Hostname: page-component-76d6cb85b7-kcxw8 Total loading time: 0 Render date: 2026-07-18T20:04:03.008Z Has data issue: false hasContentIssue false
  • English
  • Français

An exploration of spatial patterns of seasonal diarrhoeal morbidity in Thailand

Published online by Cambridge University Press:  19 September 2011

B. J. J. McCORMICK ,
W. J. ALONSO  and
M. A. MILLER
B. J. J. McCORMICK*
Affiliation:
Fogarty International Center, National Institutes of Health, Bethesda, USA
W. J. ALONSO
Affiliation:
Fogarty International Center, National Institutes of Health, Bethesda, USA
M. A. MILLER
Affiliation:
Fogarty International Center, National Institutes of Health, Bethesda, USA
*
*Author for correspondence: Dr B. J. J. McCormick, Fogarty International Center, National Institutes of Health, Bethesda, MD 20892, USA. (Email: ben.mccormick@nih.gov)
Rights & Permissions [Opens in a new window]

Summary

Studies of temporal and spatial patterns of diarrhoeal disease can suggest putative aetiological agents and environmental or socioeconomic drivers. Here, the seasonal patterns of monthly acute diarrhoeal morbidity in Thailand, where diarrhoeal morbidity is increasing, are explored. Climatic data (2003–2006) and Thai Ministry of Health annual reports (2003–2009) were used to construct a spatially weighted panel regression model. Seasonal patterns of diarrhoeal disease were generally bimodal with aetiological agents peaking at different times of the year. There is a strong association between daily mean temperature and precipitation and the incidence of hospitalization due to acute diarrhoea in Thailand leading to a distinct spatial pattern in the seasonal pattern of diarrhoea. Model performance varied across the country in relation to per capita GDP and population density. While climatic factors are likely to drive the general pattern of diarrhoeal disease in Thailand, the seasonality of diarrhoeal disease is dampened in affluent urban populations.

Keywords

Climate – impact ofdiarrhoea

Information

Type
Original Papers
Information
Epidemiology & Infection , Volume 140 , Issue 7 , July 2012 , pp. 1236 - 1243
Creative Commons
This is a work of the U.S. Government and is not subject to copyright protection in the United States
Copyright
Copyright © Cambridge University Press 2011 This is a work of the U.S. Government and is not subject to copyright protection in the United States.

INTRODUCTION

Despite therapeutic advances and improved social standards the global burden of enteric and diarrhoeal disease continues to be a leading cause of morbidity and mortality in young children worldwide [Reference Black, Morris and Bryce1–Reference Petri3]. As an illustration of this, the mortality rate due to acute diarrhoea, defined roughly as >2 watery stools within 24 h, has dropped substantially in Thailand during the past three decades. The number of hospitalizations has, however, increased from the late 1970s to the present (Fig. 1).

Fig. 1.

Trends in hospitalizations (——) and deaths ( – – –) due to acute diarrhoea in Thailand (diarrhoea data from Thai Ministry of Health, population data from the World Bank).

Different aetiological agents causing diarrhoea have been shown to circulate with distinct seasonal patterns: bacterial diarrhoea tends to occur most frequently in warmer and wetter months when environmental conditions favour the faecal–oral spread of pathogens [Reference Georges4–Reference Samal7]. In contrast, viral diarrhoea exhibits seasonal patterns similar to respiratory infections such as influenza, peaking in colder weather [Reference D'Souza, Hall and Becker8–Reference Guerrant10]. Climatic patterns have been implicated in driving the seasonal pattern of diarrhoeal disease in several temperate countries [Reference Checkley11–Reference Alonso14] although not in tropical climates where climatic seasons are less distinct [Reference Gomes5, Reference Stintzing15, Reference Cook16]. Comparable studies describing patterns of morbidity or mortality in countries with monsoon weather patterns that do not, for example, exhibit a pronounced summer and winter so much as hot and wet seasons remain rare. Previous studies of northern Thailand identified strong seasonal patterns associated with monsoon rainfall patterns, but did not comment on the seasonal patterns in the more tropical areas of the same country [Reference Pinfold, Horan and Mara17, Reference Ardkaew and Tongkumchum18].

