Table 3.
Fall risk classification models
| Author | Model | Acc (%) | Sen (%) | Spe (%) |
|---|---|---|---|---|
| Bautmans et al. [40] | logistic regression analysis, ROC | 77.0 | 78.0 | 78.0 |
| Bizovska et al. [43] | logistic regression analysis, ROC | – | 53.0 | 85.0 |
| Buisseret et al. [64] a | binary classification, ROC | 85.7 | 50.0 | 73,9 |
| Greene et al. [55] | ROC | 79.7 | 73.1 | 82.6 |
| Gietzelt et al. [36] | decision tree | 75.0 | 78.2 | 71.2 |
| Howcroft et al. [56] | support vector machine and neural networks | 80.0–84.0 | 50.0–66.7 | 89.5 |
| Hua et al. [41] | random forests | 73.7 | 81.1 | – |
| Ihlen et al. [44] | Partial Least Square Regression Analysis |
76.0 (SF) 68.0 (MF) |
71.0 (SF) 67.0 (MF) |
80.0 (SF) 69.0 (MF) |
| Ihlen et al. [49] | Partial Least Square Discriminatory Analysis | – | 59.0–88.0 | 77.0–92.0 |
| Iluz et al. [35] | Ada Boost, Support Vector Machine, Bag, Naïve Bayes | 87.1–90.6 | 83.8–89.2 | 87.2–94.4 |
| Marschollek et al. [62] | logistic regression, classification model | 70.0 | 58.0 | 78.0 |
| Marschollek et al. [61] a | classification trees | 90.0 | 57.7 | 100.0 |
| Qui et al. [50] a | logistic regression, Naïve Bayes, decision tree, boosted tree, random forest, support vector machine | 79.7–89.4 | 87.2–92.7 | 69.2–84.9 |
| Rivolta et al. [19] a | linear model, artificial neural network | – | 71.0–86.0 | 81.0–90.0 |
| Sample et al. [58] a | stepwise logistic regression, max-rescaled R2 value | – | 48.1 | 82.1 |
| Senden et al. [51] | linear regression analysis, ROC | – | 76.0 | 70.0 |
| van Schooten et al. [52] | logistic regression analysis, ROC | – | 67.9 | 66.3 |
| Weiss et al. [54] a | binary logistic regression analysis | 71.6 | 62.1 | 78.9 |
| Weiss et al. [53] | binary logistic regression analysis | 87.8 | 91.3 | 83.3 |
a These models also include data of clinical assessment (e. g. body mass index)
Acc: accuracy, Sen: sensitivity, Spe: specificity, ROC: receiver operating curve, SF: single faller, MF: multiple faller