In the present study, the following four cities located in different parts of China were selected as research sites: Changchun (in northeast China), Lanzhou (in northwest China), Zhuzhou (in central China), and Luzhou (in southwest China). A crossroad near a residential area was randomly selected in each city.

The study was conducted from 19 August 2020 to 9 September 2020, in the four selected cities. The observation time was from Monday to Sunday, 07:30–09:00 in the morning and 17:00–18:30 in the afternoon. Days with extremely bad weather such as heavy rain, hail, and gale were excluded.

By drawing on and slightly modifying the existing research [2, 21], this study divided the behaviors of pedestrian smartphone use into the following six categories, from low to high use (see Supplementary Figure S2):

1) Not observed: The pedestrians did not openly carry or use smartphones;

The pedestrians with two established behaviors (or mixed categories) simultaneously were placed in the higher use category. For example, pedestrians wearing earphones and holding smartphones simultaneously were classified as “holding”.

The researchers registered the gender and approximate age of the pedestrians based on perceived appearance. Gender was divided into the following two categories: male and female. The weather was divided into the following three categories: sunny, cloudy, and rainy. The time of day was divided into morning and afternoon. The time of the week was divided into weekdays and weekends. For the age of the pedestrians, five classes were used as follows:

1) Under 10 years old: children aged 4–10;

To study the effect of the intervention on preventing smartphone use while crossing the road, we intervened with pedestrians in the last 11 days. It should be pointed out here that the pedestrians who were study subjects on the 11 days of intervention did not necessarily coincide with those observed during the first 11 days. Considering smartphone distraction, this study proposed a combined visual and auditory measure to prevent pedestrians from using smartphones while crossing the road.

1) Safety warning signs

The “no smartphone” sign was placed at the intersections to remind pedestrians not to use smartphones while crossing the road. The safety warning sign is shown in Supplementary Figure S3.

2) Sound alerts

A broadcasting device was placed near the safety sign, repeating “do not use smartphones while crossing the road” (100 dB, loop playback) in Chinese, to remind pedestrians in an auditory way.

In this study, we had four researchers, one in each city, record the circumstances surrounding pedestrians using smartphones while crossing a road. We recorded smombie gender and age, the time and weather, and the types of smartphone use. Only the pedestrians who passed the preset zebra crossings and crossed the road in a specific direction were recorded. Pedestrians walking alone or in a group of four people or less were observed directly, while pedestrians in a group of more than four were videoed [6, 22] because of potential counting errors [1].

Researchers quickly registered the digital codes while observing. Male pedestrians were marked as 1, and female pedestrians were marked as 2. The different weather conditions were classified as 1, 2, and 3; the age groups were divided into 1, 2, 3, 4, and 5; and the different types of smartphone users were recorded as 1, 2, 3, 4, 5, and 6. For example, for a 23-year-old man wearing earphones while crossing a road on a sunny day, the code was 1231. In addition, the morning was marked as 1, while the afternoon was marked as 2; the weekday was marked as 1, while the weekend was marked as 2.

In pre-observation, Kendall’s W test was used to check the consistency of results from the four different researchers. The results verified the consistency among researchers (W > 0.7, p < 0.001). This process was performed by a supervisor other than the four researchers.

All procedures of this study were approved by the ethics committee of the affiliated hospital of Chengdu University of Traditional Chinese Medicine (ethical approval number 2016 KL-005), in accordance with the Declaration of Helsinki.

Previous studies have shown that the proportion of pedestrian smartphone use is approximately 30% [5, 6]. If δ≤ 6%, α = 0.05, the required sample size calculated by PASS software (version 15.0) is 2543. The formula is as follows: n = U α 2 p ( 1 − p ) δ 2 .

All statistical analyses were performed with SPSS (version 25.0) software. All variables were considered categorical (such as age, divided into five age groups). Bivariate and multivariate logistic regression analyses were used to analyze the impacts of the influencing factors on pedestrian smartphone use (two categories/multiple categories) and the effect of the combined visual and auditory intervention. In addition, chi-square tests were used to evaluate the significant differences with and without intervention.

A total of 25,860 pedestrians (13,086 without intervention and 12,774 with intervention) were observed in this study. The detailed descriptive statistical results of this study are shown in Table 1.

By analyzing the usage rate of smartphones without intervention, it was found that women were more likely to use smartphones while crossing roads (χ ² = 120.905, p < 0.001); smartphone use while crossing roads was most prevalent among 11 to 25-year-old people, followed by young people aged 24–44 (See Supplementary Table S1); smartphone use while crossing roads was less prevalent on rainy days in terms of weather (See Supplementary Table S2); smartphone use was more prevalent in the afternoon than in the morning (χ ² = 41.190, p < 0.001), and was more prevalent on weekdays than on weekends (χ ² = 16.218, p < 0.001). See Table 2.

Further analysis of different types of pedestrian smartphone use revealed the distribution of different conditions. The “holding” rates were higher in women than in men (χ ² = 157.348, p < 0.001). Teenagers had the highest “earphone”, “holding”, “slight smombie”, and “severe smombie” rates among all age groups, and the “talking” rates in both the teenage and young people groups were much higher than those in the other groups (See Supplementary Table S3). On rainy days, the rates of “holding” and “severe smombie” were the lowest (See Supplementary Table S4). “Talking” (χ ² = 10.544, p < 0.01) and “severe smombie” (χ ² = 19.505, p < 0.001) were more prevalent in the afternoon than in the morning. See Table 3.