In this study, the spatial trends associated with the seasonal dynamics of acute diarrhoeal morbidity are investigated across the whole of Thailand. In addition to describing the spatial pattern of seasonal diarrhoea, the model is tested against a generalized description of the climate providing an opportunity to assess the predictive power of the model. Thailand is particularly interesting as a case study with which to assess the role of climate in driving patterns of diarrhoeal disease since it not only has monsoon weather patterns in the north of the country, but also more stable tropical weather in the south thus representing a diverse range of climatic conditions that might be expected to contribute to a diverse range of diarrhoeal disease incidence.

DATA

Morbidity data

Diarrhoeal morbidity data, defined as the number of hospitalizations, due to acute diarrhoea (ICD-10: A09), enteric fever (typhoid ICD-10: A01.0; paratyphoid ICD-10: A01.1-3, unspecified) and dysentery (bacterial ICD-10: A03 and amoebic ICD-10: A06 and unspecified) were extracted from the Thai Ministry of Health annual reports, 2003–2009 [19]. Data were divided by changwat (the first sub-national administrative unit) and month resulting in 76 geographical units, each with 84 observations per causal agent. Data were input to a GIS and associated with social and physical data. The number of cases per month were standardized into cases/10 000 using the annual population per changwat and log-transformed.

Social data

Social data were taken from a combination of Thai Ministry of Statistics data hosted on the South-East Asian Health and Life Statistics Database [20] (population density/km2 and percentage of the population living in municipal areas) and the Thai Household Socioeconomic survey (2000) (per capita GDP).

Environmental data

Climatic data were extracted from the CRU TS 3.0, 0·5° global datasets [Reference Mitchell and Jones21] and the median of each pixel containing the changwat centroid and contiguous pixels were calculated for monthly mean daily temperatures and monthly precipitation for the period 2000–2006. Temperature and rainfall have both been extensively related to diarrhoeal disease [Reference Checkley11, Reference Kelly-Hope12, Reference Pinfold, Horan and Mara17, Reference Hashizume22–Reference Constantin de Magny24] and are highly correlated with other climatic variables available from the CRU data (minimum, maximum and diurnal range in temperature, the frequency of wet days, and vapour pressure).

Physical data

The median elevation of the changwat was similarly extracted using the centroid pixel and the surrounding pixels. Elevation data was supplied by the CRU.

METHOD

Diarrhoeal disease data was available from 2003 to 2009 inclusive, whereas climatic data was only available until mid-2006. A spatially weighted panel regression [Reference Baltagi25–Reference LeSage and Pace27] was therefore fitted for the period 2003–2005 (inclusive) and then used to forecast the diarrhoeal hospitalizations for 2006–2009, assuming a continuation of the climatic pattern and taking the spatial dependence from the first period.

The spatially weighted panel regression was used to compare detrended temperature and precipitation data to detrended diarrhoeal disease data. A temporal lag of 1–11 months was included for both mean daily temperature and mean monthly precipitation. To assess nonlinear relationships between climate and diarrhoeal disease quadratic terms were included in addition to an interaction term between temperature and precipitation. Following the fitting of a full model, terms were re-assessed and non-significant terms removed.

The spatial panel regression assumes a form of

{\rm ln}\lpar {\rm inc}_{it} \rpar \equals a \plus \sum\limits_{\setnum{1}}^{n} {\beta _{n} X_{nit} \plus u_{i} \plus e_{it} \comma }

where incit is the incidence of diarrhoea mortality in state i at time t; X is a series of independent variables at state i and time t; α and β are the regression parameters (global mean and predictor coefficients respectively); u i is an error term that captures the state-specific unobserved effects and e it represents all other unobserved effects (assumed to be normally distributed and independent). Spatial effects were based on a weighting, calculated using the inverse distance between changwat centroids.

In order to forecast the remaining acute diarrhoeal hospitalization data and test the fit of the spatial panel regression model, it was necessary to extend the climate data beyond mid-2006. The whole climate time-series (2000–2006) was used to fit Fourier harmonics to the temperature and precipitation data [Reference Miller and Strange28, Reference Rogers, Hay and Packer29]. A combination of an annual and bi-annual harmonics reproduced the stationary (i.e. with no overall trend) seasonal pattern in both climatic variables, generalizing the seasonal pattern as an approximation of an ‘average’ year.

