Table 3.

Statistical results comparing each wrist algorithm and metric (optimized versions) with a representative lower back algorithm with good performance (Iluz (2014) applied to the lower back).

Algorithm and metric			Mean difference		P value	Adjusted P value
Willetts (2018)
	Sensitivity	–0.317		<.001		<.001
	Specificity	–0.029		<.001		<.001
	PPVa	–0.258		<.001		<.001
	Accuracy	–0.064		<.001		<.001
	Relative number GSb error	–0.325		<.001		<.001
	Relative GS duration error	–0.055		.37		.38
Brand (2022)
	Sensitivity	–0.126		<.001		<.001
	Specificity	–0.002		.49		.49
	PPV	–0.037		.03		.04
	Accuracy	–0.014		<.001		<.001
	Relative number GS error	0.042		.19		.21
	Relative GS duration error	–0.131		<.001		<.001
Paraschiv-Ionescu (2019)
	Sensitivity	–0.277		<.001		<.001
	Specificity	–0.011		<.001		<.001
	PPV	–0.115		<.001		<.001
	Accuracy	–0.041		<.001		<.001
	Relative number GS error	0.335		<.001		<.001
	Relative GS duration error	–0.216		<.001		<.001
Iluz (2014)
	Sensitivity	–0.273		<.001		<.001
	Specificity	–0.004		.27		.29
	PPV	–0.067		<.001		<.001
	Accuracy	–0.035		<.001		<.001
	Relative number GS error	–0.500		<.001		<.001
	Relative GS duration error	–0.268		<.001		<.001
Kheirkhahan (2017)
	Sensitivity	–0.299		<.001		<.001
	Specificity	–0.005		.23		.25
	PPV	–0.063		<.001		<.001
	Accuracy	–0.038		<.001		<.001
	Relative number GS error	–0.671		<.001		<.001
	Relative GS duration error	–0.310		<.001		<.001

^aPPV: positive predictive value.

^bGS: gait sequence.