Advanced Search

Preplanned Studies: Weather Variability, Socioeconomic Factors, and Pneumonia in Children Under Five-Years Old — Bangladesh, 2012−2016

View author affiliations
  • Summary

    What is already known on this topic?

    Different socioecological factors were associated with childhood pneumonia in Bangladesh. However, previous studies did not assess spatial patterns, and socioecological factors and spatial variation have the potential to improve the accuracy and predictive ability of existing models.

    What is added by this report?

    The spatial random effects were present at the district level and were heterogeneous. Average temperature, temperature variation, and population density may influence the spatial pattern of childhood pneumonia in Bangladesh.

    What are the implications for public health practice?

    The study results will help policymakers and health managers to identify the vulnerable districts, plan further investigations, help to improve proper resource allocation, and improve health interventions.

  • loading...
  • Funding: The Queensland University of Technology Postgraduate Research Award and Queensland University of Technology Higher Degree Research International Tuition Fee Sponsorship
  • [1] DGHS. IMCI newsletter. 2017. http://www.dghs.gov.bd/images/docs/IMCI/IMCI_12_2017.pdf. [2020-11-13].http://www.dghs.gov.bd/images/docs/IMCI/IMCI_12_2017.pdf
    [2] World Health Organization. Revised WHO classification and treatment of childhood pneumonia at health facilities–evidence summaries. 2014. https://apps.who.int/iris/rest/bitstreams/611120/retrieve. [2021-1-22].https://apps.who.int/iris/rest/bitstreams/611120/retrieve
    [3] Bangladesh Bureau of Statistics. Population and household census. 2011. http://www.bbs.gov.bd/site/page/47856ad0-7e1c-4aab-bd78-892733bc06eb/Population-and-Housing-Census. [2021-2-18].http://www.bbs.gov.bd/site/page/47856ad0-7e1c-4aab-bd78-892733bc06eb/Population-and-Housing-Census
    [4] Bangladesh Bureau of Statistics, Statistics and Informatics Division, Ministry of Planning, Government of the People’s Republic of Bangladesh. Bangladesh-household income and expenditure survey. 2016. expenditure survey. 2016. https://www.academia.edu/37874731/Preliminary_Report_on_Household_Income_and_Expenditure_Survey_2016_BANGLADESH_BUREAU_OF_STATISTICS_BBS_STATISTICS_AND_INFORMATICS_DIVISION_SID. [2020-9-17].https://www.academia.edu/37874731/Preliminary_Report_on_Household_Income_and_Expenditure_Survey_2016_BANGLADESH_BUREAU_OF_STATISTICS_BBS_STATISTICS_AND_INFORMATICS_DIVISION_SID
    [5] Gullón P, Varela C, Martínez EV, Gómez-Barroso D. Association between meteorological factors and hepatitis A in Spain 2010-2014. Environ Int 2017;102:230 − 5. http://dx.doi.org/10.1016/j.envint.2017.03.008CrossRef
    [6] Thomson MC, Connor SJ, D'Alessandro U, Rowlingson B, Diggle P, Cresswell M, et al. Predicting malaria infection in Gambian children from satellite data and bed net use surveys: the importance of spatial correlation in the interpretation of results. Am J Trop Med Hyg 1999;61(1):2 − 8. http://dx.doi.org/10.4269/ajtmh.1999.61.2CrossRef
    [7] Hu WB, Clements A, Williams G, Tong SL, Mengersen K. Spatial patterns and socioecological drivers of dengue fever transmission in Queensland, Australia. Environ Health Perspect 2012;120(2):260 − 6. http://dx.doi.org/10.1289/ehp.1003270CrossRef
    [8] Duncan EW, White NM, Mengersen K. Spatial smoothing in Bayesian models: a comparison of weights matrix specifications and their impact on inference. Int J Health Geogr 2017;16:47. http://dx.doi.org/10.1186/s12942-017-0120-xCrossRef
    [9] Best N, Richardson S, Thomson A. A comparison of bayesian spatial models for disease mapping. Stat Methods Med Res 2005;14(1):35 − 59. http://dx.doi.org/10.1191/0962280205sm388oaCrossRef
  • FIGURE 1.  Posterior estimated Relative Risk of childhood pneumonia at the district level of Bangladesh from 2012 to 2016.

