Table 2.
Validity analysis of the gait and mobility measures: multiple logistic regression.
| Model | b | eb | p-values | AUC (95% CI) | AIC |
|---|---|---|---|---|---|
| h | 0.66 (0.59–0.74) | 281.7 | |||
| Age, years | 0.00 | 1.00 | |||
| Gender, male | −0.57 | 0.56 | |||
| Condition, chronic | 0.76 | 2.15 | |||
| Cognitive impairment, yes | 0.31 | 1.36 | |||
| Urinary incontinence, yes | 0.65 | 1.92 | |||
| h + GS | 0.67 (0.60–0.74) | 280.2 | |||
| Gait speed, m/s | −0.98 | 0.37 | 0.063 | ||
| h + WR | 0.67 (0.59–0.74) | 283.3 | |||
| Walk ratio, cm/steps/min | 0.97 | 2.64 | 0.565 | ||
| h + MB | 0.69 (0.62–0.76) | 276.1 | |||
| Mini-BESTest, logits | −0.29 | 0.75 | 0.006 | ||
| h + TUG | 0.68 (0.61–0.75) | 279.4 | |||
| TUG duration, s | 0.06 | 1.06 | 0.040 | ||
| h + STW | 0.67 (0.60–0.74) | 280.4 | |||
| STW duration, s | 1.31 | 3.72 | 0.069 | ||
| h + Turn | 0.69 (0.62–0.76) | 274.0 | |||
| Turn duration, s | 1.45 | 4.24 | 0.002 | ||
| h + ω | 0.68 (0.60–0.75) | 278.5 | |||
| Peak angular velocity, °/s | −0.01 | 0.99 | 0.023 |
Coefficients (b) and exponentiated coefficients (eb) from the multiple logistic regression models. h: model h, the model only includes fall risk factors from the medical history. h+: models including the fall risk factors from the medical history (i.e., the same variables included in model h) plus a single gait or mobility measure. For example, in h + GS the gait speed (GS) is added to the risk factors from model h. WR, walk ratio; MB, Mini-BESTest; TUG, total duration of the TUG test; STW, sit-to-walk duration; Turn, TUG turning duration; ω, peak of the vertical angular velocity during turning. Only the estimate of the mobility measure is reported for the h + models. p-value: type 1 error probability of the likelihood ratio test comparing model h with one of the seven h + models. p-values < 0.05 indicate that adding the mobility measure to the fall risk factors from the history increases the model’s predictive accuracy. Therefore, the added mobility measure is a significant faller predictor. The area under the curve (AUC) and the Akaike information criterion (AIC) of the h and h + models are also reported. 95% CI: 95% confidence intervals of the AUC.