Conjunction of Factors Triggering Waves of Seasonal Influenza

Ishanu Chattopadhyay, Emre Kıcıman, Joshua W. Elliott, Jeffrey Shaman, Andrey Rzhetsky

Abstract

Understanding the subtle confluence of factors triggering pan-continental, seasonal epidemics of influenza-like illness is an extremely important problem, with the potential to save tens of thousands of lives and billions of dollars every year in the US alone. Beginning with several large, longitudinal datasets on putative factors and clinical data on the disease and health status of over 150 million human subjects observed over a decade, we investigated the source and the mechanistic triggers of epidemics. Our analysis included insurance claims for a significant cross-section of the US population in the past decade, human movement patterns inferred from billions of tweets, whole-US weekly weather data covering the same time span as the medical records, data on vaccination coverage over the same period, and sequence variations of key viral proteins. We also explicitly accounted for the spatio-temporal auto-correlations of infectious waves, and a host of socioeconomic and demographic factors. We carried out multiple orthogonal statistical analyses on these diverse, large geo-temporal datasets to bolster and corroborate our findings. We conclude that the initiation of a pan-continental influenza wave emerges from the simultaneous realization of a complex set of conditions, the strongest predictor groups are as follows, ranked by importance: (1) the host population’s socio- and ethno-demographic properties; (2) weather variables pertaining to relevant area specific humidity, temperature, and solar radiation; (3) the virus’ antigenic drift over time; (4) the host population’s land-based travel habits, and; (5) the spatio-temporal dynamics’ immediate history, as reflected in the influenza wave autocorrelation. The models we infer are demonstrably predictive (area under the Receiver Operating Characteristic curve ≈ 80%) when tested with out-of-sample data, opening the door to the potential formulation of new population-level intervention and mitigation policies.

Introduction

Predicting the onset and trajectory of seasonal influenza epidemics remains a critical challenge for public health. While many studies focus on isolated variables like humidity or school schedules, real-world disease dynamics are driven by a complex interplay of environmental, social, and viral factors. This research moves beyond the ceteris paribus approach to identify the "mechanistic triggers" of pan-continental epidemics by performing an integrative joint analysis of dozens of variables simultaneously.

Methodology and Data Validation

The study leverages an unprecedented scale of data, including clinical records for over 150 million individuals, movement patterns from billions of tweets, high-resolution weather data, and viral genetic sequences. To ensure the reliability of the findings, the researchers employed three orthogonal statistical approaches: Non-parametric Granger Causality, Mixed-effect Poisson Regression, and County-matching Analysis.

Figure 1: Analysis framework and data validation. Panel A outlines the three statistical approaches used. Panel B validates the Truven clinical data against CDC prevalence records (ρ = 0.91). Panel C shows the consistency between Granger causality and mixed-effect model predictions.

The use of multiple independent methods allowed the team to map "causality streamlines," identifying the Southern and Southeastern US as the primary "sources" or cradles of national epidemics.

The "Perfect Storm" of Influenza Initiation

The research identifies a "perfect storm" of conditions required to spark a national wave. Rather than just "cold weather," the trigger often involves a period of high average humidity and temperature followed by a sharp, stressful drop in both. This environmental shift, when occurring in regions with specific socio-demographic profiles, creates the ideal conditions for a local outbreak to transition into a pan-continental wave.

Key Results and Factor Ranking

The predictive models achieved an Area Under the ROC Curve (AUC) of approximately 80% in out-of-sample tests. The strongest predictors of influenza onset, ranked by importance, are:

  1. Socio-demographics: Host population properties were the dominant predictor.
  2. Weather: Specific humidity, temperature, and solar radiation.
  3. Antigenic Drift: Viral evolution over time.
  4. Land Travel: Short-range commuting was found to be more influential than long-distance flights in propagating the wave.
  5. Autocorrelation: The immediate spatio-temporal history of the wave.
Figure 2: Relationship between viral M2 protein diversity and vaccination coverage. Diversity in the M2 protein was found to enhance the epidemic trigger, whereas diversity in other proteins like HA or NA appeared to inhibit it.

Conclusion

By identifying the Southern "trigger zones" and the specific conjunction of factors that lead to epidemic initiation, this study opens the door to more targeted public health interventions. Proactive mitigation strategies focused on these source regions could potentially disrupt the propagation of influenza waves before they reach a national scale.

Summary updated

Abstract

Abstract Understanding the subtle confluence of factors triggering pan-continental, seasonal epidemics of influenza-like illness is an extremely important problem, with the potential to save tens of thousands of lives and billions of dollars every year in the US alone. Beginning with several large, longitudinal datasets on putative factors and clinical data on the disease and health status of over 150 million human subjects observed over a decade, we investigated the source and the mechanistic triggers of epidemics. Our analysis included insurance claims for a significant cross-section of the US population in the past decade, human movement patterns inferred from billions of tweets, whole-US weekly weather data covering the same time span as the medical records, data on vaccination coverage over the same period, and sequence variations of key viral proteins. We also explicitly accounted for the spatio-temporal auto-correlations of infectious waves, and a host of socioeconomic and demographic factors. We carried out multiple orthogonal statistical analyses on these diverse, large geo-temporal datasets to bolster and corroborate our findings. We conclude that the initiation of a pan-continental influenza wave emerges from the simultaneous realization of a complex set of conditions, the strongest predictor groups are as follows, ranked by importance: (1) the host population’s socio- and ethno-demographic properties; (2) weather variables pertaining to relevant area specific humidity, temperature, and solar radiation; (3) the virus’ antigenic drift over time; (4) the host population’s land-based travel habits, and; (5) the spatio-temporal dynamics’ immediate history, as reflected in the influenza wave autocorrelation. The models we infer are demonstrably predictive (area under the Receiver Operating Characteristic curve ≈ 80%) when tested with out-of-sample data, opening the door to the potential formulation of new population-level intervention and mitigation policies.
BibTeX
@inproceedings{chattopadhyay2018conjunction,
  author = {Ishanu Chattopadhyay and Emre Kıcıman and Joshua W. Elliott and Jeffrey Shaman and Andrey Rzhetsky},
  title = {{Conjunction of Factors Triggering Waves of Seasonal Influenza}},
  booktitle = {eLife},
  year = {2018},
  doi = {10.7554/elife.30756},
}
Figure