Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model

Background and purpose — Prediction of postoperative outcomes and length of hospital stay (LOS) of patients is vital for allocation of healthcare resources. We investigated the performance of prediction models based on machinelearning algorithms compared with a previous risk stratification model usi...

Full description

Bibliographic Details
Main Authors: Katrin B Johannesdottir, Henrik Kehlet, Pelle B Petersen, Eske K Aasvang, Helge B D Sørensen, Christoffer C Jørgensen
Format: Article
Language:English
Published: Medical Journals Sweden 2022-01-01
Series:Acta Orthopaedica
Subjects:
Online Access:https://actaorthop.org/actao/article/view/843
_version_ 1831658670701674496
author Katrin B Johannesdottir
Henrik Kehlet
Pelle B Petersen
Eske K Aasvang
Helge B D Sørensen
Christoffer C Jørgensen
author_facet Katrin B Johannesdottir
Henrik Kehlet
Pelle B Petersen
Eske K Aasvang
Helge B D Sørensen
Christoffer C Jørgensen
author_sort Katrin B Johannesdottir
collection DOAJ
description Background and purpose — Prediction of postoperative outcomes and length of hospital stay (LOS) of patients is vital for allocation of healthcare resources. We investigated the performance of prediction models based on machinelearning algorithms compared with a previous risk stratification model using traditional multiple logistic regression, for predicting the risk of a LOS of > 2 days after fast-track total hip and knee replacement. Patients and methods — 3 different machine learning classifiers were trained on data from the Lundbeck Centre for Fast-track Hip and Knee Replacement Database (LCDB) collected from 9,512 patients between 2016 and 2017. The chosen classifiers were a random forest classifier (RF), a support vector machine classifier with a polynomial kernel (SVM), and a multinomial Naïve-Bayes classifier (NB). Results — Comparing performance measures of the classifiers with the traditional model revealed that all the models had a similar performance in terms of F1 score, accuracy, sensitivity, specificity, area under the receiver operating curve (AUC), and area under the precision-recall curve (AUPRC). A feature importance analysis of the RF classifier found hospital, age, use of walking aid, living alone, and joint operated on to be the most relevant input features. None of the classifiers reached a clinically relevant performance with the input data from the LCDB. Interpretation — Despite the promising prospects of machine-learning practices for disease and risk prediction, none of the machine learning models tested outperformed the traditional multiple regression model in predicting which patients in this cohort had a LOS > 2 days.
first_indexed 2024-12-19T17:47:54Z
format Article
id doaj.art-6b482f2a759f46d3906c23f3659f505d
institution Directory Open Access Journal
issn 1745-3674
1745-3682
language English
last_indexed 2024-12-19T17:47:54Z
publishDate 2022-01-01
publisher Medical Journals Sweden
record_format Article
series Acta Orthopaedica
spelling doaj.art-6b482f2a759f46d3906c23f3659f505d2022-12-21T20:12:00ZengMedical Journals SwedenActa Orthopaedica1745-36741745-36822022-01-019310.2340/17453674.2021.843Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk modelKatrin B Johannesdottir0Henrik Kehlet1Pelle B Petersen2Eske K Aasvang3Helge B D Sørensen4Christoffer C Jørgensen5Biomedical Signal Processing & AI research group, Digital Health Section, DTU Health Tech, Technical University of Denmark, LyngbySection of Surgical Pathophysiology 7621, Rigshospitalet, CopenhagenSection of Surgical Pathophysiology 7621, Rigshospitalet, CopenhagenSection of Surgical Pathophysiology 7621, Rigshospitalet, Copenhagen; Department of Anesthesiology, Center for Cancer and Organ Diseases, Copenhagen, DenmarkBiomedical Signal Processing & AI research group, Digital Health Section, DTU Health Tech, Technical University of Denmark, LyngbySection of Surgical Pathophysiology 7621, Rigshospitalet, CopenhagenBackground and purpose — Prediction of postoperative outcomes and length of hospital stay (LOS) of patients is vital for allocation of healthcare resources. We investigated the performance of prediction models based on machinelearning algorithms compared with a previous risk stratification model using traditional multiple logistic regression, for predicting the risk of a LOS of > 2 days after fast-track total hip and knee replacement. Patients and methods — 3 different machine learning classifiers were trained on data from the Lundbeck Centre for Fast-track Hip and Knee Replacement Database (LCDB) collected from 9,512 patients between 2016 and 2017. The chosen classifiers were a random forest classifier (RF), a support vector machine classifier with a polynomial kernel (SVM), and a multinomial Naïve-Bayes classifier (NB). Results — Comparing performance measures of the classifiers with the traditional model revealed that all the models had a similar performance in terms of F1 score, accuracy, sensitivity, specificity, area under the receiver operating curve (AUC), and area under the precision-recall curve (AUPRC). A feature importance analysis of the RF classifier found hospital, age, use of walking aid, living alone, and joint operated on to be the most relevant input features. None of the classifiers reached a clinically relevant performance with the input data from the LCDB. Interpretation — Despite the promising prospects of machine-learning practices for disease and risk prediction, none of the machine learning models tested outperformed the traditional multiple regression model in predicting which patients in this cohort had a LOS > 2 days.https://actaorthop.org/actao/article/view/843ArthroplastyHipKneeLength of hospital stayOutcomePrediction
spellingShingle Katrin B Johannesdottir
Henrik Kehlet
Pelle B Petersen
Eske K Aasvang
Helge B D Sørensen
Christoffer C Jørgensen
Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
Acta Orthopaedica
Arthroplasty
Hip
Knee
Length of hospital stay
Outcome
Prediction
title Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
title_full Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
title_fullStr Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
title_full_unstemmed Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
title_short Machine learning classifiers do not improve prediction of hospitalization > 2 days after fast-track hip and knee arthroplasty compared with a classical statistical risk model
title_sort machine learning classifiers do not improve prediction of hospitalization 2 days after fast track hip and knee arthroplasty compared with a classical statistical risk model
topic Arthroplasty
Hip
Knee
Length of hospital stay
Outcome
Prediction
url https://actaorthop.org/actao/article/view/843
work_keys_str_mv AT katrinbjohannesdottir machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel
AT henrikkehlet machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel
AT pellebpetersen machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel
AT eskekaasvang machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel
AT helgebdsørensen machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel
AT christoffercjørgensen machinelearningclassifiersdonotimprovepredictionofhospitalization2daysafterfasttrackhipandkneearthroplastycomparedwithaclassicalstatisticalriskmodel