A cross-sectional study was applied.

Participants were recruited from January 2021 to May 2022 at the Department of Rehabilitation Medicine of the First Affiliated Hospital of Jinan University, Guangzhou, China. Participants were primarily outpatient follow-up patients and hospital attendants. Recruitment announcements were posted on the department bulletin boards and outpatient department. Participants of interest were screened according to inclusion and exclusion criteria, and those who met the requirements were invited to participate in the study.

Inclusion criteria: (a) age 65 or older; (b) capable of independent walking for 3 min without assistance; (c) clear consciousness, able to cooperate with the assessment, Mini-mental State Examination score > 24; (d) with available informed consent. Exclusion criteria: (a) Those with foot injuries, deformities and other conditions that temporarily affect their daily walking; (b) People receiving trunk or lower limb therapy that affect lower extremity biomechanics; (c) patients with serious or unstable cardiac, pulmonary, renal and other medical diseases who can not tolerate the study; (d) patients with a history of mania, delirium and other psychiatric disorders who cannot cooperate to complete the test.

According to previous study [15] and pre-experiment results, it is expected that the combined sample standard deviation σ is 1.36, and the difference between the two groups' means δ is 1.1. Bilateral α = 0.05 is set, and power (1-β) is 80%. The sample size was calculated according to the formula (1), and n = 24 were obtained. So each group requires 24 samples.

$$n=\frac_+_\right)}^*^}^}$$

(1)

Finally, forty-eight elderly people were selected for general data collection and Berg Balance Scale (BBS) assessment, of whom 24 BBS scored ≤ 40 for the high fall risk group (HR) and the other 24 BBS scored > 40 for the low fall risk group (LR). The general data includes sex, age, height, body weight and body mass index.

The plantar pressure was detected using a wearable intelligent footwear system developed in cooperation with the Human Data Science Engineering Center of South China University of Technology and Zhongshan Super Sense Technology Co [19, 20]. As shown in Fig. 1, there are eight pressure sensors at different points in the insole of each foot. The pressure sensor has the characteristics of a short response time, large range, high sensitivity, and high wear resistance [17]. Wirelessly connected mobile phones can receive real-time datums collected by the footwear system. And the phone APP sets different colours according to the pressure grading, indicating the dynamic changes of plantar pressure, as shown in Fig. 2. It was confirmed that satisfactory accuracy, repeatability and wearing comfort was showed by this intelligent footwear system [18].

Position of eight pressure sensors. The right foot is shown as an example

Different colours according to the pressure grading. During the test, the color will change dynamically with the pressure

During the experiment, the researcher provided uniform cotton socks and the participants were asked to choose the right size of socks and intelligent shoes to ensure that their feet would not slide in the shoes while walking. Before the experiment began, the participants wore cotton socks and intelligent shoes for 1-2 min to adapt. During the formal experiment, participants need to walk for more than two minutes along a 20-m corridor at their normal gait and usual walking speed. The experiments were supervised by one investigator, with no physical contact or verbal induction. Each experiment was supervised by the same investigator. And participants went through the whole process in one day.

Based on previous studies combined with plantar mechanics, the plantar area was divided into 8 regions for analysis: 1st toe (T1), 1st metatarsal head (M1), 2nd-3rd metatarsal head (M2-3), 4th-5th metatarsal head (M4-5), medial midfoot (MMF), lateral midfoot (LMF), medial heel (MH), and lateral heel (LH), as shown in Fig. 3.

Eight regions for analysis. These 8 regions correspond to the position of the pressure sensor one by one

The raw plantar pressure data is exported from the smart terminal mobile APP background. After weight normalization and identification of valid gait cycles, for each foot region, we calculate the following parameters: peak pressure (PP), pressure-time integral (PTI), pressure gradient (maximum pressure gradient (MaxPG), minimum pressure gradient (MinPG), full width at half maximum (FWHM)), and average pressure (AP). In this study, we used the characteristic calculation method of Botros [21] and Dongran Wang [17] et al.

For each plantar region i, where i = 1, 2,…, 16 and i ∈  [1-8] for the left foot and i ∈  [9-16] for the right foot, Pi(m) is the pressure datums for each sample suiting to the region, where m = 1,2,…, M. M is the length of the sample data for one valid gait cycle. Then for each Pi(m), calculate 12 feature Fri and perform the average of individual effective gait cycles, where r = 1, 2,…, 12.

Taking the left foot as an example, with r ∈  [1-6], the 1st parameter PP can be calculated by Eq. (2).

$$}_}_\in \left[1,\mathrm\right]}^}}_}\left(\mathrm\right)_}$$

(2)

The 2nd parameter PTI can be calculated by Eqs. (3).

$$}_}= _=1}^-1}}_}\left(\mathrm\right)_}+}_}(\mathrm+1)_}).\Delta \mathrm/2$$

(3)

The 3rd parameter MaxPG and the 4th parameter MinPG are calculated by Eqs. (4) and (5), respectively.

$$}_}_\in \left[1,\mathrm\right]}^}}_}\left(\mathrm\right)_}_\in \left[1,\mathrm\right]}^}}_}\left(\mathrm\right)_}}_}\left(\mathrm-1\right)_}]/\Delta \mathrm$$

(4)

$$}_}_\in \left[1,\mathrm\right]}^}}_}(\mathrm)_}_\in \left[1,\mathrm\right]}^}}_}\left(\mathrm\right)_}}_}(\mathrm-1)_}]/\Delta \mathrm$$

(5)

The 5th parameter FWHM can be calculated by Eqs. (6) and (7), where mi1 and mi2 denote the longness of the sample datums when the pressure value is half of the PP.

$$}_}}_2}_}=}_}}_1}_}=0.5\times }_}$$

(7)

The sixth parameter AP can be calculated by Eq. (8).

$$\overline }_}}=\frac}\sum_=1}^}}_}\left(\mathrm\right)_}$$

(8)

Except for parameter calculations, all data were processed by IBM SPSS Statistics for Windows, Version 27.0 (IBM Corp, Armonk, NY, USA). Data were expressed as mean ± SD or median (Q1, Q3), as appropriate. Also, group comparison was made using two independent sample t-test or Mann-Whitney U test. Besides, paired-sample t-test and Wilcoxon test were made for within-group comparisons. The significance level was set at α = 0.05.

Approval was granted by the Medical Ethics Committee of the First Affiliated Hospital of Jinan University on December 23, 2020 (KY-2020-087). All participants were informed of the study purpose, procedure, anonymity and confidentiality of participation, and received written informed consent.