Global biogeography of human infectious diseases

Abstract

The distributions of most infectious agents causing disease in humans are poorly resolved or unknown. However, poorly known and unknown agents contribute to the global burden of disease and will underlie many future disease risks. Existing patterns of infectious disease co-occurrence could thus play a critical role in resolving or anticipating current and future disease threats. We analyzed the global occurrence patterns of 187 human infectious diseases across 225 countries and seven epidemiological classes (human-specific, zoonotic, vector-borne, non-vector-borne, bacterial, viral, and parasitic) to show that human infectious diseases exhibit distinct spatial grouping patterns at a global scale. We demonstrate, using outbreaks of Ebola virus as a test case, that this spatial structuring provides an untapped source of prior information that could be used to tighten the focus of a range of health-related research and management activities at early stages or in data-poor settings, including disease surveillance, outbreak responses, or optimizing pathogen discovery. In examining the correlates of these spatial patterns, among a range of geographic, epidemiological, environmental, and social factors, mammalian biodiversity was the strongest predictor of infectious disease co-occurrence overall and for six of the seven disease classes examined, giving rise to a striking congruence between global pathogeographic and "Wallacean" zoogeographic patterns. This clear biogeographic signal suggests that infectious disease assemblages remain fundamentally constrained in their distributions by ecological barriers to dispersal or establishment, despite the homogenizing forces of globalization. Pathogeography thus provides an overarching context in which other factors promoting infectious disease emergence and spread are set.

Keywords: biogeography; distribution; globalization; infectious disease; pathogeography.

Fig. 1.

Global human infectious disease pathogeographic patterns. Ordination analysis of β_sor of human infectious disease assemblages (n = 187 diseases).Similar colors indicate more similar disease assemblages. Separate disease classes and key to colors are presented in SI Appendix, Fig. S1.

Fig. 2.

(A) Mean number of countries positive for each disease class (black dots, box plots in gray). (B) Similarity (1 − β_sor) box plots for each disease class. (C) Turnover (β_sim) and nestedness (β_nes) components of total dissimilarity (β_sor). Turnover indicates the dissimilarity attributable to the replacement of diseases in one country relative to another. Nestedness indicates the dissimilarity attributable to diseases in one country being a subset of the diseases in another (24):$\text{Overall}\hspace{0.17em}\text{dissimilarity}=\text{Turnover}+\text{Nestedness}\left({\mathrm{\beta }}_{\text{sor}}={\mathrm{\beta }}_{\text{sim}}+{\mathrm{\beta }}_{\text{nes}}\right).$

Fig. 3.

Ebola co-zones through time (1976–2014). Additive co-zone models depict the average pairwise similarity (1 − β_sor; warmer colors indicate more similar) of zoonotic disease assemblages between countries recording primary Ebola (spillover) cases and all other countries calculated across four time points [1976 = DRC and Sudan (A), 1994 = A + Gabon and Cote d’Ivoire (B), 2000 = B + Republic of Congo and Uganda (C), and 2014 = C + Guinea (D)]. The scale is continuous but has been assigned categorical breaks to aid visualization.

Fig. 4.

Estimated relative influence of extrinsic predictors in explaining the similarity of human infectious disease assemblages among countries after accounting for the effect of geographic distance (spatial autocorrelation). Overall model: R² = 0.540. Descriptions of predictors are provided in Materials and Methods. Full model outputs are shown in Table 2 and SI Appendix, Table S1, and separate disease classes are shown in SI Appendix, Fig. S5.