    FIGURE 2.  Spatial random effects of childhood pneumonia — Bangladesh, 2012−2016.

    TABLE S1.  Spearman correlation between pneumonia and socioecological covariates in Children Under Five-Years Old — Bangladesh, 2012−2016.

    Variables12345678
    1Pneumonia
    2Temperature0.094
    3Temperature variation0.1610.235
    4Rainfall−0.0190.268*−0.146
    5Education0.008−0.011−0.1290.068
    6Internet use0.1260.0630.2230.017−0.066
    7Population density−0.148−0.25*−0.276*0.0280.0750.162
    8Poverty0.069−0.140.163−0.0770.095−0.209−0.336*
    Note: − represnt its pneumonia itself, there will be no number.
    * P<0.05.
    Download: CSV

    TABLE S2.  Descriptive statistics of childhood pneumonia and different socioecological factors — Bangladesh, 2012−2016.

    VariablesMean ± SDRange
    Pneumonia747.82 ± 245.32355.53−1612.10
    Temperature (°C)30.97 ± 0.4829.50−31.98
    Temperature variation (°C)3.63 ± 0.552.00−4.99
    Rain (mm)164.54 ± 101.373.13−386.96
    Education (%)54.66 ± 7.7537.50−73.70
    Internet use (%)0.62 ± 0.760.14−6.03
    Poverty incidence (%)27.45 ± 15.312.60−70.80
    Under five years population density (per square km)117.72 ± 87.0510.81−656.54
    Download: CSV

    TABLE S3.  Model comparison for relative risk of monthly childhood pneumonia, underlying socioecological factors, and different random effects — Bangladesh, 2012−2016.

    ModelRandom effectDeviance Information Criterion (DIC)Effective number of parameters (pD)
    Model INo 19954.2030.146
    Model IINo17382.415.001
    Model IIINo13773.0011.134
    Model IVu665.9863.940
    Model Vv665.6763.812
    Model VIu and v665.4763.719
    Download: CSV

    TABLE S4.  List of high-risk districts of Bangladesh for childhood pneumonia from 2012 to 2016.

    Name of the districtRelative Risk (95% Credible interval)Location
    Rangamati5.97 (5.63−6.31)South-eastern
    Pirojpur4.71 (4.48−4.93)South-western
    Jhalkathi4.38 (4.09−4.66)South-western
    Jaipurhut3.95 (3.70−4.19)North-eastern
    Bandarbon3.77 (3.48−4.07)South-eastern
    Meherpur3.50 (3.23−3.78)South-western
    Rajbari3.31(3.11−3.50)Central
    Khagrachari3.25 (3.02−3.49)South-eastern
    Panchagarh2.96 (2.78−3.14)Northern
    Download: CSV

    TABLE 1.  Crude and adjusted RR of different socioecological factors in Children Under Five-Years Old — Bangladesh, 2012−2016.

    VariablesCrude RR (95% CrI)Adjusted RR (95% CrI)
    Temperature*1.146 (0.929−1.432)1.161 (1.013−1.429)
    Temperature1.730 (1.694−1.763)1.529 (1.503−1.555)
    Temperature variability*1.821 (1.376−2.491)1.463 (1.170−1.839)
    Temperature variability1.623 (1.596−1.649)1.421 (1.395−1.447)
    Rainfall*1.000 (0.999−1.002)1.000 (0.999−1.002)
    Rainfall1.0007 (1.0006−1.0007)1.0001 (1.0000−1.0002)
    Population density*0.995 (0.994−0.997)0.996 (0.994−0.998)
    Population density0.9943 (0.9942−0.9945)0.9959 (0.9958−0.9961)
    Education*0.979 (0.956−1.004)0.986 (0.969−1.005)
    Education0.978 (0.977−0.979)0.9853 (0.984−0.986)
    Poverty*1.008 (0.996−1.020)1.000 (0.988−1.010)
    Poverty1.009 (1.009−1.010)1.002 (1.002−1.003)
    Internet*0.844 (0.661−1.070)0.916 (0.743−1.126)
    Internet0.874 (0.862−0.885)0.929 (0.916−0.942)
    Abbreviations: RR=relative risk; CrI=credible interval.
    * with heterogeneity (u and v).
    without heterogeneity (u and v).
    Download: CSV

