Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

From spatial ecology to spatial epidemiology: Modeling spatial distributions of different cancer types with principal coordinates of neighbor matrices

Voutilainen, A and Tolppanen, AM and Vehviläinen-Julkunen, K and Sherwood, PR (2014) From spatial ecology to spatial epidemiology: Modeling spatial distributions of different cancer types with principal coordinates of neighbor matrices. Emerging Themes in Epidemiology, 11 (1).

[img]
Preview
PDF
Published Version
Available under License : See the attached license file.

Download (1MB) | Preview
[img] Plain Text (licence)
Available under License : See the attached license file.

Download (1kB)

Abstract

Epidemiology and ecology share many fundamental research questions. Here we describe how principal coordinates of neighbor matrices (PCNM), a method from spatial ecology, can be applied to spatial epidemiology. PCNM is based on geographical distances among sites and can be applied to any set of sites providing a good coverage of a study area. In the present study, PCNM eigenvectors corresponding to positive autocorrelation were used as explanatory variables in linear regressions to model incidences of eight most common cancer types in Finnish municipalities (n = 320). The dataset was provided by the Finnish Cancer Registry and it included altogether 615,839 cases between 1953 and 2010. Results: PCNM resulted in 165 vectors with a positive eigenvalue. The first PCNM vector corresponded to the wavelength of hundreds of kilometers as it contrasted two main subareas so that municipalities located in southwestern Finland had the highest positive site scores and those located in midwestern Finland had the highest negative scores in that vector. Correspondingly, the 165thPCNM vector indicated variation mainly between the two small municipalities located in South Finland. The vectors explained 13 - 58% of the spatial variation in cancer incidences. The number of outliers having standardized residual > |3| was very low, one to six per model, and even lower, zero to two per model, according to Chauvenet's criterion. The spatial variation of prostate cancer was best captured (adjusted r 2= 0.579). Conclusions: PCNM can act as a complementary method to causal modeling to achieve a better understanding of the spatial structure of both the response and explanatory variables, and to assess the spatial importance of unmeasured explanatory factors. PCNM vectors can be used as proxies for demographics and causative agents to deal with autocorrelation, multicollinearity, and confounding variables. PCNM may help to extend spatial epidemiology to areas with limited availability of registers, improve cost-effectiveness, and aid in identifying unknown causative agents, and predict future trends in disease distributions and incidences. A large advantage of using PCNM is that it can create statistically valid reflectors of real predictors for disease incidence models with only little resources and background information.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: Article
Status: Published
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Voutilainen, A
Tolppanen, AM
Vehviläinen-Julkunen, K
Sherwood, PRprs11@pitt.eduPRS11
Date: 8 August 2014
Date Type: Publication
Journal or Publication Title: Emerging Themes in Epidemiology
Volume: 11
Number: 1
DOI or Unique Handle: 10.1186/1742-7622-11-11
Schools and Programs: School of Nursing > Nursing
Refereed: Yes
Date Deposited: 19 Sep 2016 15:56
Last Modified: 22 Jun 2021 11:55
URI: http://d-scholarship.pitt.edu/id/eprint/29502

Metrics

Monthly Views for the past 3 years

Plum Analytics

Altmetric.com


Actions (login required)

View Item View Item