Regression parameters relating the mean daily temperature and mean monthly precipitation along with the spatial coefficients based on the spatial dependence of the monthly mean of the 2003–2005 disease incidence data were applied to the extended Fourier-processed climate data to forecast the acute diarrhoeal hospitalization incidence.

Deficiencies in model fit were compared to social factors that might account for variability between changwats that is not captured by climatic variables.

Analyses, graphics, and statistical tests were conducted in Matlab (The MathWorks Inc., USA). Spatially weighted panel regressions were performed using the Spatial Econometrics toolbox [Reference LeSage and Pace27].

RESULTS

There was marked seasonality in national diarrhoeal hospitalization resulting from all the different aetiologies (Fig. 2). The timing of peak hospitalization was different between the three broad categories; however, for typhoid/paratyphoid there were dominant and unimodal peaks around July; for dysentery, peaks were bimodal, larger in May and smaller in January, although there was a trend of decreasing magnitude, which was more pronounced in the May peaks; for acute diarrhoea there were equally bimodal peaks around July and January.

Fig. 2.

Time-series of typhoid/paratyphoid (top panel), dysentery (middle panel) and acute diarrhoea (bottom panel) for the whole of Thailand, 2003–2009.

The number of hospitalizations per month for acute diarrhoea (i.e. unspecified aetiology) substantially outweighed any of the defined causes. Further analysis was only performed for acute diarrhoea, for which there was a sizable number of cases per month per changwat.

Spatial patterning was not visually observed in the mean hospitalizations for different changwats, although there was visual evidence of spatial patterns in the timing of peak incidence. First, it is worth noting that not all changwats had a similar seasonal pattern (Fig. 3), only those changwats that had an early (i.e. around January–February) peak in acute diarrhoeal hospitalizations also showed a peak around June. Changwats with peak mortality in May tended not to have secondary peaks. The unimodal changwats tended to be in the periphery of Thailand, either the extreme north or south of the country. The timing of peaks tended to be earliest in central Thailand, around Bangkok and to the north, which is the area with lowest elevations, then progressively later in the year in areas of higher altitude. By the time that the highlands in the north of Thailand had peaks (around May) the lowlands began a second wave of diarrhoeal hospitalizations.

Fig. 3.

Timing of peaks of acute diarrhoeal morbidity. (a) Timing of first peak; (b) timing of second peak. Grey areas indicate those changwats without secondary peaks.

A spatial panel regression model was made of contemporary acute diarrhoeal disease and climatic data (2003–2005). The full model, using both temperature and precipitation with lags of 1–11 months, achieved an r 2adj=0·597. While a number of both daily mean temperature and mean monthly precipitation lagged months were statistically significant, they tended to cluster around 9–10 months earlier for temperatures and 1–3 months earlier for precipitation relative to the diarrhoeal disease data (Table S1, available online). The reduced model (Table 1) achieved r 2 adj=0·544.

Table 1.

Reduced spatial panel regression model of acute diarrhoeal disease

t −n is the temporal lag of n months.

Variables in bold are significant (P<0·05).

The climatic data was projected using annual and bi-annual Fourier harmonics to match the 2006–2009 acute diarrhoeal mortality data. The average climatic profile for each changwat was projected from 2000 to 2006 up to the end of 2009. Assuming that the spatial structure was the same as the mean of the earlier period, and matching to the seasonal profile of acute diarrhoeal hospitalization data, derived from Fourier harmonics, achieved a model with an overall r 2 adj=0·462.

The overall model's goodness of fit could be partitioned to explore where the model deviated from the observed data (Fig. S1, available online). The patterning of goodness of fit was comparable between the models with contemporary and projected climatic data, and performance tended to be good (r 2 >0·7) in the north of Thailand and worst (r 2 <0·3) in the area around the Bangkok metropolis. Where the performance of the fitted and extrapolated model differed there was generally a shift in the seasonal pattern over the time period, for example, the hospitalization incidence in Bangkok shifted from a broadly unimodal peak around January to the presence of a secondary peak in June (Fig. 4). Similarly, Uthai Thani in the south of Thailand demonstrates a shifting seasonal pattern, which not only shows increasing magnitude between minimum and maximum values, but also has a broadening peak (from a peak in July to a peak extending between January and July).

Fig. 4.

