity.
Development and Internal Validation of Supervised Machine Learning Algorithms for Predicting the Adventure of Surgical Site Infection Following Minimally Invasive Transforaminal Lumbar Interbody Fusion.

Wang H, Fan T, Yang B, Lin Q, Li W, Yang M. Wang H, et al. Front end Med (Lausanne). 2021 December xx;viii:771608. doi: 10.3389/fmed.2021.771608. eCollection 2021. Front Med (Lausanne). 2021. PMID: 34988091 Free PMC article.
Machine Learning-Based Prediction of Myocardial Recovery in Patients With Left Ventricular Assistance Device Back up.

Topkara VK, Elias P, Jain R, Sayer Thou, Burkhoff D, Uriel N. Topkara VK, et al. Circ Center Neglect. 2022 Jan;15(i):e008711. doi: 10.1161/CIRCHEARTFAILURE.121.008711. Epub 2021 December 24. Circ Heart Neglect. 2022. PMID: 34949101

References

1. G. B. D. Causes of Death Collaborators . Global, regional, and national historic period‐sex activity specific bloodshed for 264 causes of death, 1980–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390: 1151–1210. - PMC - PubMed
1. Fang J, Mensah GA, Croft JB, Keenan NL. Eye failure‐related hospitalization in the U.S., 1979 to 2004. J Am Coll Cardiol 2008; 52: 428–434. - PubMed
1. Ziaeian B, Fonarow GC. Epidemiology and aetiology of heart failure. Nat Rev Cardiol 2016; thirteen: 368–378. - PMC - PubMed
1. Shameer Chiliad, Johnson KW, Glicksberg BS, Dudley JT, Sengupta PP. Car learning in cardiovascular medicine: are we at that place yet? Heart 2018; 104: 1156–1,164. - PubMed
1. Shameer Chiliad, Johnson KW, Yahi A, Miotto R, Li LI, Ricks D, Jebakaran J, Kovatch P, Sengupta PP, Gelijns Southward, Moskovitz A, Darrow B, David DL, Kasarskis A, Tatonetti NP, Pinney Due south, Dudley JT. Predictive modeling of infirmary readmission rates using electronic medical record‐broad machine learning: a case‐study using Mount Sinai Heart Failure Accomplice. Pac Symp Biocomput 2017; 22: 276–287. - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources

Full Text Sources
- Europe PubMed Central
- PubMed Fundamental
- Wiley
Medical
- MedlinePlus Health Information
Inquiry Materials
- NCI CPTC Antibody Characterization Program
Miscellaneous
- NCI CPTAC Assay Portal