Citation:

通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索
Turn off MathJax
Article Contents

Article Metrics

Article views(30304) PDF downloads(52) Cited by()

Share

Related

Weather Variability, Socioeconomic Factors, and Pneumonia in Children Under Five-Years Old — Bangladesh, 2012−2016

View author affiliations

Summary

What is already known on this topic?

Different socioecological factors were associated with childhood pneumonia in Bangladesh. However, previous studies did not assess spatial patterns, and socioecological factors and spatial variation have the potential to improve the accuracy and predictive ability of existing models.

What is added by this report?

The spatial random effects were present at the district level and were heterogeneous. Average temperature, temperature variation, and population density may influence the spatial pattern of childhood pneumonia in Bangladesh.

What are the implications for public health practice?

The study results will help policymakers and health managers to identify the vulnerable districts, plan further investigations, help to improve proper resource allocation, and improve health interventions.

  • 1. School of Public Health and Social Work, Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, Queensland, Australia
  • 2. Infectious Diseases Division, International Centre for Diarrhoeal Disease Research, Bangladesh, Mohakhali, Dhaka, Bangladesh
  • 3. Shanghai Children’s Medical Centre, School of Medicine, Shanghai Jiao Tong University, Shanghai, China
  • 4. School of Public Health, Institute of Environment and Population Health, Anhui Medical University, Hefei, Anhui, China
  • 5. Center for Global Health, School of Public Health, Nanjing Medical University, Nanjing, Jiangsu, China
  • 6. Health System and Population Studies Division, International Centre for Diarrhoeal Disease Research, Bangladesh, Mohakhali, Dhaka, Bangladesh
  • Corresponding author:

    Wenbiao Hu, w2.hu@qut.edu.au

  • Funding: The Queensland University of Technology Postgraduate Research Award and Queensland University of Technology Higher Degree Research International Tuition Fee Sponsorship
  • Online Date: July 16 2021
    Issue Date: July 16 2021
    doi: 10.46234/ccdcw2021.161
  • Pneumonia is one of the leading causes of mortality and morbidity in children aged under five years in Bangladesh. This study aimed to identify the association between weather, social factors and childhood pneumonia and identify the spatial variation of the disease. A Bayesian spatial Poisson regression model with a conditional autoregressive prior structure was developed to quantify the association between childhood pneumonia and socioecological factors and identify the spatial variation. The study results suggested that a 1 °C increase in monthly temperature and monthly temperature variation may increase the monthly associated log relative risk (RR) of childhood pneumonia by 1.161 [95% credible interval (CrI): 1.013−1.429] and 1.463 (95% CrI: 1.170−1.839), respectively. However, the population density was inversely related with pneumonia risk (RR: 0.996, CrI: 0.994−0.998). Socioecological factors may influence the spatial pattern of childhood pneumonia, and the spatial random effects were heterogeneous.

    The study was conducted in Bangladesh, which is located in the northeastern part of South Asia. Bangladesh is divided into 8 administrative divisions and 64 districts. Monthly data on under-5-years pneumonia were extracted from the District Health Information System Version 2 of the Directorate General of Health Services (DGHS) under the Ministry of Health and Family Welfare of Bangladesh from January 2012 to December 2016 (1). The pneumonia cases were diagnosed according to the World Health Organization pneumonia guidelines (2). The under-five-years population data at the district level were collected from the latest national population and household census (3). The sociodemographic data (percentage of education and internet use) at the district level were collected from socioeconomic and demographic reports (national series, volume-4) from the same census. The poverty data for each district was obtained from the Household Income and Expenditure Survey 2016 (4).

