Exposure Records Preserved by Two Users of Japan's COVID-19 Contact-Confirming Application During the Seventh Wave: A Descriptive Study

In the digital landscape of public health, a quiet experiment once took place inside the smartphones of ordinary people. During the height of the pandemic, Japan deployed a tool designed to protect its citizens without compromising their privacy. This application, known as COCOA, did not track where people went or who they met. Instead, it relied on a simple, anonymous exchange of digital signals between nearby phones. When two devices came close enough to each other, they briefly exchanged a code. If one of those people later tested positive for the virus, the app could alert the other person that they had been in the vicinity of an infection, all while keeping the identities of both parties completely hidden. The system was designed to run silently in the background, logging these close encounters as "exposure windows." For a time, these digital logs lived on the phones of millions, a silent record of proximity that could help people make safer choices. However, like many temporary measures in a crisis, the system had an expiration date. When the application stopped functioning, the data stored on the devices was lost to the void, leaving researchers with a question: what exactly did those records show while the system was still alive?

A descriptive study published recently attempts to answer this question by looking at the preserved records of just two individuals who kept their data after the app ceased operations. The researchers focused on the seventh wave of the pandemic in Japan, a period that included a significant policy shift in late September 2022 when the government changed how cases were reported. The study follows the digital footprints of two distinct observers during this time. The first observer worked at a hospital in Tokyo and commuted almost entirely by private car, a routine that likely limited their contact with others. The second observer lived in Kyoto Prefecture and spent the vast majority of their time at home. By examining the files exported from their phones, the researchers could see exactly when these two people had been near someone who might have been infected, how long those encounters lasted, and how strong the signal between the phones had been.

The records revealed a pattern that mirrored the broader rise and fall of the pandemic's seventh wave. Between early April and late September, a period spanning 176 days, the Tokyo-based observer logged 358 potential exposure windows, while the Kyoto observer recorded 261. When the researchers looked at the weekly totals, they saw that the number of recorded encounters rose sharply in July and then declined in September, tracking closely with the daily count of new cases reported in each prefecture. The connection was strong enough that the ups and downs of the two observers' records moved in step with the national trends. However, when the researchers looked at the week-to-week changes, the connection became much weaker. The records did not react instantly to every small fluctuation in the number of new cases, suggesting that the app's logs captured the general pulse of the outbreak but missed the finer, day-to-day details.

A closer look at the nature of these encounters showed that most of them were likely low-risk. The app assigns a weight to each encounter based on how strong the signal was between the two phones, which serves as a proxy for distance and duration. The vast majority of the recorded windows for both observers fell into the category with the lowest weight, indicating that the phones were either far apart or the signal was too weak to suggest a close, prolonged interaction. For the Tokyo observer, 97.2 percent of the recorded windows were in this lowest-weight category, and for the Kyoto observer, the figure was 93.1 percent. This suggests that while the app was constantly detecting proximity, most of these detections were not the kind of close, sustained contact that typically leads to transmission.

The study also captured the immediate effect of a major policy change. On September 26, 2022, Japan revised its approach to mandatory case reporting, which altered the flow of data into the app. Following this change, the number of recorded windows dropped dramatically. The Tokyo observer saw their count fall from 45 encounters in the 27 days before the revision to just three in the 27 days after. The Kyoto observer, who had recorded 25 encounters in the same pre-revision period, recorded none at all in the 23 days following the change. This sharp decline indicates that the policy shift had a tangible impact on the data the app could collect, though the researchers note that they cannot separate this effect from other factors, such as changes in how diagnosis keys were supplied, how many people continued to use the app, or updates to the phone operating systems themselves.

Ultimately, this small-scale examination of two users' data serves as a reminder of the complexity involved in analyzing digital health records after the fact. The study does not claim to prove how effective the app was at preventing the spread of the virus, nor does it offer a definitive measure of its success. Instead, it highlights that the data left on a user's device is a product of many moving parts: the definitions used to count an exposure, the dates when data was retrieved, and the history of software and policy changes that occurred over time. The records preserved by these two users show that the app was indeed capturing the rhythm of the pandemic wave, but the story they tell is nuanced, shaped by the specific circumstances of the users and the shifting rules of the system itself.