The Geriatric Depression Scale with 30 items (GDS-30) and with 15 items (GDS-15) are both valid tools for assessing depression in older adults, but their absolute values are not directly comparable. Here, we used a dataset (n=431) with GDS-30 scores from a project concerning fall-risk assessment in older adults (FARAO) to develop and validate a formula which can be used to convert GDS-15 scores into GDS-30 scores. We found that the GDS-15 score cannot simply be multiplied by 2 to obtain the GDS-30 scores. Instead, the optimal formula to estimate the GDS-30 score from the GDS-15 score was: GDS-30_estimated = 1.57 + 1.95 * GDS-15. This formula yielded an estimate of GDS-30 with an explained variance of 79%, compared to 63% when GDS-15 was simply multiplied by 2. Researchers that have used the GDS-15 and want to compare their outcomes to other studies that reported only the GDS-30 are advised to use this formula.

Background: Ambulatory measurements of trunk accelerations can provide valuable insight into the amount and quality of daily life activities. Such information has been used to create models that aim to identify individuals at high risk of falls. However, external validation of such prediction models is lacking, yet crucial for clinical implementation. We externally validated three previously described fall prediction models developed by van Schooten and colleagues (van Schooten et al., 2015a), which were based on 1) questionnaires, 2) daily life trunk accelerations, 3) a combination of both, in a new cohort of older people. Methods: Complete questionnaires and one week of trunk acceleration data were obtained in 263 community-dwelling people (mean age 71.8 years, 68.1% female). A 12-month fall record was obtained prospectively using fall diaries and monthly telephone contact, while for the consistency with original models, we only focus on 6 months of fall records. In addition, we tested if the accuracy would improve when we included only people who fell during walking. To validate models, we first used the coefficients and optimal cut-offs from original cohort, then recalibrated the original models, as well as optimized parameters based on our new cohort. Results: Among all participants, 37.6% experienced at least one fall in the past 6 months, and 39.9% experienced falls during 6-month follow-up. When using the original coefficients and cut-offs, all models showed moderate accuracy (0.50-0.75), poor precision (0.20-0.56), poor sensitivity (0.22-0.58), good specificity (0.59-0.89) and moderate model performance (AUC: 0.50-0.68). In addition, calibration of the original models did not improve model performance. Using coefficients and cut-offs optimized on the external cohort also did not improve results (accuracy: 0.53-0.66, precision: 0.27-0.53, sensitivity: 0.53-0.66, specificity: 0.53-0.66, AUC: 0.53-0.73). Last, the odds ratios in our cohort indicated that gait characteristics except for index of harmonicity ML did not improve fall prediction. Conclusions: Prediction of fall risk in our cohort was not as effective as in the original cohort. We found that all three models were overfitted, and recalibration as well as optimized model parameters resulted in limited increase in accuracy. Our results suggest that fall prediction models are highly specific to the cohort studied and highlight the need for large representative cohorts, preferably with an external validation cohort.

Background: Gait stability has been shown to be affected by age-related mobility problems, but exercise habits may reduce decline in gait stability. Our aim was to evaluate the variability and stability of feet and trunk between older healthy females and young females using inertial sensors. Method: 20 older females (OF; mean age 68.4, SD 4.1 years) and 18 young females (YF; mean age 22.3, SD 1.7 years) were asked to walk at their preferred speed, while kinematics were measured using inertial sensors on heels and lower back. Spatiotemporal parameters, acceleration characteristics and their variability, as well as trunk stability as assessed using the local divergence exponent (LDE), were calculated and compared between age groups with two-way ANOVA analyses. Results：Trunk-foot vertical acceleration attenuation, foot vertical acceleration maximum and amplitude, as well as their variability were significantly smaller in OF than in YF. In contrast, for trunk mediolateral acceleration amplitude, vertical acceleration maximum and amplitude, as well as their variability were significantly larger in OF than in YF. Moreover, OF showed lower stability (i.e. higher LDE values) in ML acceleration, ML and VT angular velocity on the trunk. Conclusion：These findings suggest that healthy older females had a lower maximum toe clearance so that were more likely to trip. Moreover, the acceleration of trunk was sensitive to the difference between healthy older and young females, both in variability and stability. Combined, although older adults had exercise habits, our metrics indicate that they were less stable, which may increase the risk of tripping and balance loss. Keywords: walking, aging, wearable system, motor control, balance