Table 4.
Pooled mean estimates of accuracy by several factors (N=19).
| Group | Number of studiesa | Sample size, n | Accuracy (%), range | Pooled mean accuracy (%), mean (95% CI) | Heterogeneity measures | Test for subgroup differences (P value) | |||||||||||||
|
|
|
|
|
|
τ2 | Q (P value) | I2 (%) |
|
|||||||||||
| Algorithm | .98 | ||||||||||||||||||
|
|
Support vector machine | 6 | 2039 | 0.65-0.94 | 0.796 (0.67-0.90) | 0.0276 | 143.0 (<.001) | 96.5 |
|
||||||||||
|
|
K-nearest neighbor | 6 | 2039 | 0.58-0.98 | 0.834 (0.71-0.93) | 0.0326 | 96.2 (<.001) | 94.8 |
|
||||||||||
|
|
Logistic regression | 4 | 1545 | 0.66-0.87 | 0.768 (0.67-0.86) | 0.0108 | 20.7 (<.001) | 85.5 |
|
||||||||||
|
|
Artificial neural network | 4 | 1686 | 0.66-0.99 | 0.811 (0.63-0.94) | 0.0436 | 58.3 (<.001) | 94.9 |
|
||||||||||
|
|
Random forest | 3 | 1215 | 0.71-0.98 | 0.829 (0.61-0.97) | 0.0456 | 43.1 (<.001) | 95.4 |
|
||||||||||
|
|
Decision tree | 2 | 494 | 0.64-0.86 | 0.762 (0.52-0.94) | 0.0319 | 29.6 (<.001) | 96.6 |
|
||||||||||
|
|
Naïve Bayes | 2 | 990 | 0.68-0.97 | 0.841 (0.50-1.00) | 0.0672 | 16.9 (<.001) | 94.1 |
|
||||||||||
| Number of stress classes | .02b | ||||||||||||||||||
|
|
2 | 25 | 9222 | 0.58-0.97 | 0.773 (0.61-0.87) | 0.0873 | 173.4 (<.001) | 97.2 |
|
||||||||||
|
|
>2 | 12 | 2203 | 0.86-0.99 | 0.944 (0.77-0.99) | 0.2110 | 302.8 (<.001) | 98.2 |
|
||||||||||
| Type of WDc | .049 | ||||||||||||||||||
|
|
Smartwatches | 11 | 2094 | 0.66-0.97 | 0.836 (0.53-0.95) | 0.1466 | 91.6 (<.001) | 96.6 |
|
||||||||||
|
|
Smart bands | 14 | 7128 | 0.58-0.77 | 0.695 (0.63-0.75) | 0.0050 | 69.2 (<.001) | 85.4 |
|
||||||||||
|
|
Electrodes | 12 | 2203 | 0.86-0.99 | 0.944 (0.77-0.99) | 0.2110 | 302.8 (<.001) | 98.2 |
|
||||||||||
| Location of WD | .02 | ||||||||||||||||||
|
|
Wrist | 25 | 9222 | 0.58-0.97 | 0.773 (0.61-0.87) | 0.0873 | 173.4 (<.001) | 97.2 |
|
||||||||||
|
|
Nonwrist | 12 | 2203 | 0.86-0.99 | 0.944 (0.77-0.99) | 0.2110 | 302.8 (<.001) | 98.2 |
|
||||||||||
| Data set size | .009 | ||||||||||||||||||
|
|
≤100 | 14 | 789 | 0.80-0.99 | 0.947 (0.81-0.99) | 0.1704 | 166.7 (<.001) | 94.1 |
|
||||||||||
|
|
>100 | 23 | 10,627 | 0.58-0.91 | 0.769 (0.62-0.86) | 0.0758 | 485.3 (<.001) | 97.4 |
|
||||||||||
| Data source | .57 | ||||||||||||||||||
|
|
WD-based | 12 | 1825 | 0.80-0.97 | 0.895 (0.87-0.91) | 0.0000 | 29.9 (<.001) | 60.5 |
|
||||||||||
|
|
WD-based data+other data | 25 | 9600 | 0.58-0.99 | 0.830 (0.50-0.95) | 0.3708 | 968.9 (<.001) | 99.4 |
|
||||||||||
| Data type | .21 | ||||||||||||||||||
|
|
Heart rate data | 13 | 5970 | 0.65-0.97 | 0.832 (0.52-0.95) | 0.1594 | 115.4 (<.001) | 98.6 |
|
||||||||||
|
|
Heart rate data+EDAd data | 12 | 2203 | 0.86-0.99 | 0.944 (0.77-0.99) | 0.2110 | 302.8 (<.001) | 98.2 |
|
||||||||||
| Stress inducers | .10 | ||||||||||||||||||
|
|
Yes | 25 | 8173 | 0.65-0.99 | 0.902 (0.75-0.96) | 0.2292 | 1137.8 (<.001) | 98.9 |
|
||||||||||
|
|
No | 12 | 3252 | 0.58-0.80 | 0.697 (0.62-0.76) | 0.0060 | 51.4 (<.001) | 80.6 |
|
||||||||||
| Ground truth | .001 | ||||||||||||||||||
|
|
Objective | 19 | 2413 | 0.80-0.99 | 0.933 (0.84-0.97) | 0.1266 | 320.2 (<.001) | 96.3 |
|
||||||||||
|
|
Subjective | 18 | 9012 | 0.58-0.80 | 0.706 (0.66-0.75) | 0.0031 | 95.0 (<.001) | 85.1 |
|
||||||||||
| Validation method | .16 | ||||||||||||||||||
|
|
K-fold | 16 | 4087 | 0.66-0.99 | 0.904 (0.64-0.98) | 0.3324 | 784.7 (<.001) | 99.0 |
|
||||||||||
|
|
Training-test split | 14 | 7128 | 0.58-0.77 | 0.695 (0.63-0.75) | 0.0050 | 69.2 (<.001) | 85.4 |
|
||||||||||
| All studies | 37 | 11,425 | 0.58-0.99 | 0.856 (0.698-0.934) | 0.2402 | 1318.2 (<.001) | 98.8 | N/Ae | |||||||||||
aMany studies were included more than once in all meta-analyses except for the meta-analysis related to algorithms given that the studies assessed the performance of more than one algorithm.
bItalics indicate statistical significance (P<.05).
cWD: wearable device.
dEDA: electrodermal activity.
eN/A: not applicable.