Considering spatial heterogeneity in the distributed lag non-linear model when analyzing spatiotemporal data

Chien, Lung-Chang, Guo, Yuming, Li, Xiao and Yu, Hwa-Lung (2016) Considering spatial heterogeneity in the distributed lag non-linear model when analyzing spatiotemporal data. Journal of Exposure Science and Environmental Epidemiology, 28 1: 13-20. doi:10.1038/jes.2016.62


Author Chien, Lung-Chang
Guo, Yuming
Li, Xiao
Yu, Hwa-Lung
Title Considering spatial heterogeneity in the distributed lag non-linear model when analyzing spatiotemporal data
Journal name Journal of Exposure Science and Environmental Epidemiology   Check publisher's open access policy
ISSN 1559-0631
1559-064X
Publication date 2016-11-16
Sub-type Article (original research)
DOI 10.1038/jes.2016.62
Open Access Status Not yet assessed
Volume 28
Issue 1
Start page 13
End page 20
Total pages 8
Place of publication London, United Kingdom
Publisher Nature Publishing Group
Language eng
Subject 2713 Epidemiology
3005 Toxicology
2310 Pollution
2739 Public Health, Environmental and Occupational Health
Abstract The distributed lag non-linear (DLNM) model has been frequently used in time series environmental health research. However, its functionality for assessing spatial heterogeneity is still restricted, especially in analyzing spatiotemporal data. This study proposed a solution to take a spatial function into account in the DLNM, and compared the influence with and without considering spatial heterogeneity in a case study. This research applied the DLNM to investigate non-linear lag effect up to 7 days in a case study about the spatiotemporal impact of fine particulate matter (PM 2.5) on preschool children's acute respiratory infection in 41 districts of northern Taiwan during 2005 to 2007. We applied two spatiotemporal methods to impute missing air pollutant data, and included the Markov random fields to analyze district boundary data in the DLNM. When analyzing the original data without a spatial function, the overall PM 2.5 effect accumulated from all lag-specific effects had a slight variation at smaller PM 2.5 measurements, but eventually decreased to relative risk significantly <1 when PM 2.5 increased. While analyzing spatiotemporal imputed data without a spatial function, the overall PM 2.5 effect did not decrease but increased in monotone as PM 2.5 increased over 20 μg/m 3. After adding a spatial function in the DLNM, spatiotemporal imputed data conducted similar results compared with the overall effect from the original data. Moreover, the spatial function showed a clear and uneven pattern in Taipei, revealing that preschool children living in 31 districts of Taipei were vulnerable to acute respiratory infection. Our findings suggest the necessity of including a spatial function in the DLNM to make a spatiotemporal analysis available and to conduct more reliable and explainable research. This study also revealed the analytical impact if spatial heterogeneity is ignored.
Formatted abstract
The distributed lag non-linear (DLNM) model has been frequently used in time series environmental health research. However, its functionality for assessing spatial heterogeneity is still restricted, especially in analyzing spatiotemporal data. This study proposed a solution to take a spatial function into account in the DLNM, and compared the influence with and without considering spatial heterogeneity in a case study. This research applied the DLNM to investigate non-linear lag effect up to 7 days in a case study about the spatiotemporal impact of fine particulate matter (PM2.5) on preschool children’s acute respiratory infection in 41 districts of northern Taiwan during 2005 to 2007. We applied two spatiotemporal methods to impute missing air pollutant data, and included the Markov random fields to analyze district boundary data in the DLNM. When analyzing the original data without a spatial function, the overall PM2.5 effect accumulated from all lag-specific effects had a slight variation at smaller PM2.5 measurements, but eventually decreased to relative risk significantly <1 when PM2.5 increased. While analyzing spatiotemporal imputed data without a spatial function, the overall PM2.5 effect did not decrease but increased in monotone as PM2.5 increased over 20 μg/m3. After adding a spatial function in the DLNM, spatiotemporal imputed data conducted similar results compared with the overall effect from the original data. Moreover, the spatial function showed a clear and uneven pattern in Taipei, revealing that preschool children living in 31 districts of Taipei were vulnerable to acute respiratory infection. Our findings suggest the necessity of including a spatial function in the DLNM to make a spatiotemporal analysis available and to conduct more reliable and explainable research. This study also revealed the analytical impact if spatial heterogeneity is ignored.
Keyword Distributed lag non-linear model
Spatial function
Spatial heterogeneity
Q-Index Code C1
Q-Index Status Provisional Code
Institutional Status UQ

Document type: Journal Article
Sub-type: Article (original research)
Collections: HERDC Pre-Audit
School of Public Health Publications
 
Versions
Version Filter Type
Citation counts: Scopus Citation Count Cited 0 times in Scopus Article
Google Scholar Search Google Scholar
Created: Sat, 19 Nov 2016, 03:54:05 EST by Yuming Guo on behalf of School of Public Health