Climatic model fitting. (a) Example of seasonal profiles for three changwats (north to south: Chiang Rai, Bangkok, Uthai Thani) showing the model fit from the contemporary climate data (circles) and projected climate (squares) along with the observed acute diarrhoea hospitalizations (black curve). Model r 2 for each changwat from (b) the contemporary and (c) projected climate indicated in the colour scale (dark to light indicating r 2=0–1).

The deficiencies of the model to accurately predict the hospitalizations due to acute diarrhoea show a geographical pattern (Fig. 4). None of the social factors proved significant in the spatially weighted regression model and did not improve the fit compared to the climate variables alone (e.g. the fitted climate model with per capita GDP achieved an r 2 adj=0·543). This result is unsurprising given that the social variables provide a single snapshot in time, whereas the climatic variables better capture the temporal dynamics of diarrhoeal disease. The changwat r 2 value was, however, significantly correlated with the per capita GDP (P<0·001 for both contemporary and projected climate models, Fig. 5a), where GDP is lower the model accuracy tends to be higher. Additionally, areas of Thailand with higher per capita GDP tended to have reduced diarrhoeal disease seasonal amplitude (P<0·001, Fig. 5b). Per capita GDP correlated positively with both population density and the proportion of the population in urban areas, although it had higher correlation with the acute diarrhoea data. The urban population of Thailand is predominantly found in the central region around and to the north of Bangkok. The highland areas to the north of the country and to the west of Bangkok are more rural and have lower GDP and it is these areas that have greater correspondence between climate and acute diarrhoeal hospitalizations. The south of Thailand can be characterized in this instance by lower population densities and higher per capita GDP than the north and these changwats are also those which tend to have poorer model accuracy.

Fig. 5.

(a) Comparison of the r 2 goodness of fit for the models with contemporary climate data (crosses and solid line) and projected climate (dots and dotted line) and the per capita GDP for each changwat. (b) Comparison of the amplitude of the seasonal pattern (2003–2009) for each changwat with the per capita GDP.

Peak acute diarrhoeal hospitalizations tended to be associated with the minimum precipitation, lagging by about 1 month (Fig. S2, online). Peak incidence preceded the peak temperatures by about 3 months. In both cases this corresponded to the significant lags in spatial panel regression.

DISCUSSION

The seasonality in the annual incidence of diarrhoeal disease in Thailand suggests a role for climatic drivers, as have been implicated in other countries [Reference Checkley11, Reference Kelly-Hope12, Reference Alonso14, Reference Kelly-Hope30]. The degree to which climate plays a role in determining the timing of seasonal diarrhoea is less well characterized in countries with a monsoon weather pattern, however, as much work has concentrated on temperate regions of the world.

There is evidence of differences in the seasonality between the bacterial and protozoan causes of diarrhoeal disease and viral causes; however, in the absence of detailed viral aetiology for Thailand it is not possible to further characterize these difference other than to conjecture that the peak hospitalizations in January categorized as acute diarrhoea do not correspond to either the peaks of typhoid/paratyphoid or of bacterial/amoebic dysentery.

A spatially weighted panel regression model using just the daily mean temperature and monthly mean precipitation performed well in predicting the incidence of hospitalization due to acute diarrhoea in Thailand. The pattern of acute diarrhoea in Thailand varies by changwat, although overall it is strongly related to the seasonal patterns in both temperature and precipitation, independently and by interaction. Diarrhoeal hospitalization was highest first around January, towards the end of the dry season when temperatures are comparatively low, and second around July at the beginning of the rainy season and when temperatures drop (after the hot season, roughly March–June). When the transition between the hot and the rainy seasons is less distinct there is merging of the two diarrhoeal peaks around May–June [e.g. in Uthai Thani, Figs 5 and S2 (online)].

Where the population was least dense or urbanized, the same places also tended to have the lowest per capita GDP, the model performance using climatic factors alone was highest. In more affluent and urbanized populations the model predictions diverged from the observed pattern of diarrhoeal disease. A possible parallel, hinted at by the relationship between the amplitude of diarrhoeal disease incidence and GDP, is that urban areas are better able to dampen the seasonal variability of diarrhoeal disease as driven by climate. Urban areas may, for example, have greater investment in sanitation and are thereby better able to dispose of faecal material [Reference Pinfold, Horan and Mara17]. The results presented here do suggest that factors other than climate begin to play an important role in determining the seasonal pattern of diarrhoeal disease in urban Thailand.

