Unconventional data, unprecedented insights: leveraging non-traditional data during a pandemic

Kaylin Bolt, Diana Gil‐González, Nuria OliverFrontiers in Public Health 2024 · DOI

This study explores how "unconventional" data—think GPS pings from your phone, Twitter sentiment, and self-reported coughs on an app—became a lifeline for public health officials in Spain and Italy during the COVID-19 pandemic. By interviewing the experts who sat at the intersection of data science and policy, the researchers identify the friction points that prevented these digital insights from being even more impactful and offer a roadmap for the next global health crisis.

TL;DR

During the COVID-19 pandemic, non-traditional data (mobility, social media, and participatory apps) filled critical gaps left by slow official statistics. However, the study finds that these tools remain underutilized due to fragmented data governance, decaying temporary infrastructure, and a lack of "data brokers" to bridge the gap between private tech and public health.

What Problem Does This Paper Address?

Traditional epidemiological data is robust but notoriously slow, often lagging by days or weeks. During a fast-moving pandemic, this delay is deadly. While "Big Data" from private sources (like telcos or social media) promised real-time insights, the transition from "cool research project" to "actionable policy" was rocky.

The authors tackle the "implementation gap": Why was it so hard to use this data? What were the specific barriers in two of the hardest-hit European countries? And most importantly, how do we stop the "innovation decay" that happens once the immediate crisis fades?

Methods

The researchers employed a qualitative semi-structured methodology. They conducted in-depth interviews with 18 multidisciplinary experts across Spain and Italy, including specialists in AI, data science, epidemiology, and high-level policymaking.

The data was analyzed using the framework analysis method, a systematic approach to coding and identifying themes across different informants. This allowed the team to compare experiences across different countries and professional backgrounds (e.g., comparing the perspective of a telco data scientist vs. a government health official).

Datasets

While this is a qualitative study about data use rather than a quantitative study using a specific dataset, it categorizes and evaluates the utility of three primary "unconventional" data types:

  • Mobility Data: Sourced from telecommunication companies (Call Detail Records) and location intelligence firms (SDK-based data).
  • Social Media/Sentiment Data: Activity (posts, likes, hashtags) from platforms like Twitter, Facebook, and Instagram.
  • Participatory Surveillance: Citizen-contributed symptom data from platform-facilitated apps.

Key Insights and Ideas

  • The "Gap-Filler" Role: Non-traditional data wasn't a replacement for official stats; it was a "complementary signal." It provided the "earlier signals" of viral spread when hospitals were too overwhelmed to report numbers accurately.
  • The "Science-to-Policy" Bridge: In regions like the Valencian Community in Spain, researchers established a direct daily feedback loop with politicians. This "unlikely collaboration" was cited as a transformative success, breaking down bureaucratic silos that usually take years to navigate.
  • The "Data Broker" Necessity: A recurring theme was the need for an institutionalized "data steward" or broker—a neutral entity that can handle the legal, ethical, and technical friction of moving data from private companies to public agencies.
  • Infrastructure Decay: Many of the "beautiful pipelines" built during the pandemic were held together by volunteerism and temporary contracts. Participants warned that these systems are now being unmaintained, meaning we might have to "start from scratch" when the next pandemic hits.

Key Results

  • Speed vs. Granularity: Non-traditional data excelled at providing rapid results but often lacked the demographic granularity (age, gender, socio-economic status) needed to address health inequities.
  • The "Black Box" Problem: Researchers expressed frustration over the lack of transparency in how private companies (like Facebook or telcos) pre-processed their data, which made it difficult to validate for scientific use.
  • Underutilization: Despite the success stories, the consensus was that these data sources remain "underutilized" in routine public health surveillance and program evaluation.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Bias: The participant pool was skewed toward men (12 vs. 6) and had a heavier representation of mobility data experts compared to social media researchers.
  • Geographic Scope: The findings are specific to the healthcare and political structures of Spain and Italy.
  • Lack of Quantitative Impact: The study does not quantify exactly how many lives were saved or how much money was saved by using this data; it focuses on the process and perception of utility.

Future Work: The authors call for the creation of permanent "data observatories" and the development of standardized data-sharing protocols (like an evolution of the GDPR) specifically for public health emergencies to ensure we aren't "building the plane while flying it" next time.

Summary updated

Abstract

While the utility of non-traditional data was demonstrated during the pandemic, opportunities exist to enhance its impact. Challenges reveal a need for data governance frameworks to guide practices and policies of use. Despite the perceived benefit of collaborations and improved data infrastructure, efforts are needed to strengthen and sustain them beyond the pandemic. Lessons from these findings can guide research institutions, multilateral organizations, governments, and public health authorities in optimizing the use of non-traditional data.
BibTeX
@inproceedings{bolt2024unconventional,
  author = {Kaylin Bolt and Diana Gil‐González and Nuria Oliver},
  title = {{Unconventional data, unprecedented insights: leveraging non-traditional data during a pandemic}},
  booktitle = {Frontiers in Public Health},
  year = {2024},
  doi = {10.3389/fpubh.2024.1350743},
}
Figure