Effects of large-scale oceanic phenomena on non-cholera vibriosis incidence in the United States: implications for climate change

Abstract

Non-cholera Vibrio (NCV) species are important causes of disease. These pathogens are thermophilic and climate change could increase the risk of NCV infection. The El Niño Southern Oscillation (ENSO) is a 'natural experiment' that may presage ocean warming effects on disease incidence. In order to evaluate possible climatic contributions to observed increases in NCV infection, we obtained NCV case counts for the United States from publicly available surveillance data. Trends and impacts of large-scale oceanic phenomena, including ENSO, were evaluated using negative binomial and distributed non-linear lag models (DNLM). Associations between latitude and changing risk were evaluated with meta-regression. Trend models demonstrated expected seasonality (P < 0.001) and a 7% (6.1%-8.1%) annual increase in incidence from 1999 to 2014. DNLM demonstrated increased vibriosis risk following ENSO conditions over the subsequent 12 months (relative risk 1.940, 95% confidence interval (CI) 1.298-2.901). The 'relative-relative risk' (RRR) of annual disease incidence increased with latitude (RRR per 10° increase 1.066, 95% CI 1.027-1.107). We conclude that NCV risk in the United States is impacted by ocean warming, which is likely to intensify with climate change, increasing NCV risk in vulnerable populations.

Keywords: Climate change; Vibrio fluvialis; Vibrio parahaemolyticus; Vibrio vulnificus; foodborne illness.

Fig. 1.

Reported and model-predicted monthly vibriosis counts, United States 1999–2014. Observed counts are represented by circles; solid curve represents predictions from negative binomial model incorporating fast Fourier transform and year terms. Date is plotted on the X-axis; counts are plotted on the Y-axis.

Fig. 2.

Association between lagged MEI and vibriosis risk, United States 1999–2014. (a) Risk surface represents modelled association between the MEI (scaled from −3, most La Niña-like, to +3, most El Niño-like, X-axis) and monthly vibriosis risk over 1–12 month lags (Y-axis). Associated RRs are plotted on the Z-axis. (b) Cross-sectional RR associated with the MEI of +3; lagged El Niño-like conditions are associated with downstream integrated RR of vibriosis (RR 1.940, 95% CI 1.298–2.900).

Fig. 3.

State-level incidence, trends and environmental influence on annual vibriosis risk in the U.S. states, 1999–2014. (a) Mean annual vibriosis incidence per 100 000 population, by state; (b) IRRs for year-on-year change in vibriosis incidence from negative binomial models, by state; (c) trends in vibriosis incidence from negative binomial models, by state; (d) IRR for vibriosis risk with a 1-unit change in the MEI, by region. Six grey-shaded states are those for which negative binomial models failed to converge.

Fig. 4.

Association between latitude and linear trend in vibriosis incidence, United States 1999–2014. Correlation between state latitude and average yearly IRR for vibriosis. Each circle represents a single U.S. state or District of Columbia, with size inversely proportional to variance in IRR estimates, corresponding to the weight assigned to each state. Circles are colour coded according to COVIS regions. Fitted lines represent the association between state latitude and IRR as predicted using univariable meta-regression (relative change in IRR per 10° increase in latitude of state centroid 1.059, 95% CI 1.020–1.099). Note that six states for which negative binomial models did not converge are excluded.