The scoping review follows the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines and adopts the Arksey and O’Malley methodology framework (11).

The primary aim of this review was to examine the classification, application, and methodological, societal, and data-related challenges and opportunities of the empirical use of mobile phone data in public health. The core research questions are:

The search strategy was developed by WJ, XC, and ZC. We conducted a search across four major citation databases: MEDLINE, PubMed, ScienceDirect, and Web of Science, covering publications in English from January 1, 2012, to June 30, 2024. The search strategy combined keywords related to mobile phone data (“Cell phone” OR “Mobile Phone”) and mobility terms (“Mobility” OR “Flow”) along with the pandemic (“COVID” OR “Pandemic”). The search strategy was refined to ensure comprehensive coverage of relevant themes (see Appendix Table 1 for details).

Inclusion criteria of the review were: (1) studies focused on population mobility, human movement, or related patterns; (2) studies clearly described the applications of mobile phone data for illustrating or measuring mobility patterns; (3) studies provided empirical evidence in the context of the pandemic. Two reviewers (XC and WJ) independently conducted the blind review process, screening titles and abstracts to exclude studies that did not pertain to mobility, pandemics, or healthcare utilization, secondary studies (e.g., review articles), theoretical studies without empirical applications, and studies lacking details on data sources. The full texts were assessed by the two reviewers based on the metrics of the mobile phone data and the study outcomes. Discrepancies when there was uncertainty (44%^†) or inconsistent decisions (9%^†) were resolved by a senior reviewer (ZC).

Data from the included studies were charted by two reviewers (XC and WJ), who identified key characteristics, including study title, research theme, study region, study population, data source, mobility metrics, main outcomes, and study findings. According to scoping review guidelines, no studies were excluded based on quality, as the aim was to identify gaps in the literature rather than assess methodological rigor (11).

A PRISMA flowchart (Figure 1) was developed to track the selection process. The findings were synthesized to identify key patterns and relationships across studies, and a Sankey diagram was used to visualize the distribution of research themes, mobility metrics, and outcome domains.

A total of 921 records were retrieved from database searches and imported into a local citation database via Zotero 6.0. After de-duplication, it yielded a list of 427 citations. Following an initial screening of titles and abstracts, 133 citations were excluded due to irrelevance. Among the remaining 294 full-text articles, 210 were excluded because of non-public health outcomes (n = 84) or incorrect data type (n = 126; specific reasons for exclusion are presented in Figure 1). Ultimately, 84 articles met the inclusion criteria and were included in the final review. A narrative and numerical summary of study characteristics was presented and tabulated in the Appendix Table 2. Four key themes with important implications emerged from our review of mobile phone data applications in pandemic control-related research.

A total of 25 studies (30%) focused on analyzing mobility patterns using mobile phone data. Jia and colleagues were the first to use real-time mobile phone data to assess the spatio-temporal dynamics of the COVID-19 spread in China (12). Using national aggregated mobile operator data, Lu et al. established the correlation between population mobility and disease transmission (1). Luo et al. analyzed the mobility patterns among different populations and how they influence the risks of disease transmission (13). Aside from China, an extensive list of research used mobile phone signaling data in public health research exists, analyzing the associations between human mobility and coronavirus spread in the US (14–19), Latin America (20), Ireland (4), Spain (6, 21), Brazil (22), Ecuador (23), Japan (24), Finland (25), and globally (26). However, in the later stages of the pandemic, the association between mobility metrics derived from mobile phone data and COVID-19 incidence growth rates gradually weakened following the lifting of initial stay-at-home orders (27). Other studies explored the socioeconomic impacts of mobility changes, linking mobility data with demographic characteristics (21, 28), economic distress (16), and health outcome indicators (29).

Thirty-six studies (43%) evaluated the effectiveness of multiple NPIs using mobile phone data, which were adapted as primary measures to restrain the spread of COVID-19 globally before the introduction of effective vaccinations. Many studies assessed social distancing (30–39) and lockdown policies (40–46) across different countries, highlighting their role in mitigating the spread of COVID-19. Research also showed that less aggressive interventions, such as remote working and closures of non-essential businesses, had varied impacts depending on the country (47–54). Notably, a study covering 135 countries found significant reductions in mobility due to travel restrictions during the first wave of the pandemic (55); however, other studies emphasized spontaneous reductions in mobility that occurred regardless of government actions and a ‘floor’ phenomenon (19).

Incorporating mobility parameters—either tuned or estimated from mobile phone data—into epidemiological models enhances their ability to simulate outbreak dynamics under varying conditions. A total of 23 studies (27%) integrated mobile phone data into epidemiological models to predict disease spread. These models helped simulate outbreak trajectories in various regions, including Chinese cities and worldwide (56–69). A notable study combined mobile phone data with genomics to track coronavirus variants in Bangladesh, revealing how large-scale human migration from urban to rural areas influenced viral diversity (70). Social factors, such as sociodemographic characteristics, were also examined, showing how mobility differences contributed to infection rates in disadvantaged groups (56, 71). Other researchers highlighted that socioeconomic factors, such as education, household size, and the proportion of the Latinx population, have consistent positive relationships with COVID-19 prevalence over time (72).

A sheer body of studies highlighted the potential of mobile phone data to analyze health inequality issues. Research on gender-specific mobility patterns (73), age-group mobility (74–76), and socioeconomic status (SES) (77–79) revealed disparities in COVID-19 transmission risks. Studies also showed that NPIs affected populations and communities differently based on SES, influencing the mobility responses and, consequently, infection rates (44, 80, 81). For instance, using mobility measures derived from mobile phone data, Carranza et al. showed that the impact of NPIs on mobility varied widely across communities, depending on their socioeconomic levels, which also contributes to disparities in infection rates between high- and low-income areas (53).

