JMIR Publications

ABSTRACT

Background:

China's anti-pandemic traveling record card — a smartphone-delivered mHealth intervention — tracked individual city-level travel histories and imposed movement restrictions on those who had visited medium- or high-risk areas during the COVID-19 pandemic. A fundamental design question underpins this system: should geographic risk areas be defined by administrative city boundaries ("city-cut"), or by a fixed radius around a confirmed case location ("radius-cut")? Despite the scale of this mHealth deployment, the empirical basis for choosing city-cut over radius-cut has never been systematically examined.

Objective:

This study aims to empirically evaluate two competing geographic containment design principles — city-cut versus radius-cut — for mHealth-based pandemic travel restriction systems. Specifically, we tested two hypotheses: (1) intra-city movements are significantly more frequent than inter-city movements across all districts; and (2) intra-city long-distance movements are significantly more frequent than inter-city movements. A secondary aim was to assess whether clustering of districts could support a finer district-cut alternative to the current city-wide approach.

Methods:

We obtained proprietary data from China Unicom, a major Chinese mobile service provider, on real-time locations of individual smartphones to restore daily movements of individuals in Shanghai. The data covered periods from October 10 to November 24, 2019 (pre-pandemic baseline) and October 10 to November 24, 2020 (post-pandemic recovery). Shanghai was divided into 2,213 grids of 2 km × 2 km. To test Hypothesis 1, intra-city versus inter-city movement proportions were calculated. To test Hypothesis 2, intra-city-long-distance movements (between nonadjacent districts) and intra-city-to-center movements (toward central districts) were calculated. To assess the secondary aim, clustering using a k-means partition of the vector of the Markov transition matrix was applied to evaluate whether a district-cut alternative is supported by observed movement patterns.

Results:

Both hypotheses were supported across all 16 districts. Intra-city movements averaged 97.2% versus 2.8% inter-city (Hypothesis 1), including in city border areas, confirming that even residents nearest to neighboring cities predominantly move within Shanghai. Intra-city-long-distance movements (average 13.7%) significantly exceeded inter-city movements in every district (P≤.01), and intra-city-to-center movements (average 12.6%) indicated extensive population mixing at city centers (Hypothesis 2). No spectral clustering partition aligned with administrative boundaries; cluster regions frequently crossed district lines and did not correspond to any administrative units or their combinations, indicating that a district-cut alternative is not empirically supported.

Conclusions:

The strong internal connectivity and compactness of Shanghai’s population movements justify the city-cut design of the anti-pandemic traveling record card system. These findings demonstrate how large-scale mobile phone data can inform the geographic calibration of mHealth-based public health interventions, offering a replicable methodology for evidence-based design of location-sensitive mHealth systems in other densely populated urban settings.