Outdoor Secondhand Smoke Exposure in a Public Smoking Area: Formative Field Study Using Passive Wi-Fi Packet Sensing

Outdoor secondhand smoke (SHS) remains a public health concern, particularly around designated outdoor smoking areas where nonsmokers may pass through or remain nearby. Although prior studies have quantified outdoor SHS concentrations, fewer have examined how many people may be present within a plausible exposure setting. Estimating exposure opportunity level requires methods that are feasible, scalable, and minimally intrusive.

This study aimed to evaluate the feasibility of using passive Wi-Fi packet sensing, calibrated with brief onsite observation, to estimate the number of smokers and passersby within a plausible SHS exposure range at a public outdoor smoking area in Japan.

We conducted a formative field study at a designated outdoor smoking area of the Asia Pacific Trade Center in Osaka, Japan. A passive Wi-Fi packet sensor collected timestamps, anonymized device identifiers, organizationally unique identifiers, and received signal strength indicator values from October 13 to 29, 2023. The main analysis focused on October 28, 2023, a high football event day selected for direct calibration. Episodes were classified using empirically derived received signal strength indicator thresholds and class specific calibration ratios were applied to estimate day level counts.

Of 128,313 anonymized detections recorded on October 28, 115,950 occurred during business hours. Among these, 11,068 identifiers were detected more than once, dwell time could be calculated for 1,817 identifiers, and 659 eligible presence episodes remained after preprocessing. During a 30 minutes validation window, smokers and passersby were counted manually within a 25 m radius. During the validation window, 6,230 signal records formed 104 stays, with a mean stay duration of 9.89 minutes (SD 7.89). During the validation window, direct observation recorded 14 smokers and 207 passersby within the 25 m radius. Applying the rule based classification and calibration ratios to business hours data yielded estimated day totals of 262 smokers and 3,907 passersby within the plausible SHS exposure range. Estimated smoker counts showed 2 peaks, around 12:00 and 16:00, whereas passerby volume peaked around midday. In an exploratory analysis, a random forest model using stay duration, mean received signal strength indicator, and received signal strength indicator variability achieved an accuracy of 0.95, sensitivity of 0.75, specificity of 0.97, and area under the receiver operating characteristic curve of 0.99.

This formative field study suggests that passive Wi-Fi packet sensing, combined with brief on site observation, can be used to estimate population level exposure opportunity around an outdoor smoking area. The　method identified substantial numbers of potentially exposed passersby in a high footfall public setting. Although the findings are site specific and preliminary, they indicate that exposure count metrics may complement concentration based and survey based SHS research. Further studies incorporating repeated validation, direct pollutant monitoring, and multiple sites are needed to refine the method and strengthen its usefulness for tobacco control and public health decision making.