Use of electronic health data to identify patients with moderate-to-severe osteoarthritis of the hip and/or knee and inadequate response to pain medications

Table 5 Performance of machine learning algorithms to identify patients with moderate-to-severe OA of the hip and/or knee with inadequate response to two or more pain-related medications applied to claims data

Performance Metric	Logistic Regression	Classification and Regression Tree	Random Forest
Positive Predictive Value (SD)^a	0.88 (0.02)	0.89 (0.03)	0.92 (0.02)
Negative Predictive Value	0.47	0.48	0.62
Sensitivity	0.77	0.78	0.86
Specificity	0.66	0.67	0.75
Accuracy	0.75	0.75	0.83
Area Under the Curve	0.72	0.73	0.81
F1^b	0.82	0.83	0.89

^aThe mean and SD are estimated from the outer folds of the nested cross validation
^bF1 score is calculated as the weighted average of PPV and sensitivity. A value between 0 and 1 with 1 being the highest (most accurate)
Abbreviations: OA Osteoarthritis, SD Standard deviation

ISSN: 1471-2288