A novel algorithm may be accurate at identifying subgroups of patients with osteoarthritis and predicting patients who may experience a worse quality of life, according to a study published by Lu et al in BMC Medical Research Methodology. Researchers used chart data and electronic medical records to identify the presence or absence of three osteoarthritis-related characteristics—including osteoarthritis of the hip and/or knee, moderate-to-severe disease, and failure to respond to pain medications—among 571 patients. They then developed two sets of algorithms: predefined algorithms based on both literature reviews, and clinical inputs and machine-learning algorithms. They found that despite demonstrating high positive predictive values (≥ 0.83), the predefined algorithms had low negative predictive values (range = 0.16–0.54) and sometimes low sensitivity when identifying the characteristics. Further, the predefined algorithms had a sensitivity and specificity of 0.95 and 0.26, respectively, when identifying those who had all three of the characteristics. However, the machine-learning algorithms provided positive predictive values of 0.88 to 0.92, negative predictive values of 0.47 to 0.62, sensitivity of 0.77 to 0.86, and specificity of 0.66 to 0.75. The researchers highlighted that the machine-learning algorithms were more capable of differentiating between patients on the basis of disease severity and response to pain medications compared with the predefined algorithms.
July 20, 2023