Table 4.
Pooled mean estimates of highest sensitivity by several factors.
| Group | Studies, Na | Sample size, N | Sensitivity (%), range | Pooled mean sensitivity, mean (95% CIb) | Heterogeneity measures | Test for subgroup differences (P value) | ||||||||||||
|
|
|
|
|
|
τ2 | Q (P value) | I2 (%) |
|
||||||||||
| Algorithms | .53 | |||||||||||||||||
|
|
Random forest | 5 | 10,424 | 0.57-0.99 | 0.78 (0.56-0.94) | 0.0638 | 539.7 (<.001) | 99.3 |
|
|||||||||
|
|
Support vector machine | 3 | 37,807 | 0.47-1.00 | 0.84 (0.45-1.00) | 0.1434 | 520.6 (<.001) | 99.6 |
|
|||||||||
|
|
Decision tree | 3 | 10,149 | 0.58-0.98 | 0.87 (0.57-1.00) | 0.0884 | 427.7 (<.001) | 99.5 |
|
|||||||||
|
|
Multilayer perceptron | 3 | 206 | 0.60-0.90 | 0.76 (0.54-0.93) | 0.0333 | 15.99 (<.001) | 87.4 |
|
|||||||||
|
|
Logistic regression | 2 | 47 | 0.63-0.71 | 0.66 (0.52-0.79) | 0.0000 | 0.2 (.64) | 0.0 |
|
|||||||||
|
|
XGBoost | 2 | 359 | 0.21-0.85 | 0.52 (0.01-1.00) | 0.2192 | 44.1 (<.001) | 97.7 |
|
|||||||||
| Aims of AIc | .70 | |||||||||||||||||
|
|
Detectionb | 17 | 95,770 | 0.47-1.00 | 0.82 (0.54-0.93) | 0.4634 | 7418.4 (<.001) | 99.9 |
|
|||||||||
|
|
Prediction | 7 | 2041 | 0.21-0.85 | 0.69 (0.01-0.93) | 0.2090 | 58.5 (<.001) | 99.9 |
|
|||||||||
| Status of WDd | .74 | |||||||||||||||||
|
|
Commercialb | 20 | 97,299 | 0.21-1.00 | 0.78 (0.46-0.92) | 0.4808 | 16,064.4 (<.001) | 99.9 |
|
|||||||||
|
|
Noncommercialb | 4 | 495 | 0.75-0.93 | 0.87 (0.50-0.97) | 0.0864 | 7.3 (.06) | 91.2 |
|
|||||||||
| Data sources | .86 | |||||||||||||||||
|
|
WD-basedb | 15 | 95,313 | 0.47-1.00 | 0.80 (0.51-0.93) | 0.4507 | 6773.9 (<.001) | 100 |
|
|||||||||
|
|
WD-based and othersb | 9 | 2481 | 0.21-0.93 | 0.77 (0.01-0.97) | 0.4715 | 416.5 (<.001) | 99.2 |
|
|||||||||
| Reference standards | .36 | |||||||||||||||||
|
|
STAIb,e | 7 | 153 | 0.60-0.83 | 0.72 (0.59-0.81) | 0.0000 | 4.4 (.62) | 2.8 |
|
|||||||||
|
|
CIDIf | 2 | 46 | 0.70-0.85 | 0.78 (0.63-0.91) | 0.0036 | 1.3 (.25) | 24.9 |
|
|||||||||
| Validation methods | .34 | |||||||||||||||||
|
|
K-foldb | 18 | 97,045 | 0.21-1.00 | 0.83 (0.48-0.95) | 0.5614 | 14.910.3 (<.001) | 100 |
|
|||||||||
|
|
Leave-one-out | 2 | 254 | 0.47-0.57 | 0.50 (0.40-0.59) | 0.0018 | 1.5 (.22) | 33.3 |
|
|||||||||
| All studiesb | 24 | 97,794 | 0.21-1.00 | 0.79 (0.57-0.91) | 0.4039 | 16,735.8 (<.001) | 99.9 | N/Ag | ||||||||||
aMany studies were included more than once in each meta-analysis given that they assessed the performance of more than one algorithm.
bSensitivity was pooled using the multilevel meta-analysis method.
cAI: artificial intelligence.
dWD: wearable device.
eSTAI: State-Trait Anxiety Inventory.
fCIDI: Composite International Diagnostic Interview.
gN/A: not applicable.