JMIR Publications

ABSTRACT

Background:

Early identification of knee osteoarthritis (OA) can improve treatment outcomes and reduce medical costs. However, there are major limitations amongst existing classification or prediction models including abstract data processing and complicated dataset attributes, which hinder their applications in clinical practice.

Objective:

Develop a Bayesian network (BN)-based classification model to classify people with knee OA. The proposed model can be treated as a prescreening tool which can provide decision support for health professionals.

Methods:

The proposed model’s structure is based on a three-level BN structure, and then retrained by the Bayesian Search (BS) learning algorithm. The model’s parameters are determined by the Expectation-Maximization (EM) algorithm. The utilized dataset includes backgrounds (the basic characteristics of subjects), the target disease (namely the knee OA), and predictors (the scores of physical fitness tests). The performance of the model is evaluated based on classification accuracy, area under a curve (AUC), specificity, sensitivity, positive predictive value (PPV) and negative predictive value (NPV), and is also compared with other well-known classification models. A test is also performed to explore whether physical fitness tests could improve the performance of the proposed model.

Results:

249 elderly people between 60 and 80 years of age, living in the Kongjiang community (shanghai) were recruited from April to September 2007. A total of 157 instances are adopted as the dataset after data preprocessing. The experimental results show that the proposed model’s results are higher than, or equal to, the mean scores of the other classification models: 0.754 for accuracy, 0.78 for AUC, 0.78 for specificity and 0.73 for sensitivity. The proposed model can provide 0.45 for PPV and 0.92 for NPV at the prevalence of 20%. The proposed model also shows a significant improvement when compared to the traditional BN model: 6.35% increase in accuracy (from 0.709 to 0.754), 4.00% increase in AUC (from 0.75 to 0.78), 6.85% increase in specificity (from 0.73 to 0.78), 5.80% increase in sensitivity (from 0.69 to 0.73), 15.38% increase in PPV (from 0.39 to 0.45) and 2.22% increase in NPV (from 0.90 to 0.92). Furthermore, the test results show that the performance of the proposed model could be largely enhanced through physical fitness tests in three evaluation indexes: 10.56% increase in accuracy (from 0.682 to 0.754), 16.42% increase in AUC (from 0.67 to 0.78) and 30.00% increase in specificity (from 0.60 to 0.78).

Conclusions:

The proposed model presents a promising method to classify people with knee OA when compared to other classification models and the traditional BN model. The proposed model could be implemented in clinical practice as a prescreening tool for knee OA, which could, not only improve the quality of healthcare for elderly people, but also reduce overall medical costs.