There were 4,289 pedestrians who used smartphones in the first 11 days of the study without intervention, accounting for 32.8% of the observed people. In the next 11 days, with the visual and auditory intervention, 3,579 pedestrians were found to have used smartphones, accounting for 28.1% of the observed people. Through chi-square tests (χ ² = 69.120, p < 0.001), smartphone usage rates with visual and auditory intervention were statistically lower than without intervention. See Supplementary Table S5.

The results on the effect of visual and auditory intervention on six categories showed that in comparison to those without intervention, the pedestrians with the intervention were more resistant to smartphones while crossing the road. The rates of “holding” (χ ² = 49.228, p < 0.001), “slight smombie” (χ ² = 9.047, p < 0.01), and “severe smombie” (χ ² = 153.297, p < 0.001) decreased significantly. See Table 4.

The distribution of smartphone usage with intervention showed that usage rates were decreased both in women (χ ² = 37.449, p < 0.001) and men (χ ² = 33.414, p < 0.001) in comparison to those without intervention. The smartphone usage rates were decreased in all age groups (<10 years: χ ² = 21.333, p < 0.001; 11–25 years: χ ² = 33.758, p < 0.001; 45–59 years: χ ² = 7.565, p < 0.001; >60 years: χ ² = 5.629, p < 0.05), except for young people aged 26–45 (χ ² = 0.253, p = 0.615). Smartphone usage rates were also found to decrease on sunny days (χ ² = 66.992, p < 0.001). The usage rates with intervention in the morning (χ ² = 17.524, p < 0.001) and afternoon (χ ² = 49.818, p < 0.001) also decreased, as well as those on weekdays (χ ² = 46.707, p < 0.001) and weekends (χ ² = 46.707, p < 0.001). See Figure 1.

The “severe smombie” rates and the “holding” rates with the intervention were decreased in almost all conditions. The “slight smombie” rates were decreased both in women (χ ² = 4.914, p < 0.05) and men (χ ² = 4.144, p < 0.05), young people (χ ² = 7.208, p < 0.01), on sunny (χ ² = 10.901, p < 0.01) and rainy days (χ ² = 6.045, p < 0.05), in the afternoon (χ ² = 12.126, p < 0.001), and on weekdays (χ ² = 4.751, p < 0.05) and weekends (χ ² = 5.038, p < 0.05). The “talking” rates showed no significant change in all conditions. See Figure 2.

With visual and auditory intervention, the smartphone usage rate decreased to 28.1%, which was significantly different from that without intervention. Specifically, the usage rates on “holding”, “slight smombie”, and “severe smombie” were significantly decreased compared to without intervention. Further analysis showed that our intervention was effective regardless of gender, age (except for young people), and time of day and week, suggesting that the combined visual and auditory intervention was generally available.

Through logistic regression analysis with and without intervention, we further confirmed the effectiveness of visual and auditory intervention in reducing smartphone use behavior while crossing roads. Because of distracted concerns, a single visual or auditory stimulus might be ignored by smombies [18, 19]. Therefore, we used a combined visual and auditory intervention, and the results demonstrated the preventive effect of visual and auditory intervention on “holding,” “slight smombie,” and “severe smombie.”

It is important to note that there was no significant difference in the category of “talking” with and without intervention. This might be related to the instantaneous nature of talking [5, 32]. Even when they were reminded of the danger of talking while crossing the road, the pedestrians continued ignoring their own safety to satisfy their needs for instant communication. This might be why our intervention was most unsuccessful in young people, who had the highest “talking” rate.

Unexpectedly, the usage of “earphones” by pedestrians increased rather than decreased with intervention. We thought that might be related to the intervention we performed. In pedestrians’ opinion, wearing earphones was relatively safe [20]. Thus, after being reminded not to use smartphones, the pedestrians put on earphones instead of holding phones in their hand.

To the best of our knowledge, this study is the first large-sample, multicenter observation study on smartphone use while crossing roads. This study could reflect the current status of pedestrian smartphone use in China. Bivariate and multivariate logistic regression analyses were used to explore the influencing factors, such as gender, age, weather, and time. Moreover, visual and auditory intervention was used to prevent this behavior, and demonstrated an effect in reducing the use of smartphones while crossing the road.

There are also some limitations to our study. First due to the potential bias in estimating some variables (such as age), the results might have been influenced, even though we converted the continuous variables to categorical variables. Second, there was only one observer in each city, and the results might be biased. Moreover, this study did not conduct a statistical analysis of the regional effects. Then, there was no unified standard for the classification of smartphone types, and the devices that earphones were connecting to were unknown. In addition, the findings might be influenced by the seasons we conducted this study in. Moreover, the exclusion of pedestrians wearing watches in this study might not be easily distinguishable for observers, and therefore, the actual percentage of smartphone use might have been higher. In addition, the visual and auditory intervention was proven effective in the short term, but the long-term effects were not studied. Last, although visual and auditory intervention effectively reduced smartphone usage, more than a quarter of the pedestrians still used their smartphones while crossing the road. Thus, other interventions should be considered in the future, such as pedestrian bridges or dedicated “smartphone sidewalks”.

In summary, through this observational study, we found that smartphone use while crossing roads is common in four cities in China. Gender, age, weather, and time were the main factors influencing pedestrian smartphone use while crossing roads. Moreover, our findings also verified the effect of the combined visual and auditory intervention on smartphone use. To our knowledge, this study is the first large-sample, multicenter observational study on pedestrian smartphone use and the intervention for it, which might provide a reference for future research.