Premium
A spatial structural equation modelling framework for health count responses
Author(s) -
Congdon Peter,
Almog Michael,
Curtis Sarah,
Ellerman Raymond
Publication year - 2007
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.2921
Subject(s) - construct (python library) , structural equation modeling , set (abstract data type) , poisson regression , residual , econometrics , referral , service (business) , computer science , statistics , medicine , mathematics , environmental health , economics , population , algorithm , economy , family medicine , programming language
A structural equation model is proposed for the impact on area health referral counts of spatially correlated latent constructs. One type of construct is indicator based and represents the underlying morbidity or health need; such constructs are derived in a normal errors measurement model involving a set of observed socio‐economic indicators. Another set of residual constructs represents particularities of service configuration or spatially correlated risks that cannot be proxied by observed indicators. The structural model relates the referral outcomes to both types of construct in a Poisson regression. While multiple spatial factors based on measured socio‐economic indicators have already been proposed, the extension to include both indicator‐based and residual factors is novel. Further novel features are to represent spatially structured heterogeneity and nonlinearity in the impact of the indicator‐based constructs. A case study considers the modelling of a single indicator‐based morbidity construct (taken to represent a need for care) in 62 New York counties, with responses being psychiatric referrals to ambulatory and hospital care. Copyright © 2007 John Wiley & Sons, Ltd.