The “migratory case” definition is fundamentally flawed—you assume different onset vs. reporting city equals healthcare seeking, but it could just be people getting sick while traveling for work or holidays. You never validate this, yet it’s the entire basis for your disparity argument.
You kept 2020 data despite admitting COVID lockdowns massively distorted travel and disease patterns. Averaging 2016, 2020 as one homogeneous period is indefensible. Run it without 2020, bet your network structures change.
Your 10-case threshold for community detection is completely arbitrary. For hep B it’s nothing, for shigellosis it’s a brutal filter. You’re deleting edges for rare diseases and calling it “denoising.” Your “sensitivity analysis” at 20 cases is just another random number.
Figure 2 is visually useless. Fourteen diseases, 337 cities, illegible colors, no legend. I cannot see the patterns you describe. A Shiny website doesn’t fix a broken main figure.
Your SHAP analysis is circular, using AMAP mobility data to predict patient movement. Yes, they correlate; people move, patients move. The actual question is what drives patient-specific movement beyond general mobility, and your SHAP values say… almost nothing. All <10%.
Network density correlates with case count at r=0.87. This is trivial: more cases fill more city-pairs. This isn’t epidemiology, it’s arithmetic.
Your CFR comparisons are confounded and you ignore it. Migratory flu/pertussis have higher CFR; migratory TB/HIV have lower CFR. You spin narratives about referrals and healthier migrants. But maybe it’s just detection bias: acute severe cases seek care while traveling; chronic cases get screened earlier in cities. You don’t control for stage at diagnosis. Without that, these are just stories.