Background
Neglected tropical diseases (NTDs) have shown evidently seasonal patterns with various timings, amplitudes and types [
1‐
3]. NTDs remain socio-economic burdens in developing countries, especially in low-income communities in Asia and Africa. The considerable seasonality of these NTDs is affected by seasonal changes of both environmental and social risk factors [
4,
5]. Understanding the regularity or irregularity of NTDs’ outbreaks is necessary to obtain optimal control over the disease spread [
6], to reduce their burdens and to reach the UN sustainable development goals for health [
7].
Amongst others, the infections of
Coxsackievirus A16 (CA16) and human
Enterovirus 71 (EV71) that mainly cause hand-foot-mouth disease (HFMD) in children under 5 years of age (U5s) have been recognized as an emerging public health problem in North East and South East Asian countries [
8,
9]. So far, studies of the seasonal variations of childhood HFMD infections have mainly focused on climatic driving factors such as temperature, humidity and rainfall [
9]. The effects of seasonally changing social contacts on the dynamics of childhood HFMD for small areas have been overlooked, even though HFMD viruses are mainly transmitted by directly physical contacts between infected and non-infected children [
10]. Additional research has been done for northeastern Asian countries [
11‐
13], whereas some studies were also done for southeastern Asian regions [
14,
15]. Despite the increasing application of mathematical models to understand the epidemiology of infectious diseases, spatial-temporal statistical methods (STS) have not yet been well recognized and leveraged for studying seasonality of NTDs, i.e. to quantitatively explain the spatial-temporal dynamics of the disease outbreaks [
16], especially for small geographical areas.
In this research, we study the seasonality of childhood HFMD in Viet Nam and its geographical variations at small areas as affected by seasonally changing weather and the social contacts. The study of Horby et al. in Viet Nam in 2011 [
17] shows that the physical contact amongst U5s themselves is the most intensive, especially during the period at preschool. Hence, the social contact amongst U5s is measured by the annually cumulated period of time they spend at preschool since their first day at preschool. We distinguish between the trend and the seasonality. The trend establishes the long-term changing pattern of the mean level, whereas the seasonality represents yearly periodic variations. We propose a two stage STS analysis to extract the seasonal patterns and to identify the driving factors of the incidences of childhood HFMD infections in Da Nang city, Viet Nam as an application. In so doing, we aimed to identify and understand the effects of seasonally changing social contacts of U5s on the dynamics of HFMD outbreaks at multiple district level.
Discussion
In this section, we discuss our main findings, the consistency of the seasonal patterns extracted from \( {IR}_{ij}^s \) using our methods compared to other methods and we also highlight the limitations of this study.
In this study, we have decomposed the monthly time series of the adjusted incidence proportions of childhood HFMD infections for the spatial-temporal strata of the mainland of Da Nang city to reveal the trend and the seasonality. The yearly larger outbreaks simultaneously happened in all three spatial strata in April. The variation of the seasonality of the second smaller outbreaks in T3 is intriguing. This variation suggests the effects of the season of increasing social contacts among U5s when they go to preschool on the onset of HFMD infection.
We have shown statistical evidences of these effects during the preschooling period. The positive values of the spatial-temporal regression coefficients imply the simultaneity of the peaks and vice versa [
21]. The mismatch between the maximum incidence proportions and the maximum average temperature indicates that the increase of the average temperature was the driving factor, not its maximum values. In the last two quarters of all years, the average temperature was not the predominant risk factor. The contrary was evident in T
1. At small areas of multiple districts in the study area, the increasing social contact has been shown to be the important driving factor of the geographical variation of the seasonality of the childhood HFMD outbreaks.
The estimated spatial-temporal regression parameters of (4) in T
3 are interestingly more informative than those in T
1 because all the smaller peaks occurred in T
3. As the period at the preschools in T
3 cumulatively increased from August to December, the negative values of
\( {\beta}_{T2}^s \) indicate the occurrence of the smaller peaks at the beginning of the preschooling period. Recall from the study of Horby et al. [
17] that the physical contacts of U5s are most intense within this age group in Viet Nam. Given that the physical contact is the prominent means of passing HFMD viruses, the nature, frequency and duration of the contacts of those children during the preschooling period explain the occurrence and the geographical dynamics of these smaller outbreaks.
The trends show the occurrences of the large outbreaks at the beginning and at the end of the observed period. Study of a longer period of time would show a long-term oscillation of the large outbreaks with the frequency of every two to 3 years in this city. The possible reasons can be the required lapse of time to cumulate the critical susceptible population [
31] or the existence of multiple viral strains in the study area [
32,
33].
The decomposed trends and seasonal components were highly correlated with those fitted by the method of seasonal and trend decomposition using Loess [
34]. The correlation coefficients of the trends from both methods at lag zero varied from 0.85 to 0.90. Those of the seasonality varied from 0.83 to 0.97. The similar results from both decomposition methods imply the consistency of the seasonality derived from the Fourier decomposition method. This method provides a semi-parametric approach to extract different components of a time series including the long-term and short-term irregular and regular fluctuations, taking into account effects of possible confounders [
35,
36]. Using this approach instead of the common seasonal ARIMA, the drawbacks of stationarization by differencing and statistical transformation can be avoided [
37]. Moreover, Fourier decomposition in combination with the spatial-temporal autoregressive model allow both spatial and temporal autocorrelation existing in the data to be included into the model calibration. This reduces the biases in the estimation of the associations between the incidence proportions and the risk factors. Notwithstanding, from a statistical point of view, the small number of spatial strata in the study area places one of the limitations of the study to include the effects of the spatial auto-correlation. In addition, the separation the temporal trend from the time series was based mainly upon expert judgments. In other words, the efficiency of the Fourier decomposition in many cases relies on the expert’s familiarity with the understudied phenomena.
Conclusions
HFMD has been becoming one of the most important pediatric NTDs in the Northeastern and Southeastern Asian countries. By applying spatial-temporal statistical analyses, the results have shown that at multiple district level, the social contact is the important driving factor of the spatial variation of the disease’s outbreaks. This study provides statistical evidences of the effects of the seasonality of increasing social contact amongst U5s on geographical dynamics of HFMD infection outbreaks. Our findings contribute to extend the understanding of the underlying driving factors of the disease dynamics at small areas. This contribution is necessary to inform the next insightful research into the spatial-temporal dynamics of the transmission of the NTDs within the local population.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (
http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.