A combination of annual and nonannual forces drive respiratory disease in the tropics

Abstract

Background: It is well known that influenza and other respiratory viruses are wintertime-seasonal in temperate regions. However, respiratory disease seasonality in the tropics remains elusive. In this study, we aimed to characterize the seasonality of influenza-like illness (ILI) and influenza virus in Ho Chi Minh City (HCMC), Vietnam.

Methods: We monitored the daily number of ILI patients in 89 outpatient clinics from January 2010 to December 2019. We collected nasal swabs and tested for influenza from a subset of clinics from May 2012 to December 2019. We used spectral analysis to describe the periodicities in the system. We evaluated the contribution of these periodicities to predicting ILI and influenza patterns through lognormal and gamma hurdle models.

Findings: During ten years of community surveillance, 66,799 ILI reports were collected covering 2.9 million patient visits; 2604 nasal swabs were collected 559 of which were PCR-positive for influenza virus. Both annual and nonannual cycles were detected in the ILI time series, with the annual cycle showing 8.9% lower ILI activity (95% CI: 8.8%-9.0%) from February 24 to May 15. Nonannual cycles had substantial explanatory power for ILI trends (ΔAIC = 183) compared to all annual covariates (ΔAIC = 263). Near-annual signals were observed for PCR-confirmed influenza but were not consistent along in time or across influenza (sub)types.

Interpretation: Our study reveals a unique pattern of respiratory disease dynamics in a tropical setting influenced by both annual and nonannual drivers. Timing of vaccination campaigns and hospital capacity planning may require a complex forecasting approach.

Keywords: Vietnam; influenza; influenza-like illness; mHealth; participatory epidemiology; syndromic surveillance; tropics.

Figure 1.

Daily ILI ζ- score (grey line) and 7-day smoothed ILI ζ- score (black line) from 2010-01-01 to 2019-12-31. The mean of ILI ζ- score in each year is shown as a blue horizontal dashed line.

Figure. 2.

Nonannual and annual cycles in ILI ζ- score. (A) The Pearson autocorrelation function (ACF) of ILI ζ- score time series, split across different study periods. Horizontal blue dashed line labels the region where ACF is significantly different from 0 (p = 0.05). Vertical dashed line labels the day lag when the ACF is the highest between 150 to 450 days. Annual cycle is labeled with a black dot. Periods are inclusive so “2010–2015” spans six years. (B) The shift of ACF value from nonannual cycle to annual cycle. The x-axis denotes the last year included in the time series, and the y-axis shows the ACF value. ACF values of annual cycles (circles) and nonannual cycles (triangles) are shown. The nonannual cycle showed stronger signal at the through 2017.

Figure 3.

Nonannual and annual cycles in ILI ζ- score. (A) The AIC of cyclic step functions given ILI ζ- score among different number of steps and different cycle lengths. AIC is lowest when cycle is 210 days and close to annual. (B) Contribution of each predictor is calculated as the AIC difference when removing the predictor from the full model. (C) The predicted values of ILI ζ- score (red) and the 95% prediction intervals (orange) of the full model including both cycles and the observed ILI ζ- score (gray).

Figure 4.

ILI+ time series and periodic signals. (A) The 7-day smoothed overall ILI+, stacked by subtypes. (B) ACF plot of the entire time series of overall ILI+. Vertical dashed line labels show ACF peaks at 358-day lag and the subsequent cycles. Black points show annual cycles. (C) The peak lag of ACF varied between 338 and 377 days in varying length of included ILI+ time series. (D) Discrete Fourier transform of the entire time series of overall ILI+. Black circle labels the dominant 324-day cycle.

Figure 5.

Cycles in ILI+.(A) AIC heatmap for cycle length and number of steps allowed in each cycle. Near-annual cycle lengths (330-day and 385-day) have the best fits. (B) AIC contribution of each predictor in gamma hurdle model, measured as the AIC difference when removing the predictor from the final model. (C) The predicted ILI+ (red) with 95% prediction interval (orange) is shown with observed ILI+ (gray).