Evaluating Doctor Performance: Ordinal Regression-Based Approach

BACKGROUND: Doctor’s performance evaluation is an important task in mobile health (mHealth), which aims to evaluate the overall quality of online diagnosis and patient outcomes so that customer satisfaction and loyalty can be attained. However, most patients tend not to rate doctors’ performance, therefore, it is imperative to develop a model to make doctor’s performance evaluation automatic. When evaluating doctors’ performance, we rate it into a score label that is as close as possible to the true one. OBJECTIVE: This study aims to perform automatic doctor’s performance evaluation from online textual consultations between doctors and patients by way of a novel machine learning method. METHODS: We propose a solution that models doctor’s performance evaluation as an ordinal regression problem. In doing so, a support vector machine combined with an ordinal partitioning model (SVMOP), along with an innovative predictive function will be developed to capture the hidden preferences of the ordering labels over doctor’s performance evaluation. When engineering the basic text features, eight customized features (extracted from over 70,000 medical entries) were added and further boosted by the Gradient Boosting Decision Tree algorithm. RESULTS: Real data sets from one of the largest mobile doctor/patient communication platforms in China are used in our study. Statistically, 64% of data on mHealth platforms lack the evaluation labels from patients. Experimental results reveal that our approach can support an automatic doctor performance evaluation. Compared with other auto-evaluation models, SVMOP improves mean absolute error (MAE) by 0.1, mean square error (MSE) by 0.5, pairwise accuracy (PAcc) by 5%; the suggested customized features improve MAE by 0.1, MSE by 0.2, PAcc by 3%. After boosting, performance is further improved. Based on SVMOP, predictive features like politeness and sentiment words can be mined, which can be further applied to guide the development of mHealth platforms. CONCLUSIONS: The initial modelling of doctor performance evaluation is an ordinal regression problem. Experiments show that the performance of our proposed model with revised prediction function is better than many other machine learning methods on MAE, MSE, as well as PAcc. With this model, the mHealth platform could not only make an online auto-evaluation of physician performance, but also obtain the most effective features, thereby guiding physician performance and the development of mHealth platforms.