Researchers train machine model to predict depression in patients with knee OA
Click Here to Manage Email Alerts
TAMPA, Fla. — External and internal validation of six trained machine learning models showed they performed acceptably in predicting depression in patients with knee osteoarthritis, according to a presenter.
“With the exception of decision tree, our models achieved clinically acceptable performance in predicting depression in patients with knee osteoarthritis,” Zuzanna Nowinka, of the department of surgery and cancer at Imperial College London, said in a recorded presentation at the Orthopaedic Research Society Annual Meeting, here.
Nowinka and colleagues studied machine learning models to predict depression in patient with knee OA because the role of depression in contributing to increased severity of symptoms is often overlooked, she said.
Tools are needed to aid patients who are at risk of developing depression, Nowinka said.
For the training model and internal test samples, researchers included data for 2,969 patients and 742 patients, respectively, from the Osteoarthritis Initiative Study. Patient clinical characteristics were used to train Lasso, logistic regression, decision tree, ridge, random forest and gradient boost models.
Additionally, external validation testing of the machine learning models was done using data for 2,236 patients from the Multicenter Osteoarthritis Study. Researchers evaluated the performance of all models using the area under the receiver operating characteristic curve (AUC).
AUC results for internal test sets were from 0.673 to 0.869, according to the abstract. This was similar to the AUC on external validation, which ranged from 0.720 to 0.876.
“The Lasso model had consistently high performance with 0.869 AUC on the internal validation data set and an AUC of 0.876 on the external validation data set,” Nowinka said.
In terms of the features of OA that emerged from the study as most important related to depression and outcome, “those were blood pressure, baseline depression score, knee pain and stiffness, and quality of life,” she said.
Collapse
Nowinka Z, et al. Paper 12. Presented at: Orthopaedic Research Society Annual Meeting. Feb. 4-8, 2022; Tampa, Fla.
Read more about
Click Here to Manage Email Alerts