Table 2.

Validation values for the algorithm, based on increasing turn magnitudes, from the training cohort.

Turning angle	Cohen’s kappa	Accuracy	Sens	Spec	PPV	NPV	True positive turns	False positive turns
≥45°	0.15	0.66	0.92	0.21	0.67	0.59	694	341
≥60°	0.53	0.78	0.91	0.61	0.76	0.82	627	194
≥70°	0.64	0.83	0.91	0.71	0.82	0.85	607	148
≥80°	0.68	0.84	0.90	0.77	0.83	0.87	586	123
≥90°	0.72	0.86	0.90	0.82	0.85	0.88	565	98
≥100°	0.76	0.88	0.90	0.86	0.87	0.89	529	82
≥110°	0.77	0.89	0.89	0.88	0.88	0.90	511	72

758 turns ≥45° defined by video observation were included.

As expected, validation values improved with increasing magnitude of turns, at the expense of number of turns included in the analysis. Based on these values, we decided to use a threshold magnitude of 90° for validation purposes.

NPV, negative predictive value; PPV, positive predictive value, Sens, sensitivity; Spec, specificity.