    Climate data (temperature and rainfall) were obtained from the National Environmental Satellite, Data and Information service (https://www7.ncdc.noaa.gov/CDO/cdoselect.cmd?datasetabbv=GSOD), which is publicly available and widely used in previous studies (5-6). Poisson regression models in a Bayesian framework were developed for pneumonia cases at the district level. These models assume that the observed counts of childhood pneumonia cases (Ok) for the kth district (k=1···64) follow a Poisson distribution with mean µk:

    $${{\rm{O}}_k} \sim {\rm{ Poisson }}\left( {{{\rm{\mu }}_k}} \right)$$ (1)
    $${\rm{log }}\left( {{{\rm{\mu }}_k}} \right) = {\rm{ log }}\left( {{{\rm{E}}_k}} \right) + {{\rm{\theta }}_k}$$ (2)

    where Ek (the expected number of cases in Districtk) is an offset to control population size and θk is the associated log RR.

    Prior to this analysis, we examined multicollinearity among the different covariates but did not find sufficiently strong associations to warrant exclusion or other treatment of any variables (Supplementary Table S1). As a consequence, a total of 6 models were developed (Supplementary Material). The model which incorporated all socioecological covariates with both structured and unstructured random effects were selected for the final analysis.

    Variables12345678
    1Pneumonia
    2Temperature0.094
    3Temperature variation0.1610.235
    4Rainfall−0.0190.268*−0.146
    5Education0.008−0.011−0.1290.068
    6Internet use0.1260.0630.2230.017−0.066
    7Population density−0.148−0.25*−0.276*0.0280.0750.162
    8Poverty0.069−0.140.163−0.0770.095−0.209−0.336*
    Note: − represnt its pneumonia itself, there will be no number.
    * P<0.05.

    Table S1.  Spearman correlation between pneumonia and socioecological covariates in Children Under Five-Years Old — Bangladesh, 2012−2016.

    The expected log relative risk θk was represented as follows:

    $$\begin{aligned} {{\rm{\theta }}_k} = & {\rm{ \alpha }} + \left( {{\rm{Tem}}{{\rm{p}}_k}} \right){{\rm{\beta }}_1} + \left( {{\rm{Tempv}}{{\rm{a}}_k}} \right){{\rm{\beta }}_2} + \left( {{\rm{Rai}}{{\rm{n}}_k}} \right){{\rm{\beta }}_3} \\ &+ \left( {{\rm{Ed}}{{\rm{u}}_k}} \right){{\rm{\beta }}_4}+ \left( {{\rm{In}}{{\rm{t}}_k}} \right){{\rm{\beta }}_5} + \left( {{\rm{pov}}{{\rm{i}}_k}} \right){{\rm{\beta }}_6}\\ & + \left( {{\rm{po}}{{\rm{p}}_k}} \right){{\rm{\beta }}_7} + {{\rm{u}}_k} + {{\rm{v}}_k} \end{aligned} $$

