Using a machine learning approach to predict outcome after surgery for degenerative cervical myelopathy

Abstract

Degenerative cervical myelopathy (DCM) is a spinal cord condition that results in progressive non-traumatic compression of the cervical spinal cord. Spine surgeons must consider a large quantity of information relating to disease presentation, imaging features, and patient characteristics to determine if a patient will benefit from surgery for DCM. We applied a supervised machine learning approach to develop a classification model to predict individual patient outcome after surgery for DCM. Patients undergoing surgery for DCM as a part of the AOSpine CSM-NA or CSM-I prospective, multi-centre studies were included in the analysis. Out of 757 patients 605, 583, and 539 patients had complete follow-up information at 6, 12, and 24 months respectively and were included in the analysis. The primary outcome was improvement in the SF-6D quality of life indicator score by the minimum clinically important difference (MCID). The secondary outcome was improvement in the modified Japanese Orthopedic Association (mJOA) score by the MCID. Predictor variables reflected information about pre-operative disease severity, disease presentation, patient demographics, and comorbidities. A machine learning approach of feature engineering, data pre-processing, and model optimization was used to create the most accurate predictive model of outcome after surgery for DCM. Following data pre-processing 48, 108, and 101 features were chosen for model training at 6, 12, and 24 months respectively. The best performing predictive model used a random forest structure and had an average area under the curve (AUC) of 0.70, classification accuracy of 77%, and sensitivity of 78% when evaluated on a testing cohort that was not used for model training. Worse pre-operative disease severity, longer duration of DCM symptoms, older age, higher body weight, and current smoking status were associated with worse surgical outcomes. We developed a model that predicted positive surgical outcome for DCM with good accuracy at the individual patient level on an independent testing cohort. Our analysis demonstrates the applicability of machine-learning to predictive modeling in spine surgery.

Trial registration: ClinicalTrials.gov NCT00285337 NCT00565734.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1

Results of the recursive feature elimination algorithm applied to 6-month follow-up (A), 12-month follow-up (B), and 24-month follow-up (C). The figures demonstrate the change in root mean squared error (RMSE) as features were iteratively added to the model. As greater number of features were added to the model the RMSE decreased to a minimum value, demonstrating best model fit, then began to increase as greater numbers of ‘distracting’ features were added. The set of features that achieved the minimum RMSE were used for model training (shown by vertical black line).

Fig 2. Receiver operating characteristic curves for the random forest model at all follow-up points on the training/validation dataset.

The blue lines represent each cross validation fold.

Fig 3. Density plots for the top 6 most important predictive features selected by the random forest model.

These density plots demonstrate the distribution of the key features between the patients who did (blue) and did not (red) show improvement in SF6D at 1-year follow-up. In all key features there is overlap of the curves, demonstrating that there is no one singe feature that can alone predict if a patient with DCM will improve with surgery.