Previous models of diarrhoeal disease in Thailand have suggested that there is considerable underreporting of cases [Reference Ardkaew and Tongkumchum18, Reference Yotthanoo and Choonpradub31]. This finding is not surprising as diarrhoeal disease can become routine in the local population such that it is not necessarily reported to national surveillance programmes [Reference Black32]. The decreasing mortality from diarrhoeal disease in Thailand may also suggest that severe cases have also diminished even though incidence has not, thereby further reducing the reporting of cases. The method of correcting for underreporting used in those models [Reference Ardkaew and Tongkumchum18, Reference Yotthanoo and Choonpradub31] was to re-fit the model using interpolated values rather than supposedly spurious records. In this study such an iterative procedure was not used, a possible consequence being that the model predictions are ‘less wrong’ than they appear compared to questionable observation data. This is particularly the case with very low observed incidence, for which the model generally predicted higher hospitalization rates. While zero incidences are improbable, the regularly low number of cases between October and December is unlikely to be underreporting as much as less favourable conditions for diarrhoeal disease transmission [Reference Yotthanoo and Choonpradub31].

There is little evidence to directly relate the climatic conditions at large spatial scales with causal mechanisms of diarrhoeal disease. There is, however, evidence of the effects of micro-climatic temperature and humidity on the replication and survival of bacterial pathogens [Reference Kramer, Schwebke and Kampf33, Reference Wathes, Howard and Webster34] and thereby the risk of exposure to pathogens [Reference Vu Nguyen6, Reference Barrell and Rowland35]. The indirect effects of rainfall have been suggested to be both beneficial and detrimental to bacterial transmission [Reference Pinfold, Horan and Mara17]; e.g. a combination of increased bacterial survival during periods of high humidity, and also improved faecal disposal and hygiene practices when water is more abundant [Reference Vu Nguyen6, Reference Esrey, Feachem and Hughes36]. Extrapolating an assumed relationship between climatic factors and diarrhoeal disease burden is subject to the compounding errors of aggregating heterogeneous data [Reference Leyk, McCormick and Nuckols37], although identifying putative determinants at a course scale is likely to increase the targeted research necessary to bridge the gap in resolution.

Extending the understanding of possible causal mechanisms for disease patterns is a useful step to devising appropriately targeted interventions. In Thailand the increasing morbidity associated with diarrhoeal disease is a cause for concern and this study extends the previously observed spatial patterning from identification of ‘season’ as a predictor of disease burden [Reference Ardkaew and Tongkumchum18, Reference Yotthanoo and Choonpradub31] to a possible early warning forecast using meteorological conditions.

NOTE

Supplementary material accompanies this paper on the Journal's website (http://journals.cambridge.org/hyg).

ACKNOWLEDGEMENTS

We thank Ramya Ambikapathi for her assistance in locating appropriate disease data and Cynthia Schuck-Paim for insightful comments on the manuscript. This study was funded as part of the Malnutrition and Enteric Diseases (Mal-ED) project, sponsored by the Bill and Melinda Gates Foundation and the Fogarty International Center of the National Institutes of Health.

DECLARATION OF INTEREST

None.