    where α is a constant; β1 is the coefficient for temperature, β2 is the coefficient for temperature variation, β3 is the coefficient for rainfall, β4 is the coefficient for percentage of education at the district level, β5 is the coefficient for percentage of internet user at the district level, β6 is percentage of poverty at the district level, and β7 is the population density per square kilometer; vk is a spatially unstructured random effect that is assumed to be normally distributed with mean zero and variance σv2 and uk is the spatially structured random effect that was modeled using a conditional autoregressive (CAR) prior $ {{\rm{u}}_k} \sim $$ {\rm{ N}}\left( {{{\bar {\rm{u}}}_{ \sim k}},{\rm{ }}{{\rm{\sigma }}_u}^2/{{\rm{n}}_k}} \right)$, where ~k denotes the neighbors of the kth district based on a simple adjacency matrix and nk is the corresponding number of neighbors (7). WinBUGS software (version 1.4.3, MRC Biostatistics Unit, Cambridge, and Imperial College School of Medicine, London) was used to fit the Bayesian Poisson regression models. In the Markov chain Monte Carlo analysis, a 30,000 iteration “burn-in” was followed by 100,000 iteration sample collection. In every case, the Monte Carlo error was <5% of the overall standard deviation, indicating sufficient iterations of the model had been run after convergence. Model comparison was performed using the Deviance Information Criterion (DIC). Best fitted models were indicated by smaller DIC values (8).

    The mean monthly number of pneumonia cases in children <5 years was 747.82. The mean monthly temperature, temperature variation, and rainfall were 30.97 °C, 3.63 °C, and 164.54 mm, respectively. Among the social factors, the mean percentages of education, internet use, poverty (per 100 population) and <5 years children density (per square kilometer) were 54.66%, 0.62%, 24.45%, and 117.72, respectively, at the district level (Supplementary Table S2). The average monthly temperature was higher in the western region, while the monthly temperature variation was higher in most of the hilly areas located in the southern part of the country and two districts (Bhola and Pirojpur) of the coastal region of Bangladesh. The distributions of higher monthly average rainfall were scattered in different regions. Inclusion of spatial autocorrelation in the model was important. The model which included both structured and unstructured random effects had the smallest DIC (665.47 and 13,773.00 for models with and without random effects, respectively) (Supplementary Table S3). The highest RRs were observed in the southeastern part (Rangamati district) and southern part (Pirojpur district) of the country (Figure 1). Supplementary Table S4) shows the list of districts with the higher RR.

    VariablesMean ± SDRange
    Pneumonia747.82 ± 245.32355.53−1612.10
    Temperature (°C)30.97 ± 0.4829.50−31.98
    Temperature variation (°C)3.63 ± 0.552.00−4.99
    Rain (mm)164.54 ± 101.373.13−386.96
    Education (%)54.66 ± 7.7537.50−73.70
    Internet use (%)0.62 ± 0.760.14−6.03
    Poverty incidence (%)27.45 ± 15.312.60−70.80
    Under five years population density (per square km)117.72 ± 87.0510.81−656.54

    Table S2.  Descriptive statistics of childhood pneumonia and different socioecological factors — Bangladesh, 2012−2016.

    ModelRandom effectDeviance Information Criterion (DIC)Effective number of parameters (pD)
    Model INo 19954.2030.146
    Model IINo17382.415.001
    Model IIINo13773.0011.134
    Model IVu665.9863.940
    Model Vv665.6763.812
    Model VIu and v665.4763.719

    Table S3.  Model comparison for relative risk of monthly childhood pneumonia, underlying socioecological factors, and different random effects — Bangladesh, 2012−2016.

    Name of the districtRelative Risk (95% Credible interval)Location
    Rangamati5.97 (5.63−6.31)South-eastern
    Pirojpur4.71 (4.48−4.93)South-western
    Jhalkathi4.38 (4.09−4.66)South-western
    Jaipurhut3.95 (3.70−4.19)North-eastern
    Bandarbon3.77 (3.48−4.07)South-eastern
    Meherpur3.50 (3.23−3.78)South-western
    Rajbari3.31(3.11−3.50)Central
    Khagrachari3.25 (3.02−3.49)South-eastern
    Panchagarh2.96 (2.78−3.14)Northern

    Table S4.  List of high-risk districts of Bangladesh for childhood pneumonia from 2012 to 2016.

    Figure 1. 

    Posterior estimated Relative Risk of childhood pneumonia at the district level of Bangladesh from 2012 to 2016.