Mobile phone data enable the measurement of human mobility through diverse metrics, supporting multidisciplinary research. As illustrated in the Sankey diagram (Figure 2), we summarize mobility metrics into four major categories. Population flow metrics, such as origin–destination flow indicators, can track the movement of infected individuals across regions, revealing transmission pathways (1, 6, 62). Population density metrics, including the number of visits or hourly population density to POIs, help evaluate the implementation and effectiveness of pandemic control policies across different population groups (56, 82). Geographical movement indicators, such as the radius of gyration or entropy of movement, can be used to refine epidemiological models and simulate disease transmission (9, 59). Time- or distance-related metrics—such as duration spent at specific locations, distance traveled, travel frequency, and activity space—capture behavioral response to public health interventions (31, 40, 48). Additional information is presented in Appendix Table 3.

However, the diversity of mobility metrics also introduces challenges. Comparisons between studies become complicated, and the predictive power of different mobility indicators in empirical models varied substantially across regions and time period (15, 83). The reliability of mobility metrics derived from mobile phone data as proxies for disease transmission is inconsistent, temporally and geographically. For instance, the correlation between alternative mobility metrics and the effective reproductive number of the coronavirus fluctuated during the early stages of COVID-19 (59). These findings underscore the importance for researchers of carefully selecting appropriate mobility indicators and ensuring their robustness when interpreting results or informing health policymaking.

The procedures and algorithms used to collect and process mobile phone raw data are often opaque to public health and social science researchers. Many studies do not provide a detailed description of how the raw data is generated. For example, the frequency of mobile signal pings used to define a device’s location (e.g., every 5 min versus every 10 min) can significantly impact mobility estimates and the inferred effect of interventions (59). Data generation is also influenced by factors such as cell tower distribution and the types of mobile applications used, which can vary from urban to rural areas and across countries (35). This complexity and lack of transparency in data generation hinder the ability to link mobility data to human behavior (59), which may reduce the reliability and trustworthiness of subsequent analyses.

As the large-scale use of mobile phone data becomes more common, protecting individual privacy remains a central concern (84). To minimize privacy risks, data providers often aggregate and preprocess data through methods such as statistical thresholds, differential privacy, and appropriate security controls (35), providing researchers with spatially and temporally aggregated datasets. While these approaches safeguard privacy, they reduce data granularity and can introduce uncertainties in data quality (7), which may limit the comparability and generalizability of results across studies.

Mobile phone data enable rapid, large-scale, and context-specific data collection (85), but representativeness requires careful consideration. Representativeness of mobile phone data depends on market share, user demographics, and geographic coverage. Data from a single operator or regions with uneven network coverage may systematically underrepresent children, older adults, and rural or low-income populations, limiting the generalizability of results and potentially biasing conclusions about behavioral responses to NPIs (84–86). During the COVID-19 pandemic, mobility reductions were found to be smaller in low-income communities due to occupational and structural constraints, suggesting that operator coverage biases may distort inferences on health equity and assessments of policy effectiveness (53, 80). To mitigate these issues, recent studies have recommended corrective strategies such as reweighting using census covariates, small-area estimation to improve spatial representativeness, and integrating multiple mobility and demographic data sources (14, 84, 85, 87). Strengthening such representativeness adjustments is crucial to ensure that mobility-derived evidence makes a meaningful contribution to equity-oriented public health research.

Selecting the appropriate spatial and temporal scales is critical for striking a balance between privacy protection and analytical utility. Spatial units may include grids or an administrative area, which should reflect population density differences between urban and rural regions. Temporal scales should align with study designs and research objectives; for example, daily measures can better indicate short-term disease transmission, whereas weekly measures can better reflect seasonal or migration patterns (35). Flexible aggregation ensures mobility metrics are informative, actionable, and ethically responsible.

A major challenge is the limited understanding of how these data are collected, processed, and transformed into usable metrics. Current preprocessing methods for mobile phone mobility data often operate as black boxes for researchers, which may lead to biases in analyses with ambiguous directions (84). Researchers must critically examine the assumptions underlying default conditions and thoroughly understand the critical decision rules (35), such as the spatial boundaries of POI, the minimum interactions required to register activity, and the criteria defining “stay” or “pass-by” actions. Transparent reporting of these technical details is crucial for replicating, validating, and scaling findings across academia, industry, and policy contexts. A list of recommended reporting items is in Appendix Table 4.

Current uses of mobile phone datasets are typically anonymized and aggregated to preserve privacy. Although useful for broad mobility trends, such aggregates may be insufficient for targeted public health responses, especially in the post-pandemic era, where sustained surveillance is required (84, 85). In addition, there is a growing need to access finer, sub-population-level mobility data to capture the heterogeneous transmission patterns and disproportionate epidemic burdens, as our review identified. Therefore, there is a need to bolster open discussions and collaborations among mobile operators, policymakers, and researchers to develop frameworks that ensure privacy while enabling legally compliant extraction of detailed, actionable mobility information.

A new and growing body of literature explores the integration of mobile phone data with complementary datasets, such as census demographics (20, 21, 79), social media (33, 82), internet search volume (16), or genomics (70). With its extensive coverage and spatiotemporal scale, integrating mobile phone data with other data sources can overcome user-based selection bias, provide a more comprehensive picture of population activity (84), and enhance understanding of the profound social and health equity implications. Linking mobility data with socioeconomic and behavioral information at an appropriate geographic scale (e.g., census block group level) enables a more comprehensive view of transmission patterns, disparities, and intervention effectiveness (7), supporting more effective and informed public health decision-making.