References

REFERENCES

1.Black, RE, Morris, SS, Bryce, J. Where and why are 10 million children dying every year? Lancet 2003; 361: 2226–2234.CrossRefGoogle ScholarPubMed
2.Bryce, J, et al. WHO estimates of the causes of death in children. Lancet 2005; 365: 1147–1152.CrossRefGoogle ScholarPubMed
3.Petri, WA, et al. Enteric infections, diarrhea, and their impact on function and development. Journal of Clinical Investigation 2008; 118: 1277–1290.CrossRefGoogle ScholarPubMed
4.Georges, MC, et al. Parasitic, bacterial, and viral enteric pathogens associated with diarrhea in the Central African Republic. Journal of Clinical Microbiology 1984; 19: 571–575.CrossRefGoogle ScholarPubMed
5.Gomes, TAT, et al. Enteropathogens associated with acute diarrheal disease in urban infants in Sao Paulo, Brazil. Journal of Infectious Diseases 1991; 164: 331–337.CrossRefGoogle ScholarPubMed
6.Vu Nguyen, T, et al. Etiology and epidemiology of diarrhea in children in Hanoi, Vietnam. International Journal of Infectious Diseases 2006; 10: 298–308.CrossRefGoogle ScholarPubMed
7.Samal, SK, et al. Incidence of bacterial enteropathogens among hospitalized diarrhea patients from Orissa, India. Japanese Journal of Infectious Diseases 2008; 61: 350–355.CrossRefGoogle ScholarPubMed
8.D'Souza, RM, Hall, G, Becker, NG. Climatic factors associated with hospitalizations for rotavirus diarrhoea in children under 5 years of age. Epidemiology and Infection 2008; 136: 56–64.CrossRefGoogle ScholarPubMed
9.David-West, TS, Cooke, AR. Laboratory and clinical investigation of the 1974 influenza epidemic in Nigeria. Bulletin of the World Health Organization 1974; 51: 103–105.Google ScholarPubMed
10.Guerrant, RL, et al. Prospective study of diarrheal illnesses in northeastern Brazil – patterns of disease, nutritional impact, etiologies, and risk-factors. Journal of Infectious Diseases 1983; 148: 986–997.CrossRefGoogle ScholarPubMed
11.Checkley, W, et al. Effects of El Nino and ambient temperature on hospital admissions for diarrhoeal diseases in Peruvian children. Lancet 2000; 355: 442–450.Google ScholarPubMed
12.Kelly-Hope, LA, et al. Temporal trends and climatic factors associated with bacterial enteric diseases in Vietnam, 1991–2001. Environmental Health Perspectives 2008; 116: 7–12.CrossRefGoogle ScholarPubMed
13.Kale, PL, Fernandes, C, Nobre, FF. Temporal pattern of diarrhea hospitalizations and deaths in children, 1995 to 1998, Brazil [in Portuguese]. Revista de Saúde Pública 2004; 38: 30–37.CrossRefGoogle ScholarPubMed
14.Alonso, WJ, et al. Spatio-temporal patterns of diarrhoeal mortality in Mexico. Epidemiology and Infection. Published online: 14 April 2011. doi:10.1017/S0950268811000562.Google Scholar
15.Stintzing, G, et al. Seasonal fluctuations in the occurrence of enterotoxigenic bacteria and rotavirus in paediatric diarrhoea in Addis Ababa. Bulletin of the World Health Organization 1981; 59: 67–73.Google ScholarPubMed
16.Cook, SM, et al. Global seasonality of rotavirus infections. Bulletin of the World Health Organization 1990; 68: 171–177.Google ScholarPubMed
17.Pinfold, JV, Horan, NJ, Mara, DD. Seasonal effects on the reported incidence of acute diarrheal disease in northeast Thailand. International Journal of Epidemiology 1991; 20: 777–786.CrossRefGoogle ScholarPubMed
18.Ardkaew, J, Tongkumchum, P. Statistical modelling of childhood diarrhea in northeastern Thailand. Southeast Asian Journal of Tropical Medicine and Public Health 2009; 40: 807–815.Google ScholarPubMed
19.Bureau of Epidemiology, Ministry of Public Health, Thailand. Annual reports, 2010 (http://203.157.15.4/). Accessed 9 June 2010.Google Scholar
20.Southeast Asian Health and Life Statistics Database. (http://www.sls-db.org/). Accessed 23 March 2010.Google Scholar
21.Mitchell, TD, Jones, PD. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. International Journal of Climatology 2005; 25: 693–712.CrossRefGoogle Scholar
22.Hashizume, M, et al. Rotavirus infections and climate variability in Dhaka, Bangladesh: a time-series analysis. Epidemiology and Infection 2008; 136: 1281–1289.CrossRefGoogle ScholarPubMed
23.Levy, K, Hubbard, AE, Eisenberg, JNS. Seasonality of rotavirus disease in the tropics: a systematic review and meta-analysis. International Journal of Epidemiology 2009; 38: 1487–1496.CrossRefGoogle ScholarPubMed
24.Constantin de Magny, G, et al. From the Cover: Environmental signatures associated with cholera epidemics. Proceedings of the National Academy of Sciences USA 2008; 105: 17676–17681.CrossRefGoogle Scholar
25.Baltagi, HB. Econometric Analysis of Panel Data. Chichester: John Wiley & Sons, 1995.Google Scholar
26.Elhorst, JP. Specification and estimation of spatial panel data models. International Regional Science Review 2003; 26: 244–268.CrossRefGoogle Scholar
27.LeSage, J, Pace, RK. Introduction to Spatial Econometrics. CRC Press: Taylor & Francis Group (http://www.spatial-econometrics.com/), 2009.CrossRefGoogle Scholar
28.Miller, M, Strange, P. Use of Fourier models for analysis and interpretation of continuous glucose monitoring glucose profiles. Journal of Diabetes Science and Technology 2007; 1: 630–638.CrossRefGoogle ScholarPubMed
29.Rogers, DJ, Hay, SI, Packer, MJ. Predicting the distribution of tsetse flies in West Africa using temporal Fourier processed meteorological satellite data. Annals of Tropical Medicine and Parasitology 1996; 90: 225–242.CrossRefGoogle ScholarPubMed
30.Kelly-Hope, LA, et al. Geographical distribution and risk factors associated with enteric diseases in Vietnam. American Journal of Tropical Medicine and Hygiene 2007; 76: 709–712.CrossRefGoogle ScholarPubMed
31.Yotthanoo, S, Choonpradub, C. A statistical method for estimating under-reported incidence rates with application to child diarrhea in Thai provinces bordering Cambodia. Southeast Asian Journal of Tropical Medicine and Public Health 2010; 41: 203–214.Google Scholar
32.Black, RE. Diarrheal diseases and child morbidity and mortality. Population and Development Review 1984; 10: 141–161.CrossRefGoogle Scholar
33.Kramer, A, Schwebke, I, Kampf, G. How long do nosocomial pathogens persist on inanimate surfaces? A systematic review. BMC Infectious Diseases 2006; 6: 130–138.CrossRefGoogle Scholar
34.Wathes, CM, Howard, K, Webster, AJF. The survival of Escherichia coli in an aerosol at air temperatures of 15°C and 30°C and a range of humidities. Epidemiology and Infection 1986; 97: 489–496.Google Scholar
35.Barrell, RAE, Rowland, MGM. Infant foods as a potential source of diarrhoeal illness in rural West Africa. Transactions of the Royal Society of Tropical Medicine and Hygiene 1979; 73: 85–90.CrossRefGoogle ScholarPubMed
36.Esrey, SA, Feachem, RG, Hughes, JM. Interventions for the control of diarrhoeal diseases among young children: improving water supplies and excreta disposal facilities. Bulletin of the World Health Organization 1985; 63: 757–772.Google ScholarPubMed
37.Leyk, S, McCormick, BJJ, Nuckols, JR. Effects of varying temporal scale on spatial models of mortality patterns attributed to pediatric diarrhea. Spatial and Spatio-temporal. Epidemiology 2011; 2: 91–101.CrossRefGoogle ScholarPubMed
Figure 0