    Figure 2 depicts the distribution of spatial random effects (structured heterogeneity) of pneumonia in Bangladesh. The districts with darker color (red color) had relatively high spatial variation. These districts with high spatial variation might have some unknown factors that may have had effects on the incidence of childhood pneumonia but that we did not consider in the models (e.g., incomplete measurement of variables, lack of geocoding, and generalization of geographic features).

    Figure 2. 

    Spatial random effects of childhood pneumonia — Bangladesh, 2012−2016.

    Our study results suggested that a rise of 1 °C average monthly temperature and temperature variation was associated with RR estimates of childhood pneumonia of 1.161 (95% CrI: 1.012–1.428) and 1.463 (95% CrI: 1.169–1.838), respectively. The density of children under five years in population was negatively associated with pneumonia (RR: 0.996, 95% CrI: 0.994–0.998) (Table 1). Additionally, no significant associations were found between childhood pneumonia and rainfall, education, internet use, or poverty since the corresponding 95% CrIs for the RR of each factor included 1.

    VariablesCrude RR (95% CrI)Adjusted RR (95% CrI)
    Temperature*1.146 (0.929−1.432)1.161 (1.013−1.429)
    Temperature1.730 (1.694−1.763)1.529 (1.503−1.555)
    Temperature variability*1.821 (1.376−2.491)1.463 (1.170−1.839)
    Temperature variability1.623 (1.596−1.649)1.421 (1.395−1.447)
    Rainfall*1.000 (0.999−1.002)1.000 (0.999−1.002)
    Rainfall1.0007 (1.0006−1.0007)1.0001 (1.0000−1.0002)
    Population density*0.995 (0.994−0.997)0.996 (0.994−0.998)
    Population density0.9943 (0.9942−0.9945)0.9959 (0.9958−0.9961)
    Education*0.979 (0.956−1.004)0.986 (0.969−1.005)
    Education0.978 (0.977−0.979)0.9853 (0.984−0.986)
    Poverty*1.008 (0.996−1.020)1.000 (0.988−1.010)
    Poverty1.009 (1.009−1.010)1.002 (1.002−1.003)
    Internet*0.844 (0.661−1.070)0.916 (0.743−1.126)
    Internet0.874 (0.862−0.885)0.929 (0.916−0.942)
    Abbreviations: RR=relative risk; CrI=credible interval.
    * with heterogeneity (u and v).
    without heterogeneity (u and v).

    Table 1.  Crude and adjusted RR of different socioecological factors in Children Under Five-Years Old — Bangladesh, 2012−2016.

  • In young children, the thermoregulation system is not yet matured and makes the children more vulnerable to temperature variation. This study describes the spatial pattern of childhood pneumonia and their socioecological factors in Bangladesh. Identifying the spatial variation of childhood pneumonia and important socioecological determinants can help target high-risk communities with evidence-based effective preventative measures.

    Mapping of the spatially structured random effects indicated the spatial variation after controlling socioecological factors and spatial autocorrelation in the model. The Bayesian CAR model included unknown parameters as random effects, which incorporated the spatially correlated random effects (9). This approach can account for the residual variability resulting from spatial variation in effects that were not included in the models. The districts containing higher spatial random effects or variation may have some other risk factors remaining after adjustment of socioecological factors and spatial correlation.

    This study was subject to some limitations. First, in this study, we used data from monthly reports of the DGHS. This represented the number of patients that attended different levels of health facilities in Bangladesh for pneumonia treatment. However, there might be some patients in the community who did not attend any health facilities and who received treatment from village doctors or spiritual healers, especially in the rural areas. Therefore, there was a chance of measurement and information biases. Second, the unit of analysis was at the group level rather than at the individual level, so the results may be prone to the ecological fallacy.

    The findings of this study could help policymakers better understand that childhood pneumonia has a heterogeneous spatial pattern and that socioecological factors may play a significant role in describing this pattern.

    Acknowledgements: The Management Information System, The Director General of Health Services, Ministry of Health, and Family Welfare, Bangladesh.

Reference (9)

Citation:

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return