Clinical predictive modelling of post-surgical recovery in individuals with cervical radiculopathy: a machine learning approach

Bernard X W Liew,Johanna Wibault,Peter Zsigmond,Deborah Falla,Asa Dedering,David Rugamer,Anneli Peolsson,Hakan Löfgren

doi:10.1038/s41598-020-73740-7

Abstract

Prognostic models play an important role in the clinical management of cervical radiculopathy (CR). No study has compared the performance of modern machine learning techniques, against more traditional stepwise regression techniques, when developing prognostic models in individuals with CR. We analysed a prospective cohort dataset of 201 individuals with CR. Four modelling techniques (stepwise regression, least absolute shrinkage and selection operator [LASSO], boosting, and multivariate adaptive regression splines [MuARS]) were each used to form a prognostic model for each of four outcomes obtained at a 12 month follow-up (disability—neck disability index [NDI]), quality of life (EQ5D), present neck pain intensity, and present arm pain intensity). For all four outcomes, the differences in mean performance between all four models were small (difference of NDI < 1 point; EQ5D < 0.1 point; neck and arm pain < 2 points). Given that the predictive accuracy of all four modelling methods were clinically similar, the optimal modelling method may be selected based on the parsimony of predictors. Some of the most parsimonious models were achieved using MuARS, a non-linear technique. Modern machine learning methods may be used to probe relationships along different regions of the predictor space.

Highlights

Prognostic models play an important role in the clinical management of cervical radiculopathy (CR)
For the outcome neck disability index (NDI), stepwise regression was the most accurate technique compared to least absolute shrinkage and selection operator regression (LASSO) (p = 0.028), boosting (p = 0.008), and multivariate adaptive regression splines (MuARS) (p = 0.002)
For EQ5D, stepwise regression was the most accurate compared to LASSO (p = 0.001) and boosting (p = 0.042); whilst MuARS was the least accurate compared to stepwise regression (p < 0.001), LASSO (p < 0.001), and boosting (p < 0.001)

Summary

Introduction

Prognostic models play an important role in the clinical management of cervical radiculopathy (CR). No study has compared the performance of modern machine learning techniques, against more traditional stepwise regression techniques, when developing prognostic models in individuals with CR. Prognostic models play an important role in the clinical prediction of future health outcomes, and identifying the most influential predictors that could inform either clinical management or lead to the development of novel therapeutic interventions[6]. Developing a prognostic model with both self-reported and physical measures could result in a model where the number of predictors exceed sample size and in this case, the model cannot be estimated with traditional fitting methods (e.g. maximum likelihood for simple regression) without additional penalisation as the corresponding algorithm for parameter estimation suffers from identifiability issues. Biased regression coefficients will result in the ensuing model having variable predictive performances when applied to different datasets

Objectives

Methods

Results

Discussion

Conclusion