Machine learning methods are comparable to logistic regression techniques in predicting severe walking limitation follow
- PDF / 1,018,657 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 22 Downloads / 182 Views
KNEE
Machine learning methods are comparable to logistic regression techniques in predicting severe walking limitation following total knee arthroplasty Yong‑Hao Pua1 · Hakmook Kang2 · Julian Thumboo3 · Ross Allan Clark4 · Eleanor Shu‑Xian Chew1 · Cheryl Lian‑Li Poon1 · Hwei‑Chi Chong5 · Seng‑Jin Yeo6 Received: 16 October 2019 / Accepted: 5 December 2019 © European Society of Sports Traumatology, Knee Surgery, Arthroscopy (ESSKA) 2019
Abstract Purpose Machine-learning methods are flexible prediction algorithms with potential advantages over conventional regression. This study aimed to use machine learning methods to predict post-total knee arthroplasty (TKA) walking limitation, and to compare their performance with that of logistic regression. Methods From the department’s clinical registry, a cohort of 4026 patients who underwent elective, primary TKA between July 2013 and July 2017 was identified. Candidate predictors included demographics and preoperative clinical, psychosocial, and outcome measures. The primary outcome was severe walking limitation at 6 months post-TKA, defined as a maximum walk time ≤ 15 min. Eight common regression (logistic, penalized logistic, and ordinal logistic with natural splines) and ensemble machine learning (random forest, extreme gradient boosting, and SuperLearner) methods were implemented to predict the probability of severe walking limitation. Models were compared on discrimination and calibration metrics. Results At 6 months post-TKA, 13% of patients had severe walking limitation. Machine learning and logistic regression models performed moderately [mean area under the ROC curves (AUC) 0.73–0.75]. Overall, the ordinal logistic regression model performed best while the SuperLearner performed best among machine learning methods, with negligible differences between them (Brier score difference, 30 min, (2) 16–30 min, (3) 5–15 min, and (4) around the house only. Severe walking limitation was defined as a maximum walk time of ≤ 15 min (severe walking limitation = 1
13
Knee Surgery, Sports Traumatology, Arthroscopy
for those who were in categories 3 and 4 and severe walking limitation = 0 otherwise).
Predictor variables Predictor variables were selected based on clinical expertise, literature review [6, 16, 38], and data availability in the department’s databases. To improve the practicality of the prediction models, variables which were less equipment dependent and were routinely and easily measured in the clinical setting were considered. Altogether, 25 predictors were identified and they included demographics and preoperative clinical, psychosocial, and outcome measures (Table 1). Of note, these clinical, psychosocial, and outcome measures were mainly derived from the Short Form 36 (SF36) health survey, Oxford Knee Questionnaire, and Knee Society Clinical Rating Scale, and previous studies [7, 23, 26] have demonstrated good test–retest reliability for these instruments in patients with TKA (intraclass correlation coefficients 0.80–0.92).
Model development Apart from t
Data Loading...