Fig. 1. Trends in hospitalizations (——) and deaths ( – – –) due to acute diarrhoea in Thailand (diarrhoea data from Thai Ministry of Health, population data from the World Bank).

Figure 1

Fig. 2. Time-series of typhoid/paratyphoid (top panel), dysentery (middle panel) and acute diarrhoea (bottom panel) for the whole of Thailand, 2003–2009.

Figure 2

Fig. 3. Timing of peaks of acute diarrhoeal morbidity. (a) Timing of first peak; (b) timing of second peak. Grey areas indicate those changwats without secondary peaks.

Figure 3

Table 1. Reduced spatial panel regression model of acute diarrhoeal disease

Figure 4

Fig. 4. Climatic model fitting. (a) Example of seasonal profiles for three changwats (north to south: Chiang Rai, Bangkok, Uthai Thani) showing the model fit from the contemporary climate data (circles) and projected climate (squares) along with the observed acute diarrhoea hospitalizations (black curve). Model r2 for each changwat from (b) the contemporary and (c) projected climate indicated in the colour scale (dark to light indicating r2=0–1).

Figure 5

Fig. 5. (a) Comparison of the r2 goodness of fit for the models with contemporary climate data (crosses and solid line) and projected climate (dots and dotted line) and the per capita GDP for each changwat. (b) Comparison of the amplitude of the seasonal pattern (2003–2009) for each changwat with the per capita GDP.

Supplementary material: File

McCormick Supplementary Material

McCormick Supplementary Material

Download McCormick Supplementary Material(File)
File 182.3 KB